Original Research

Take-Home Points

• The incidence of geriatric hip fractures is rising nationally.
• Costs associated with hip fracture care have risen significantly.
• CMN and SHS are effective for stable, intertrochanteric hip fractures.
• Current trends show increased utilization of CMN compared to SHS for stable introchanteric hip fractures.
• Surgeon awareness of implant cost is a critical factor in delivering cost-effective, evidence-based surgical care.

The continuing increase in the population of patients older than 65 years in the United States is well known. For orthopedic surgeons, this trend highlights the importance of effective geriatric fracture care, particularly hip fracture care. Hip fractures in the elderly are expected to increase 50% by 2025 and to number 500,000 by 2040.¹ The growing burden of hip fracture cases is accompanied by increasing costs of care. In 2012, 88% of the healthcare dollars spent on these patients were for direct fracture care, a significant increase from 60% in 2009.² Although fewer than 1 in 5 fractures in the elderly are hip fractures, these injuries account for 72% of the total cost of geriatric fracture care, more than the total cost of all other osteoporosis-related injuries combined.¹ Currently, the direct cost of hip fracture care ranges from $8358 to $32,195 and is expected, in total, to reach $25 billion by 2025.^2,3

About 50% of geriatric hip fractures are extracapsular intertrochanteric or pertrochanteric.⁴ Several studies have compared sliding hip screws (SHSs) with cephalomedullary nails (CMNs) in the effective management of stable intertrochanteric fractures.^5-11 Although these implants have shown an increased risk for peri-implant fracture and subsequent reoperation, markers such as mortality, medical complications, and restoration of prefracture function have all been equivocal relative to SHSs.¹² Ultimately, one implant cannot be definitively recommended over the other for management of stable intertrochanteric hip fractures.^13,14 Nevertheless, the current trend increasingly favors CMNs over SHSs.^4,15 Most orthopedic surgeons are unaware of or underestimate the cost difference between these implants—a fact even more pronounced for newer implants.^4,16 Considering the ever growing cost burden of hip fractures in the United States, orthopedists must consider not only the efficacy of care but the cost of delivery.

We conducted a study to determine the effect that surgeon knowledge of implant cost had on rates of use of SHSs and CMNs in the management of stable intertrochanteric hip fractures. 

Patients and Methods

On May 1, 2012, all 9 attending orthopedic surgeons in a private practice group serving a suburban level II trauma center met to discuss implant prices and implant-related costs for the $850 Versafix SHS, the $1950 short Gamma3 nail (SGN), and the $2900 long Gamma3 nail (LGN), all manufactured by Stryker. All surgeons denied previous knowledge of the costs of these implants. During the discussion, no particular implant was recommended for management of any specific type of fracture. Surgeons were not directly instructed to consider implant cost in subsequent hip fracture surgeries and were not informed of our upcoming study of implant utilization.

After obtaining Institutional Review Board approval, we performed a retrospective chart and radiologic review of all hip fractures (Current Procedural Terminology [CPT] code 27244 or 27245) managed with fixation at our institution between May 1, 2011 and April 30, 2013. Two hundred six patients were identified (Figure 1).

One year later, surgeons were again shown their respective hip fracture radiographs, with patient identifying data removed. They were asked to reclassify their respective cases using the OTA system and then indicate the implant they would use for operative fixation in each of their cases.

Patient age, sex, injury side, fracture types, and utilization rates of the SHS, SGN, and LGN implants were compared between the groups. For each eligible patient, implant cost and other financial data were obtained from the hospital’s finance department. Statistical analyses were performed with SPSS (Statistical Package for the Social Sciences) Version 20 for Macintosh. Independent 2-sample t test was used for parametric comparisons, and Fisher exact test was used for nonparametric comparisons.

Results

Examination of implant use per fracture classification revealed an interesting change. In the before group, SHS was the implant most commonly used for 31-A1.1 fractures (7/16, 43.8%), 31-A1.2 fractures (8/18, 44.4%), and 31-A2.1 fractures (10/25, 40.0%), and LGN was used in 66.7% (8/12) of 31-A1.3 fractures. By contrast, in the after group, SHS was most commonly used only for 31-A1.2 fractures (7/12, 58.3%), SGN was used in 90% (9/10) of 31-A1.1 fractures, and LGN was used in 42.1% (8/19) of 31-A2.1 fractures. In addition, 85.7% (6/7) of 31-A1.3 fractures were managed with a version of the Gamma nail.

Reclassification resulted in more A2.1 fractures (42.0% vs 37.0%) and fewer A1.3 fractures (10.1% vs 16.0%). About the same numbers of fractures were classified A1.1 (21.0% vs 21.8%) and A1.2 (26.9% vs 25.2%). SHS was favored for A1.1 fractures (92.0%) and A1.2 fractures (65.6%). SGN was favored for A1.3 fractures (75.0%). Gamma nails of both sizes were favored for A2.1 fractures (88.0%).

Discussion

Comparisons of SHS/plate and CMN constructs in the management of stable intertrochanteric hip fractures have long been discussed in the orthopedic literature. The major concern with CMNs (vs SHSs) is a statistically significantly higher rate of revision surgery, most often for peri-implant fracture. Rates of previous revision surgery for peri- implant fracture have ranged from 2.4% to 6% for CMNs and from 0.6% to 4% for SHSs.^5-7,9 In a Cochrane review of 22 studies (3749 patients), Parker and Handoll¹² compared CMN and SHS outcomes in 23 categories and found a statistically significant difference only in postoperative fracture rate. However, in a meta-analysis of studies conducted between 2000 and 2005, Bhandari and colleagues⁸ found no statistically significant difference in risk of femoral shaft fracture between CMNs (0.6%) and SHSs (0.1%). In addition, Utrilla and colleagues¹⁰ reported no postoperative fractures with use of Gamma3 CMNs. These recent studies may indicate that newer CMN designs can reduce the incidence of postoperative peri-implant fracture.^8,10 Other outcome measures, such as 1-year mortality, functional outcome, and medical complication rate, have shown no statistically significant differences between the 2 implants.^10-12 Ultimately, the current recommendation for fixation of stable intertrochanteric hip fractures is either SHS or CMN.^13,14

Of our study patients, 78.9% (before group) and 64.6% (after group) were female, and 49.3% (before group) and 47.9% (after group) were between 80 and 89 years of age. Similarly, a review of hip fracture Medicare claims made between 1999 and 2002 revealed that >75% of the patients were females and 48% to 49% were octogenarians.^4,18 However, our rates of different fracture types differed from those of Adams and colleagues.⁵ In a 1-year single-institution study, they found that, for both CMNs and SHSs, the most common stable intertrochanteric fractures were 31-A1.1 fractures; in our study’s before and after groups, more than one-third of injuries were 31-A2.1 fractures. Least common were 31-A1.3 fractures, both in the study by Adams and colleagues⁵ and in our before (16.9%) and after (14.6%) groups. Although our fracture rates differ from those of previous studies, all 4 classification categories fall under the umbrella of stable intertrochanteric hip fracture, which is the sole focus of this study.¹⁴ 

We hypothesized that cost would be a significant driver of implant choice in the management of these injuries. Given that SHS costs $1186.91 less than SGN and $1625.88 less than LGN at our institution, we expected that the before- discussion group’s overall SHS use rate of 38.0% would increase after discussion. Instead, SGN became the favored implant, with use in almost half of all fractures in the after group. Although the change in overall implant use rate was notable, these findings were not statistically significant. Examination of individual fracture patterns revealed 2 areas of interest. First, SHS was assumed to be the implant of choice in the management of the relatively simple 31-A1.1 fractures. Although this assumption was verified in the before group (SHS use in 43.8% of fractures), SGN was used in almost every case (90%) in the after group. However, when surgeons were asked 1 year later to recommend an implant for A1.1 fractures, 92% suggested SHS. The more cost-effective SHS construct may be the most amenable for use in these injury types given all intertrochanteric hip fracture patterns, though this has not been studied.

On the other hand, for 31-A2.1 fractures, perhaps the most complicated of the stable patterns, LGN became the implant of choice (42.1%). Despite surgeons’ awareness of the cost differences, management of these fractures shifted in the after group to the most expensive implant, indicative of surgeon concern about eventual loss of reduction with SHS and surgeon comfort with a particular procedure. This trend held when surgeons were asked to reclassify fractures 1 year later, as CMNs were recommended for 88% of 31-A2.1 fractures. Although both SHS and CMN were acceptable in 97% of the fractures included in this study, SGNs or LGNs were preferred for almost every fracture pattern involving the lesser trochanter. All 9 attending surgeons described involvement of the lesser trochanter as an indicator of posteromedial calcar injury. Surgeons became particularly concerned when this fracture pattern occurred in patients with significant osteopenia; SHS fixation, in their opinion, would be poor in the setting of a combination of greater posteromedial instability and poor bone quality. In a level I prospective, randomized trial, Barton and colleagues⁷ found no difference in outcomes between LGN and SHS fixation for 31-A2 proximal femur fractures and recommended SHS implants for the cost savings. In the clinical experience of this group, however, A1.3 and A2.1 fractures were at especially high risk for failure with SHS use, which necessitated greater implant stability through CMN fixation. On the other hand, for simpler fracture patterns, most surgeons suggested SHS implants. In their opinion, SGN and LGN implants offered no additional benefit of stability without evidence of posteromedial injury, even in the setting of osteopenia. For A1.2 fractures, posteromedial involvement was judged on the basis of size of the inferomedial spike or the extent of the inferomedial fracture line. Two surgeons preferred CMN for simple fractures, one because of the increased comfort with the implants and the other because of the minimally invasive surgical technique. Overall, our results indicate that knowledge of implant cost is not a strong enough factor to change surgeon behavior in selecting fixation for uncomplicated stable intertrochanteric hip fractures in previously ambulatory elderly patients. 

The lack of effect could be a consequence of surgeons’ training and comfort with various implants, especially among younger attending surgeons. Most of the attendings in the practice are under age 50 years, which correlates with a preference for CMN fixation.¹⁹ Case loads of >80 intertrochanteric hip fractures per calendar year, as in the after group, also correlates with more CMN use.¹⁹ However, the before group had more intertrochanteric hip fractures, and yet SHS was the implant of choice. Resident physician experience and comfort with various implants may play a role too, as teaching hospitals with resident assistance also correlate with CMN use.¹⁹ However, no major change in resident physician involvement was undertaken during this period. The institution studied is near a major metropolis in the Northeast, a region that has disfavored SHS in recent years.¹⁸ The change from before to after fits an overall trend in changing implant use. Anglen and colleagues¹⁵ found a significant decrease in SHS use, from 97% in 1999 to 33% in 2006, for intertrochanteric fracture fixation. Simultaneously, CMN use increased from 3% to 67%.

This study had several limitations. First, its overall sample size was small, and therefore any data fluctuations may be exaggerated. Furthermore, changes in utilization rates were compared over 2 years, which may not be long enough to show a changing trend in implant selection. Post hoc analysis of the sample size determined a power of 0.76 for an α of 0.05 and an effect size of 0.50. Second, radiologic classification was performed in a retrospective review, not officially by the operative surgeon. Fractures that we considered stable may have been considered unstable by the operative surgeon, influencing implant selection. Third, patients were selected from only one hospital, and orthopedic surgeons from other institutions may be more sensitive to cost considerations, changing implant selection more quickly. Fourth, initial selection of patients by CPT code might not have captured all those who satisfied the inclusion criteria. Fifth, only a single intervention was used, and follow-up meetings certainly could have increased the effectiveness of the intervention. Last, this and other retrospective studies are inherently weaker because of possible bias.

Conclusion

Our study results showed that implant cost is not a significant factor in implant selection for uncomplicated stable intertrochanteric hip fractures in previously ambulatory elderly patients. By itself, knowledge of implant cost may not be a strong enough force to change surgeon behavior or preference secondary to consequences of failure or comfort with particular implants. In an economic climate in which healthcare is scrutinized for both its medical effectiveness and cost-effectiveness, further study of forces that could influence orthopedic surgeons to select a more cost-effective implant is warranted.

Take-Home Points

• The incidence of geriatric hip fractures is rising nationally.
• Costs associated with hip fracture care have risen significantly.
• CMN and SHS are effective for stable, intertrochanteric hip fractures.
• Current trends show increased utilization of CMN compared to SHS for stable introchanteric hip fractures.
• Surgeon awareness of implant cost is a critical factor in delivering cost-effective, evidence-based surgical care.

The continuing increase in the population of patients older than 65 years in the United States is well known. For orthopedic surgeons, this trend highlights the importance of effective geriatric fracture care, particularly hip fracture care. Hip fractures in the elderly are expected to increase 50% by 2025 and to number 500,000 by 2040.¹ The growing burden of hip fracture cases is accompanied by increasing costs of care. In 2012, 88% of the healthcare dollars spent on these patients were for direct fracture care, a significant increase from 60% in 2009.² Although fewer than 1 in 5 fractures in the elderly are hip fractures, these injuries account for 72% of the total cost of geriatric fracture care, more than the total cost of all other osteoporosis-related injuries combined.¹ Currently, the direct cost of hip fracture care ranges from $8358 to $32,195 and is expected, in total, to reach $25 billion by 2025.^2,3

About 50% of geriatric hip fractures are extracapsular intertrochanteric or pertrochanteric.⁴ Several studies have compared sliding hip screws (SHSs) with cephalomedullary nails (CMNs) in the effective management of stable intertrochanteric fractures.^5-11 Although these implants have shown an increased risk for peri-implant fracture and subsequent reoperation, markers such as mortality, medical complications, and restoration of prefracture function have all been equivocal relative to SHSs.¹² Ultimately, one implant cannot be definitively recommended over the other for management of stable intertrochanteric hip fractures.^13,14 Nevertheless, the current trend increasingly favors CMNs over SHSs.^4,15 Most orthopedic surgeons are unaware of or underestimate the cost difference between these implants—a fact even more pronounced for newer implants.^4,16 Considering the ever growing cost burden of hip fractures in the United States, orthopedists must consider not only the efficacy of care but the cost of delivery.

We conducted a study to determine the effect that surgeon knowledge of implant cost had on rates of use of SHSs and CMNs in the management of stable intertrochanteric hip fractures. 

Patients and Methods

On May 1, 2012, all 9 attending orthopedic surgeons in a private practice group serving a suburban level II trauma center met to discuss implant prices and implant-related costs for the $850 Versafix SHS, the $1950 short Gamma3 nail (SGN), and the $2900 long Gamma3 nail (LGN), all manufactured by Stryker. All surgeons denied previous knowledge of the costs of these implants. During the discussion, no particular implant was recommended for management of any specific type of fracture. Surgeons were not directly instructed to consider implant cost in subsequent hip fracture surgeries and were not informed of our upcoming study of implant utilization.

After obtaining Institutional Review Board approval, we performed a retrospective chart and radiologic review of all hip fractures (Current Procedural Terminology [CPT] code 27244 or 27245) managed with fixation at our institution between May 1, 2011 and April 30, 2013. Two hundred six patients were identified (Figure 1).

One year later, surgeons were again shown their respective hip fracture radiographs, with patient identifying data removed. They were asked to reclassify their respective cases using the OTA system and then indicate the implant they would use for operative fixation in each of their cases.

Patient age, sex, injury side, fracture types, and utilization rates of the SHS, SGN, and LGN implants were compared between the groups. For each eligible patient, implant cost and other financial data were obtained from the hospital’s finance department. Statistical analyses were performed with SPSS (Statistical Package for the Social Sciences) Version 20 for Macintosh. Independent 2-sample t test was used for parametric comparisons, and Fisher exact test was used for nonparametric comparisons.

Results

Examination of implant use per fracture classification revealed an interesting change. In the before group, SHS was the implant most commonly used for 31-A1.1 fractures (7/16, 43.8%), 31-A1.2 fractures (8/18, 44.4%), and 31-A2.1 fractures (10/25, 40.0%), and LGN was used in 66.7% (8/12) of 31-A1.3 fractures. By contrast, in the after group, SHS was most commonly used only for 31-A1.2 fractures (7/12, 58.3%), SGN was used in 90% (9/10) of 31-A1.1 fractures, and LGN was used in 42.1% (8/19) of 31-A2.1 fractures. In addition, 85.7% (6/7) of 31-A1.3 fractures were managed with a version of the Gamma nail.

Reclassification resulted in more A2.1 fractures (42.0% vs 37.0%) and fewer A1.3 fractures (10.1% vs 16.0%). About the same numbers of fractures were classified A1.1 (21.0% vs 21.8%) and A1.2 (26.9% vs 25.2%). SHS was favored for A1.1 fractures (92.0%) and A1.2 fractures (65.6%). SGN was favored for A1.3 fractures (75.0%). Gamma nails of both sizes were favored for A2.1 fractures (88.0%).

Discussion

Comparisons of SHS/plate and CMN constructs in the management of stable intertrochanteric hip fractures have long been discussed in the orthopedic literature. The major concern with CMNs (vs SHSs) is a statistically significantly higher rate of revision surgery, most often for peri-implant fracture. Rates of previous revision surgery for peri- implant fracture have ranged from 2.4% to 6% for CMNs and from 0.6% to 4% for SHSs.^5-7,9 In a Cochrane review of 22 studies (3749 patients), Parker and Handoll¹² compared CMN and SHS outcomes in 23 categories and found a statistically significant difference only in postoperative fracture rate. However, in a meta-analysis of studies conducted between 2000 and 2005, Bhandari and colleagues⁸ found no statistically significant difference in risk of femoral shaft fracture between CMNs (0.6%) and SHSs (0.1%). In addition, Utrilla and colleagues¹⁰ reported no postoperative fractures with use of Gamma3 CMNs. These recent studies may indicate that newer CMN designs can reduce the incidence of postoperative peri-implant fracture.^8,10 Other outcome measures, such as 1-year mortality, functional outcome, and medical complication rate, have shown no statistically significant differences between the 2 implants.^10-12 Ultimately, the current recommendation for fixation of stable intertrochanteric hip fractures is either SHS or CMN.^13,14

Of our study patients, 78.9% (before group) and 64.6% (after group) were female, and 49.3% (before group) and 47.9% (after group) were between 80 and 89 years of age. Similarly, a review of hip fracture Medicare claims made between 1999 and 2002 revealed that >75% of the patients were females and 48% to 49% were octogenarians.^4,18 However, our rates of different fracture types differed from those of Adams and colleagues.⁵ In a 1-year single-institution study, they found that, for both CMNs and SHSs, the most common stable intertrochanteric fractures were 31-A1.1 fractures; in our study’s before and after groups, more than one-third of injuries were 31-A2.1 fractures. Least common were 31-A1.3 fractures, both in the study by Adams and colleagues⁵ and in our before (16.9%) and after (14.6%) groups. Although our fracture rates differ from those of previous studies, all 4 classification categories fall under the umbrella of stable intertrochanteric hip fracture, which is the sole focus of this study.¹⁴ 

We hypothesized that cost would be a significant driver of implant choice in the management of these injuries. Given that SHS costs $1186.91 less than SGN and $1625.88 less than LGN at our institution, we expected that the before- discussion group’s overall SHS use rate of 38.0% would increase after discussion. Instead, SGN became the favored implant, with use in almost half of all fractures in the after group. Although the change in overall implant use rate was notable, these findings were not statistically significant. Examination of individual fracture patterns revealed 2 areas of interest. First, SHS was assumed to be the implant of choice in the management of the relatively simple 31-A1.1 fractures. Although this assumption was verified in the before group (SHS use in 43.8% of fractures), SGN was used in almost every case (90%) in the after group. However, when surgeons were asked 1 year later to recommend an implant for A1.1 fractures, 92% suggested SHS. The more cost-effective SHS construct may be the most amenable for use in these injury types given all intertrochanteric hip fracture patterns, though this has not been studied.

On the other hand, for 31-A2.1 fractures, perhaps the most complicated of the stable patterns, LGN became the implant of choice (42.1%). Despite surgeons’ awareness of the cost differences, management of these fractures shifted in the after group to the most expensive implant, indicative of surgeon concern about eventual loss of reduction with SHS and surgeon comfort with a particular procedure. This trend held when surgeons were asked to reclassify fractures 1 year later, as CMNs were recommended for 88% of 31-A2.1 fractures. Although both SHS and CMN were acceptable in 97% of the fractures included in this study, SGNs or LGNs were preferred for almost every fracture pattern involving the lesser trochanter. All 9 attending surgeons described involvement of the lesser trochanter as an indicator of posteromedial calcar injury. Surgeons became particularly concerned when this fracture pattern occurred in patients with significant osteopenia; SHS fixation, in their opinion, would be poor in the setting of a combination of greater posteromedial instability and poor bone quality. In a level I prospective, randomized trial, Barton and colleagues⁷ found no difference in outcomes between LGN and SHS fixation for 31-A2 proximal femur fractures and recommended SHS implants for the cost savings. In the clinical experience of this group, however, A1.3 and A2.1 fractures were at especially high risk for failure with SHS use, which necessitated greater implant stability through CMN fixation. On the other hand, for simpler fracture patterns, most surgeons suggested SHS implants. In their opinion, SGN and LGN implants offered no additional benefit of stability without evidence of posteromedial injury, even in the setting of osteopenia. For A1.2 fractures, posteromedial involvement was judged on the basis of size of the inferomedial spike or the extent of the inferomedial fracture line. Two surgeons preferred CMN for simple fractures, one because of the increased comfort with the implants and the other because of the minimally invasive surgical technique. Overall, our results indicate that knowledge of implant cost is not a strong enough factor to change surgeon behavior in selecting fixation for uncomplicated stable intertrochanteric hip fractures in previously ambulatory elderly patients. 

The lack of effect could be a consequence of surgeons’ training and comfort with various implants, especially among younger attending surgeons. Most of the attendings in the practice are under age 50 years, which correlates with a preference for CMN fixation.¹⁹ Case loads of >80 intertrochanteric hip fractures per calendar year, as in the after group, also correlates with more CMN use.¹⁹ However, the before group had more intertrochanteric hip fractures, and yet SHS was the implant of choice. Resident physician experience and comfort with various implants may play a role too, as teaching hospitals with resident assistance also correlate with CMN use.¹⁹ However, no major change in resident physician involvement was undertaken during this period. The institution studied is near a major metropolis in the Northeast, a region that has disfavored SHS in recent years.¹⁸ The change from before to after fits an overall trend in changing implant use. Anglen and colleagues¹⁵ found a significant decrease in SHS use, from 97% in 1999 to 33% in 2006, for intertrochanteric fracture fixation. Simultaneously, CMN use increased from 3% to 67%.

This study had several limitations. First, its overall sample size was small, and therefore any data fluctuations may be exaggerated. Furthermore, changes in utilization rates were compared over 2 years, which may not be long enough to show a changing trend in implant selection. Post hoc analysis of the sample size determined a power of 0.76 for an α of 0.05 and an effect size of 0.50. Second, radiologic classification was performed in a retrospective review, not officially by the operative surgeon. Fractures that we considered stable may have been considered unstable by the operative surgeon, influencing implant selection. Third, patients were selected from only one hospital, and orthopedic surgeons from other institutions may be more sensitive to cost considerations, changing implant selection more quickly. Fourth, initial selection of patients by CPT code might not have captured all those who satisfied the inclusion criteria. Fifth, only a single intervention was used, and follow-up meetings certainly could have increased the effectiveness of the intervention. Last, this and other retrospective studies are inherently weaker because of possible bias.

Conclusion

Our study results showed that implant cost is not a significant factor in implant selection for uncomplicated stable intertrochanteric hip fractures in previously ambulatory elderly patients. By itself, knowledge of implant cost may not be a strong enough force to change surgeon behavior or preference secondary to consequences of failure or comfort with particular implants. In an economic climate in which healthcare is scrutinized for both its medical effectiveness and cost-effectiveness, further study of forces that could influence orthopedic surgeons to select a more cost-effective implant is warranted.

Take-Home Points

• The incidence of geriatric hip fractures is rising nationally.
• Costs associated with hip fracture care have risen significantly.
• CMN and SHS are effective for stable, intertrochanteric hip fractures.
• Current trends show increased utilization of CMN compared to SHS for stable introchanteric hip fractures.
• Surgeon awareness of implant cost is a critical factor in delivering cost-effective, evidence-based surgical care.

The continuing increase in the population of patients older than 65 years in the United States is well known. For orthopedic surgeons, this trend highlights the importance of effective geriatric fracture care, particularly hip fracture care. Hip fractures in the elderly are expected to increase 50% by 2025 and to number 500,000 by 2040.¹ The growing burden of hip fracture cases is accompanied by increasing costs of care. In 2012, 88% of the healthcare dollars spent on these patients were for direct fracture care, a significant increase from 60% in 2009.² Although fewer than 1 in 5 fractures in the elderly are hip fractures, these injuries account for 72% of the total cost of geriatric fracture care, more than the total cost of all other osteoporosis-related injuries combined.¹ Currently, the direct cost of hip fracture care ranges from $8358 to $32,195 and is expected, in total, to reach $25 billion by 2025.^2,3

About 50% of geriatric hip fractures are extracapsular intertrochanteric or pertrochanteric.⁴ Several studies have compared sliding hip screws (SHSs) with cephalomedullary nails (CMNs) in the effective management of stable intertrochanteric fractures.^5-11 Although these implants have shown an increased risk for peri-implant fracture and subsequent reoperation, markers such as mortality, medical complications, and restoration of prefracture function have all been equivocal relative to SHSs.¹² Ultimately, one implant cannot be definitively recommended over the other for management of stable intertrochanteric hip fractures.^13,14 Nevertheless, the current trend increasingly favors CMNs over SHSs.^4,15 Most orthopedic surgeons are unaware of or underestimate the cost difference between these implants—a fact even more pronounced for newer implants.^4,16 Considering the ever growing cost burden of hip fractures in the United States, orthopedists must consider not only the efficacy of care but the cost of delivery.

We conducted a study to determine the effect that surgeon knowledge of implant cost had on rates of use of SHSs and CMNs in the management of stable intertrochanteric hip fractures. 

Patients and Methods

On May 1, 2012, all 9 attending orthopedic surgeons in a private practice group serving a suburban level II trauma center met to discuss implant prices and implant-related costs for the $850 Versafix SHS, the $1950 short Gamma3 nail (SGN), and the $2900 long Gamma3 nail (LGN), all manufactured by Stryker. All surgeons denied previous knowledge of the costs of these implants. During the discussion, no particular implant was recommended for management of any specific type of fracture. Surgeons were not directly instructed to consider implant cost in subsequent hip fracture surgeries and were not informed of our upcoming study of implant utilization.

After obtaining Institutional Review Board approval, we performed a retrospective chart and radiologic review of all hip fractures (Current Procedural Terminology [CPT] code 27244 or 27245) managed with fixation at our institution between May 1, 2011 and April 30, 2013. Two hundred six patients were identified (Figure 1).

One year later, surgeons were again shown their respective hip fracture radiographs, with patient identifying data removed. They were asked to reclassify their respective cases using the OTA system and then indicate the implant they would use for operative fixation in each of their cases.

Patient age, sex, injury side, fracture types, and utilization rates of the SHS, SGN, and LGN implants were compared between the groups. For each eligible patient, implant cost and other financial data were obtained from the hospital’s finance department. Statistical analyses were performed with SPSS (Statistical Package for the Social Sciences) Version 20 for Macintosh. Independent 2-sample t test was used for parametric comparisons, and Fisher exact test was used for nonparametric comparisons.

Results

Examination of implant use per fracture classification revealed an interesting change. In the before group, SHS was the implant most commonly used for 31-A1.1 fractures (7/16, 43.8%), 31-A1.2 fractures (8/18, 44.4%), and 31-A2.1 fractures (10/25, 40.0%), and LGN was used in 66.7% (8/12) of 31-A1.3 fractures. By contrast, in the after group, SHS was most commonly used only for 31-A1.2 fractures (7/12, 58.3%), SGN was used in 90% (9/10) of 31-A1.1 fractures, and LGN was used in 42.1% (8/19) of 31-A2.1 fractures. In addition, 85.7% (6/7) of 31-A1.3 fractures were managed with a version of the Gamma nail.

Reclassification resulted in more A2.1 fractures (42.0% vs 37.0%) and fewer A1.3 fractures (10.1% vs 16.0%). About the same numbers of fractures were classified A1.1 (21.0% vs 21.8%) and A1.2 (26.9% vs 25.2%). SHS was favored for A1.1 fractures (92.0%) and A1.2 fractures (65.6%). SGN was favored for A1.3 fractures (75.0%). Gamma nails of both sizes were favored for A2.1 fractures (88.0%).

Discussion

Comparisons of SHS/plate and CMN constructs in the management of stable intertrochanteric hip fractures have long been discussed in the orthopedic literature. The major concern with CMNs (vs SHSs) is a statistically significantly higher rate of revision surgery, most often for peri-implant fracture. Rates of previous revision surgery for peri- implant fracture have ranged from 2.4% to 6% for CMNs and from 0.6% to 4% for SHSs.^5-7,9 In a Cochrane review of 22 studies (3749 patients), Parker and Handoll¹² compared CMN and SHS outcomes in 23 categories and found a statistically significant difference only in postoperative fracture rate. However, in a meta-analysis of studies conducted between 2000 and 2005, Bhandari and colleagues⁸ found no statistically significant difference in risk of femoral shaft fracture between CMNs (0.6%) and SHSs (0.1%). In addition, Utrilla and colleagues¹⁰ reported no postoperative fractures with use of Gamma3 CMNs. These recent studies may indicate that newer CMN designs can reduce the incidence of postoperative peri-implant fracture.^8,10 Other outcome measures, such as 1-year mortality, functional outcome, and medical complication rate, have shown no statistically significant differences between the 2 implants.^10-12 Ultimately, the current recommendation for fixation of stable intertrochanteric hip fractures is either SHS or CMN.^13,14

Of our study patients, 78.9% (before group) and 64.6% (after group) were female, and 49.3% (before group) and 47.9% (after group) were between 80 and 89 years of age. Similarly, a review of hip fracture Medicare claims made between 1999 and 2002 revealed that >75% of the patients were females and 48% to 49% were octogenarians.^4,18 However, our rates of different fracture types differed from those of Adams and colleagues.⁵ In a 1-year single-institution study, they found that, for both CMNs and SHSs, the most common stable intertrochanteric fractures were 31-A1.1 fractures; in our study’s before and after groups, more than one-third of injuries were 31-A2.1 fractures. Least common were 31-A1.3 fractures, both in the study by Adams and colleagues⁵ and in our before (16.9%) and after (14.6%) groups. Although our fracture rates differ from those of previous studies, all 4 classification categories fall under the umbrella of stable intertrochanteric hip fracture, which is the sole focus of this study.¹⁴ 

We hypothesized that cost would be a significant driver of implant choice in the management of these injuries. Given that SHS costs $1186.91 less than SGN and $1625.88 less than LGN at our institution, we expected that the before- discussion group’s overall SHS use rate of 38.0% would increase after discussion. Instead, SGN became the favored implant, with use in almost half of all fractures in the after group. Although the change in overall implant use rate was notable, these findings were not statistically significant. Examination of individual fracture patterns revealed 2 areas of interest. First, SHS was assumed to be the implant of choice in the management of the relatively simple 31-A1.1 fractures. Although this assumption was verified in the before group (SHS use in 43.8% of fractures), SGN was used in almost every case (90%) in the after group. However, when surgeons were asked 1 year later to recommend an implant for A1.1 fractures, 92% suggested SHS. The more cost-effective SHS construct may be the most amenable for use in these injury types given all intertrochanteric hip fracture patterns, though this has not been studied.

On the other hand, for 31-A2.1 fractures, perhaps the most complicated of the stable patterns, LGN became the implant of choice (42.1%). Despite surgeons’ awareness of the cost differences, management of these fractures shifted in the after group to the most expensive implant, indicative of surgeon concern about eventual loss of reduction with SHS and surgeon comfort with a particular procedure. This trend held when surgeons were asked to reclassify fractures 1 year later, as CMNs were recommended for 88% of 31-A2.1 fractures. Although both SHS and CMN were acceptable in 97% of the fractures included in this study, SGNs or LGNs were preferred for almost every fracture pattern involving the lesser trochanter. All 9 attending surgeons described involvement of the lesser trochanter as an indicator of posteromedial calcar injury. Surgeons became particularly concerned when this fracture pattern occurred in patients with significant osteopenia; SHS fixation, in their opinion, would be poor in the setting of a combination of greater posteromedial instability and poor bone quality. In a level I prospective, randomized trial, Barton and colleagues⁷ found no difference in outcomes between LGN and SHS fixation for 31-A2 proximal femur fractures and recommended SHS implants for the cost savings. In the clinical experience of this group, however, A1.3 and A2.1 fractures were at especially high risk for failure with SHS use, which necessitated greater implant stability through CMN fixation. On the other hand, for simpler fracture patterns, most surgeons suggested SHS implants. In their opinion, SGN and LGN implants offered no additional benefit of stability without evidence of posteromedial injury, even in the setting of osteopenia. For A1.2 fractures, posteromedial involvement was judged on the basis of size of the inferomedial spike or the extent of the inferomedial fracture line. Two surgeons preferred CMN for simple fractures, one because of the increased comfort with the implants and the other because of the minimally invasive surgical technique. Overall, our results indicate that knowledge of implant cost is not a strong enough factor to change surgeon behavior in selecting fixation for uncomplicated stable intertrochanteric hip fractures in previously ambulatory elderly patients. 

The lack of effect could be a consequence of surgeons’ training and comfort with various implants, especially among younger attending surgeons. Most of the attendings in the practice are under age 50 years, which correlates with a preference for CMN fixation.¹⁹ Case loads of >80 intertrochanteric hip fractures per calendar year, as in the after group, also correlates with more CMN use.¹⁹ However, the before group had more intertrochanteric hip fractures, and yet SHS was the implant of choice. Resident physician experience and comfort with various implants may play a role too, as teaching hospitals with resident assistance also correlate with CMN use.¹⁹ However, no major change in resident physician involvement was undertaken during this period. The institution studied is near a major metropolis in the Northeast, a region that has disfavored SHS in recent years.¹⁸ The change from before to after fits an overall trend in changing implant use. Anglen and colleagues¹⁵ found a significant decrease in SHS use, from 97% in 1999 to 33% in 2006, for intertrochanteric fracture fixation. Simultaneously, CMN use increased from 3% to 67%.

This study had several limitations. First, its overall sample size was small, and therefore any data fluctuations may be exaggerated. Furthermore, changes in utilization rates were compared over 2 years, which may not be long enough to show a changing trend in implant selection. Post hoc analysis of the sample size determined a power of 0.76 for an α of 0.05 and an effect size of 0.50. Second, radiologic classification was performed in a retrospective review, not officially by the operative surgeon. Fractures that we considered stable may have been considered unstable by the operative surgeon, influencing implant selection. Third, patients were selected from only one hospital, and orthopedic surgeons from other institutions may be more sensitive to cost considerations, changing implant selection more quickly. Fourth, initial selection of patients by CPT code might not have captured all those who satisfied the inclusion criteria. Fifth, only a single intervention was used, and follow-up meetings certainly could have increased the effectiveness of the intervention. Last, this and other retrospective studies are inherently weaker because of possible bias.

Conclusion

Our study results showed that implant cost is not a significant factor in implant selection for uncomplicated stable intertrochanteric hip fractures in previously ambulatory elderly patients. By itself, knowledge of implant cost may not be a strong enough force to change surgeon behavior or preference secondary to consequences of failure or comfort with particular implants. In an economic climate in which healthcare is scrutinized for both its medical effectiveness and cost-effectiveness, further study of forces that could influence orthopedic surgeons to select a more cost-effective implant is warranted.

Lichen planus (LP) is a chronic inflammatory dermatosis of unknown origin that involves the skin and mucous membranes, and lichenoid drug eruption (LDE) is an uncommon cutaneous adverse reaction to a medication.¹ The manifestations resemble each other clinically, and sometimes it is difficult to differentiate between them on histology. The pathogenesis still is not well characterized, especially the key initiating event that leads to the development of LP or LDE postimmunization. There have been reports of LP or LDEs after certain vaccines, especially the hepatitis B and influenza vaccines.^2-4 Both vaccines are routinely administered in the United States; more than 100 million individuals have received the hepatitis B vaccine in the United States since it became available in 1982,⁵ and the Advisory Committee on Immunization Practices of the Centers for Disease Control and Prevention (CDC) recommends that all individuals 6 months or older receive an influenza vaccine every year.⁶ Currently, influenza vaccine coverage among adults 18 years or older reaches approximately 40% annually in the United States.⁶

Although certain viral infections (eg, hepatitis C virus) seem to play a role in the development of LP,^7,8 the link between LP and hepatitis B vaccination is less well recognized. Reports of LP and LDE after vaccination have been largely limited to case reports and case series.^2-4,9,10 Therefore, we aimed to characterize and review cases of LP and LDE following vaccination by analyzing the Vaccine Adverse Event Reporting System (VAERS) database.

Methods

The VAERS is a national vaccine safety surveillance database maintained jointly by the CDC and the US Food and Drug Administration to analyze adverse events (AEs) following immunizations. Serious AEs and deaths recorded in the VAERS were followed up periodically by VAERS staff. Information on vaccine-associated LP or LDE was retrieved from the VAERS database using the CDC WONDER online interface (http://wonder.cdc.gov/vaers.html). To examine if LP or LDE after vaccination occurred more frequently in patients with certain demographic risk factors, all reported cases of LP and LDE associated with vaccines administered from July 1990 to November 2014 were identified in the symptoms section of the VAERS system using the search terms lichen planus, oral lichen planus, and lichenoid drug eruption. Characteristics such as age, gender, time to onset, type of vaccine, method of diagnosis, and clinical outcome were collected.

The statistical package for social sciences (SPSS version 22) was utilized for the descriptive analysis. Fisher exact and χ² tests were used to evaluate statistical significance. A 2-sided P value of <.05 was considered statistically significant.

Results

There were 434,943 reported AEs following vaccination in the VAERS database from July 1990 to November 2014; among them, 33 cases involved LP or LDE. Of these vaccine-associated AEs, LP was diagnosed in 23 (69.7%) cases, while LDE and oral LP were diagnosed in 6 (18.2%) and 4 (12.1%) cases, respectively. Females represented slightly more than half (57.6% [19/33]) of the total cases. The median age of onset was 47 years. Approximately two-thirds of the identified cases were confirmed on skin biopsy and histology, while the rest were diagnosed either by a dermatologist or a primary care physician. The time to onset of symptoms ranged from 1 to 297 days after vaccination, with a median time of 14 days.

Patients with LP or LDE were significantly older compared to the reported AEs overall (P<.001); the median age of onset was 47 years for LP or LDE compared to 24 years for all reported AEs. Table 1 shows the various vaccines associated with LP or LDE. The hepatitis B, influenza, and herpes zoster vaccines were the 3 most common types of vaccines associated with these conditions. The hepatitis B vaccine accounted for 24.2% (8/33) of the reported events, followed by influenza (18.2% [6/33]) and herpes zoster (15.2% [5/33]) vaccines. In addition, there were 3 cases of cutaneous reaction after receiving the combination hepatitis A and hepatitis B vaccine. Table 2 presents details of the reported events associated with hepatitis B, influenza, herpes zoster, combination hepatitis A and hepatitis B, and hepatitis A vaccination.

Of 8 AEs associated with hepatitis B vaccination, 1 AE resulted in permanent disability and required hospitalization. Of 5 cases of AEs associated with herpes zoster vaccination, 1 AE resulted in permanent disability.

Comment

The estimated prevalence of LP ranges from 0.22% to 5% worldwide,^11-15 with an incidence of 0.032% to 0.037%.¹⁶ Although rare, LP and LDE can occur from certain medications or vaccines. Cases of LP have been reported after hepatitis B and influenza vaccinations. The first case of LP following hepatitis B vaccination was described by Ciaccio and Rebora¹⁷ in 1990. Since then, a total of 50 similar cases have been reported worldwide.² There also have been reports of LP following influenza, tetanus-diphtheria-pertussis, measles-mumps-rubella, and inactivated polio vaccines.^3,4,9,10 Table 3 summarizes cases of LP following various vaccinations.

The key initiating event of the pathogenesis for both LP and LDE is not completely understood. Both conditions share similar immunologic mechanisms of persistently activated CD8 autocytotoxic T lymphocytes against epidermal cells.¹⁸ These cells can induce apoptosis of basal epidermal keratinocytes and generate various cytokines (eg, IFN-γ, IL-5) to enhance expression of class II MHC molecules and antigen presentation to CD4 T cells.^19-22 It is conceivable that one of the initiating factors may be related to components in vaccines.

Hepatitis B, influenza, and herpes zoster vaccines were the 3 most common vaccines implicated in postimmunization LP or LDEs in our study. The excipients of these vaccines were compared based on the product inserts to identify any common components. It was found that all 3 vaccines contain either yeast protein or egg protein with various forms of phosphate buffers, while the hepatitis A and herpes zoster vaccines share Medical Research Council cell strain 5 (human diploid) cells as well as other cellular components.²³ Sato et al⁴ suggested that specific vaccine components, such as the vaccine itself or egg proteins, could have contributed to the development of LP following vaccination. It has been postulated that the protein S fraction of hepatitis B surface antigen plays a crucial role in the pathogenesis of both LP and LDE after hepatitis B vaccination.^2,24 It is likely that protein S shares common epitopes on keratinocytes that are recognized by the immune system, thus activating cytotoxic T lymphocytes and inducing apoptosis.^2,24

In this study, the median time to onset of vaccine-related LP was 14 days, which is consistent with a case series by Sato et al,⁴ suggesting that adverse reactions mainly occurred within 2 weeks after influenza vaccination. Onset of symptoms within 2 weeks of vaccination would therefore be a crucial clue for diagnosing possible vaccine-related LP or LDE. On the other hand, at least 4 patients in our study had onset of LP and LDE more than 1 month after vaccination; 2 of 4 cases even reported symptom onset at 175 and 297 days after hepatitis B vaccination, which were much longer than the 120 days reported by Tarakji et al.² It is not known if these cases constitute true vaccine-associated LP or LDE or if unmeasured confounding factors such as concurrent medications or comorbidities may have contributed to the development of these AEs.

It also is interesting to note that LP and LDE affected mainly middle-aged women. An increased risk of autoimmunity in female adults partly explains this observation.²⁵ Some vaccines, such as herpes zoster and influenza vaccines, generally are recommended for older adults who also are more likely to have multiple comorbidities or take multiple medications/supplements, which can potentially skew the prevalence of AEs toward an older age group. It should be noted, however, that LP and LDE were relatively uncommon AEs following vaccination in the current study. In this study, LP and LDE consisted of only 0.01% (N=42,230) of all AEs after hepatitis B vaccination, while the more common AEs such as pyrexia, nonspecific rashes, nonspecific gastrointestinal symptoms, and headache contributed to approximately 66.5% of all reported events.

One of the strengths of our study is that up to two-thirds of cases were confirmed histologically and all patients were seen and followed up by dermatologists or physicians. The VAERS is an easily accessible, up-to-date, and live reporting system that collects all AEs associated with vaccines in the United States. Important clinical and laboratory information usually is available in the database; however, the main limitation is that this study can only demonstrate a possible association but not a causal relationship between vaccination and LP or LDE. There can be various sources of biases such as underreporting, overreporting, or inaccurate reporting.^26,27 Pertinent clinical information (eg, new medications, new dental fillings/implants) that could potentially misrepresent the actual relationship between vaccination and development of AEs also was not available in the VAERS database. A cohort study with long-term follow-up or a large-scale case-control study would be useful in evaluating such associations.

Conclusion

Lichen planus and LDE can occur, albeit rarely, after vaccination, especially following hepatitis B vaccination. When middle-aged adults present to the clinic with LP or LDE, it is important to inquire about recent vaccination history in addition to a detailed medication history.

Lichen planus (LP) is a chronic inflammatory dermatosis of unknown origin that involves the skin and mucous membranes, and lichenoid drug eruption (LDE) is an uncommon cutaneous adverse reaction to a medication.¹ The manifestations resemble each other clinically, and sometimes it is difficult to differentiate between them on histology. The pathogenesis still is not well characterized, especially the key initiating event that leads to the development of LP or LDE postimmunization. There have been reports of LP or LDEs after certain vaccines, especially the hepatitis B and influenza vaccines.^2-4 Both vaccines are routinely administered in the United States; more than 100 million individuals have received the hepatitis B vaccine in the United States since it became available in 1982,⁵ and the Advisory Committee on Immunization Practices of the Centers for Disease Control and Prevention (CDC) recommends that all individuals 6 months or older receive an influenza vaccine every year.⁶ Currently, influenza vaccine coverage among adults 18 years or older reaches approximately 40% annually in the United States.⁶

Although certain viral infections (eg, hepatitis C virus) seem to play a role in the development of LP,^7,8 the link between LP and hepatitis B vaccination is less well recognized. Reports of LP and LDE after vaccination have been largely limited to case reports and case series.^2-4,9,10 Therefore, we aimed to characterize and review cases of LP and LDE following vaccination by analyzing the Vaccine Adverse Event Reporting System (VAERS) database.

Methods

The VAERS is a national vaccine safety surveillance database maintained jointly by the CDC and the US Food and Drug Administration to analyze adverse events (AEs) following immunizations. Serious AEs and deaths recorded in the VAERS were followed up periodically by VAERS staff. Information on vaccine-associated LP or LDE was retrieved from the VAERS database using the CDC WONDER online interface (http://wonder.cdc.gov/vaers.html). To examine if LP or LDE after vaccination occurred more frequently in patients with certain demographic risk factors, all reported cases of LP and LDE associated with vaccines administered from July 1990 to November 2014 were identified in the symptoms section of the VAERS system using the search terms lichen planus, oral lichen planus, and lichenoid drug eruption. Characteristics such as age, gender, time to onset, type of vaccine, method of diagnosis, and clinical outcome were collected.

The statistical package for social sciences (SPSS version 22) was utilized for the descriptive analysis. Fisher exact and χ² tests were used to evaluate statistical significance. A 2-sided P value of <.05 was considered statistically significant.

Results

There were 434,943 reported AEs following vaccination in the VAERS database from July 1990 to November 2014; among them, 33 cases involved LP or LDE. Of these vaccine-associated AEs, LP was diagnosed in 23 (69.7%) cases, while LDE and oral LP were diagnosed in 6 (18.2%) and 4 (12.1%) cases, respectively. Females represented slightly more than half (57.6% [19/33]) of the total cases. The median age of onset was 47 years. Approximately two-thirds of the identified cases were confirmed on skin biopsy and histology, while the rest were diagnosed either by a dermatologist or a primary care physician. The time to onset of symptoms ranged from 1 to 297 days after vaccination, with a median time of 14 days.

Patients with LP or LDE were significantly older compared to the reported AEs overall (P<.001); the median age of onset was 47 years for LP or LDE compared to 24 years for all reported AEs. Table 1 shows the various vaccines associated with LP or LDE. The hepatitis B, influenza, and herpes zoster vaccines were the 3 most common types of vaccines associated with these conditions. The hepatitis B vaccine accounted for 24.2% (8/33) of the reported events, followed by influenza (18.2% [6/33]) and herpes zoster (15.2% [5/33]) vaccines. In addition, there were 3 cases of cutaneous reaction after receiving the combination hepatitis A and hepatitis B vaccine. Table 2 presents details of the reported events associated with hepatitis B, influenza, herpes zoster, combination hepatitis A and hepatitis B, and hepatitis A vaccination.

Of 8 AEs associated with hepatitis B vaccination, 1 AE resulted in permanent disability and required hospitalization. Of 5 cases of AEs associated with herpes zoster vaccination, 1 AE resulted in permanent disability.

Comment

The estimated prevalence of LP ranges from 0.22% to 5% worldwide,^11-15 with an incidence of 0.032% to 0.037%.¹⁶ Although rare, LP and LDE can occur from certain medications or vaccines. Cases of LP have been reported after hepatitis B and influenza vaccinations. The first case of LP following hepatitis B vaccination was described by Ciaccio and Rebora¹⁷ in 1990. Since then, a total of 50 similar cases have been reported worldwide.² There also have been reports of LP following influenza, tetanus-diphtheria-pertussis, measles-mumps-rubella, and inactivated polio vaccines.^3,4,9,10 Table 3 summarizes cases of LP following various vaccinations.

The key initiating event of the pathogenesis for both LP and LDE is not completely understood. Both conditions share similar immunologic mechanisms of persistently activated CD8 autocytotoxic T lymphocytes against epidermal cells.¹⁸ These cells can induce apoptosis of basal epidermal keratinocytes and generate various cytokines (eg, IFN-γ, IL-5) to enhance expression of class II MHC molecules and antigen presentation to CD4 T cells.^19-22 It is conceivable that one of the initiating factors may be related to components in vaccines.

Hepatitis B, influenza, and herpes zoster vaccines were the 3 most common vaccines implicated in postimmunization LP or LDEs in our study. The excipients of these vaccines were compared based on the product inserts to identify any common components. It was found that all 3 vaccines contain either yeast protein or egg protein with various forms of phosphate buffers, while the hepatitis A and herpes zoster vaccines share Medical Research Council cell strain 5 (human diploid) cells as well as other cellular components.²³ Sato et al⁴ suggested that specific vaccine components, such as the vaccine itself or egg proteins, could have contributed to the development of LP following vaccination. It has been postulated that the protein S fraction of hepatitis B surface antigen plays a crucial role in the pathogenesis of both LP and LDE after hepatitis B vaccination.^2,24 It is likely that protein S shares common epitopes on keratinocytes that are recognized by the immune system, thus activating cytotoxic T lymphocytes and inducing apoptosis.^2,24

In this study, the median time to onset of vaccine-related LP was 14 days, which is consistent with a case series by Sato et al,⁴ suggesting that adverse reactions mainly occurred within 2 weeks after influenza vaccination. Onset of symptoms within 2 weeks of vaccination would therefore be a crucial clue for diagnosing possible vaccine-related LP or LDE. On the other hand, at least 4 patients in our study had onset of LP and LDE more than 1 month after vaccination; 2 of 4 cases even reported symptom onset at 175 and 297 days after hepatitis B vaccination, which were much longer than the 120 days reported by Tarakji et al.² It is not known if these cases constitute true vaccine-associated LP or LDE or if unmeasured confounding factors such as concurrent medications or comorbidities may have contributed to the development of these AEs.

It also is interesting to note that LP and LDE affected mainly middle-aged women. An increased risk of autoimmunity in female adults partly explains this observation.²⁵ Some vaccines, such as herpes zoster and influenza vaccines, generally are recommended for older adults who also are more likely to have multiple comorbidities or take multiple medications/supplements, which can potentially skew the prevalence of AEs toward an older age group. It should be noted, however, that LP and LDE were relatively uncommon AEs following vaccination in the current study. In this study, LP and LDE consisted of only 0.01% (N=42,230) of all AEs after hepatitis B vaccination, while the more common AEs such as pyrexia, nonspecific rashes, nonspecific gastrointestinal symptoms, and headache contributed to approximately 66.5% of all reported events.

One of the strengths of our study is that up to two-thirds of cases were confirmed histologically and all patients were seen and followed up by dermatologists or physicians. The VAERS is an easily accessible, up-to-date, and live reporting system that collects all AEs associated with vaccines in the United States. Important clinical and laboratory information usually is available in the database; however, the main limitation is that this study can only demonstrate a possible association but not a causal relationship between vaccination and LP or LDE. There can be various sources of biases such as underreporting, overreporting, or inaccurate reporting.^26,27 Pertinent clinical information (eg, new medications, new dental fillings/implants) that could potentially misrepresent the actual relationship between vaccination and development of AEs also was not available in the VAERS database. A cohort study with long-term follow-up or a large-scale case-control study would be useful in evaluating such associations.

Conclusion

Lichen planus and LDE can occur, albeit rarely, after vaccination, especially following hepatitis B vaccination. When middle-aged adults present to the clinic with LP or LDE, it is important to inquire about recent vaccination history in addition to a detailed medication history.

Lichen planus (LP) is a chronic inflammatory dermatosis of unknown origin that involves the skin and mucous membranes, and lichenoid drug eruption (LDE) is an uncommon cutaneous adverse reaction to a medication.¹ The manifestations resemble each other clinically, and sometimes it is difficult to differentiate between them on histology. The pathogenesis still is not well characterized, especially the key initiating event that leads to the development of LP or LDE postimmunization. There have been reports of LP or LDEs after certain vaccines, especially the hepatitis B and influenza vaccines.^2-4 Both vaccines are routinely administered in the United States; more than 100 million individuals have received the hepatitis B vaccine in the United States since it became available in 1982,⁵ and the Advisory Committee on Immunization Practices of the Centers for Disease Control and Prevention (CDC) recommends that all individuals 6 months or older receive an influenza vaccine every year.⁶ Currently, influenza vaccine coverage among adults 18 years or older reaches approximately 40% annually in the United States.⁶

Although certain viral infections (eg, hepatitis C virus) seem to play a role in the development of LP,^7,8 the link between LP and hepatitis B vaccination is less well recognized. Reports of LP and LDE after vaccination have been largely limited to case reports and case series.^2-4,9,10 Therefore, we aimed to characterize and review cases of LP and LDE following vaccination by analyzing the Vaccine Adverse Event Reporting System (VAERS) database.

Methods

The VAERS is a national vaccine safety surveillance database maintained jointly by the CDC and the US Food and Drug Administration to analyze adverse events (AEs) following immunizations. Serious AEs and deaths recorded in the VAERS were followed up periodically by VAERS staff. Information on vaccine-associated LP or LDE was retrieved from the VAERS database using the CDC WONDER online interface (http://wonder.cdc.gov/vaers.html). To examine if LP or LDE after vaccination occurred more frequently in patients with certain demographic risk factors, all reported cases of LP and LDE associated with vaccines administered from July 1990 to November 2014 were identified in the symptoms section of the VAERS system using the search terms lichen planus, oral lichen planus, and lichenoid drug eruption. Characteristics such as age, gender, time to onset, type of vaccine, method of diagnosis, and clinical outcome were collected.

The statistical package for social sciences (SPSS version 22) was utilized for the descriptive analysis. Fisher exact and χ² tests were used to evaluate statistical significance. A 2-sided P value of <.05 was considered statistically significant.

Results

There were 434,943 reported AEs following vaccination in the VAERS database from July 1990 to November 2014; among them, 33 cases involved LP or LDE. Of these vaccine-associated AEs, LP was diagnosed in 23 (69.7%) cases, while LDE and oral LP were diagnosed in 6 (18.2%) and 4 (12.1%) cases, respectively. Females represented slightly more than half (57.6% [19/33]) of the total cases. The median age of onset was 47 years. Approximately two-thirds of the identified cases were confirmed on skin biopsy and histology, while the rest were diagnosed either by a dermatologist or a primary care physician. The time to onset of symptoms ranged from 1 to 297 days after vaccination, with a median time of 14 days.

Patients with LP or LDE were significantly older compared to the reported AEs overall (P<.001); the median age of onset was 47 years for LP or LDE compared to 24 years for all reported AEs. Table 1 shows the various vaccines associated with LP or LDE. The hepatitis B, influenza, and herpes zoster vaccines were the 3 most common types of vaccines associated with these conditions. The hepatitis B vaccine accounted for 24.2% (8/33) of the reported events, followed by influenza (18.2% [6/33]) and herpes zoster (15.2% [5/33]) vaccines. In addition, there were 3 cases of cutaneous reaction after receiving the combination hepatitis A and hepatitis B vaccine. Table 2 presents details of the reported events associated with hepatitis B, influenza, herpes zoster, combination hepatitis A and hepatitis B, and hepatitis A vaccination.

Of 8 AEs associated with hepatitis B vaccination, 1 AE resulted in permanent disability and required hospitalization. Of 5 cases of AEs associated with herpes zoster vaccination, 1 AE resulted in permanent disability.

Comment

The estimated prevalence of LP ranges from 0.22% to 5% worldwide,^11-15 with an incidence of 0.032% to 0.037%.¹⁶ Although rare, LP and LDE can occur from certain medications or vaccines. Cases of LP have been reported after hepatitis B and influenza vaccinations. The first case of LP following hepatitis B vaccination was described by Ciaccio and Rebora¹⁷ in 1990. Since then, a total of 50 similar cases have been reported worldwide.² There also have been reports of LP following influenza, tetanus-diphtheria-pertussis, measles-mumps-rubella, and inactivated polio vaccines.^3,4,9,10 Table 3 summarizes cases of LP following various vaccinations.

The key initiating event of the pathogenesis for both LP and LDE is not completely understood. Both conditions share similar immunologic mechanisms of persistently activated CD8 autocytotoxic T lymphocytes against epidermal cells.¹⁸ These cells can induce apoptosis of basal epidermal keratinocytes and generate various cytokines (eg, IFN-γ, IL-5) to enhance expression of class II MHC molecules and antigen presentation to CD4 T cells.^19-22 It is conceivable that one of the initiating factors may be related to components in vaccines.

Hepatitis B, influenza, and herpes zoster vaccines were the 3 most common vaccines implicated in postimmunization LP or LDEs in our study. The excipients of these vaccines were compared based on the product inserts to identify any common components. It was found that all 3 vaccines contain either yeast protein or egg protein with various forms of phosphate buffers, while the hepatitis A and herpes zoster vaccines share Medical Research Council cell strain 5 (human diploid) cells as well as other cellular components.²³ Sato et al⁴ suggested that specific vaccine components, such as the vaccine itself or egg proteins, could have contributed to the development of LP following vaccination. It has been postulated that the protein S fraction of hepatitis B surface antigen plays a crucial role in the pathogenesis of both LP and LDE after hepatitis B vaccination.^2,24 It is likely that protein S shares common epitopes on keratinocytes that are recognized by the immune system, thus activating cytotoxic T lymphocytes and inducing apoptosis.^2,24

In this study, the median time to onset of vaccine-related LP was 14 days, which is consistent with a case series by Sato et al,⁴ suggesting that adverse reactions mainly occurred within 2 weeks after influenza vaccination. Onset of symptoms within 2 weeks of vaccination would therefore be a crucial clue for diagnosing possible vaccine-related LP or LDE. On the other hand, at least 4 patients in our study had onset of LP and LDE more than 1 month after vaccination; 2 of 4 cases even reported symptom onset at 175 and 297 days after hepatitis B vaccination, which were much longer than the 120 days reported by Tarakji et al.² It is not known if these cases constitute true vaccine-associated LP or LDE or if unmeasured confounding factors such as concurrent medications or comorbidities may have contributed to the development of these AEs.

It also is interesting to note that LP and LDE affected mainly middle-aged women. An increased risk of autoimmunity in female adults partly explains this observation.²⁵ Some vaccines, such as herpes zoster and influenza vaccines, generally are recommended for older adults who also are more likely to have multiple comorbidities or take multiple medications/supplements, which can potentially skew the prevalence of AEs toward an older age group. It should be noted, however, that LP and LDE were relatively uncommon AEs following vaccination in the current study. In this study, LP and LDE consisted of only 0.01% (N=42,230) of all AEs after hepatitis B vaccination, while the more common AEs such as pyrexia, nonspecific rashes, nonspecific gastrointestinal symptoms, and headache contributed to approximately 66.5% of all reported events.

One of the strengths of our study is that up to two-thirds of cases were confirmed histologically and all patients were seen and followed up by dermatologists or physicians. The VAERS is an easily accessible, up-to-date, and live reporting system that collects all AEs associated with vaccines in the United States. Important clinical and laboratory information usually is available in the database; however, the main limitation is that this study can only demonstrate a possible association but not a causal relationship between vaccination and LP or LDE. There can be various sources of biases such as underreporting, overreporting, or inaccurate reporting.^26,27 Pertinent clinical information (eg, new medications, new dental fillings/implants) that could potentially misrepresent the actual relationship between vaccination and development of AEs also was not available in the VAERS database. A cohort study with long-term follow-up or a large-scale case-control study would be useful in evaluating such associations.

Conclusion

Lichen planus and LDE can occur, albeit rarely, after vaccination, especially following hepatitis B vaccination. When middle-aged adults present to the clinic with LP or LDE, it is important to inquire about recent vaccination history in addition to a detailed medication history.

Take-Home Points

• Injuries to the medial knee are the most common knee ligament injuries, and often occur in the athletic population.
• Complete posteromedial corner injuries require surgical treatment to restore joint stability and biomechanics.
• Biomechanical evidence has demonstrated an important load-sharing distribution between the sMCL and the POL.
• Valgus instability caused by a medial side injury, can lead to both ACL/posterior cruciate ligament reconstruction graft failure if the medial sided injury is not concurrently repaired or reconstructed.
• Anatomic posteromedial corner reconstruction yields excellent biomechanical and patient-reported outcomes.

Most injuries of the medial structures of the knee are treated conservatively.^1-3 In severe acute injuries and chronic symptomatic instabilities, however, surgical treatment is needed to restore knee stability and to prevent degenerative changes secondary to instability.⁴ Three structures involved in medial stability are the superficial medial collateral ligament (sMCL), which is the primary valgus restraint; the posterior oblique ligament (POL), which is the primary restraint to internal rotation and the secondary valgus restraint; and the semimembranosus.^5,6

Surgical techniques for posteromedial knee reconstruction include direct repair,⁷ repair with augmentation,^8,9 advancement of the tibial insertion of the sMCL,¹⁰ and transfer of the pes anserine tendons.¹¹ In anatomical reconstruction of the posteromedial corner, which has been described before, the sMCL and the POL are reconstructed to reproduce the native motion and stability of the knee.¹² Clinically, repair and reconstruction have similar patient-reported outcomes and medial opening evaluations over the short term.

These approaches require large incisions and extensive dissection of soft tissue on the medial aspect of the knee.⁵ Given these drawbacks, it is reasonable to consider less invasive options. Minimally invasive surgery has the advantages of reduced scarring and blood loss, less disruption of surrounding tissue, faster recovery, and improved aesthetics.⁴

We conducted a study of a minimally invasive technique for reconstructing the posteromedial structures of the knee. We compared medial compartment stability measured on valgus stress radiographs in intact, sectioned, and reconstructed states in cadaveric knees. We hypothesized that a minimally invasive technique using autogenous hamstring graft in the appropriate anatomical location would return valgus stability to its nearly native state.

Materials and Methods

This study was conducted at the Buenos Aires British Hospital in Buenos Aires, Argentina, and at the University of Colorado Hospital in Aurora. Ten fresh-frozen cadaveric knees with no evidence of ligamentous injuries, osteoarthritis, or previous surgery were used. Mean donor age was 69.4 years (range, 45-87 years). Each specimen was maintained at room temperature for 24 hours before use. The femur was sectioned 20 cm proximal to the knee joint. The tibia was sectioned 12.5 cm distal to the knee joint.

Identification and Sectioning of Posteromedial Structures

After intact-state evaluation, each knee’s sMCL, dMCL, and POL were sectioned at their tibial insertion. Valgus stress radiograph was repeated and medial compartment gap was remeasured for comparison of the sectioned state with the intact and reconstructed states.

Anatomical Reconstruction With Mini-Invasive Technique

After sectioning of medial stabilizing structures, minimally invasive reconstruction was performed through 2 small incisions on the medial aspect of each of the 10 knees, as follows. First, the semitendinosus tendon was identified through the oblique incision that had been used for sectioning. Then, an open-ended tendon stripper was placed around the circumference of the semitendinosus and was passed proximomedially, transecting the tendon at its musculotendinous junction. While the tendon stripper was being passed, care was taken to maintain the nearby tibial insertion of the sartorius fascia (Figures 1D-1F).

With the semitendinosus tendon looped around the wire, isometricity was tested by pulling the suture within the tendon and moving the knee through a full range of motion. The isometric point was confirmed by tendon migration of <2 mm.¹³ Migration was measured by marking the graft 2 mm from its insertion; the graft was then pulled to ensure correct isometric point position. An 18-mm cannulated spiked screw and washer (Arthrex) were then passed over the wire and partially secured to the femur—the attachment point for the proximal sMCL portion of the semitendinosus graft. The semitendinosus tendon was then secured beneath the spiked washer with the knee in 20° of flexion with neutral rotation, recreating the sMCL.

Posteriorly, the distal insertion site of the POL was identified at the posteromedial aspect of the tibia through the oblique incision previously described. A 7-mm tunnel was drilled starting posteromedial (10 mm under tibial articular surface) and exiting just distal and medial to the Gerdy tubercle.

After final fixation, the medial knee was openly dissected to assess the inverted-V ligament reconstruction for anatomical placement and avoidance of crucial structures.

Stability Testing

Per International Knee Documentation Committee guidelines for stressing the medial compartment,¹⁴ valgus stress radiographs were obtained for all specimens at 0° and 20° of flexion in intact, sectioned, and reconstructed states.

The medial gap formed by the femoral condyle and its corresponding tibial plateau (at site of maximal separation) was tested in all 3 state conditions (intact, sectioned, reconstructed). Distances were digitally measured with a picture archiving and communication system viewer (Imagecast; IDX Systems Corporation). Medial gap was measured by taking the shortest distance between the subchondral bone surface of the most distal aspect of the medial femoral condyle and the corresponding medial tibial plateau. Three independent examiners took all the measurements; each examiner was blinded to the others’ measurements.

Statistics

Paired Student t tests were used to compare the 3 conditions, and the Shapiro-Wilk test was used to check for a normally distributed population. Statistical significance was set at P < .05. Statistical analyses were performed with GraphPad software.

Results

In all 10 specimens, the sMCL, the dMCL, and the POL were successfully identified and sectioned through a medial oblique incision over the distal insertion of the structures.

During all valgus testing states, there was no loss of graft fixation, and there was no gross graft slippage. In addition, all grafts remained in continuity with no evidence of failure, and there were no failures or breakages of the proximal or distal screw.

After posteromedial sectioning, mean medial gap was statistically significantly larger (P = .0002) at full extension (11 mm vs 3.3 mm) and at 20° of flexion (12.6 mm vs 3.8 mm). There was no statistically significant difference between the value of the intact state and the value after minimally invasive reconstruction at 0° (P = .56) or 20° (P = .102) of flexion.

Discussion

In this article, we describe a minimally invasive technique for anatomical posteromedial reconstruction of the knee in a cadaveric model. This technique restores the knee’s native valgus stability without causing extensive damage to the surrounding soft tissues and thereby potentially prevents scar formation and reduces blood loss.

Superficial MCL injury, one of the most common knee ligament injuries, is often associated with POL injury.⁷ Although most sMCL injuries are treated nonoperatively, with good results,³ surgical treatment is needed for severe (grade III) instabilities, symptomatic chronic instabilities, and knee dislocations.^12,17 Most posteromedial reconstruction techniques require an extensive approach that causes damage to surrounding soft tissue,^6,7,9,10 which in turn may compromise healing and positive patient outcomes. Surgical techniques include direct repair with sutures or anchors,¹⁸ capsular procedures,¹⁹ augmentations,⁹ internal bracing,⁶ and complete reconstruction of the posteromedial corner.²⁰

LaPrade and Wijdicks¹² have previously described anatomical reconstruction of the posteromedial corner. In their technique, a split semitendinosus autograft is used to reconstruct the sMCL and the POL separately, using 4 implants and reproducing each ligament’s anatomical attachment site. In this proposed technique, the distal attachment of the semitendinosus insertion is left intact, and uses 1 attachment point on the distal femur and 1 on the proximal tibia, allowing use of only 2 implants. In addition, it is performed with a minimally invasive approach, reduces cost, limits surgical exposure, and with experience may shorten operative time. To reduce the graft failure rate, the technique of LaPrade and Wijdicks¹² positions the sMCL tibial attachment as posterior as possible, which can be performed with this minimally invasive approach as well.

To reduce the graft failure rate, the technique of LaPrade and Wijdicks¹² positions the sMCL as posterior as possible. Despite the potential for increased graft stress with an anterior position, as in our modified technique, our group of 10 knees had no graft fixation failures in isolated valgus stress testing in either extension or flexion. Our minimally invasive posteromedial knee reconstruction significantly improved knee stability over the sectioned state as well as medial compartment gapping with valgus stress. There was no significant difference in medial compartment gapping between the intact and reconstructed states.

Our technique was built on open procedures (described by Kim and colleagues¹³) that carefully identify the isometric point of the graft. In addition, it adopted the modification (proposed by Lind and colleagues²¹) in which a fixation point is added at the distal insertion of the POL instead of being sutured to the direct arm of the semitendinosus tendon.

Furthermore, our technique, despite being similar to those described by Dong and colleagues²² and Borden and colleagues,²³  has the advantages of minimally invasive surgery and reduced disruption of soft tissues. Dong and colleagues²² reported on 64 patients with a mean follow-up of 34 months; patients’ medial opening measurements were significantly decreased at follow-up and fell within the normal range. 

The present study had several limitations. First, the age of our specimens was higher than the mean age of patients with knee ligament injury, potentially leading to firmer or more fibrotic tendons less susceptible to elongation. Second, we did not evaluate the knees’ rotational stability, and anterior cruciate ligaments (ACLs) were intact. As most posteromedial injuries co-occur with ACL injuries, a more realistic situation would have been reproduced by assessing rotational stability while performing both ACL reconstruction and the proposed posteromedial reconstruction. Third, static specimen measurements do not reflect the dynamic function of the posteromedial corner. Prospective clinical studies are needed to assess the true effectiveness of the posteromedial corner in the clinical scenario.

Knowledge of the anatomy of the medial aspect of the knee is vital to reconstruction of the medial side of the knee. Our results suggest that a minimally invasive technique can restore valgus stability without the need for extensive dissection and disruption of surrounding soft tissues. More research is needed to determine the results of this technique in vivo.

Take-Home Points

• Injuries to the medial knee are the most common knee ligament injuries, and often occur in the athletic population.
• Complete posteromedial corner injuries require surgical treatment to restore joint stability and biomechanics.
• Biomechanical evidence has demonstrated an important load-sharing distribution between the sMCL and the POL.
• Valgus instability caused by a medial side injury, can lead to both ACL/posterior cruciate ligament reconstruction graft failure if the medial sided injury is not concurrently repaired or reconstructed.
• Anatomic posteromedial corner reconstruction yields excellent biomechanical and patient-reported outcomes.

Most injuries of the medial structures of the knee are treated conservatively.^1-3 In severe acute injuries and chronic symptomatic instabilities, however, surgical treatment is needed to restore knee stability and to prevent degenerative changes secondary to instability.⁴ Three structures involved in medial stability are the superficial medial collateral ligament (sMCL), which is the primary valgus restraint; the posterior oblique ligament (POL), which is the primary restraint to internal rotation and the secondary valgus restraint; and the semimembranosus.^5,6

Surgical techniques for posteromedial knee reconstruction include direct repair,⁷ repair with augmentation,^8,9 advancement of the tibial insertion of the sMCL,¹⁰ and transfer of the pes anserine tendons.¹¹ In anatomical reconstruction of the posteromedial corner, which has been described before, the sMCL and the POL are reconstructed to reproduce the native motion and stability of the knee.¹² Clinically, repair and reconstruction have similar patient-reported outcomes and medial opening evaluations over the short term.

These approaches require large incisions and extensive dissection of soft tissue on the medial aspect of the knee.⁵ Given these drawbacks, it is reasonable to consider less invasive options. Minimally invasive surgery has the advantages of reduced scarring and blood loss, less disruption of surrounding tissue, faster recovery, and improved aesthetics.⁴

We conducted a study of a minimally invasive technique for reconstructing the posteromedial structures of the knee. We compared medial compartment stability measured on valgus stress radiographs in intact, sectioned, and reconstructed states in cadaveric knees. We hypothesized that a minimally invasive technique using autogenous hamstring graft in the appropriate anatomical location would return valgus stability to its nearly native state.

Materials and Methods

This study was conducted at the Buenos Aires British Hospital in Buenos Aires, Argentina, and at the University of Colorado Hospital in Aurora. Ten fresh-frozen cadaveric knees with no evidence of ligamentous injuries, osteoarthritis, or previous surgery were used. Mean donor age was 69.4 years (range, 45-87 years). Each specimen was maintained at room temperature for 24 hours before use. The femur was sectioned 20 cm proximal to the knee joint. The tibia was sectioned 12.5 cm distal to the knee joint.

Identification and Sectioning of Posteromedial Structures

After intact-state evaluation, each knee’s sMCL, dMCL, and POL were sectioned at their tibial insertion. Valgus stress radiograph was repeated and medial compartment gap was remeasured for comparison of the sectioned state with the intact and reconstructed states.

Anatomical Reconstruction With Mini-Invasive Technique

After sectioning of medial stabilizing structures, minimally invasive reconstruction was performed through 2 small incisions on the medial aspect of each of the 10 knees, as follows. First, the semitendinosus tendon was identified through the oblique incision that had been used for sectioning. Then, an open-ended tendon stripper was placed around the circumference of the semitendinosus and was passed proximomedially, transecting the tendon at its musculotendinous junction. While the tendon stripper was being passed, care was taken to maintain the nearby tibial insertion of the sartorius fascia (Figures 1D-1F).

With the semitendinosus tendon looped around the wire, isometricity was tested by pulling the suture within the tendon and moving the knee through a full range of motion. The isometric point was confirmed by tendon migration of <2 mm.¹³ Migration was measured by marking the graft 2 mm from its insertion; the graft was then pulled to ensure correct isometric point position. An 18-mm cannulated spiked screw and washer (Arthrex) were then passed over the wire and partially secured to the femur—the attachment point for the proximal sMCL portion of the semitendinosus graft. The semitendinosus tendon was then secured beneath the spiked washer with the knee in 20° of flexion with neutral rotation, recreating the sMCL.

Posteriorly, the distal insertion site of the POL was identified at the posteromedial aspect of the tibia through the oblique incision previously described. A 7-mm tunnel was drilled starting posteromedial (10 mm under tibial articular surface) and exiting just distal and medial to the Gerdy tubercle.

After final fixation, the medial knee was openly dissected to assess the inverted-V ligament reconstruction for anatomical placement and avoidance of crucial structures.

Stability Testing

Per International Knee Documentation Committee guidelines for stressing the medial compartment,¹⁴ valgus stress radiographs were obtained for all specimens at 0° and 20° of flexion in intact, sectioned, and reconstructed states.

The medial gap formed by the femoral condyle and its corresponding tibial plateau (at site of maximal separation) was tested in all 3 state conditions (intact, sectioned, reconstructed). Distances were digitally measured with a picture archiving and communication system viewer (Imagecast; IDX Systems Corporation). Medial gap was measured by taking the shortest distance between the subchondral bone surface of the most distal aspect of the medial femoral condyle and the corresponding medial tibial plateau. Three independent examiners took all the measurements; each examiner was blinded to the others’ measurements.

Statistics

Paired Student t tests were used to compare the 3 conditions, and the Shapiro-Wilk test was used to check for a normally distributed population. Statistical significance was set at P < .05. Statistical analyses were performed with GraphPad software.

Results

In all 10 specimens, the sMCL, the dMCL, and the POL were successfully identified and sectioned through a medial oblique incision over the distal insertion of the structures.

During all valgus testing states, there was no loss of graft fixation, and there was no gross graft slippage. In addition, all grafts remained in continuity with no evidence of failure, and there were no failures or breakages of the proximal or distal screw.

After posteromedial sectioning, mean medial gap was statistically significantly larger (P = .0002) at full extension (11 mm vs 3.3 mm) and at 20° of flexion (12.6 mm vs 3.8 mm). There was no statistically significant difference between the value of the intact state and the value after minimally invasive reconstruction at 0° (P = .56) or 20° (P = .102) of flexion.

Discussion

In this article, we describe a minimally invasive technique for anatomical posteromedial reconstruction of the knee in a cadaveric model. This technique restores the knee’s native valgus stability without causing extensive damage to the surrounding soft tissues and thereby potentially prevents scar formation and reduces blood loss.

Superficial MCL injury, one of the most common knee ligament injuries, is often associated with POL injury.⁷ Although most sMCL injuries are treated nonoperatively, with good results,³ surgical treatment is needed for severe (grade III) instabilities, symptomatic chronic instabilities, and knee dislocations.^12,17 Most posteromedial reconstruction techniques require an extensive approach that causes damage to surrounding soft tissue,^6,7,9,10 which in turn may compromise healing and positive patient outcomes. Surgical techniques include direct repair with sutures or anchors,¹⁸ capsular procedures,¹⁹ augmentations,⁹ internal bracing,⁶ and complete reconstruction of the posteromedial corner.²⁰

LaPrade and Wijdicks¹² have previously described anatomical reconstruction of the posteromedial corner. In their technique, a split semitendinosus autograft is used to reconstruct the sMCL and the POL separately, using 4 implants and reproducing each ligament’s anatomical attachment site. In this proposed technique, the distal attachment of the semitendinosus insertion is left intact, and uses 1 attachment point on the distal femur and 1 on the proximal tibia, allowing use of only 2 implants. In addition, it is performed with a minimally invasive approach, reduces cost, limits surgical exposure, and with experience may shorten operative time. To reduce the graft failure rate, the technique of LaPrade and Wijdicks¹² positions the sMCL tibial attachment as posterior as possible, which can be performed with this minimally invasive approach as well.

To reduce the graft failure rate, the technique of LaPrade and Wijdicks¹² positions the sMCL as posterior as possible. Despite the potential for increased graft stress with an anterior position, as in our modified technique, our group of 10 knees had no graft fixation failures in isolated valgus stress testing in either extension or flexion. Our minimally invasive posteromedial knee reconstruction significantly improved knee stability over the sectioned state as well as medial compartment gapping with valgus stress. There was no significant difference in medial compartment gapping between the intact and reconstructed states.

Our technique was built on open procedures (described by Kim and colleagues¹³) that carefully identify the isometric point of the graft. In addition, it adopted the modification (proposed by Lind and colleagues²¹) in which a fixation point is added at the distal insertion of the POL instead of being sutured to the direct arm of the semitendinosus tendon.

Furthermore, our technique, despite being similar to those described by Dong and colleagues²² and Borden and colleagues,²³  has the advantages of minimally invasive surgery and reduced disruption of soft tissues. Dong and colleagues²² reported on 64 patients with a mean follow-up of 34 months; patients’ medial opening measurements were significantly decreased at follow-up and fell within the normal range. 

The present study had several limitations. First, the age of our specimens was higher than the mean age of patients with knee ligament injury, potentially leading to firmer or more fibrotic tendons less susceptible to elongation. Second, we did not evaluate the knees’ rotational stability, and anterior cruciate ligaments (ACLs) were intact. As most posteromedial injuries co-occur with ACL injuries, a more realistic situation would have been reproduced by assessing rotational stability while performing both ACL reconstruction and the proposed posteromedial reconstruction. Third, static specimen measurements do not reflect the dynamic function of the posteromedial corner. Prospective clinical studies are needed to assess the true effectiveness of the posteromedial corner in the clinical scenario.

Knowledge of the anatomy of the medial aspect of the knee is vital to reconstruction of the medial side of the knee. Our results suggest that a minimally invasive technique can restore valgus stability without the need for extensive dissection and disruption of surrounding soft tissues. More research is needed to determine the results of this technique in vivo.

Take-Home Points

• Injuries to the medial knee are the most common knee ligament injuries, and often occur in the athletic population.
• Complete posteromedial corner injuries require surgical treatment to restore joint stability and biomechanics.
• Biomechanical evidence has demonstrated an important load-sharing distribution between the sMCL and the POL.
• Valgus instability caused by a medial side injury, can lead to both ACL/posterior cruciate ligament reconstruction graft failure if the medial sided injury is not concurrently repaired or reconstructed.
• Anatomic posteromedial corner reconstruction yields excellent biomechanical and patient-reported outcomes.

Most injuries of the medial structures of the knee are treated conservatively.^1-3 In severe acute injuries and chronic symptomatic instabilities, however, surgical treatment is needed to restore knee stability and to prevent degenerative changes secondary to instability.⁴ Three structures involved in medial stability are the superficial medial collateral ligament (sMCL), which is the primary valgus restraint; the posterior oblique ligament (POL), which is the primary restraint to internal rotation and the secondary valgus restraint; and the semimembranosus.^5,6

Surgical techniques for posteromedial knee reconstruction include direct repair,⁷ repair with augmentation,^8,9 advancement of the tibial insertion of the sMCL,¹⁰ and transfer of the pes anserine tendons.¹¹ In anatomical reconstruction of the posteromedial corner, which has been described before, the sMCL and the POL are reconstructed to reproduce the native motion and stability of the knee.¹² Clinically, repair and reconstruction have similar patient-reported outcomes and medial opening evaluations over the short term.

These approaches require large incisions and extensive dissection of soft tissue on the medial aspect of the knee.⁵ Given these drawbacks, it is reasonable to consider less invasive options. Minimally invasive surgery has the advantages of reduced scarring and blood loss, less disruption of surrounding tissue, faster recovery, and improved aesthetics.⁴

We conducted a study of a minimally invasive technique for reconstructing the posteromedial structures of the knee. We compared medial compartment stability measured on valgus stress radiographs in intact, sectioned, and reconstructed states in cadaveric knees. We hypothesized that a minimally invasive technique using autogenous hamstring graft in the appropriate anatomical location would return valgus stability to its nearly native state.

Materials and Methods

This study was conducted at the Buenos Aires British Hospital in Buenos Aires, Argentina, and at the University of Colorado Hospital in Aurora. Ten fresh-frozen cadaveric knees with no evidence of ligamentous injuries, osteoarthritis, or previous surgery were used. Mean donor age was 69.4 years (range, 45-87 years). Each specimen was maintained at room temperature for 24 hours before use. The femur was sectioned 20 cm proximal to the knee joint. The tibia was sectioned 12.5 cm distal to the knee joint.

Identification and Sectioning of Posteromedial Structures

After intact-state evaluation, each knee’s sMCL, dMCL, and POL were sectioned at their tibial insertion. Valgus stress radiograph was repeated and medial compartment gap was remeasured for comparison of the sectioned state with the intact and reconstructed states.

Anatomical Reconstruction With Mini-Invasive Technique

After sectioning of medial stabilizing structures, minimally invasive reconstruction was performed through 2 small incisions on the medial aspect of each of the 10 knees, as follows. First, the semitendinosus tendon was identified through the oblique incision that had been used for sectioning. Then, an open-ended tendon stripper was placed around the circumference of the semitendinosus and was passed proximomedially, transecting the tendon at its musculotendinous junction. While the tendon stripper was being passed, care was taken to maintain the nearby tibial insertion of the sartorius fascia (Figures 1D-1F).

With the semitendinosus tendon looped around the wire, isometricity was tested by pulling the suture within the tendon and moving the knee through a full range of motion. The isometric point was confirmed by tendon migration of <2 mm.¹³ Migration was measured by marking the graft 2 mm from its insertion; the graft was then pulled to ensure correct isometric point position. An 18-mm cannulated spiked screw and washer (Arthrex) were then passed over the wire and partially secured to the femur—the attachment point for the proximal sMCL portion of the semitendinosus graft. The semitendinosus tendon was then secured beneath the spiked washer with the knee in 20° of flexion with neutral rotation, recreating the sMCL.

Posteriorly, the distal insertion site of the POL was identified at the posteromedial aspect of the tibia through the oblique incision previously described. A 7-mm tunnel was drilled starting posteromedial (10 mm under tibial articular surface) and exiting just distal and medial to the Gerdy tubercle.

After final fixation, the medial knee was openly dissected to assess the inverted-V ligament reconstruction for anatomical placement and avoidance of crucial structures.

Stability Testing

Per International Knee Documentation Committee guidelines for stressing the medial compartment,¹⁴ valgus stress radiographs were obtained for all specimens at 0° and 20° of flexion in intact, sectioned, and reconstructed states.

The medial gap formed by the femoral condyle and its corresponding tibial plateau (at site of maximal separation) was tested in all 3 state conditions (intact, sectioned, reconstructed). Distances were digitally measured with a picture archiving and communication system viewer (Imagecast; IDX Systems Corporation). Medial gap was measured by taking the shortest distance between the subchondral bone surface of the most distal aspect of the medial femoral condyle and the corresponding medial tibial plateau. Three independent examiners took all the measurements; each examiner was blinded to the others’ measurements.

Statistics

Paired Student t tests were used to compare the 3 conditions, and the Shapiro-Wilk test was used to check for a normally distributed population. Statistical significance was set at P < .05. Statistical analyses were performed with GraphPad software.

Results

In all 10 specimens, the sMCL, the dMCL, and the POL were successfully identified and sectioned through a medial oblique incision over the distal insertion of the structures.

During all valgus testing states, there was no loss of graft fixation, and there was no gross graft slippage. In addition, all grafts remained in continuity with no evidence of failure, and there were no failures or breakages of the proximal or distal screw.

After posteromedial sectioning, mean medial gap was statistically significantly larger (P = .0002) at full extension (11 mm vs 3.3 mm) and at 20° of flexion (12.6 mm vs 3.8 mm). There was no statistically significant difference between the value of the intact state and the value after minimally invasive reconstruction at 0° (P = .56) or 20° (P = .102) of flexion.

Discussion

In this article, we describe a minimally invasive technique for anatomical posteromedial reconstruction of the knee in a cadaveric model. This technique restores the knee’s native valgus stability without causing extensive damage to the surrounding soft tissues and thereby potentially prevents scar formation and reduces blood loss.

Superficial MCL injury, one of the most common knee ligament injuries, is often associated with POL injury.⁷ Although most sMCL injuries are treated nonoperatively, with good results,³ surgical treatment is needed for severe (grade III) instabilities, symptomatic chronic instabilities, and knee dislocations.^12,17 Most posteromedial reconstruction techniques require an extensive approach that causes damage to surrounding soft tissue,^6,7,9,10 which in turn may compromise healing and positive patient outcomes. Surgical techniques include direct repair with sutures or anchors,¹⁸ capsular procedures,¹⁹ augmentations,⁹ internal bracing,⁶ and complete reconstruction of the posteromedial corner.²⁰

LaPrade and Wijdicks¹² have previously described anatomical reconstruction of the posteromedial corner. In their technique, a split semitendinosus autograft is used to reconstruct the sMCL and the POL separately, using 4 implants and reproducing each ligament’s anatomical attachment site. In this proposed technique, the distal attachment of the semitendinosus insertion is left intact, and uses 1 attachment point on the distal femur and 1 on the proximal tibia, allowing use of only 2 implants. In addition, it is performed with a minimally invasive approach, reduces cost, limits surgical exposure, and with experience may shorten operative time. To reduce the graft failure rate, the technique of LaPrade and Wijdicks¹² positions the sMCL tibial attachment as posterior as possible, which can be performed with this minimally invasive approach as well.

To reduce the graft failure rate, the technique of LaPrade and Wijdicks¹² positions the sMCL as posterior as possible. Despite the potential for increased graft stress with an anterior position, as in our modified technique, our group of 10 knees had no graft fixation failures in isolated valgus stress testing in either extension or flexion. Our minimally invasive posteromedial knee reconstruction significantly improved knee stability over the sectioned state as well as medial compartment gapping with valgus stress. There was no significant difference in medial compartment gapping between the intact and reconstructed states.

Our technique was built on open procedures (described by Kim and colleagues¹³) that carefully identify the isometric point of the graft. In addition, it adopted the modification (proposed by Lind and colleagues²¹) in which a fixation point is added at the distal insertion of the POL instead of being sutured to the direct arm of the semitendinosus tendon.

Furthermore, our technique, despite being similar to those described by Dong and colleagues²² and Borden and colleagues,²³  has the advantages of minimally invasive surgery and reduced disruption of soft tissues. Dong and colleagues²² reported on 64 patients with a mean follow-up of 34 months; patients’ medial opening measurements were significantly decreased at follow-up and fell within the normal range. 

The present study had several limitations. First, the age of our specimens was higher than the mean age of patients with knee ligament injury, potentially leading to firmer or more fibrotic tendons less susceptible to elongation. Second, we did not evaluate the knees’ rotational stability, and anterior cruciate ligaments (ACLs) were intact. As most posteromedial injuries co-occur with ACL injuries, a more realistic situation would have been reproduced by assessing rotational stability while performing both ACL reconstruction and the proposed posteromedial reconstruction. Third, static specimen measurements do not reflect the dynamic function of the posteromedial corner. Prospective clinical studies are needed to assess the true effectiveness of the posteromedial corner in the clinical scenario.

Knowledge of the anatomy of the medial aspect of the knee is vital to reconstruction of the medial side of the knee. Our results suggest that a minimally invasive technique can restore valgus stability without the need for extensive dissection and disruption of surrounding soft tissues. More research is needed to determine the results of this technique in vivo.

Take-Home Points

• There is a 41% rate of AVN or PTOA after operatively managed talus fracture.
• Surgical timing does not affect development of AVN or PTOA.
• Open fractures are associated with development of AVN and PTOA.
• Quality of reduction is likely more important than timing of reduction.
• Urgent surgical treatment is necessary for threatened soft tissue or neurovascular compromise.

Talus fractures are rare injuries that present a significant treatment dilemma.^1-12 These fractures represent <1% of all fractures⁴ and are second only to calcaneus fractures in fractures of the hindfoot. Talus fractures with associated dislocations are even rarer and may provide treating surgeons with a significant surgical quandary.^6,13-16

Talus fractures historically have been characterized by their anatomical location: head, neck, or body. Two systems are commonly used to classify talus fractures: Hawkins and AO/OTA (Arbeitsgemeinschaft für Osteosynthesefragen/Orthopaedic Trauma Association). The first, developed by Hawkins⁷ and modified by Canale and Kelly² and Vallier and colleagues,¹ identifies 4 basic fracture types with associated dislocations. The other system, published in 1996¹⁷ and republished in 2007,¹⁸ uses the combined methods of AO and OTA to systematically describe talus fractures. Although these classification systems accurately describe talus fractures with associated dislocation, both have difficulty predicting clinical outcomes.^1,19,20

Talus fractures commonly result in avascular necrosis (AVN) of the talus and posttraumatic osteoarthritis (PTOA) of the tibiotalar and subtalar joints.^{3,8,9,12,14-16} Hawkins⁷ initially described subchondral lucency as indicating revascularization of the talus after injury. AVN and PTOA rates traditionally have been thought to be related to a blood supply disruption, given the prognostic value of the Hawkins sign.^1,7,12,21 New methods, including a dual-incision approach and expedited transfer to foot and ankle surgeons or orthopedic traumatologists, have improved reduction quality^21-24 but not patient outcomes.^{3,5,8,9,12,14}

Recently, time from injury to surgical intervention has been a topic of much discussion, and there have been studies on the specific effects of timing with respect to outcome.^1,15,16 Vallier and colleagues,¹ who wanted to identify injury characteristics predictive of osteonecrosis, found that delaying reduction and surgical fixation did not increase the risk of AVN. Another study found that urgent reduction of fracture-dislocation with delayed open reduction and internal fixation (ORIF) using a dual approach may improve clinical outcomes.²¹

In this vein, we conducted a study to evaluate the effect of time to surgical reduction of talus fractures and talus fracture-dislocations on the development of AVN and PTOA. We hypothesized that time to surgical reduction of talus fracture-dislocation as classified with the AO/OTA system would have no effect of the development of AVN/PTOA.

Methods

After this study received Institutional Review Board approval, we retrospectively reviewed the records on talus fractures surgically managed at a level I trauma center during the 10-year period 2003 to 2013. Of the 119 potential cases identified using Current Procedural Terminology code 28445 (ORIF of talus), 13 were excluded (12 for inaccurate coding or missing documentation, 1 for being a pediatric case), leaving 106 for analysis. Using the Hawkins and AO/OTA systems, 3 independent reviewers classified the injuries on plain radiographs.

Injury dates and times were obtained from the medical records. Operating room start times were also obtained. Surgical timing was defined as time from injury to operating room start. For cases without an injury time, time of presentation to emergency department was used.

Open fracture-dislocations were managed with intravenous antibiotics, urgent surgical irrigation, débridement, and immediate fixation or temporizing external fixation after reduction. All fractures were definitively managed with standard ORIF with an anteromedial, anterolateral, or dual approach and mini-fragment implants. After fixation, weight-bearing typically was restricted for 6 to 12 weeks.

Follow-up radiographs were evaluated. Presence or absence of Hawkins sign⁷ was noted on radiographs at 6 or 8 weeks, and all follow-up radiographs were evaluated for AVN as defined by increased radiographic density within the talar dome or collapse of the articular surface. All radiographs were evaluated for PTOA as defined by loss of joint space within the tibiotalar, subtalar, or talonavicular joint on follow-up radiographs.

Clinical outcomes were analyzed for development of AVN, PTOA, or secondary corrective surgery or arthrodesis. Continuous variables were evaluated with the t test, and the χ² test was used to compare distributions of categorical variables. The Wilcoxon rank sum test was used to compare non-normally distributed variables. Significance was set at P < .05.

Classification Analysis (Table 1)

Subject Analysis (Table 2)

The mechanisms of injury were motor vehicle accident (70/106; 66%), fall from height (25; 24%), misstep (4), sports related (2), object falling on ankle (2), and not reported (3). 

Of the 106 patients, 45 (42%) had isolated talus injuries, 35 had concomitant ipsilateral lower extremity injuries, 25 had concomitant contralateral lower extremity injuries, and 1 had a concomitant upper extremity injury. 

Smoking status was everyday (14 patients), past (10), never (34), and unreported (48). Five patients reported a history of alcohol abuse, and 4 patients reported illicit drug use. Two had a history of atrial fibrillation, 9 had hypertension, 3 had hyperlipidemia, 3 had renal disease, 3 had heart disease, 4 had diabetes, 3 had lung disease, and 1 had a history of lung cancer.

Overall Analysis of AVN/PTOA (Table 3)

Analysis of AVN/PTOA in 81-B3 Fracture-Dislocations (Table 4)

Discussion

Our results showed that time from talus fracture-dislocation to surgical reduction had no effect on development of AVN/PTOA. The findings in this largest series to date agree with earlier findings^1,8,15,16,24 and add to the volume of literature suggesting that time to surgical reduction of talus fractures and talus fracture-dislocations does not markedly affect outcome.

Talus fractures continue to present a significant treatment dilemma. Despite recent improvements in surgical techniques and overall management of these injuries, rates of AVN and PTOA have not significantly decreased.^1,16,23 At most treating facilities, talus fracture-dislocations are considered surgical emergencies/urgencies, and every effort is made to reduce and surgically address these injuries as soon as possible.^1,13

In this study, rates of AVN/PTOA were 41% (all talus fractures) and 50% (displaced talar neck fractures), and the difference was not significant (Table 3). These rates are higher but consistent with previously reported rates (range, 14%-49%).^{1,2,7-9,12,14,24} There was no difference in surgical timing for development of AVN/PTOA. We analyzed the cases of all patients who had talus fractures and developed AVN/PTOA (43/106). Within this group, there were no significant differences in surgical timing, age, sex, polytrauma, or BMI between patients who developed AVN/PTOA and those who did not. Compared with patients who did not develop AVN/PTOA, those who developed AVN/PTOA were significantly more likely to have open injuries. This finding, consistent with those in other reports^9,12,13 (Table 3), indicates outcome is more likely related to injury severity and not necessarily injury class.

We retrospectively analyzed talus fractures and talus fracture-dislocations to determine if urgent surgical management affects outcomes. Current practice at our institution is to routinely reduce and surgically address these fractures urgently, often during the middle of the night, when orthopedic resources are reduced. Our study found a significant difference in surgical timing for patients with talus fracture-dislocations and patients with talus fractures without dislocations (Table 2). Given our findings, urgent surgical reduction and fixation are not indicated to preserve the talus blood supply and prevent AVN/PTOA, though we still recommend urgent surgical management in the setting of an open wound, skin necrosis, or soft-tissue/neurovascular compromise. 

This study had several limitations, primarily related to its retrospective nature. Surgical timing was defined as time from injury, as noted in the medical record, to operating room start. In some instances, time of injury was not noted in the medical record, and time of presentation to emergency room was used instead. Thus, surgical timing for these patients may have been longer than identified. In addition, given the rare injury pattern and the retrospective design, this study was susceptible to type II error and may have been underpowered to detect whether time to surgical reduction predicted complications. Also, the study did not address functional outcome as measured by validated outcome scores. Outcome measures were obtained in many but not all cases, making functional outcome measurement difficult. Similarly, the quality of the anatomical reductions was not assessed, potentially affecting complication rates. Postoperative reduction assessment, possibly performed with computed tomography, is an avenue of further study.

Strengths of this study include its large sample size (this was one of the largest studies of talus fractures), long follow-up (mean, 150 weeks), and novel use of AO/OTA classification.

We postulate that development of AVN/PTOA is not necessarily related to the urgency or timing of surgical reduction and fixation and is more likely related to injury severity. This idea is supported by the finding that development of AVN/PTOA was significantly correlated to open injuries in all talus fractures, including talus fracture-dislocations and isolated talus fractures.

Conclusion 

Talus fracture-dislocations are devastating injuries with high rates of complications. In this study, open talus fractures, and fractures with associated tibiotalar or subtalar dislocations, had higher complication rates. Given the evidence presented, we recommend basing surgical timing on injury severity, not necessarily for AVN/PTOA prevention. Specifically, in the absence of an open wound, skin necrosis, or soft-tissue/neurovascular compromise, talus fracture-dislocations can be surgically reduced and stabilized when optimal resources are available.

Take-Home Points

• There is a 41% rate of AVN or PTOA after operatively managed talus fracture.
• Surgical timing does not affect development of AVN or PTOA.
• Open fractures are associated with development of AVN and PTOA.
• Quality of reduction is likely more important than timing of reduction.
• Urgent surgical treatment is necessary for threatened soft tissue or neurovascular compromise.

Talus fractures are rare injuries that present a significant treatment dilemma.^1-12 These fractures represent <1% of all fractures⁴ and are second only to calcaneus fractures in fractures of the hindfoot. Talus fractures with associated dislocations are even rarer and may provide treating surgeons with a significant surgical quandary.^6,13-16

Talus fractures historically have been characterized by their anatomical location: head, neck, or body. Two systems are commonly used to classify talus fractures: Hawkins and AO/OTA (Arbeitsgemeinschaft für Osteosynthesefragen/Orthopaedic Trauma Association). The first, developed by Hawkins⁷ and modified by Canale and Kelly² and Vallier and colleagues,¹ identifies 4 basic fracture types with associated dislocations. The other system, published in 1996¹⁷ and republished in 2007,¹⁸ uses the combined methods of AO and OTA to systematically describe talus fractures. Although these classification systems accurately describe talus fractures with associated dislocation, both have difficulty predicting clinical outcomes.^1,19,20

Talus fractures commonly result in avascular necrosis (AVN) of the talus and posttraumatic osteoarthritis (PTOA) of the tibiotalar and subtalar joints.^{3,8,9,12,14-16} Hawkins⁷ initially described subchondral lucency as indicating revascularization of the talus after injury. AVN and PTOA rates traditionally have been thought to be related to a blood supply disruption, given the prognostic value of the Hawkins sign.^1,7,12,21 New methods, including a dual-incision approach and expedited transfer to foot and ankle surgeons or orthopedic traumatologists, have improved reduction quality^21-24 but not patient outcomes.^{3,5,8,9,12,14}

Recently, time from injury to surgical intervention has been a topic of much discussion, and there have been studies on the specific effects of timing with respect to outcome.^1,15,16 Vallier and colleagues,¹ who wanted to identify injury characteristics predictive of osteonecrosis, found that delaying reduction and surgical fixation did not increase the risk of AVN. Another study found that urgent reduction of fracture-dislocation with delayed open reduction and internal fixation (ORIF) using a dual approach may improve clinical outcomes.²¹

In this vein, we conducted a study to evaluate the effect of time to surgical reduction of talus fractures and talus fracture-dislocations on the development of AVN and PTOA. We hypothesized that time to surgical reduction of talus fracture-dislocation as classified with the AO/OTA system would have no effect of the development of AVN/PTOA.

Methods

After this study received Institutional Review Board approval, we retrospectively reviewed the records on talus fractures surgically managed at a level I trauma center during the 10-year period 2003 to 2013. Of the 119 potential cases identified using Current Procedural Terminology code 28445 (ORIF of talus), 13 were excluded (12 for inaccurate coding or missing documentation, 1 for being a pediatric case), leaving 106 for analysis. Using the Hawkins and AO/OTA systems, 3 independent reviewers classified the injuries on plain radiographs.

Injury dates and times were obtained from the medical records. Operating room start times were also obtained. Surgical timing was defined as time from injury to operating room start. For cases without an injury time, time of presentation to emergency department was used.

Open fracture-dislocations were managed with intravenous antibiotics, urgent surgical irrigation, débridement, and immediate fixation or temporizing external fixation after reduction. All fractures were definitively managed with standard ORIF with an anteromedial, anterolateral, or dual approach and mini-fragment implants. After fixation, weight-bearing typically was restricted for 6 to 12 weeks.

Follow-up radiographs were evaluated. Presence or absence of Hawkins sign⁷ was noted on radiographs at 6 or 8 weeks, and all follow-up radiographs were evaluated for AVN as defined by increased radiographic density within the talar dome or collapse of the articular surface. All radiographs were evaluated for PTOA as defined by loss of joint space within the tibiotalar, subtalar, or talonavicular joint on follow-up radiographs.

Clinical outcomes were analyzed for development of AVN, PTOA, or secondary corrective surgery or arthrodesis. Continuous variables were evaluated with the t test, and the χ² test was used to compare distributions of categorical variables. The Wilcoxon rank sum test was used to compare non-normally distributed variables. Significance was set at P < .05.

Classification Analysis (Table 1)

Subject Analysis (Table 2)

The mechanisms of injury were motor vehicle accident (70/106; 66%), fall from height (25; 24%), misstep (4), sports related (2), object falling on ankle (2), and not reported (3). 

Of the 106 patients, 45 (42%) had isolated talus injuries, 35 had concomitant ipsilateral lower extremity injuries, 25 had concomitant contralateral lower extremity injuries, and 1 had a concomitant upper extremity injury. 

Smoking status was everyday (14 patients), past (10), never (34), and unreported (48). Five patients reported a history of alcohol abuse, and 4 patients reported illicit drug use. Two had a history of atrial fibrillation, 9 had hypertension, 3 had hyperlipidemia, 3 had renal disease, 3 had heart disease, 4 had diabetes, 3 had lung disease, and 1 had a history of lung cancer.

Overall Analysis of AVN/PTOA (Table 3)

Analysis of AVN/PTOA in 81-B3 Fracture-Dislocations (Table 4)

Discussion

Our results showed that time from talus fracture-dislocation to surgical reduction had no effect on development of AVN/PTOA. The findings in this largest series to date agree with earlier findings^1,8,15,16,24 and add to the volume of literature suggesting that time to surgical reduction of talus fractures and talus fracture-dislocations does not markedly affect outcome.

Talus fractures continue to present a significant treatment dilemma. Despite recent improvements in surgical techniques and overall management of these injuries, rates of AVN and PTOA have not significantly decreased.^1,16,23 At most treating facilities, talus fracture-dislocations are considered surgical emergencies/urgencies, and every effort is made to reduce and surgically address these injuries as soon as possible.^1,13

In this study, rates of AVN/PTOA were 41% (all talus fractures) and 50% (displaced talar neck fractures), and the difference was not significant (Table 3). These rates are higher but consistent with previously reported rates (range, 14%-49%).^{1,2,7-9,12,14,24} There was no difference in surgical timing for development of AVN/PTOA. We analyzed the cases of all patients who had talus fractures and developed AVN/PTOA (43/106). Within this group, there were no significant differences in surgical timing, age, sex, polytrauma, or BMI between patients who developed AVN/PTOA and those who did not. Compared with patients who did not develop AVN/PTOA, those who developed AVN/PTOA were significantly more likely to have open injuries. This finding, consistent with those in other reports^9,12,13 (Table 3), indicates outcome is more likely related to injury severity and not necessarily injury class.

We retrospectively analyzed talus fractures and talus fracture-dislocations to determine if urgent surgical management affects outcomes. Current practice at our institution is to routinely reduce and surgically address these fractures urgently, often during the middle of the night, when orthopedic resources are reduced. Our study found a significant difference in surgical timing for patients with talus fracture-dislocations and patients with talus fractures without dislocations (Table 2). Given our findings, urgent surgical reduction and fixation are not indicated to preserve the talus blood supply and prevent AVN/PTOA, though we still recommend urgent surgical management in the setting of an open wound, skin necrosis, or soft-tissue/neurovascular compromise. 

This study had several limitations, primarily related to its retrospective nature. Surgical timing was defined as time from injury, as noted in the medical record, to operating room start. In some instances, time of injury was not noted in the medical record, and time of presentation to emergency room was used instead. Thus, surgical timing for these patients may have been longer than identified. In addition, given the rare injury pattern and the retrospective design, this study was susceptible to type II error and may have been underpowered to detect whether time to surgical reduction predicted complications. Also, the study did not address functional outcome as measured by validated outcome scores. Outcome measures were obtained in many but not all cases, making functional outcome measurement difficult. Similarly, the quality of the anatomical reductions was not assessed, potentially affecting complication rates. Postoperative reduction assessment, possibly performed with computed tomography, is an avenue of further study.

Strengths of this study include its large sample size (this was one of the largest studies of talus fractures), long follow-up (mean, 150 weeks), and novel use of AO/OTA classification.

We postulate that development of AVN/PTOA is not necessarily related to the urgency or timing of surgical reduction and fixation and is more likely related to injury severity. This idea is supported by the finding that development of AVN/PTOA was significantly correlated to open injuries in all talus fractures, including talus fracture-dislocations and isolated talus fractures.

Conclusion 

Talus fracture-dislocations are devastating injuries with high rates of complications. In this study, open talus fractures, and fractures with associated tibiotalar or subtalar dislocations, had higher complication rates. Given the evidence presented, we recommend basing surgical timing on injury severity, not necessarily for AVN/PTOA prevention. Specifically, in the absence of an open wound, skin necrosis, or soft-tissue/neurovascular compromise, talus fracture-dislocations can be surgically reduced and stabilized when optimal resources are available.

Take-Home Points

• There is a 41% rate of AVN or PTOA after operatively managed talus fracture.
• Surgical timing does not affect development of AVN or PTOA.
• Open fractures are associated with development of AVN and PTOA.
• Quality of reduction is likely more important than timing of reduction.
• Urgent surgical treatment is necessary for threatened soft tissue or neurovascular compromise.

Talus fractures are rare injuries that present a significant treatment dilemma.^1-12 These fractures represent <1% of all fractures⁴ and are second only to calcaneus fractures in fractures of the hindfoot. Talus fractures with associated dislocations are even rarer and may provide treating surgeons with a significant surgical quandary.^6,13-16

Talus fractures historically have been characterized by their anatomical location: head, neck, or body. Two systems are commonly used to classify talus fractures: Hawkins and AO/OTA (Arbeitsgemeinschaft für Osteosynthesefragen/Orthopaedic Trauma Association). The first, developed by Hawkins⁷ and modified by Canale and Kelly² and Vallier and colleagues,¹ identifies 4 basic fracture types with associated dislocations. The other system, published in 1996¹⁷ and republished in 2007,¹⁸ uses the combined methods of AO and OTA to systematically describe talus fractures. Although these classification systems accurately describe talus fractures with associated dislocation, both have difficulty predicting clinical outcomes.^1,19,20

Talus fractures commonly result in avascular necrosis (AVN) of the talus and posttraumatic osteoarthritis (PTOA) of the tibiotalar and subtalar joints.^{3,8,9,12,14-16} Hawkins⁷ initially described subchondral lucency as indicating revascularization of the talus after injury. AVN and PTOA rates traditionally have been thought to be related to a blood supply disruption, given the prognostic value of the Hawkins sign.^1,7,12,21 New methods, including a dual-incision approach and expedited transfer to foot and ankle surgeons or orthopedic traumatologists, have improved reduction quality^21-24 but not patient outcomes.^{3,5,8,9,12,14}

Recently, time from injury to surgical intervention has been a topic of much discussion, and there have been studies on the specific effects of timing with respect to outcome.^1,15,16 Vallier and colleagues,¹ who wanted to identify injury characteristics predictive of osteonecrosis, found that delaying reduction and surgical fixation did not increase the risk of AVN. Another study found that urgent reduction of fracture-dislocation with delayed open reduction and internal fixation (ORIF) using a dual approach may improve clinical outcomes.²¹

In this vein, we conducted a study to evaluate the effect of time to surgical reduction of talus fractures and talus fracture-dislocations on the development of AVN and PTOA. We hypothesized that time to surgical reduction of talus fracture-dislocation as classified with the AO/OTA system would have no effect of the development of AVN/PTOA.

Methods

After this study received Institutional Review Board approval, we retrospectively reviewed the records on talus fractures surgically managed at a level I trauma center during the 10-year period 2003 to 2013. Of the 119 potential cases identified using Current Procedural Terminology code 28445 (ORIF of talus), 13 were excluded (12 for inaccurate coding or missing documentation, 1 for being a pediatric case), leaving 106 for analysis. Using the Hawkins and AO/OTA systems, 3 independent reviewers classified the injuries on plain radiographs.

Injury dates and times were obtained from the medical records. Operating room start times were also obtained. Surgical timing was defined as time from injury to operating room start. For cases without an injury time, time of presentation to emergency department was used.

Open fracture-dislocations were managed with intravenous antibiotics, urgent surgical irrigation, débridement, and immediate fixation or temporizing external fixation after reduction. All fractures were definitively managed with standard ORIF with an anteromedial, anterolateral, or dual approach and mini-fragment implants. After fixation, weight-bearing typically was restricted for 6 to 12 weeks.

Follow-up radiographs were evaluated. Presence or absence of Hawkins sign⁷ was noted on radiographs at 6 or 8 weeks, and all follow-up radiographs were evaluated for AVN as defined by increased radiographic density within the talar dome or collapse of the articular surface. All radiographs were evaluated for PTOA as defined by loss of joint space within the tibiotalar, subtalar, or talonavicular joint on follow-up radiographs.

Clinical outcomes were analyzed for development of AVN, PTOA, or secondary corrective surgery or arthrodesis. Continuous variables were evaluated with the t test, and the χ² test was used to compare distributions of categorical variables. The Wilcoxon rank sum test was used to compare non-normally distributed variables. Significance was set at P < .05.

Classification Analysis (Table 1)

Subject Analysis (Table 2)

The mechanisms of injury were motor vehicle accident (70/106; 66%), fall from height (25; 24%), misstep (4), sports related (2), object falling on ankle (2), and not reported (3). 

Of the 106 patients, 45 (42%) had isolated talus injuries, 35 had concomitant ipsilateral lower extremity injuries, 25 had concomitant contralateral lower extremity injuries, and 1 had a concomitant upper extremity injury. 

Smoking status was everyday (14 patients), past (10), never (34), and unreported (48). Five patients reported a history of alcohol abuse, and 4 patients reported illicit drug use. Two had a history of atrial fibrillation, 9 had hypertension, 3 had hyperlipidemia, 3 had renal disease, 3 had heart disease, 4 had diabetes, 3 had lung disease, and 1 had a history of lung cancer.

Overall Analysis of AVN/PTOA (Table 3)

Analysis of AVN/PTOA in 81-B3 Fracture-Dislocations (Table 4)

Discussion

Our results showed that time from talus fracture-dislocation to surgical reduction had no effect on development of AVN/PTOA. The findings in this largest series to date agree with earlier findings^1,8,15,16,24 and add to the volume of literature suggesting that time to surgical reduction of talus fractures and talus fracture-dislocations does not markedly affect outcome.

Talus fractures continue to present a significant treatment dilemma. Despite recent improvements in surgical techniques and overall management of these injuries, rates of AVN and PTOA have not significantly decreased.^1,16,23 At most treating facilities, talus fracture-dislocations are considered surgical emergencies/urgencies, and every effort is made to reduce and surgically address these injuries as soon as possible.^1,13

In this study, rates of AVN/PTOA were 41% (all talus fractures) and 50% (displaced talar neck fractures), and the difference was not significant (Table 3). These rates are higher but consistent with previously reported rates (range, 14%-49%).^{1,2,7-9,12,14,24} There was no difference in surgical timing for development of AVN/PTOA. We analyzed the cases of all patients who had talus fractures and developed AVN/PTOA (43/106). Within this group, there were no significant differences in surgical timing, age, sex, polytrauma, or BMI between patients who developed AVN/PTOA and those who did not. Compared with patients who did not develop AVN/PTOA, those who developed AVN/PTOA were significantly more likely to have open injuries. This finding, consistent with those in other reports^9,12,13 (Table 3), indicates outcome is more likely related to injury severity and not necessarily injury class.

We retrospectively analyzed talus fractures and talus fracture-dislocations to determine if urgent surgical management affects outcomes. Current practice at our institution is to routinely reduce and surgically address these fractures urgently, often during the middle of the night, when orthopedic resources are reduced. Our study found a significant difference in surgical timing for patients with talus fracture-dislocations and patients with talus fractures without dislocations (Table 2). Given our findings, urgent surgical reduction and fixation are not indicated to preserve the talus blood supply and prevent AVN/PTOA, though we still recommend urgent surgical management in the setting of an open wound, skin necrosis, or soft-tissue/neurovascular compromise. 

This study had several limitations, primarily related to its retrospective nature. Surgical timing was defined as time from injury, as noted in the medical record, to operating room start. In some instances, time of injury was not noted in the medical record, and time of presentation to emergency room was used instead. Thus, surgical timing for these patients may have been longer than identified. In addition, given the rare injury pattern and the retrospective design, this study was susceptible to type II error and may have been underpowered to detect whether time to surgical reduction predicted complications. Also, the study did not address functional outcome as measured by validated outcome scores. Outcome measures were obtained in many but not all cases, making functional outcome measurement difficult. Similarly, the quality of the anatomical reductions was not assessed, potentially affecting complication rates. Postoperative reduction assessment, possibly performed with computed tomography, is an avenue of further study.

Strengths of this study include its large sample size (this was one of the largest studies of talus fractures), long follow-up (mean, 150 weeks), and novel use of AO/OTA classification.

We postulate that development of AVN/PTOA is not necessarily related to the urgency or timing of surgical reduction and fixation and is more likely related to injury severity. This idea is supported by the finding that development of AVN/PTOA was significantly correlated to open injuries in all talus fractures, including talus fracture-dislocations and isolated talus fractures.

Conclusion 

Talus fracture-dislocations are devastating injuries with high rates of complications. In this study, open talus fractures, and fractures with associated tibiotalar or subtalar dislocations, had higher complication rates. Given the evidence presented, we recommend basing surgical timing on injury severity, not necessarily for AVN/PTOA prevention. Specifically, in the absence of an open wound, skin necrosis, or soft-tissue/neurovascular compromise, talus fracture-dislocations can be surgically reduced and stabilized when optimal resources are available.

Take-Home Points

• Cellular healing response differs between bony and soft tissue biceps tenodesis.
• Bony tenodesis incites an inflammatory healing response.
• Bony tenodesis healing occurs at the tendon-bone interface.
• Intrasseous bony fixation leads to tendon degeneration within the bone.
• Tendon-to-tendon tenodesis may result in regenerative tendon healing.

The long head of the biceps tendon (LHBT) is a well-established pain generator of the anterior shoulder^1,2 and may be surgically addressed in refractory cases.³ According to a recent study of 44,932 cases, biceps tenodesis rates increased 80% over just 3 years (2008-2011).⁴ Nevertheless, optimal tenodesis location and technique remain controversial. Proximal and distal tenodesis, including numerous soft-tissue and bony techniques, have been described.^5-7 Several studies have focused on the biomechanical strength of various fixation modalities.^8-14 These data highlight the ongoing evolution of our understanding of biceps-labrum complex (BLC) disease.

Over the years, tenodesis location has proved to be an important factor in outcomes.^3,15-20 Several recent studies have elucidated the role of the extra-articular LHBT and the limited capabilities of diagnostic arthroscopy.^15-17,20,21 Taylor and colleagues¹⁷ defined the bicipital tunnel as the extra-articular segment of LHBT and its fibro-osseous enclosure. The tunnel extends from the articular margin through the subpectoral region and can be divided into 3 zones: Zone 1 goes from the articular margin to the inferior margin of the subscapularis, zone 2 goes from the inferior margin of the subscapularis to the proximal margin of the pectoralis major tendon, and zone 3 is the subpectoral region. Zone 2 is often referred to as “no man’s land” for its relative invisibility from arthroscopy above and open exposure below.^17,21 Notably, a recent study reported a 47% prevalence of hidden tunnel lesions in patients with chronic BLC disease symptoms.¹⁸ Other studies have shown that standard proximal tenodesis methods often fail to address LHBT pathology in this area, leading to residual symptoms.^9,22 It is evident that tenodesis location and technique play important roles in patient outcomes. Sanders and colleagues¹⁶ found that the revision rate was significantly higher among patients who underwent biceps tenodesis without release of the bicipital tunnel sheath than among patients who underwent tenodesis with the release. Dr. O’Brien developed an alternative option: soft-tissue tenodesis with transfer of the LHBT to the conjoint tendon within the subdeltoid space.^23,24 This technique addresses intra-articular and extra-articular tunnel disease while mitigating the complications associated with bony tenodesis. Early and midterm studies have shown this to be an effective intervention for chronically symptomatic BLC disease.^25,26

Despite the abundance of literature on tenodesis techniques, no one has histologically evaluated the location-dependent healing and inflammatory responses. We conducted a study to determine the impact of tenodesis location on healing and inflammation in a rat model. We hypothesized that, compared with tendon-to-bone techniques, soft-tissue tenodesis would minimize inflammatory response and optimize healing.

Methods 

The study was approved by the Institutional Animal Care and Use Committee at the Hospital for Special Surgery. 

Animals

Biceps tenodesis was performed at 1 of 3 locations in 36 thirteen-week-old Sprague-Dawley rats (Charles River Laboratories). All rats were prepared for surgery by an experienced veterinary technician. Sedation was induced with isoflurane gas through a nose cone. 

Surgical Procedure

Animals were randomly assigned to 3 different tenodesis groups: tendon-to-bone in the bicipital groove (metaphyseal, M); tendon-to-bone in the subpectoral region (diaphyseal, D); and soft tissue-to-soft tissue transfer to the conjoint tendon (T). A standard deltopectoral approach was used to expose the biceps tendon. The tendon was tagged with a 5-0 polypropylene suture and tenotomized at the level of the bicipital groove (zone 1). All wounds were irrigated and closed with 4-0 nylon suture.

For animals undergoing tendon-to-bone metaphyseal tenodesis, a 0.045-mm Kirschner wire was used to drill bicortically into the intertubercular sulcus. Wire positioning distal to the physeal plate was confirmed with fluoroscopy. A locking stitch of 5-0 polypropylene suture was run along the free edge of the tendon. The tendon was then passed through the bone tunnel in an anterior-to-posterior direction, and the limbs of the suture were tied around the lateral cortex. 

The process was repeated for animals undergoing diaphyseal tenodesis; only the tenodesis location was different. The inferior border of the pectoralis major was identified, and a bicortical tunnel was made in the center of the diaphyseal bone. The tendon was then prepared and tenodesed to bone using the method already described.

In soft-tissue tenodesis, the conjoint tendon was identified and carefully dissected from surrounding tissues. The LHBT was then tenodesed to the attached conjoint tendon with interrupted simple stitches of 5-0 polypropylene suture.

The animals were allowed to bear weight on the operative limb immediately after surgery and without immobilization.

Specimen Harvest and Preparation

Four animals from each group were sacrificed at 6, 12, and 24 weeks. Harvested specimens were fixed in 10% neutral-buffered formalin solution. Bony specimens consisted of the upper half of the humerus and the tenodesed biceps tendon, and soft-tissue specimens consisted of the tenodesed LHBT-conjoint tendon complex. Bony specimens were decalcified in 10% ethylenediaminetetraacetic acid. All specimens were paraffin-embedded and sectioned at 7 microns.

Analysis of Cellularity

Sections were stained with hematoxylin-eosin. Overall cellularity at the tenodesis interface was quantified by averaging the nuclei count within 3 separate standardized ×20 magnification high power fields. Only nucleated cells were included in the cell count. Immunohistochemical staining with tenomodulin (Santa Cruz Laboratories, sc-49324) was performed to characterize the cell population at the interface. Deparaffinized sections underwent antigen retrieval with pronase for 30 minutes at 37°C and were incubated overnight with the anti-tenomodulin goat monoclonal antibody diluted to 1:200 in 1% phosphate-buffered saline. The prepared slides were then counterstained with methyl green. Specimens treated with tenomodulin were evaluated for presence or absence of a positive reaction at the tenodesis interface. 

Analysis of Inflammation

Inflammation at the interface was evaluated with the CD68 macrophage marker (ABcam, ab31630). Deparaffinized sections underwent antigen retrieval with pronase for 30 minutes at 37°C and were incubated overnight with anti-CD68 mouse monoclonal antibodies diluted to 1:200 in 1% phosphate-buffered saline. The prepared slides were then counterstained with neutral red. Inflammation was quantified by averaging the number of reactive cells within 3 separate standardized ×20 magnification high power fields.

Statistical Analysis

Descriptive statistics were calculated for cell and macrophage counts for each group at every time point. Two-way analysis of variance was used to compare the cell and macrophage counts between groups at each time point as well as the count differences within each group between time points. P values were Bonferroni-corrected to account for the multiple comparisons between groups. P < .05 was used to signify statistical significance.

Results

All 36 animals survived to their designated harvest time without complications. Twelve specimens were successfully harvested at 6 weeks and another 12 at 24 weeks. At 12 weeks, tenodesis failure occurred in 1 animal in group D, leaving 11 specimens for analysis.

Cellularity

Within-group analysis revealed a trend of increasing cellularity at 12 weeks followed by a decrease at 24 weeks in all 3 groups (Table 2).

Inflammatory Response 

During specimen processing, 1 group D specimen was severely degraded after pronase treatment, leaving 3 specimens for evaluation. Descriptive statistics for each group are listed in Table 3A.

At 6 weeks, mean CD68 cell count was significantly higher in group M than in group D (P = .011) and group T (P < .001) (Table 3B). Likewise, CD68 count was significantly higher in group D than in group T (P < .001). There were no differences in CD68 counts between the 2 bony tenodesis groups at 12 weeks (P = .486) or 24 weeks (P = .315). Both bony tenodesis groups, however, had persistently higher CD68 counts at 12 weeks when compared with group T (group M, P = .002; group D, P < .001). In these specimens, an inflammatory milieu characterized by a large accumulation of lymphocytes and giant cells was noted at the bone-tendon interface.

Tissue-Specific Staining

At 6 weeks, antigen retrieval resulted in severe degradation of 2 group M specimens, 2 group D specimens, and 1 group T specimen. The most notable tenomodulin reaction occurred in group T at the 6- and 12-week harvests, with the 6-week group having the most robust reaction. There was scant reaction in this group at 24 weeks.

Discussion

In this study, the healing response differed between bony and soft-tissue tenodesis techniques in a rat model. Tendon-to-bone tenodesis, both diaphyseal and metaphyseal, appeared to incite an inflammatory degenerative response, whereas tendon-to-tendon healing occurred in a more quiescent and perhaps even regenerative manner.

The early inflammatory response that occurred in the bony tenodesis groups is not unlike what occurs in fracture healing.²⁷ The reaction was even more robust at 12 weeks, signifying an ongoing inflammatory process. In this context, tendon degeneration may plausibly explain the consistent absence of mature tendon within the tunnels at all 3 time points. Some tendon degeneration may be explained by the vascular damage that occurred during surgery, but this damage was a constant factor in all 3 study groups. Interestingly, group M showed the highest early CD68 counts, consistent with this being the more biologically active region of bone.²⁸

Group T had significantly lower cell and macrophage counts throughout the study period, possibly indicating improved healing—an observation supported by a study in which the impact of macrophage depletion on bone-tendon interface healing was evaluated.²⁹ The authors found that, in suppressing macrophage activity, the morphologic and biomechanical properties at the healing interface were significantly improved.²⁹ These findings are consistent with Dr. O’Brien’s anecdotal experience with patients who previously underwent the biceps transfer; on second-look arthroscopy, there was complete seamless integration of tendon and conjoint tendon (Figure 4). 

Studies have found that the inflammatory process is closely associated with pain, and pain syndromes such as fibromyalgia.^30,31 Persistent inflammation, as seen in our bony tenodesis group, could explain the recalcitrant anterior shoulder pain that often occurs in patients after bony tenodesis of the LHBT.^2,6,19,32 

Studies have also suggested that osteoclasts at the bone-tendon interface—osteoclasts share a cell lineage with macrophages—may contribute to bone loss and tunnel widening.^33,34 Osteoclasts are expected at the bone tunnel, as fracture healing occurs at the bone-tendon interface. These osteoclasts could have contributed to the strong CD68 reaction in our bony tenodesis groups. However, CD68 historically has been described as the classic macrophage marker.³⁵ We specifically selected CD68 for this reason: Macrophages are the primary inflammatory cells involved in early healing and are key to the inflammatory process.³⁶

Results of the tenomodulin analysis suggested 2 different healing processes are occurring in the bony and tendon groups. Tenomodulin is a known tenocyte marker for developing and mature tendon in both rats and humans.^37,38 In our study, only group T had a positive tenomodulin reaction. Notably, the reaction occurred only at 6 and 12 weeks. This finding may indicate that a regenerative healing pattern becomes quiescent by 24 weeks. Indeed, it has been suggested that tenomodulin is a key regulator of tenocyte proliferation and tendon maturation.³⁹

The complete absence of tenomodulin reaction in our bony tenodesis groups in the setting of significant inflammation further supports our theory of tendon degeneration within the tunnel. One potential explanation for this finding may be that as the tendon heals to the surface of the bone, the intra-osseous tendon is no longer load-bearing and is resorbed by the body through an inflammatory response. This finding differs from those in previous studies, which have described viable tendon within the bone tunnel at all time points up to 26 weeks.⁴⁰ More recently, it has been suggested that callus formation at the external cortical tendon-bone interface is critical for healing and mechanical strength.^41,42 In addition, recent studies have found a predominantly fibroblastic healing process at the midtunnel, potentially leading to the formation of loose fibrovascular tissue at the tendon-bone interface.⁴³ These data, in concert with ours, call into question the rationale for performing intra-osseous tenodesis through bone tunnels.

Our study results, if confirmed in humans, will have significant clinical implications. If a similar effect can be confirmed in the human shoulder, one could argue that soft-tissue tenodesis may result in decreased postoperative shoulder pain. In addition, if tendon degeneration does occur within the intramedullary tunnel, surface fixation may be the better, safer alternative. Although older studies reported suboptimal strength with this type of fixation,^8,44 more recent studies have found surface fixation strength equivalent to screw fixation strength.^45,46 Such a shift in the treatment paradigm would obviate the need for violation of the humeral cortex, eliminating potential stress risers associated with screw fixation,⁴⁷ and effectively eliminating the risk of iatrogenic fracture.^48,49 It would be interesting to investigate what occurs histologically at the bone-tendon interface in surface fixation (ie, suture anchors). Would the inflammatory response at the surface be similar to the inflammatory intramedullary healing, or would it be similar to the quieter tendon-tendon healing? Answers to such questions have the potential to streamline the treatment algorithm for patients who require tenodesis.

Study Limitations

Our study had several limitations. First, as this was a basic science study using a rat model, its conclusions can only be extrapolated to humans. Second, given the nonspecific nature of the cellular analysis, we cannot draw any definitive conclusions about the cell population at the bone-tendon interface. For example, although tenomodulin is expressed by tenocytes, it is not an established specific marker for tenocytes and may be expressed by other fibroblastic cells. Still, our results provide insight into the local microenvironment and identify important differences between the tenodesis methods. Similarly, the complete absence of tendon within the bone tunnels suggests that an analysis of osteoclastic activity at the tenodesis interface may have been a valuable addition to the study. This finding, however, was unexpected, and we did not have the foresight to include it in our methods. A third limitation is that our fixation method essentially uses the suspension tenodesis method. This fixation method differs from the common fixation techniques used in the clinical setting. Testing of other fixation constructs would require a larger animal model. Furthermore, in suspension- type constructs, micromotion within the bone tunnel may independently elicit an inflammatory response. Inert suture was used in our fixation in order to reduce the risk of an iatrogenic inflammatory response. Last, it would have been valuable to perform a biomechanical analysis of the strength of each tenodesis construct. This was explored with our institution’s biomechanics team, but specimen size precluded successful analysis.

Conclusion

Our results indicated that, compared with tendon-to-tendon fixation, tendon-to-bone tenodesis produces a significantly greater inflammatory response at the tenodesis interface. An inflammatory milieu in the absence of tendon within the bony tunnel suggests intraosseous tendon degeneration. Tendon-to-tendon tenodesis, on the other hand, seems to limit the inflammatory response. In addition, a robust tenomodulin reaction in the early phases of tendon-to-tendon healing suggests regenerative healing. Our results showed a fundamental difference in the healing response between the 2 tenodesis methods. Further study is needed to evaluate the validity and applicability of our findings to the human patient population. Most important, our results underscore the need for more study to elucidate optimal tenodesis location and encourage orthopedic surgeons to reexamine current clinical practice patterns.

Take-Home Points

• Cellular healing response differs between bony and soft tissue biceps tenodesis.
• Bony tenodesis incites an inflammatory healing response.
• Bony tenodesis healing occurs at the tendon-bone interface.
• Intrasseous bony fixation leads to tendon degeneration within the bone.
• Tendon-to-tendon tenodesis may result in regenerative tendon healing.

The long head of the biceps tendon (LHBT) is a well-established pain generator of the anterior shoulder^1,2 and may be surgically addressed in refractory cases.³ According to a recent study of 44,932 cases, biceps tenodesis rates increased 80% over just 3 years (2008-2011).⁴ Nevertheless, optimal tenodesis location and technique remain controversial. Proximal and distal tenodesis, including numerous soft-tissue and bony techniques, have been described.^5-7 Several studies have focused on the biomechanical strength of various fixation modalities.^8-14 These data highlight the ongoing evolution of our understanding of biceps-labrum complex (BLC) disease.

Over the years, tenodesis location has proved to be an important factor in outcomes.^3,15-20 Several recent studies have elucidated the role of the extra-articular LHBT and the limited capabilities of diagnostic arthroscopy.^15-17,20,21 Taylor and colleagues¹⁷ defined the bicipital tunnel as the extra-articular segment of LHBT and its fibro-osseous enclosure. The tunnel extends from the articular margin through the subpectoral region and can be divided into 3 zones: Zone 1 goes from the articular margin to the inferior margin of the subscapularis, zone 2 goes from the inferior margin of the subscapularis to the proximal margin of the pectoralis major tendon, and zone 3 is the subpectoral region. Zone 2 is often referred to as “no man’s land” for its relative invisibility from arthroscopy above and open exposure below.^17,21 Notably, a recent study reported a 47% prevalence of hidden tunnel lesions in patients with chronic BLC disease symptoms.¹⁸ Other studies have shown that standard proximal tenodesis methods often fail to address LHBT pathology in this area, leading to residual symptoms.^9,22 It is evident that tenodesis location and technique play important roles in patient outcomes. Sanders and colleagues¹⁶ found that the revision rate was significantly higher among patients who underwent biceps tenodesis without release of the bicipital tunnel sheath than among patients who underwent tenodesis with the release. Dr. O’Brien developed an alternative option: soft-tissue tenodesis with transfer of the LHBT to the conjoint tendon within the subdeltoid space.^23,24 This technique addresses intra-articular and extra-articular tunnel disease while mitigating the complications associated with bony tenodesis. Early and midterm studies have shown this to be an effective intervention for chronically symptomatic BLC disease.^25,26

Despite the abundance of literature on tenodesis techniques, no one has histologically evaluated the location-dependent healing and inflammatory responses. We conducted a study to determine the impact of tenodesis location on healing and inflammation in a rat model. We hypothesized that, compared with tendon-to-bone techniques, soft-tissue tenodesis would minimize inflammatory response and optimize healing.

Methods 

The study was approved by the Institutional Animal Care and Use Committee at the Hospital for Special Surgery. 

Animals

Biceps tenodesis was performed at 1 of 3 locations in 36 thirteen-week-old Sprague-Dawley rats (Charles River Laboratories). All rats were prepared for surgery by an experienced veterinary technician. Sedation was induced with isoflurane gas through a nose cone. 

Surgical Procedure

Animals were randomly assigned to 3 different tenodesis groups: tendon-to-bone in the bicipital groove (metaphyseal, M); tendon-to-bone in the subpectoral region (diaphyseal, D); and soft tissue-to-soft tissue transfer to the conjoint tendon (T). A standard deltopectoral approach was used to expose the biceps tendon. The tendon was tagged with a 5-0 polypropylene suture and tenotomized at the level of the bicipital groove (zone 1). All wounds were irrigated and closed with 4-0 nylon suture.

For animals undergoing tendon-to-bone metaphyseal tenodesis, a 0.045-mm Kirschner wire was used to drill bicortically into the intertubercular sulcus. Wire positioning distal to the physeal plate was confirmed with fluoroscopy. A locking stitch of 5-0 polypropylene suture was run along the free edge of the tendon. The tendon was then passed through the bone tunnel in an anterior-to-posterior direction, and the limbs of the suture were tied around the lateral cortex. 

The process was repeated for animals undergoing diaphyseal tenodesis; only the tenodesis location was different. The inferior border of the pectoralis major was identified, and a bicortical tunnel was made in the center of the diaphyseal bone. The tendon was then prepared and tenodesed to bone using the method already described.

In soft-tissue tenodesis, the conjoint tendon was identified and carefully dissected from surrounding tissues. The LHBT was then tenodesed to the attached conjoint tendon with interrupted simple stitches of 5-0 polypropylene suture.

The animals were allowed to bear weight on the operative limb immediately after surgery and without immobilization.

Specimen Harvest and Preparation

Four animals from each group were sacrificed at 6, 12, and 24 weeks. Harvested specimens were fixed in 10% neutral-buffered formalin solution. Bony specimens consisted of the upper half of the humerus and the tenodesed biceps tendon, and soft-tissue specimens consisted of the tenodesed LHBT-conjoint tendon complex. Bony specimens were decalcified in 10% ethylenediaminetetraacetic acid. All specimens were paraffin-embedded and sectioned at 7 microns.

Analysis of Cellularity

Sections were stained with hematoxylin-eosin. Overall cellularity at the tenodesis interface was quantified by averaging the nuclei count within 3 separate standardized ×20 magnification high power fields. Only nucleated cells were included in the cell count. Immunohistochemical staining with tenomodulin (Santa Cruz Laboratories, sc-49324) was performed to characterize the cell population at the interface. Deparaffinized sections underwent antigen retrieval with pronase for 30 minutes at 37°C and were incubated overnight with the anti-tenomodulin goat monoclonal antibody diluted to 1:200 in 1% phosphate-buffered saline. The prepared slides were then counterstained with methyl green. Specimens treated with tenomodulin were evaluated for presence or absence of a positive reaction at the tenodesis interface. 

Analysis of Inflammation

Inflammation at the interface was evaluated with the CD68 macrophage marker (ABcam, ab31630). Deparaffinized sections underwent antigen retrieval with pronase for 30 minutes at 37°C and were incubated overnight with anti-CD68 mouse monoclonal antibodies diluted to 1:200 in 1% phosphate-buffered saline. The prepared slides were then counterstained with neutral red. Inflammation was quantified by averaging the number of reactive cells within 3 separate standardized ×20 magnification high power fields.

Statistical Analysis

Descriptive statistics were calculated for cell and macrophage counts for each group at every time point. Two-way analysis of variance was used to compare the cell and macrophage counts between groups at each time point as well as the count differences within each group between time points. P values were Bonferroni-corrected to account for the multiple comparisons between groups. P < .05 was used to signify statistical significance.

Results

All 36 animals survived to their designated harvest time without complications. Twelve specimens were successfully harvested at 6 weeks and another 12 at 24 weeks. At 12 weeks, tenodesis failure occurred in 1 animal in group D, leaving 11 specimens for analysis.

Cellularity

Within-group analysis revealed a trend of increasing cellularity at 12 weeks followed by a decrease at 24 weeks in all 3 groups (Table 2).

Inflammatory Response 

During specimen processing, 1 group D specimen was severely degraded after pronase treatment, leaving 3 specimens for evaluation. Descriptive statistics for each group are listed in Table 3A.

At 6 weeks, mean CD68 cell count was significantly higher in group M than in group D (P = .011) and group T (P < .001) (Table 3B). Likewise, CD68 count was significantly higher in group D than in group T (P < .001). There were no differences in CD68 counts between the 2 bony tenodesis groups at 12 weeks (P = .486) or 24 weeks (P = .315). Both bony tenodesis groups, however, had persistently higher CD68 counts at 12 weeks when compared with group T (group M, P = .002; group D, P < .001). In these specimens, an inflammatory milieu characterized by a large accumulation of lymphocytes and giant cells was noted at the bone-tendon interface.

Tissue-Specific Staining

At 6 weeks, antigen retrieval resulted in severe degradation of 2 group M specimens, 2 group D specimens, and 1 group T specimen. The most notable tenomodulin reaction occurred in group T at the 6- and 12-week harvests, with the 6-week group having the most robust reaction. There was scant reaction in this group at 24 weeks.

Discussion

In this study, the healing response differed between bony and soft-tissue tenodesis techniques in a rat model. Tendon-to-bone tenodesis, both diaphyseal and metaphyseal, appeared to incite an inflammatory degenerative response, whereas tendon-to-tendon healing occurred in a more quiescent and perhaps even regenerative manner.

The early inflammatory response that occurred in the bony tenodesis groups is not unlike what occurs in fracture healing.²⁷ The reaction was even more robust at 12 weeks, signifying an ongoing inflammatory process. In this context, tendon degeneration may plausibly explain the consistent absence of mature tendon within the tunnels at all 3 time points. Some tendon degeneration may be explained by the vascular damage that occurred during surgery, but this damage was a constant factor in all 3 study groups. Interestingly, group M showed the highest early CD68 counts, consistent with this being the more biologically active region of bone.²⁸

Group T had significantly lower cell and macrophage counts throughout the study period, possibly indicating improved healing—an observation supported by a study in which the impact of macrophage depletion on bone-tendon interface healing was evaluated.²⁹ The authors found that, in suppressing macrophage activity, the morphologic and biomechanical properties at the healing interface were significantly improved.²⁹ These findings are consistent with Dr. O’Brien’s anecdotal experience with patients who previously underwent the biceps transfer; on second-look arthroscopy, there was complete seamless integration of tendon and conjoint tendon (Figure 4). 

Studies have found that the inflammatory process is closely associated with pain, and pain syndromes such as fibromyalgia.^30,31 Persistent inflammation, as seen in our bony tenodesis group, could explain the recalcitrant anterior shoulder pain that often occurs in patients after bony tenodesis of the LHBT.^2,6,19,32 

Studies have also suggested that osteoclasts at the bone-tendon interface—osteoclasts share a cell lineage with macrophages—may contribute to bone loss and tunnel widening.^33,34 Osteoclasts are expected at the bone tunnel, as fracture healing occurs at the bone-tendon interface. These osteoclasts could have contributed to the strong CD68 reaction in our bony tenodesis groups. However, CD68 historically has been described as the classic macrophage marker.³⁵ We specifically selected CD68 for this reason: Macrophages are the primary inflammatory cells involved in early healing and are key to the inflammatory process.³⁶

Results of the tenomodulin analysis suggested 2 different healing processes are occurring in the bony and tendon groups. Tenomodulin is a known tenocyte marker for developing and mature tendon in both rats and humans.^37,38 In our study, only group T had a positive tenomodulin reaction. Notably, the reaction occurred only at 6 and 12 weeks. This finding may indicate that a regenerative healing pattern becomes quiescent by 24 weeks. Indeed, it has been suggested that tenomodulin is a key regulator of tenocyte proliferation and tendon maturation.³⁹

The complete absence of tenomodulin reaction in our bony tenodesis groups in the setting of significant inflammation further supports our theory of tendon degeneration within the tunnel. One potential explanation for this finding may be that as the tendon heals to the surface of the bone, the intra-osseous tendon is no longer load-bearing and is resorbed by the body through an inflammatory response. This finding differs from those in previous studies, which have described viable tendon within the bone tunnel at all time points up to 26 weeks.⁴⁰ More recently, it has been suggested that callus formation at the external cortical tendon-bone interface is critical for healing and mechanical strength.^41,42 In addition, recent studies have found a predominantly fibroblastic healing process at the midtunnel, potentially leading to the formation of loose fibrovascular tissue at the tendon-bone interface.⁴³ These data, in concert with ours, call into question the rationale for performing intra-osseous tenodesis through bone tunnels.

Our study results, if confirmed in humans, will have significant clinical implications. If a similar effect can be confirmed in the human shoulder, one could argue that soft-tissue tenodesis may result in decreased postoperative shoulder pain. In addition, if tendon degeneration does occur within the intramedullary tunnel, surface fixation may be the better, safer alternative. Although older studies reported suboptimal strength with this type of fixation,^8,44 more recent studies have found surface fixation strength equivalent to screw fixation strength.^45,46 Such a shift in the treatment paradigm would obviate the need for violation of the humeral cortex, eliminating potential stress risers associated with screw fixation,⁴⁷ and effectively eliminating the risk of iatrogenic fracture.^48,49 It would be interesting to investigate what occurs histologically at the bone-tendon interface in surface fixation (ie, suture anchors). Would the inflammatory response at the surface be similar to the inflammatory intramedullary healing, or would it be similar to the quieter tendon-tendon healing? Answers to such questions have the potential to streamline the treatment algorithm for patients who require tenodesis.

Study Limitations

Our study had several limitations. First, as this was a basic science study using a rat model, its conclusions can only be extrapolated to humans. Second, given the nonspecific nature of the cellular analysis, we cannot draw any definitive conclusions about the cell population at the bone-tendon interface. For example, although tenomodulin is expressed by tenocytes, it is not an established specific marker for tenocytes and may be expressed by other fibroblastic cells. Still, our results provide insight into the local microenvironment and identify important differences between the tenodesis methods. Similarly, the complete absence of tendon within the bone tunnels suggests that an analysis of osteoclastic activity at the tenodesis interface may have been a valuable addition to the study. This finding, however, was unexpected, and we did not have the foresight to include it in our methods. A third limitation is that our fixation method essentially uses the suspension tenodesis method. This fixation method differs from the common fixation techniques used in the clinical setting. Testing of other fixation constructs would require a larger animal model. Furthermore, in suspension- type constructs, micromotion within the bone tunnel may independently elicit an inflammatory response. Inert suture was used in our fixation in order to reduce the risk of an iatrogenic inflammatory response. Last, it would have been valuable to perform a biomechanical analysis of the strength of each tenodesis construct. This was explored with our institution’s biomechanics team, but specimen size precluded successful analysis.

Conclusion

Our results indicated that, compared with tendon-to-tendon fixation, tendon-to-bone tenodesis produces a significantly greater inflammatory response at the tenodesis interface. An inflammatory milieu in the absence of tendon within the bony tunnel suggests intraosseous tendon degeneration. Tendon-to-tendon tenodesis, on the other hand, seems to limit the inflammatory response. In addition, a robust tenomodulin reaction in the early phases of tendon-to-tendon healing suggests regenerative healing. Our results showed a fundamental difference in the healing response between the 2 tenodesis methods. Further study is needed to evaluate the validity and applicability of our findings to the human patient population. Most important, our results underscore the need for more study to elucidate optimal tenodesis location and encourage orthopedic surgeons to reexamine current clinical practice patterns.

Take-Home Points

• Cellular healing response differs between bony and soft tissue biceps tenodesis.
• Bony tenodesis incites an inflammatory healing response.
• Bony tenodesis healing occurs at the tendon-bone interface.
• Intrasseous bony fixation leads to tendon degeneration within the bone.
• Tendon-to-tendon tenodesis may result in regenerative tendon healing.

The long head of the biceps tendon (LHBT) is a well-established pain generator of the anterior shoulder^1,2 and may be surgically addressed in refractory cases.³ According to a recent study of 44,932 cases, biceps tenodesis rates increased 80% over just 3 years (2008-2011).⁴ Nevertheless, optimal tenodesis location and technique remain controversial. Proximal and distal tenodesis, including numerous soft-tissue and bony techniques, have been described.^5-7 Several studies have focused on the biomechanical strength of various fixation modalities.^8-14 These data highlight the ongoing evolution of our understanding of biceps-labrum complex (BLC) disease.

Over the years, tenodesis location has proved to be an important factor in outcomes.^3,15-20 Several recent studies have elucidated the role of the extra-articular LHBT and the limited capabilities of diagnostic arthroscopy.^15-17,20,21 Taylor and colleagues¹⁷ defined the bicipital tunnel as the extra-articular segment of LHBT and its fibro-osseous enclosure. The tunnel extends from the articular margin through the subpectoral region and can be divided into 3 zones: Zone 1 goes from the articular margin to the inferior margin of the subscapularis, zone 2 goes from the inferior margin of the subscapularis to the proximal margin of the pectoralis major tendon, and zone 3 is the subpectoral region. Zone 2 is often referred to as “no man’s land” for its relative invisibility from arthroscopy above and open exposure below.^17,21 Notably, a recent study reported a 47% prevalence of hidden tunnel lesions in patients with chronic BLC disease symptoms.¹⁸ Other studies have shown that standard proximal tenodesis methods often fail to address LHBT pathology in this area, leading to residual symptoms.^9,22 It is evident that tenodesis location and technique play important roles in patient outcomes. Sanders and colleagues¹⁶ found that the revision rate was significantly higher among patients who underwent biceps tenodesis without release of the bicipital tunnel sheath than among patients who underwent tenodesis with the release. Dr. O’Brien developed an alternative option: soft-tissue tenodesis with transfer of the LHBT to the conjoint tendon within the subdeltoid space.^23,24 This technique addresses intra-articular and extra-articular tunnel disease while mitigating the complications associated with bony tenodesis. Early and midterm studies have shown this to be an effective intervention for chronically symptomatic BLC disease.^25,26

Despite the abundance of literature on tenodesis techniques, no one has histologically evaluated the location-dependent healing and inflammatory responses. We conducted a study to determine the impact of tenodesis location on healing and inflammation in a rat model. We hypothesized that, compared with tendon-to-bone techniques, soft-tissue tenodesis would minimize inflammatory response and optimize healing.

Methods 

The study was approved by the Institutional Animal Care and Use Committee at the Hospital for Special Surgery. 

Animals

Biceps tenodesis was performed at 1 of 3 locations in 36 thirteen-week-old Sprague-Dawley rats (Charles River Laboratories). All rats were prepared for surgery by an experienced veterinary technician. Sedation was induced with isoflurane gas through a nose cone. 

Surgical Procedure

Animals were randomly assigned to 3 different tenodesis groups: tendon-to-bone in the bicipital groove (metaphyseal, M); tendon-to-bone in the subpectoral region (diaphyseal, D); and soft tissue-to-soft tissue transfer to the conjoint tendon (T). A standard deltopectoral approach was used to expose the biceps tendon. The tendon was tagged with a 5-0 polypropylene suture and tenotomized at the level of the bicipital groove (zone 1). All wounds were irrigated and closed with 4-0 nylon suture.

For animals undergoing tendon-to-bone metaphyseal tenodesis, a 0.045-mm Kirschner wire was used to drill bicortically into the intertubercular sulcus. Wire positioning distal to the physeal plate was confirmed with fluoroscopy. A locking stitch of 5-0 polypropylene suture was run along the free edge of the tendon. The tendon was then passed through the bone tunnel in an anterior-to-posterior direction, and the limbs of the suture were tied around the lateral cortex. 

The process was repeated for animals undergoing diaphyseal tenodesis; only the tenodesis location was different. The inferior border of the pectoralis major was identified, and a bicortical tunnel was made in the center of the diaphyseal bone. The tendon was then prepared and tenodesed to bone using the method already described.

In soft-tissue tenodesis, the conjoint tendon was identified and carefully dissected from surrounding tissues. The LHBT was then tenodesed to the attached conjoint tendon with interrupted simple stitches of 5-0 polypropylene suture.

The animals were allowed to bear weight on the operative limb immediately after surgery and without immobilization.

Specimen Harvest and Preparation

Four animals from each group were sacrificed at 6, 12, and 24 weeks. Harvested specimens were fixed in 10% neutral-buffered formalin solution. Bony specimens consisted of the upper half of the humerus and the tenodesed biceps tendon, and soft-tissue specimens consisted of the tenodesed LHBT-conjoint tendon complex. Bony specimens were decalcified in 10% ethylenediaminetetraacetic acid. All specimens were paraffin-embedded and sectioned at 7 microns.

Analysis of Cellularity

Sections were stained with hematoxylin-eosin. Overall cellularity at the tenodesis interface was quantified by averaging the nuclei count within 3 separate standardized ×20 magnification high power fields. Only nucleated cells were included in the cell count. Immunohistochemical staining with tenomodulin (Santa Cruz Laboratories, sc-49324) was performed to characterize the cell population at the interface. Deparaffinized sections underwent antigen retrieval with pronase for 30 minutes at 37°C and were incubated overnight with the anti-tenomodulin goat monoclonal antibody diluted to 1:200 in 1% phosphate-buffered saline. The prepared slides were then counterstained with methyl green. Specimens treated with tenomodulin were evaluated for presence or absence of a positive reaction at the tenodesis interface. 

Analysis of Inflammation

Inflammation at the interface was evaluated with the CD68 macrophage marker (ABcam, ab31630). Deparaffinized sections underwent antigen retrieval with pronase for 30 minutes at 37°C and were incubated overnight with anti-CD68 mouse monoclonal antibodies diluted to 1:200 in 1% phosphate-buffered saline. The prepared slides were then counterstained with neutral red. Inflammation was quantified by averaging the number of reactive cells within 3 separate standardized ×20 magnification high power fields.

Statistical Analysis

Descriptive statistics were calculated for cell and macrophage counts for each group at every time point. Two-way analysis of variance was used to compare the cell and macrophage counts between groups at each time point as well as the count differences within each group between time points. P values were Bonferroni-corrected to account for the multiple comparisons between groups. P < .05 was used to signify statistical significance.

Results

All 36 animals survived to their designated harvest time without complications. Twelve specimens were successfully harvested at 6 weeks and another 12 at 24 weeks. At 12 weeks, tenodesis failure occurred in 1 animal in group D, leaving 11 specimens for analysis.

Cellularity

Within-group analysis revealed a trend of increasing cellularity at 12 weeks followed by a decrease at 24 weeks in all 3 groups (Table 2).

Inflammatory Response 

During specimen processing, 1 group D specimen was severely degraded after pronase treatment, leaving 3 specimens for evaluation. Descriptive statistics for each group are listed in Table 3A.

At 6 weeks, mean CD68 cell count was significantly higher in group M than in group D (P = .011) and group T (P < .001) (Table 3B). Likewise, CD68 count was significantly higher in group D than in group T (P < .001). There were no differences in CD68 counts between the 2 bony tenodesis groups at 12 weeks (P = .486) or 24 weeks (P = .315). Both bony tenodesis groups, however, had persistently higher CD68 counts at 12 weeks when compared with group T (group M, P = .002; group D, P < .001). In these specimens, an inflammatory milieu characterized by a large accumulation of lymphocytes and giant cells was noted at the bone-tendon interface.

Tissue-Specific Staining

At 6 weeks, antigen retrieval resulted in severe degradation of 2 group M specimens, 2 group D specimens, and 1 group T specimen. The most notable tenomodulin reaction occurred in group T at the 6- and 12-week harvests, with the 6-week group having the most robust reaction. There was scant reaction in this group at 24 weeks.

Discussion

In this study, the healing response differed between bony and soft-tissue tenodesis techniques in a rat model. Tendon-to-bone tenodesis, both diaphyseal and metaphyseal, appeared to incite an inflammatory degenerative response, whereas tendon-to-tendon healing occurred in a more quiescent and perhaps even regenerative manner.

The early inflammatory response that occurred in the bony tenodesis groups is not unlike what occurs in fracture healing.²⁷ The reaction was even more robust at 12 weeks, signifying an ongoing inflammatory process. In this context, tendon degeneration may plausibly explain the consistent absence of mature tendon within the tunnels at all 3 time points. Some tendon degeneration may be explained by the vascular damage that occurred during surgery, but this damage was a constant factor in all 3 study groups. Interestingly, group M showed the highest early CD68 counts, consistent with this being the more biologically active region of bone.²⁸

Group T had significantly lower cell and macrophage counts throughout the study period, possibly indicating improved healing—an observation supported by a study in which the impact of macrophage depletion on bone-tendon interface healing was evaluated.²⁹ The authors found that, in suppressing macrophage activity, the morphologic and biomechanical properties at the healing interface were significantly improved.²⁹ These findings are consistent with Dr. O’Brien’s anecdotal experience with patients who previously underwent the biceps transfer; on second-look arthroscopy, there was complete seamless integration of tendon and conjoint tendon (Figure 4). 

Studies have found that the inflammatory process is closely associated with pain, and pain syndromes such as fibromyalgia.^30,31 Persistent inflammation, as seen in our bony tenodesis group, could explain the recalcitrant anterior shoulder pain that often occurs in patients after bony tenodesis of the LHBT.^2,6,19,32 

Studies have also suggested that osteoclasts at the bone-tendon interface—osteoclasts share a cell lineage with macrophages—may contribute to bone loss and tunnel widening.^33,34 Osteoclasts are expected at the bone tunnel, as fracture healing occurs at the bone-tendon interface. These osteoclasts could have contributed to the strong CD68 reaction in our bony tenodesis groups. However, CD68 historically has been described as the classic macrophage marker.³⁵ We specifically selected CD68 for this reason: Macrophages are the primary inflammatory cells involved in early healing and are key to the inflammatory process.³⁶

Results of the tenomodulin analysis suggested 2 different healing processes are occurring in the bony and tendon groups. Tenomodulin is a known tenocyte marker for developing and mature tendon in both rats and humans.^37,38 In our study, only group T had a positive tenomodulin reaction. Notably, the reaction occurred only at 6 and 12 weeks. This finding may indicate that a regenerative healing pattern becomes quiescent by 24 weeks. Indeed, it has been suggested that tenomodulin is a key regulator of tenocyte proliferation and tendon maturation.³⁹

The complete absence of tenomodulin reaction in our bony tenodesis groups in the setting of significant inflammation further supports our theory of tendon degeneration within the tunnel. One potential explanation for this finding may be that as the tendon heals to the surface of the bone, the intra-osseous tendon is no longer load-bearing and is resorbed by the body through an inflammatory response. This finding differs from those in previous studies, which have described viable tendon within the bone tunnel at all time points up to 26 weeks.⁴⁰ More recently, it has been suggested that callus formation at the external cortical tendon-bone interface is critical for healing and mechanical strength.^41,42 In addition, recent studies have found a predominantly fibroblastic healing process at the midtunnel, potentially leading to the formation of loose fibrovascular tissue at the tendon-bone interface.⁴³ These data, in concert with ours, call into question the rationale for performing intra-osseous tenodesis through bone tunnels.

Our study results, if confirmed in humans, will have significant clinical implications. If a similar effect can be confirmed in the human shoulder, one could argue that soft-tissue tenodesis may result in decreased postoperative shoulder pain. In addition, if tendon degeneration does occur within the intramedullary tunnel, surface fixation may be the better, safer alternative. Although older studies reported suboptimal strength with this type of fixation,^8,44 more recent studies have found surface fixation strength equivalent to screw fixation strength.^45,46 Such a shift in the treatment paradigm would obviate the need for violation of the humeral cortex, eliminating potential stress risers associated with screw fixation,⁴⁷ and effectively eliminating the risk of iatrogenic fracture.^48,49 It would be interesting to investigate what occurs histologically at the bone-tendon interface in surface fixation (ie, suture anchors). Would the inflammatory response at the surface be similar to the inflammatory intramedullary healing, or would it be similar to the quieter tendon-tendon healing? Answers to such questions have the potential to streamline the treatment algorithm for patients who require tenodesis.

Study Limitations

Our study had several limitations. First, as this was a basic science study using a rat model, its conclusions can only be extrapolated to humans. Second, given the nonspecific nature of the cellular analysis, we cannot draw any definitive conclusions about the cell population at the bone-tendon interface. For example, although tenomodulin is expressed by tenocytes, it is not an established specific marker for tenocytes and may be expressed by other fibroblastic cells. Still, our results provide insight into the local microenvironment and identify important differences between the tenodesis methods. Similarly, the complete absence of tendon within the bone tunnels suggests that an analysis of osteoclastic activity at the tenodesis interface may have been a valuable addition to the study. This finding, however, was unexpected, and we did not have the foresight to include it in our methods. A third limitation is that our fixation method essentially uses the suspension tenodesis method. This fixation method differs from the common fixation techniques used in the clinical setting. Testing of other fixation constructs would require a larger animal model. Furthermore, in suspension- type constructs, micromotion within the bone tunnel may independently elicit an inflammatory response. Inert suture was used in our fixation in order to reduce the risk of an iatrogenic inflammatory response. Last, it would have been valuable to perform a biomechanical analysis of the strength of each tenodesis construct. This was explored with our institution’s biomechanics team, but specimen size precluded successful analysis.

Conclusion

Our results indicated that, compared with tendon-to-tendon fixation, tendon-to-bone tenodesis produces a significantly greater inflammatory response at the tenodesis interface. An inflammatory milieu in the absence of tendon within the bony tunnel suggests intraosseous tendon degeneration. Tendon-to-tendon tenodesis, on the other hand, seems to limit the inflammatory response. In addition, a robust tenomodulin reaction in the early phases of tendon-to-tendon healing suggests regenerative healing. Our results showed a fundamental difference in the healing response between the 2 tenodesis methods. Further study is needed to evaluate the validity and applicability of our findings to the human patient population. Most important, our results underscore the need for more study to elucidate optimal tenodesis location and encourage orthopedic surgeons to reexamine current clinical practice patterns.

Take-Home Points

• TTO specifics depend on anatomy, radiographic alignment characteristics, and presence of chondral defects.
• Osteotomy and movement of the tibial tubercle can include anteriorization, anteromedialization, proximalization, medialization, or distalization.
• TTO was most commonly performed for isolated patellar instability in the presence of knee pain.
• Young women with prior surgery on the affected knee made up the primary patient population for this procedure.
• While TTO significantly improves knee pain and clinical outcome scores, >1 in 5 patients required reoperation for hardware removal.

Patellofemoral pain and patellofemoral instability are common orthopedic problems. Studies have found that 30% of patients 13 to 19 years old have patellofemoral pain and that 29 in 100,000 patients 10 to 17 years old have patellofemoral instability.^1-3 The reported rate of recurrence after nonoperative management of patellofemoral instability is 33%.⁴ Tibial tubercle osteotomy (TTO), first described by Hauser⁵ in 1938, is an effective treatment option for many patellofemoral disorders.

TTO indications include patellofemoral maltracking or malalignment, patellar instability, patellofemoral arthritis, and focal patellofemoral chondral defects.⁶ With TTO, the goal is to move the tibial tubercle in a direction that will either improve patellar tracking or offload the medial or lateral patellar facet to improve pain and function.^7,8 This action typically involves anterior, medial, lateral, or distal translation of the tibial tubercle, as posteriorization can lead to increased contact forces across the patellofemoral joint, resulting in accelerated patellofemoral wear and increased pain.⁹

We systematically reviewed the TTO literature to identify indications, clinical outcomes, complications, and reoperations. We hypothesized that the overall complication rate and the overall reoperation rate would both be <10%.

Clinical Evaluation of Patellofemoral Pathology

Patients with patellofemoral pain often report anterior knee pain, which typically begins gradually and is often activity related. Several symptoms may be present: pain with prolonged sitting with knees bent; pain on rising from a seated position; pain or crepitus with climbing stairs; and pain during repetitive activity such as running, squatting, or jumping. Location, duration, and onset of symptoms should be elicited. Patellofemoral instability can be described as dislocation events or subluxation events; number of events, mechanisms of injury, and resulting need for reduction should be documented. As age, sex, body mass index, and physical fitness are relevant to risk of recurrence, the physician should ask about general ligamentous laxity, other joint dislocations, and prior surgical intervention. Swelling or mechanical symptoms may indicate patellofemoral joint pathology.^6,10

Physical examination of patients with patellofemoral pathology begins with assessment for overall limb alignment (including resting position of patella and corresponding quadriceps angle [Q-angle]), generalized ligamentous laxity (including hypermobile joints, evaluated with Brighton criteria), overall peri-knee muscle tone and strength, effusion, and gait pattern.

Common TTO Procedures

TTO specifics depend on anatomy, radiographic alignment characteristics, and presence of chondral defects. Essentially, the patella is translated to offload the affected areas. Osteotomy and movement of the tibial tubercle can include anteriorization, anteromedialization, proximalization, medialization, or distalization.

Methods

Search Strategy and Data Collection

We searched the PubMed (Medline) database for all English-language TTO studies published between database inception and April 9, 2015. After PROSPERO registration, and following PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines, we used the algorithm (“tibial” AND “tubercle” AND “osteotomy”) NOT (“total” AND “knee” AND “arthroplasty”) to search the literature. Inclusion criteria included level I-IV studies on TTO indications, operative findings, and outcomes. Exclusion criteria were non-English studies, unpublished studies, level V evidence, letters to the editor, editorials, review articles, basic science articles, technique articles, revision procedures, articles without clinical outcomes, and conference proceeding abstracts. Studies that reported on duplicate populations were included only with the most recent available clinical outcomes. All abstracts were reviewed in duplicate by Dr. Levy and Dr. Rao and assessed with respect to the criteria outlined. Then the same authors performed full-text reviews of eligible studies before including these studies in the systematic review.

Assessment of Study Quality

The quality of each TTO study in the review was assessed with a modified Coleman methodology score (MCMS), which ranges from 0 to 100. A study with an MCMS of <55 points is considered a poor-quality study.¹¹

Data Synthesis and Statistical Analysis

Given that most of the included studies were level IV, a formal meta-analysis was not indicated. In this article, we report categorical data as frequencies with percentages and preoperative and postoperative continuous data as means (SDs), with weighted means based on number of patients in each study, where applicable. We used 2-tailed t tests for comparisons made with the free Meta-Analysis Calculator and Grapher (http://www.healthstrategy.com/meta/meta.pl ). Statistical significance was set at P < .05. 

Results

Search Results and Included Studies

Only 1 study provided preoperative body mass index (27 kg/m2). There were 55.35% of patients who had prior surgery on the affected knee (6 studies reporting).

Preoperative Data

Preoperative pathologic, radiographic, and clinical scoring data were scarcely reported and nonuniform (Table 2). The most common pathology treated with TTO was isolated patellofemoral instability (746/1055 patients, 70.7%). The other pathologies addressed were isolated patellofemoral osteoarthritis/chondromalacia patellae (143, 13.6%), patellofemoral instability with patella alta (61, 5.8%), patellofemoral instability with patellofemoral osteoarthritis (45, 4.3%), isolated patella baja (41, 3.9%), isolated patella alta (19, 1.8%), and patellofemoral osteoarthritis with patella baja (2, 0.2%). Five hundred fifty-five patients (53%) had a preoperative complaint that included knee pain, and 809 (77%) reported preoperative patellar laxity or instability events. The imaging data reported were Q-angle, Insall-Salvati ratio, Caton-Deschamps index, Blackburne-Peel ratio, Outerbridge osteoarthritis grade, and TT-TG distance. Preoperative clinical scoring data most prominently included a visual analog scale (VAS) score of 70.50 (4 studies reporting), a Lysholm score of 59.19 (5 studies), and a Kujala score of 41.16 (4 studies). Shelbourne-Trumper and Cox-Insall scores were reported in 1 and 2 studies, respectively.

Operative Characteristics

Of the 21 studies, 12 reported only on patients who had TTO performed in isolation; in the other 9 studies, cohorts included patients who underwent concurrent procedures. In the 17 studies (856 patients) that listed numbers of patients who underwent specific concomitant procedures, 715 patients (83.5%) underwent an isolated TTO procedure, and the other 141 (16.5%) underwent either concomitant lateral femoral trochleoplasty, arthroscopic drilling of chondral lesions, patellar shaving chondroplasty, partial meniscectomy or concomitant meniscal repair, intra-articular loose body removal, and/or lateral release with or without medial plication. 

Postoperative Data

There was a cumulative total of 79 complications (8% of cohort): 17 recurrent patellar dislocations (1.9%), 4 recurrent patellar subluxations (0.4%), 10 wound complications (1.0%), 2 intraoperative complications (0.2%), 14 tibial tubercle fractures (1.3%), 19 proximal tibia fractures (1.8%), 4 cases of anterior knee pain (0.4%), 4 cases of neuropraxia (0.4%), and 5 infections (0.5%). Of note, 219 knees (21%) required reoperation, but 170 (16.3%) of these were for painful hardware removal. Sixteen knees (1.5%) required revision TTO, 1 (0.1%) required subsequent high tibial osteotomy, 2 (0.2%) underwent patellofemoral arthroplasty for advanced arthritic changes, and 5 (0.5%) underwent total knee arthroplasty for advanced arthritic changes.

Studies With TTO Performed in Isolation

Twelve studies reported outcomes of isolated TTO procedures. In the 638 patients who underwent isolated TTO, the pathologies addressed were instability/laxity (429 patients, 67%), patellofemoral osteoarthritis (74, 12%), patella alta with instability (61, 10%), patellofemoral osteoarthritis with instability (31, 5%), patella baja (24, 4%), and patella alta (19, 3%). Pain was a preoperative issue in 289 (45%) of these patients and instability in 472 (74%).

Only 2.8% of patients experienced postoperative patellar dislocation events. Of the 12 studies, 2 reported VAS scores (34-point weighted mean improvement, 65 points before surgery to 31 after surgery), 3 reported Lysholm scores (30-point improvement, from 60 to 90), and 2 reported Kujala scores (21-point improvement, from 46 to 67).

Complication rates for this isolated-TTO pooled cohort of patients were 1.2% for revision TTOs, 0.5% for wound complications, 0.8% for tibial tubercle fractures, and 1.9% for proximal tibia fractures. In total, 16% of patients required hardware removal after surgery. 

Discussion

This study found that TTO improved patient pain and clinical outcome scores despite having a high (16%) rate of reoperation for painful hardware in patients with preoperative pain or instability, or with patellofemoral osteoarthritis or aberrant patellar anatomy. This reoperation rate and the overall complication rate both exceeded our hypothesized 10% cumulative rate. However, <1% of patients required conversion to a definitive end-stage surgery (patellofemoral arthroplasty or total knee arthroplasty) by final follow-up, and the rates of comorbidities (anterior knee pain, wound infection, recurrent patellar subluxation/dislocation, tibial fracture) were relatively low.

Patellofemoral disorders are common in the general population and a frequent primary complaint on presentation to orthopedic offices. Having a thorough understanding of knee joint biomechanics is imperative when trying to determine whether surgery is appropriate for these complaints and how to proceed. Extensor mechanism abnormalities, including high lateral force vectors (or larger TT-TG distances) and excessive patellar tilt, can affect alignment and increase the risk for patellofemoral dislocations, patellofemoral anterior- based knee pains, and chondral lesions. Patella alta, an elevated patella, risks increased contact stresses between the patella and the trochlear groove³³ and decreases the osseous constraints that inhibit dislocation of the patella with physiologic flexion of the joint.³⁴ With TTO, the change in tuberosity position can alter angles in the extensor mechanism and thereby decrease joint reaction forces and patellofemoral contact area forces.^35,36

Although its use began as an option for combating patellar instability events in patients with predisposed patellofemoral kinematics,⁵ TTO has evolved in its therapeutic uses to include offloading patellar and trochlear focal chondral lesions and slowing progression of patellofemoral arthritis. Multiple iterations and modifications of the procedure have involved distal and medial transfer of the tibial tuberosity, medialization alone, concurrent anterior and medial elevation of the tuberosity, and proximal or distal transfers, depending on the pathology being corrected. Although TTO is highly versatile in treating multiple patellofemoral joint pathologies, this study found that its primary indication continues to be patellar instability, with anteromedialization as the most common direction of tubercle transfer in support of the medial structures providing the medial force vector that keeps the patella in place. These medial structures include the medial patellofemoral ligament, the vastus medialis obliquus, the medial patellotibial ligament, and the medial retinaculum. 

Also notable was the relatively high rate of reoperation after TTO. However, >75% of reoperations were performed to remove painful hardware, and the need for reoperation seemed to have no effect on the statistically significant overall preoperative-to-postoperative improvement in VAS, Lysholm, and Kujala scores. Rates of definitive surgery for end-stage patellofemoral changes, including patellofemoral arthroplasty and total knee arthroplasty, were quite low at the weighted mean follow-up of several years after surgery, suggesting a role for TTO in avoiding arthroplasty. Although the infection rate was <1%, the rate of tibial tubercle or proximal tibia fractures was a cumulative 3.1%. Patients should be counseled on this complication risk, as treatment can require cast immobilization and weight-bearing limitations.²⁴

The 69% proportion of women in the overall cohort and the mean (SD) age of 27.68 (10.45) years highlight the primary patient population that undergoes TTO. Compared with men, young women are more likely to have aberrant patellofemoral biomechanics, owing to their native anatomy, including their relatively larger Q-angle and TT-TG distance and thus increased lateral translational force vectors on the patella.³⁷ In addition, more than half of patients who are having TTO underwent previous surgery on the affected knee—an indication that TTO is still not universally considered first-line in addressing patellofemoral pathology.

Limitations of the Analysis

The limitations of this analysis derive from the limitations of the included studies, which were mostly retrospective case series with relatively short follow-up. The low MCMS (<55) of all 21 studies highlights their low quality as well. These studies showed considerable heterogeneity in their reporting of specific preoperative, intraoperative, and postoperative radiographic, physical examination, and clinical outcome scores, which may be indicative of the relatively low rate of use of TTO, a procedure originally described decades ago. These studies also showed ample heterogeneity in the specific radiographic parameters or outcome scales they used to present their data. We were therefore limited in our ability to cohesively summarize and provide cumulative data points from the patients as a unified cohort. There was substantial variety in the procedures performed, surgical techniques used, concomitant pathologies addressed at time of surgery, and diagnoses treated—indicating a performance bias. This additionally precluded any significant meta-analysis within the patient cohort. A higher quality study, a randomized controlled trial, is needed to answer more definitively and completely the questions we left unanswered, including the effect on radiographic parameters, additional clinical outcomes, and patient satisfaction.

Conclusion

TTO is most commonly performed for isolated patellar instability in the presence of knee pain. Other pathologies addressed are patellofemoral osteoarthritis, and patella alta and patella baja with and without associated knee pain. TTO significantly improves knee pain and clinical outcome scores, though 21% of patients (>1 in 5) require reoperation for hardware removal. Young women with prior surgery on the affected knee are the primary patient population.

Take-Home Points

• TTO specifics depend on anatomy, radiographic alignment characteristics, and presence of chondral defects.
• Osteotomy and movement of the tibial tubercle can include anteriorization, anteromedialization, proximalization, medialization, or distalization.
• TTO was most commonly performed for isolated patellar instability in the presence of knee pain.
• Young women with prior surgery on the affected knee made up the primary patient population for this procedure.
• While TTO significantly improves knee pain and clinical outcome scores, >1 in 5 patients required reoperation for hardware removal.

Patellofemoral pain and patellofemoral instability are common orthopedic problems. Studies have found that 30% of patients 13 to 19 years old have patellofemoral pain and that 29 in 100,000 patients 10 to 17 years old have patellofemoral instability.^1-3 The reported rate of recurrence after nonoperative management of patellofemoral instability is 33%.⁴ Tibial tubercle osteotomy (TTO), first described by Hauser⁵ in 1938, is an effective treatment option for many patellofemoral disorders.

TTO indications include patellofemoral maltracking or malalignment, patellar instability, patellofemoral arthritis, and focal patellofemoral chondral defects.⁶ With TTO, the goal is to move the tibial tubercle in a direction that will either improve patellar tracking or offload the medial or lateral patellar facet to improve pain and function.^7,8 This action typically involves anterior, medial, lateral, or distal translation of the tibial tubercle, as posteriorization can lead to increased contact forces across the patellofemoral joint, resulting in accelerated patellofemoral wear and increased pain.⁹

We systematically reviewed the TTO literature to identify indications, clinical outcomes, complications, and reoperations. We hypothesized that the overall complication rate and the overall reoperation rate would both be <10%.

Clinical Evaluation of Patellofemoral Pathology

Patients with patellofemoral pain often report anterior knee pain, which typically begins gradually and is often activity related. Several symptoms may be present: pain with prolonged sitting with knees bent; pain on rising from a seated position; pain or crepitus with climbing stairs; and pain during repetitive activity such as running, squatting, or jumping. Location, duration, and onset of symptoms should be elicited. Patellofemoral instability can be described as dislocation events or subluxation events; number of events, mechanisms of injury, and resulting need for reduction should be documented. As age, sex, body mass index, and physical fitness are relevant to risk of recurrence, the physician should ask about general ligamentous laxity, other joint dislocations, and prior surgical intervention. Swelling or mechanical symptoms may indicate patellofemoral joint pathology.^6,10

Physical examination of patients with patellofemoral pathology begins with assessment for overall limb alignment (including resting position of patella and corresponding quadriceps angle [Q-angle]), generalized ligamentous laxity (including hypermobile joints, evaluated with Brighton criteria), overall peri-knee muscle tone and strength, effusion, and gait pattern.

Common TTO Procedures

TTO specifics depend on anatomy, radiographic alignment characteristics, and presence of chondral defects. Essentially, the patella is translated to offload the affected areas. Osteotomy and movement of the tibial tubercle can include anteriorization, anteromedialization, proximalization, medialization, or distalization.

Methods

Search Strategy and Data Collection

We searched the PubMed (Medline) database for all English-language TTO studies published between database inception and April 9, 2015. After PROSPERO registration, and following PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines, we used the algorithm (“tibial” AND “tubercle” AND “osteotomy”) NOT (“total” AND “knee” AND “arthroplasty”) to search the literature. Inclusion criteria included level I-IV studies on TTO indications, operative findings, and outcomes. Exclusion criteria were non-English studies, unpublished studies, level V evidence, letters to the editor, editorials, review articles, basic science articles, technique articles, revision procedures, articles without clinical outcomes, and conference proceeding abstracts. Studies that reported on duplicate populations were included only with the most recent available clinical outcomes. All abstracts were reviewed in duplicate by Dr. Levy and Dr. Rao and assessed with respect to the criteria outlined. Then the same authors performed full-text reviews of eligible studies before including these studies in the systematic review.

Assessment of Study Quality

The quality of each TTO study in the review was assessed with a modified Coleman methodology score (MCMS), which ranges from 0 to 100. A study with an MCMS of <55 points is considered a poor-quality study.¹¹

Data Synthesis and Statistical Analysis

Given that most of the included studies were level IV, a formal meta-analysis was not indicated. In this article, we report categorical data as frequencies with percentages and preoperative and postoperative continuous data as means (SDs), with weighted means based on number of patients in each study, where applicable. We used 2-tailed t tests for comparisons made with the free Meta-Analysis Calculator and Grapher (http://www.healthstrategy.com/meta/meta.pl ). Statistical significance was set at P < .05. 

Results

Search Results and Included Studies

Only 1 study provided preoperative body mass index (27 kg/m2). There were 55.35% of patients who had prior surgery on the affected knee (6 studies reporting).

Preoperative Data

Preoperative pathologic, radiographic, and clinical scoring data were scarcely reported and nonuniform (Table 2). The most common pathology treated with TTO was isolated patellofemoral instability (746/1055 patients, 70.7%). The other pathologies addressed were isolated patellofemoral osteoarthritis/chondromalacia patellae (143, 13.6%), patellofemoral instability with patella alta (61, 5.8%), patellofemoral instability with patellofemoral osteoarthritis (45, 4.3%), isolated patella baja (41, 3.9%), isolated patella alta (19, 1.8%), and patellofemoral osteoarthritis with patella baja (2, 0.2%). Five hundred fifty-five patients (53%) had a preoperative complaint that included knee pain, and 809 (77%) reported preoperative patellar laxity or instability events. The imaging data reported were Q-angle, Insall-Salvati ratio, Caton-Deschamps index, Blackburne-Peel ratio, Outerbridge osteoarthritis grade, and TT-TG distance. Preoperative clinical scoring data most prominently included a visual analog scale (VAS) score of 70.50 (4 studies reporting), a Lysholm score of 59.19 (5 studies), and a Kujala score of 41.16 (4 studies). Shelbourne-Trumper and Cox-Insall scores were reported in 1 and 2 studies, respectively.

Operative Characteristics

Of the 21 studies, 12 reported only on patients who had TTO performed in isolation; in the other 9 studies, cohorts included patients who underwent concurrent procedures. In the 17 studies (856 patients) that listed numbers of patients who underwent specific concomitant procedures, 715 patients (83.5%) underwent an isolated TTO procedure, and the other 141 (16.5%) underwent either concomitant lateral femoral trochleoplasty, arthroscopic drilling of chondral lesions, patellar shaving chondroplasty, partial meniscectomy or concomitant meniscal repair, intra-articular loose body removal, and/or lateral release with or without medial plication. 

Postoperative Data

There was a cumulative total of 79 complications (8% of cohort): 17 recurrent patellar dislocations (1.9%), 4 recurrent patellar subluxations (0.4%), 10 wound complications (1.0%), 2 intraoperative complications (0.2%), 14 tibial tubercle fractures (1.3%), 19 proximal tibia fractures (1.8%), 4 cases of anterior knee pain (0.4%), 4 cases of neuropraxia (0.4%), and 5 infections (0.5%). Of note, 219 knees (21%) required reoperation, but 170 (16.3%) of these were for painful hardware removal. Sixteen knees (1.5%) required revision TTO, 1 (0.1%) required subsequent high tibial osteotomy, 2 (0.2%) underwent patellofemoral arthroplasty for advanced arthritic changes, and 5 (0.5%) underwent total knee arthroplasty for advanced arthritic changes.

Studies With TTO Performed in Isolation

Twelve studies reported outcomes of isolated TTO procedures. In the 638 patients who underwent isolated TTO, the pathologies addressed were instability/laxity (429 patients, 67%), patellofemoral osteoarthritis (74, 12%), patella alta with instability (61, 10%), patellofemoral osteoarthritis with instability (31, 5%), patella baja (24, 4%), and patella alta (19, 3%). Pain was a preoperative issue in 289 (45%) of these patients and instability in 472 (74%).

Only 2.8% of patients experienced postoperative patellar dislocation events. Of the 12 studies, 2 reported VAS scores (34-point weighted mean improvement, 65 points before surgery to 31 after surgery), 3 reported Lysholm scores (30-point improvement, from 60 to 90), and 2 reported Kujala scores (21-point improvement, from 46 to 67).

Complication rates for this isolated-TTO pooled cohort of patients were 1.2% for revision TTOs, 0.5% for wound complications, 0.8% for tibial tubercle fractures, and 1.9% for proximal tibia fractures. In total, 16% of patients required hardware removal after surgery. 

Discussion

This study found that TTO improved patient pain and clinical outcome scores despite having a high (16%) rate of reoperation for painful hardware in patients with preoperative pain or instability, or with patellofemoral osteoarthritis or aberrant patellar anatomy. This reoperation rate and the overall complication rate both exceeded our hypothesized 10% cumulative rate. However, <1% of patients required conversion to a definitive end-stage surgery (patellofemoral arthroplasty or total knee arthroplasty) by final follow-up, and the rates of comorbidities (anterior knee pain, wound infection, recurrent patellar subluxation/dislocation, tibial fracture) were relatively low.

Patellofemoral disorders are common in the general population and a frequent primary complaint on presentation to orthopedic offices. Having a thorough understanding of knee joint biomechanics is imperative when trying to determine whether surgery is appropriate for these complaints and how to proceed. Extensor mechanism abnormalities, including high lateral force vectors (or larger TT-TG distances) and excessive patellar tilt, can affect alignment and increase the risk for patellofemoral dislocations, patellofemoral anterior- based knee pains, and chondral lesions. Patella alta, an elevated patella, risks increased contact stresses between the patella and the trochlear groove³³ and decreases the osseous constraints that inhibit dislocation of the patella with physiologic flexion of the joint.³⁴ With TTO, the change in tuberosity position can alter angles in the extensor mechanism and thereby decrease joint reaction forces and patellofemoral contact area forces.^35,36

Although its use began as an option for combating patellar instability events in patients with predisposed patellofemoral kinematics,⁵ TTO has evolved in its therapeutic uses to include offloading patellar and trochlear focal chondral lesions and slowing progression of patellofemoral arthritis. Multiple iterations and modifications of the procedure have involved distal and medial transfer of the tibial tuberosity, medialization alone, concurrent anterior and medial elevation of the tuberosity, and proximal or distal transfers, depending on the pathology being corrected. Although TTO is highly versatile in treating multiple patellofemoral joint pathologies, this study found that its primary indication continues to be patellar instability, with anteromedialization as the most common direction of tubercle transfer in support of the medial structures providing the medial force vector that keeps the patella in place. These medial structures include the medial patellofemoral ligament, the vastus medialis obliquus, the medial patellotibial ligament, and the medial retinaculum. 

Also notable was the relatively high rate of reoperation after TTO. However, >75% of reoperations were performed to remove painful hardware, and the need for reoperation seemed to have no effect on the statistically significant overall preoperative-to-postoperative improvement in VAS, Lysholm, and Kujala scores. Rates of definitive surgery for end-stage patellofemoral changes, including patellofemoral arthroplasty and total knee arthroplasty, were quite low at the weighted mean follow-up of several years after surgery, suggesting a role for TTO in avoiding arthroplasty. Although the infection rate was <1%, the rate of tibial tubercle or proximal tibia fractures was a cumulative 3.1%. Patients should be counseled on this complication risk, as treatment can require cast immobilization and weight-bearing limitations.²⁴

The 69% proportion of women in the overall cohort and the mean (SD) age of 27.68 (10.45) years highlight the primary patient population that undergoes TTO. Compared with men, young women are more likely to have aberrant patellofemoral biomechanics, owing to their native anatomy, including their relatively larger Q-angle and TT-TG distance and thus increased lateral translational force vectors on the patella.³⁷ In addition, more than half of patients who are having TTO underwent previous surgery on the affected knee—an indication that TTO is still not universally considered first-line in addressing patellofemoral pathology.

Limitations of the Analysis

The limitations of this analysis derive from the limitations of the included studies, which were mostly retrospective case series with relatively short follow-up. The low MCMS (<55) of all 21 studies highlights their low quality as well. These studies showed considerable heterogeneity in their reporting of specific preoperative, intraoperative, and postoperative radiographic, physical examination, and clinical outcome scores, which may be indicative of the relatively low rate of use of TTO, a procedure originally described decades ago. These studies also showed ample heterogeneity in the specific radiographic parameters or outcome scales they used to present their data. We were therefore limited in our ability to cohesively summarize and provide cumulative data points from the patients as a unified cohort. There was substantial variety in the procedures performed, surgical techniques used, concomitant pathologies addressed at time of surgery, and diagnoses treated—indicating a performance bias. This additionally precluded any significant meta-analysis within the patient cohort. A higher quality study, a randomized controlled trial, is needed to answer more definitively and completely the questions we left unanswered, including the effect on radiographic parameters, additional clinical outcomes, and patient satisfaction.

Conclusion

TTO is most commonly performed for isolated patellar instability in the presence of knee pain. Other pathologies addressed are patellofemoral osteoarthritis, and patella alta and patella baja with and without associated knee pain. TTO significantly improves knee pain and clinical outcome scores, though 21% of patients (>1 in 5) require reoperation for hardware removal. Young women with prior surgery on the affected knee are the primary patient population.

Take-Home Points

• TTO specifics depend on anatomy, radiographic alignment characteristics, and presence of chondral defects.
• Osteotomy and movement of the tibial tubercle can include anteriorization, anteromedialization, proximalization, medialization, or distalization.
• TTO was most commonly performed for isolated patellar instability in the presence of knee pain.
• Young women with prior surgery on the affected knee made up the primary patient population for this procedure.
• While TTO significantly improves knee pain and clinical outcome scores, >1 in 5 patients required reoperation for hardware removal.

Patellofemoral pain and patellofemoral instability are common orthopedic problems. Studies have found that 30% of patients 13 to 19 years old have patellofemoral pain and that 29 in 100,000 patients 10 to 17 years old have patellofemoral instability.^1-3 The reported rate of recurrence after nonoperative management of patellofemoral instability is 33%.⁴ Tibial tubercle osteotomy (TTO), first described by Hauser⁵ in 1938, is an effective treatment option for many patellofemoral disorders.

TTO indications include patellofemoral maltracking or malalignment, patellar instability, patellofemoral arthritis, and focal patellofemoral chondral defects.⁶ With TTO, the goal is to move the tibial tubercle in a direction that will either improve patellar tracking or offload the medial or lateral patellar facet to improve pain and function.^7,8 This action typically involves anterior, medial, lateral, or distal translation of the tibial tubercle, as posteriorization can lead to increased contact forces across the patellofemoral joint, resulting in accelerated patellofemoral wear and increased pain.⁹

We systematically reviewed the TTO literature to identify indications, clinical outcomes, complications, and reoperations. We hypothesized that the overall complication rate and the overall reoperation rate would both be <10%.

Clinical Evaluation of Patellofemoral Pathology

Patients with patellofemoral pain often report anterior knee pain, which typically begins gradually and is often activity related. Several symptoms may be present: pain with prolonged sitting with knees bent; pain on rising from a seated position; pain or crepitus with climbing stairs; and pain during repetitive activity such as running, squatting, or jumping. Location, duration, and onset of symptoms should be elicited. Patellofemoral instability can be described as dislocation events or subluxation events; number of events, mechanisms of injury, and resulting need for reduction should be documented. As age, sex, body mass index, and physical fitness are relevant to risk of recurrence, the physician should ask about general ligamentous laxity, other joint dislocations, and prior surgical intervention. Swelling or mechanical symptoms may indicate patellofemoral joint pathology.^6,10

Physical examination of patients with patellofemoral pathology begins with assessment for overall limb alignment (including resting position of patella and corresponding quadriceps angle [Q-angle]), generalized ligamentous laxity (including hypermobile joints, evaluated with Brighton criteria), overall peri-knee muscle tone and strength, effusion, and gait pattern.

Common TTO Procedures

TTO specifics depend on anatomy, radiographic alignment characteristics, and presence of chondral defects. Essentially, the patella is translated to offload the affected areas. Osteotomy and movement of the tibial tubercle can include anteriorization, anteromedialization, proximalization, medialization, or distalization.

Methods

Search Strategy and Data Collection

We searched the PubMed (Medline) database for all English-language TTO studies published between database inception and April 9, 2015. After PROSPERO registration, and following PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines, we used the algorithm (“tibial” AND “tubercle” AND “osteotomy”) NOT (“total” AND “knee” AND “arthroplasty”) to search the literature. Inclusion criteria included level I-IV studies on TTO indications, operative findings, and outcomes. Exclusion criteria were non-English studies, unpublished studies, level V evidence, letters to the editor, editorials, review articles, basic science articles, technique articles, revision procedures, articles without clinical outcomes, and conference proceeding abstracts. Studies that reported on duplicate populations were included only with the most recent available clinical outcomes. All abstracts were reviewed in duplicate by Dr. Levy and Dr. Rao and assessed with respect to the criteria outlined. Then the same authors performed full-text reviews of eligible studies before including these studies in the systematic review.

Assessment of Study Quality

The quality of each TTO study in the review was assessed with a modified Coleman methodology score (MCMS), which ranges from 0 to 100. A study with an MCMS of <55 points is considered a poor-quality study.¹¹

Data Synthesis and Statistical Analysis

Given that most of the included studies were level IV, a formal meta-analysis was not indicated. In this article, we report categorical data as frequencies with percentages and preoperative and postoperative continuous data as means (SDs), with weighted means based on number of patients in each study, where applicable. We used 2-tailed t tests for comparisons made with the free Meta-Analysis Calculator and Grapher (http://www.healthstrategy.com/meta/meta.pl ). Statistical significance was set at P < .05. 

Results

Search Results and Included Studies

Only 1 study provided preoperative body mass index (27 kg/m2). There were 55.35% of patients who had prior surgery on the affected knee (6 studies reporting).

Preoperative Data

Preoperative pathologic, radiographic, and clinical scoring data were scarcely reported and nonuniform (Table 2). The most common pathology treated with TTO was isolated patellofemoral instability (746/1055 patients, 70.7%). The other pathologies addressed were isolated patellofemoral osteoarthritis/chondromalacia patellae (143, 13.6%), patellofemoral instability with patella alta (61, 5.8%), patellofemoral instability with patellofemoral osteoarthritis (45, 4.3%), isolated patella baja (41, 3.9%), isolated patella alta (19, 1.8%), and patellofemoral osteoarthritis with patella baja (2, 0.2%). Five hundred fifty-five patients (53%) had a preoperative complaint that included knee pain, and 809 (77%) reported preoperative patellar laxity or instability events. The imaging data reported were Q-angle, Insall-Salvati ratio, Caton-Deschamps index, Blackburne-Peel ratio, Outerbridge osteoarthritis grade, and TT-TG distance. Preoperative clinical scoring data most prominently included a visual analog scale (VAS) score of 70.50 (4 studies reporting), a Lysholm score of 59.19 (5 studies), and a Kujala score of 41.16 (4 studies). Shelbourne-Trumper and Cox-Insall scores were reported in 1 and 2 studies, respectively.

Operative Characteristics

Of the 21 studies, 12 reported only on patients who had TTO performed in isolation; in the other 9 studies, cohorts included patients who underwent concurrent procedures. In the 17 studies (856 patients) that listed numbers of patients who underwent specific concomitant procedures, 715 patients (83.5%) underwent an isolated TTO procedure, and the other 141 (16.5%) underwent either concomitant lateral femoral trochleoplasty, arthroscopic drilling of chondral lesions, patellar shaving chondroplasty, partial meniscectomy or concomitant meniscal repair, intra-articular loose body removal, and/or lateral release with or without medial plication. 

Postoperative Data

There was a cumulative total of 79 complications (8% of cohort): 17 recurrent patellar dislocations (1.9%), 4 recurrent patellar subluxations (0.4%), 10 wound complications (1.0%), 2 intraoperative complications (0.2%), 14 tibial tubercle fractures (1.3%), 19 proximal tibia fractures (1.8%), 4 cases of anterior knee pain (0.4%), 4 cases of neuropraxia (0.4%), and 5 infections (0.5%). Of note, 219 knees (21%) required reoperation, but 170 (16.3%) of these were for painful hardware removal. Sixteen knees (1.5%) required revision TTO, 1 (0.1%) required subsequent high tibial osteotomy, 2 (0.2%) underwent patellofemoral arthroplasty for advanced arthritic changes, and 5 (0.5%) underwent total knee arthroplasty for advanced arthritic changes.

Studies With TTO Performed in Isolation

Twelve studies reported outcomes of isolated TTO procedures. In the 638 patients who underwent isolated TTO, the pathologies addressed were instability/laxity (429 patients, 67%), patellofemoral osteoarthritis (74, 12%), patella alta with instability (61, 10%), patellofemoral osteoarthritis with instability (31, 5%), patella baja (24, 4%), and patella alta (19, 3%). Pain was a preoperative issue in 289 (45%) of these patients and instability in 472 (74%).

Only 2.8% of patients experienced postoperative patellar dislocation events. Of the 12 studies, 2 reported VAS scores (34-point weighted mean improvement, 65 points before surgery to 31 after surgery), 3 reported Lysholm scores (30-point improvement, from 60 to 90), and 2 reported Kujala scores (21-point improvement, from 46 to 67).

Complication rates for this isolated-TTO pooled cohort of patients were 1.2% for revision TTOs, 0.5% for wound complications, 0.8% for tibial tubercle fractures, and 1.9% for proximal tibia fractures. In total, 16% of patients required hardware removal after surgery. 

Discussion

This study found that TTO improved patient pain and clinical outcome scores despite having a high (16%) rate of reoperation for painful hardware in patients with preoperative pain or instability, or with patellofemoral osteoarthritis or aberrant patellar anatomy. This reoperation rate and the overall complication rate both exceeded our hypothesized 10% cumulative rate. However, <1% of patients required conversion to a definitive end-stage surgery (patellofemoral arthroplasty or total knee arthroplasty) by final follow-up, and the rates of comorbidities (anterior knee pain, wound infection, recurrent patellar subluxation/dislocation, tibial fracture) were relatively low.

Patellofemoral disorders are common in the general population and a frequent primary complaint on presentation to orthopedic offices. Having a thorough understanding of knee joint biomechanics is imperative when trying to determine whether surgery is appropriate for these complaints and how to proceed. Extensor mechanism abnormalities, including high lateral force vectors (or larger TT-TG distances) and excessive patellar tilt, can affect alignment and increase the risk for patellofemoral dislocations, patellofemoral anterior- based knee pains, and chondral lesions. Patella alta, an elevated patella, risks increased contact stresses between the patella and the trochlear groove³³ and decreases the osseous constraints that inhibit dislocation of the patella with physiologic flexion of the joint.³⁴ With TTO, the change in tuberosity position can alter angles in the extensor mechanism and thereby decrease joint reaction forces and patellofemoral contact area forces.^35,36

Although its use began as an option for combating patellar instability events in patients with predisposed patellofemoral kinematics,⁵ TTO has evolved in its therapeutic uses to include offloading patellar and trochlear focal chondral lesions and slowing progression of patellofemoral arthritis. Multiple iterations and modifications of the procedure have involved distal and medial transfer of the tibial tuberosity, medialization alone, concurrent anterior and medial elevation of the tuberosity, and proximal or distal transfers, depending on the pathology being corrected. Although TTO is highly versatile in treating multiple patellofemoral joint pathologies, this study found that its primary indication continues to be patellar instability, with anteromedialization as the most common direction of tubercle transfer in support of the medial structures providing the medial force vector that keeps the patella in place. These medial structures include the medial patellofemoral ligament, the vastus medialis obliquus, the medial patellotibial ligament, and the medial retinaculum. 

Also notable was the relatively high rate of reoperation after TTO. However, >75% of reoperations were performed to remove painful hardware, and the need for reoperation seemed to have no effect on the statistically significant overall preoperative-to-postoperative improvement in VAS, Lysholm, and Kujala scores. Rates of definitive surgery for end-stage patellofemoral changes, including patellofemoral arthroplasty and total knee arthroplasty, were quite low at the weighted mean follow-up of several years after surgery, suggesting a role for TTO in avoiding arthroplasty. Although the infection rate was <1%, the rate of tibial tubercle or proximal tibia fractures was a cumulative 3.1%. Patients should be counseled on this complication risk, as treatment can require cast immobilization and weight-bearing limitations.²⁴

The 69% proportion of women in the overall cohort and the mean (SD) age of 27.68 (10.45) years highlight the primary patient population that undergoes TTO. Compared with men, young women are more likely to have aberrant patellofemoral biomechanics, owing to their native anatomy, including their relatively larger Q-angle and TT-TG distance and thus increased lateral translational force vectors on the patella.³⁷ In addition, more than half of patients who are having TTO underwent previous surgery on the affected knee—an indication that TTO is still not universally considered first-line in addressing patellofemoral pathology.

Limitations of the Analysis

The limitations of this analysis derive from the limitations of the included studies, which were mostly retrospective case series with relatively short follow-up. The low MCMS (<55) of all 21 studies highlights their low quality as well. These studies showed considerable heterogeneity in their reporting of specific preoperative, intraoperative, and postoperative radiographic, physical examination, and clinical outcome scores, which may be indicative of the relatively low rate of use of TTO, a procedure originally described decades ago. These studies also showed ample heterogeneity in the specific radiographic parameters or outcome scales they used to present their data. We were therefore limited in our ability to cohesively summarize and provide cumulative data points from the patients as a unified cohort. There was substantial variety in the procedures performed, surgical techniques used, concomitant pathologies addressed at time of surgery, and diagnoses treated—indicating a performance bias. This additionally precluded any significant meta-analysis within the patient cohort. A higher quality study, a randomized controlled trial, is needed to answer more definitively and completely the questions we left unanswered, including the effect on radiographic parameters, additional clinical outcomes, and patient satisfaction.

Conclusion

TTO is most commonly performed for isolated patellar instability in the presence of knee pain. Other pathologies addressed are patellofemoral osteoarthritis, and patella alta and patella baja with and without associated knee pain. TTO significantly improves knee pain and clinical outcome scores, though 21% of patients (>1 in 5) require reoperation for hardware removal. Young women with prior surgery on the affected knee are the primary patient population.

Combat veterans who have served in Iraq and Afghanistan in the post-9/11 era face unique reintegration challenges, one being the transition from driving in combat zones to driving at home.¹ Relative to previous conflicts, post-9/11 combat involves increased participation in road patrols and convoys along with more prevalent threats of improvised explosive devices (IEDs).^1,2 Roadside ambushes designed to destroy or stop vehicles became a common warfare strategy, meaning that driving became an inherently dangerous combat maneuver.³

The modern combat driving framework includes cognitive tools (eg, targeted aggression and tactical awareness) combined with specific behaviors (eg, driving unpredictably fast, using rapid lane changes, and keeping other vehicles at a distance to avoid IEDs).⁴ This framework is adaptive and lifesaving in combat zones, but it can be maladaptive and dangerous in civilian environments. Service members face difficulty in updating this cognitive framework after leaving combat zones and may continue to experience specific cognitions (eg, “the world is dangerous”; “that car holds an IED”) while driving on civilian roads.^3,5-8

The high prevalence of posttraumatic stress disorder (PTSD) and traumatic brain injury (TBI) in post-9/11 veterans may complicate reintegration. Both PTSD and TBI are considered signature wounds of these conflicts.^8-11 Traumatic brain injury may be sustained as a result of blast injury or other mechanism, including a closed head injury or penetrating brain injury.¹⁰ Previous literature indicated that both PTSD and TBI across all severities are related to deficits in executive functioning, attention, and memory.^12-16

In addition to cognitive deficits, veterans with PTSD also may experience cognitive misappraisal, in which they are more likely to perceive ambiguous stimuli as threatening because of an inability to suppress trauma-related schema and associations.^5,17,18 Examples of roadside-specific trauma triggers include busy highways, traffic, loud or distracting noises, and vehicles of similar make and model as those commonly rigged with IEDs in Iraq or Afghanistan.^2,7

Blast injury, often from IEDs, is the most common cause of TBI reported in U.S. service members, so veterans that have experienced such an injury may become hyperaware of vehicles that may appear to hide IEDs.^7,19 Cognitive dysfunction and misappraisal of neutral stimuli may have an additive effect on behaviors and experiences behind the wheel.^7,15,20 As a result, veterans with comorbid PTSD and TBI may drive unsafely, self-restrict driving time, or avoid driving completely.^5,8,18

Prior research suggests that veterans with PTSD and/or TBI experience significantly higher levels of anxiety in response to common roadside stimuli (ie, an overpass or stop sign) while driving than do veterans without either PTSD or TBI.³ Cognitive behavioral therapy (CBT) interventions have been developed and systematically evaluated for treating anxiety.²¹ The goal of CBT is to identify and change dysfunctional cognitions that result in biased information processing. Cognitive restructuring, the process by which problematic cognitions (negative automatic thoughts) are identified and examined for distortions, is one method of accomplishing this goal. Distortions then are disputed and rebutted with assistance from the clinician.²² A strategy for restructuring negative automatic thoughts is coping self-instruction, which centers on identifying when negative automatic thoughts are focused on others’ behavior, accepting that their behavior cannot be changed, and using positive coping behaviors to minimize negative automatic thoughts.²³

The link between history of comorbid PTSD and TBI and combat driving, current driving anxiety, and coping strategies has not yet been extensively studied in veterans. Thus, the aim of the current study is to determine whether veterans with comorbid PTSD and TBI utilize coping self-instruction behind the wheel. Driving-specific coping self-instruction involves generating thoughts that are adaptive and accepting of others’ driving behaviors (eg, “Just turn up the radio and tune them out”). It was hypothesized that veterans with comorbid PTSD and TBI would endorse fewer coping self-instruction thoughts than would veterans without either PTSD or TBI.

Methods

The current project is part of a larger study that examines driving behaviors of post-9/11 combat veterans at the Michael J. Crescenz Veterans Affairs Medical Center in Philadelphia, Pennsylvania. Thirty-two male veterans aged between 22 and 48 years (M = 31.6, SD = 6.9) were included in the sample. Twenty-three were diagnosed with comorbid PTSD and TBI and 9 veterans with no major psychiatric or physical

Assessment

All participants completed a battery of questionnaires, including the Driver’s Angry Thoughts Questionnaire (DATQ).²³ The DATQ was used to investigate the specific thoughts that veterans experienced while driving.²³ Participants indicated on a Likert scale from 1 (not at all) to 5 (all the time) how often they experienced any of 65 thoughts while driving. Each item was categorized into 1 of 5 distinct subscales (Table 2). A frequency score was generated for each of the 5 subscales. Each subscale had good internal consistency and convergent, divergent, and predictive validity. The Coping Self-Instruction subscale, which is defined as engaging in relaxing thoughts to accept others’ driving behaviors, was of primary interest. It is a 9-item scale (frequency score can range from 9 to 45) with good reliability (α = .83).²³

Given the small and unequal sample sizes, nonparametric independent samples Mann-Whitney U-tests were selected to compare frequency of driving-related thoughts across veterans with comorbid PTSD and TBI and those of veterans without either PTSD or TBI.

Results

Descriptive statistics and results for each DATQ subscale are reported in Table 3. Group comparisons revealed that veterans with comorbid PTSD and TBI endorsed statistically significantly fewer coping self-instruction thoughts while driving (M = 11.5, SD = 7.2) than did combat veterans without either PTSD or TBI (M = 18.1, SD = 6.9; U = 56.0, P = .05). Conversely, frequency of angry thoughts were statistically significant in their difference as a function of PTSD or TBI diagnostic status.

Discussion

While driving, veterans with PTSD or TBI endorsed statistically significantly fewer coping self-instruction thoughts than did veterans without either PTSD or TBI. Prior research suggests that veterans with PTSD or TBI experience greater anxiety than do veterans without either condition while driving.^2,3 Taken together, this suggests that veterans with PTSD or TBI may lack efficient cognitive coping strategies related to the anxiety they experience while driving. Furthermore, the groups did not significantly differ in frequency of angry thoughts behind the wheel. This result was expected based on prior analyses that suggested that veterans with and without PTSD or TBI endorsed feelings of aggression, impatience, and frustration while driving at similar frequencies.³

Because all veterans in the current sample were exposed to combat, these results help to parse out the unique contribution of PTSD and TBI diagnoses on driving in civilian environments. Exposure to combat plus diagnoses of PTSD or TBI may be related to veterans’ ability to cope with typical driving situations at home. In the context of prior literature, results suggest that veterans with PTSD or TBI automatically may perceive neutral roadside stimuli as threatening, feel anxious in response to this perceived threat, and be ill-equipped to cope with this anxiety.^3,5,17,18 According to CBT models, negative automatic thoughts play a critical role in maintaining anxiety.²⁴ Particular cognitive distortions associated with PTSD symptomatology and combat driving experiences, such as misperceiving ambiguous stimuli as threatening because of an inability to suppress trauma-related schema and associations, may therefore maintain driving anxiety following military separation.

Research on CBT interventions suggests that cognitive restructuring, including coping self-instruction, are effective treatments to reduce anxiety.^22,24 The current findings suggest that combat veterans with PTSD and TBI who experience driving anxiety endorse significantly fewer coping self-instruction thoughts than do controls in response to anxiety-provoking driving situations. In fact, prior research suggests that a majority of veterans experiencing driving-related anxiety do not seek help for their symptoms, and many of those who do prefer to reach out to friends rather than mental health professionals.² However, due to their high levels of anxiety, these veterans likely would benefit from CBT interventions specifically targeted to coping strategies for civilian driving. These coping strategies should focus on recognizing that common roadside stimuli are not necessarily threatening in civilian environments. This type of cognitive restructuring may help veterans better manage anxiety while driving.

Limitations

The current study is limited by its small and unequal sample sizes and lack of a noncombat exposure comparison group. Additionally, while this study highlights a potential relationship between reduced cognitive coping strategies and behind-the-wheel anxiety in veterans with PTSD or TBI, causal inferences cannot be made. It is possible that individuals without coping strategies who are deployed to combat are more likely to develop PTSD or TBI. Being equipped with few coping strategies may then lead these veterans to experience greater anxiety while driving. Conversely, PTSD and TBI symptoms may prevent veterans from developing coping strategies over time.

Furthermore, the comorbid PTSD and TBI group was separated from the military for significantly longer than was the control group. Future studies using a longitudinal design could better examine the potential causal relationship between comorbid PTSD and TBI and coping and determine whether endorsement of coping self-instruction changes as a function of time since military separation.

Veterans in the current study report a variety of deployment experiences and locations. Methods of combat, type of vehicle, driving terrain, and prevalence of IEDs changed over the multiple post-9/11 military campaigns. Veterans who were deployed to Iraq in the mid-2000s were instructed to drive quickly and erratically to avoid IEDs and mortars, whereas veterans deployed in later years were taught to drive slowly and carefully to hunt for IEDs in heavily armored vehicles.³ Seventy-five percent of the veterans with PTSD or TBI in the current sample were deployed to Iraq in the early to mid-2000s, compared with 33% of the veterans without PTSD or TBI. Thus, the 2 groups in the current sample may have experienced different combat environments, which could impact how they perceived roadside stimuli. Future studies should recruit a larger and more balanced sample to better determine whether specific combat experiences impact coping strategies while driving.

Conclusion

To the best of the authors’ knowledge, the current study is the first to examine specific types of thoughts that veterans with and without PTSD or TBI experience while driving on civilian roads. Veterans with PTSD or TBI are not engaging in as many coping self-instruction thoughts behind the wheel, despite experiencing greater anxiety than that of veterans without either PTSD or TBI. Cognitive behavioral therapy interventions for anxiety include engaging in coping self-instruction during anxiety-provoking situations.²² Therefore, veterans with PTSD or TBI may benefit from learning targeted coping self-instruction thoughts that they can utilize when anxiety-provoking situations arise behind the wheel. Results suggest that clinicians should work with veterans with comorbid PTSD and TBI to develop specific coping self-instruction statements that they can utilize internally when faced with anxiety-provoking driving situations.

Acknowledgments
This study is the result of work supported by the Council on Brain Injury (grant #260472). The authors thank Dr. Rosette Biester for her guidance.

Combat veterans who have served in Iraq and Afghanistan in the post-9/11 era face unique reintegration challenges, one being the transition from driving in combat zones to driving at home.¹ Relative to previous conflicts, post-9/11 combat involves increased participation in road patrols and convoys along with more prevalent threats of improvised explosive devices (IEDs).^1,2 Roadside ambushes designed to destroy or stop vehicles became a common warfare strategy, meaning that driving became an inherently dangerous combat maneuver.³

The modern combat driving framework includes cognitive tools (eg, targeted aggression and tactical awareness) combined with specific behaviors (eg, driving unpredictably fast, using rapid lane changes, and keeping other vehicles at a distance to avoid IEDs).⁴ This framework is adaptive and lifesaving in combat zones, but it can be maladaptive and dangerous in civilian environments. Service members face difficulty in updating this cognitive framework after leaving combat zones and may continue to experience specific cognitions (eg, “the world is dangerous”; “that car holds an IED”) while driving on civilian roads.^3,5-8

The high prevalence of posttraumatic stress disorder (PTSD) and traumatic brain injury (TBI) in post-9/11 veterans may complicate reintegration. Both PTSD and TBI are considered signature wounds of these conflicts.^8-11 Traumatic brain injury may be sustained as a result of blast injury or other mechanism, including a closed head injury or penetrating brain injury.¹⁰ Previous literature indicated that both PTSD and TBI across all severities are related to deficits in executive functioning, attention, and memory.^12-16

In addition to cognitive deficits, veterans with PTSD also may experience cognitive misappraisal, in which they are more likely to perceive ambiguous stimuli as threatening because of an inability to suppress trauma-related schema and associations.^5,17,18 Examples of roadside-specific trauma triggers include busy highways, traffic, loud or distracting noises, and vehicles of similar make and model as those commonly rigged with IEDs in Iraq or Afghanistan.^2,7

Blast injury, often from IEDs, is the most common cause of TBI reported in U.S. service members, so veterans that have experienced such an injury may become hyperaware of vehicles that may appear to hide IEDs.^7,19 Cognitive dysfunction and misappraisal of neutral stimuli may have an additive effect on behaviors and experiences behind the wheel.^7,15,20 As a result, veterans with comorbid PTSD and TBI may drive unsafely, self-restrict driving time, or avoid driving completely.^5,8,18

Prior research suggests that veterans with PTSD and/or TBI experience significantly higher levels of anxiety in response to common roadside stimuli (ie, an overpass or stop sign) while driving than do veterans without either PTSD or TBI.³ Cognitive behavioral therapy (CBT) interventions have been developed and systematically evaluated for treating anxiety.²¹ The goal of CBT is to identify and change dysfunctional cognitions that result in biased information processing. Cognitive restructuring, the process by which problematic cognitions (negative automatic thoughts) are identified and examined for distortions, is one method of accomplishing this goal. Distortions then are disputed and rebutted with assistance from the clinician.²² A strategy for restructuring negative automatic thoughts is coping self-instruction, which centers on identifying when negative automatic thoughts are focused on others’ behavior, accepting that their behavior cannot be changed, and using positive coping behaviors to minimize negative automatic thoughts.²³

The link between history of comorbid PTSD and TBI and combat driving, current driving anxiety, and coping strategies has not yet been extensively studied in veterans. Thus, the aim of the current study is to determine whether veterans with comorbid PTSD and TBI utilize coping self-instruction behind the wheel. Driving-specific coping self-instruction involves generating thoughts that are adaptive and accepting of others’ driving behaviors (eg, “Just turn up the radio and tune them out”). It was hypothesized that veterans with comorbid PTSD and TBI would endorse fewer coping self-instruction thoughts than would veterans without either PTSD or TBI.

Methods

The current project is part of a larger study that examines driving behaviors of post-9/11 combat veterans at the Michael J. Crescenz Veterans Affairs Medical Center in Philadelphia, Pennsylvania. Thirty-two male veterans aged between 22 and 48 years (M = 31.6, SD = 6.9) were included in the sample. Twenty-three were diagnosed with comorbid PTSD and TBI and 9 veterans with no major psychiatric or physical

Assessment

All participants completed a battery of questionnaires, including the Driver’s Angry Thoughts Questionnaire (DATQ).²³ The DATQ was used to investigate the specific thoughts that veterans experienced while driving.²³ Participants indicated on a Likert scale from 1 (not at all) to 5 (all the time) how often they experienced any of 65 thoughts while driving. Each item was categorized into 1 of 5 distinct subscales (Table 2). A frequency score was generated for each of the 5 subscales. Each subscale had good internal consistency and convergent, divergent, and predictive validity. The Coping Self-Instruction subscale, which is defined as engaging in relaxing thoughts to accept others’ driving behaviors, was of primary interest. It is a 9-item scale (frequency score can range from 9 to 45) with good reliability (α = .83).²³

Given the small and unequal sample sizes, nonparametric independent samples Mann-Whitney U-tests were selected to compare frequency of driving-related thoughts across veterans with comorbid PTSD and TBI and those of veterans without either PTSD or TBI.

Results

Descriptive statistics and results for each DATQ subscale are reported in Table 3. Group comparisons revealed that veterans with comorbid PTSD and TBI endorsed statistically significantly fewer coping self-instruction thoughts while driving (M = 11.5, SD = 7.2) than did combat veterans without either PTSD or TBI (M = 18.1, SD = 6.9; U = 56.0, P = .05). Conversely, frequency of angry thoughts were statistically significant in their difference as a function of PTSD or TBI diagnostic status.

Discussion

While driving, veterans with PTSD or TBI endorsed statistically significantly fewer coping self-instruction thoughts than did veterans without either PTSD or TBI. Prior research suggests that veterans with PTSD or TBI experience greater anxiety than do veterans without either condition while driving.^2,3 Taken together, this suggests that veterans with PTSD or TBI may lack efficient cognitive coping strategies related to the anxiety they experience while driving. Furthermore, the groups did not significantly differ in frequency of angry thoughts behind the wheel. This result was expected based on prior analyses that suggested that veterans with and without PTSD or TBI endorsed feelings of aggression, impatience, and frustration while driving at similar frequencies.³

Because all veterans in the current sample were exposed to combat, these results help to parse out the unique contribution of PTSD and TBI diagnoses on driving in civilian environments. Exposure to combat plus diagnoses of PTSD or TBI may be related to veterans’ ability to cope with typical driving situations at home. In the context of prior literature, results suggest that veterans with PTSD or TBI automatically may perceive neutral roadside stimuli as threatening, feel anxious in response to this perceived threat, and be ill-equipped to cope with this anxiety.^3,5,17,18 According to CBT models, negative automatic thoughts play a critical role in maintaining anxiety.²⁴ Particular cognitive distortions associated with PTSD symptomatology and combat driving experiences, such as misperceiving ambiguous stimuli as threatening because of an inability to suppress trauma-related schema and associations, may therefore maintain driving anxiety following military separation.

Research on CBT interventions suggests that cognitive restructuring, including coping self-instruction, are effective treatments to reduce anxiety.^22,24 The current findings suggest that combat veterans with PTSD and TBI who experience driving anxiety endorse significantly fewer coping self-instruction thoughts than do controls in response to anxiety-provoking driving situations. In fact, prior research suggests that a majority of veterans experiencing driving-related anxiety do not seek help for their symptoms, and many of those who do prefer to reach out to friends rather than mental health professionals.² However, due to their high levels of anxiety, these veterans likely would benefit from CBT interventions specifically targeted to coping strategies for civilian driving. These coping strategies should focus on recognizing that common roadside stimuli are not necessarily threatening in civilian environments. This type of cognitive restructuring may help veterans better manage anxiety while driving.

Limitations

The current study is limited by its small and unequal sample sizes and lack of a noncombat exposure comparison group. Additionally, while this study highlights a potential relationship between reduced cognitive coping strategies and behind-the-wheel anxiety in veterans with PTSD or TBI, causal inferences cannot be made. It is possible that individuals without coping strategies who are deployed to combat are more likely to develop PTSD or TBI. Being equipped with few coping strategies may then lead these veterans to experience greater anxiety while driving. Conversely, PTSD and TBI symptoms may prevent veterans from developing coping strategies over time.

Furthermore, the comorbid PTSD and TBI group was separated from the military for significantly longer than was the control group. Future studies using a longitudinal design could better examine the potential causal relationship between comorbid PTSD and TBI and coping and determine whether endorsement of coping self-instruction changes as a function of time since military separation.

Veterans in the current study report a variety of deployment experiences and locations. Methods of combat, type of vehicle, driving terrain, and prevalence of IEDs changed over the multiple post-9/11 military campaigns. Veterans who were deployed to Iraq in the mid-2000s were instructed to drive quickly and erratically to avoid IEDs and mortars, whereas veterans deployed in later years were taught to drive slowly and carefully to hunt for IEDs in heavily armored vehicles.³ Seventy-five percent of the veterans with PTSD or TBI in the current sample were deployed to Iraq in the early to mid-2000s, compared with 33% of the veterans without PTSD or TBI. Thus, the 2 groups in the current sample may have experienced different combat environments, which could impact how they perceived roadside stimuli. Future studies should recruit a larger and more balanced sample to better determine whether specific combat experiences impact coping strategies while driving.

Conclusion

To the best of the authors’ knowledge, the current study is the first to examine specific types of thoughts that veterans with and without PTSD or TBI experience while driving on civilian roads. Veterans with PTSD or TBI are not engaging in as many coping self-instruction thoughts behind the wheel, despite experiencing greater anxiety than that of veterans without either PTSD or TBI. Cognitive behavioral therapy interventions for anxiety include engaging in coping self-instruction during anxiety-provoking situations.²² Therefore, veterans with PTSD or TBI may benefit from learning targeted coping self-instruction thoughts that they can utilize when anxiety-provoking situations arise behind the wheel. Results suggest that clinicians should work with veterans with comorbid PTSD and TBI to develop specific coping self-instruction statements that they can utilize internally when faced with anxiety-provoking driving situations.

Acknowledgments
This study is the result of work supported by the Council on Brain Injury (grant #260472). The authors thank Dr. Rosette Biester for her guidance.

Combat veterans who have served in Iraq and Afghanistan in the post-9/11 era face unique reintegration challenges, one being the transition from driving in combat zones to driving at home.¹ Relative to previous conflicts, post-9/11 combat involves increased participation in road patrols and convoys along with more prevalent threats of improvised explosive devices (IEDs).^1,2 Roadside ambushes designed to destroy or stop vehicles became a common warfare strategy, meaning that driving became an inherently dangerous combat maneuver.³

The modern combat driving framework includes cognitive tools (eg, targeted aggression and tactical awareness) combined with specific behaviors (eg, driving unpredictably fast, using rapid lane changes, and keeping other vehicles at a distance to avoid IEDs).⁴ This framework is adaptive and lifesaving in combat zones, but it can be maladaptive and dangerous in civilian environments. Service members face difficulty in updating this cognitive framework after leaving combat zones and may continue to experience specific cognitions (eg, “the world is dangerous”; “that car holds an IED”) while driving on civilian roads.^3,5-8

The high prevalence of posttraumatic stress disorder (PTSD) and traumatic brain injury (TBI) in post-9/11 veterans may complicate reintegration. Both PTSD and TBI are considered signature wounds of these conflicts.^8-11 Traumatic brain injury may be sustained as a result of blast injury or other mechanism, including a closed head injury or penetrating brain injury.¹⁰ Previous literature indicated that both PTSD and TBI across all severities are related to deficits in executive functioning, attention, and memory.^12-16

In addition to cognitive deficits, veterans with PTSD also may experience cognitive misappraisal, in which they are more likely to perceive ambiguous stimuli as threatening because of an inability to suppress trauma-related schema and associations.^5,17,18 Examples of roadside-specific trauma triggers include busy highways, traffic, loud or distracting noises, and vehicles of similar make and model as those commonly rigged with IEDs in Iraq or Afghanistan.^2,7

Blast injury, often from IEDs, is the most common cause of TBI reported in U.S. service members, so veterans that have experienced such an injury may become hyperaware of vehicles that may appear to hide IEDs.^7,19 Cognitive dysfunction and misappraisal of neutral stimuli may have an additive effect on behaviors and experiences behind the wheel.^7,15,20 As a result, veterans with comorbid PTSD and TBI may drive unsafely, self-restrict driving time, or avoid driving completely.^5,8,18

Prior research suggests that veterans with PTSD and/or TBI experience significantly higher levels of anxiety in response to common roadside stimuli (ie, an overpass or stop sign) while driving than do veterans without either PTSD or TBI.³ Cognitive behavioral therapy (CBT) interventions have been developed and systematically evaluated for treating anxiety.²¹ The goal of CBT is to identify and change dysfunctional cognitions that result in biased information processing. Cognitive restructuring, the process by which problematic cognitions (negative automatic thoughts) are identified and examined for distortions, is one method of accomplishing this goal. Distortions then are disputed and rebutted with assistance from the clinician.²² A strategy for restructuring negative automatic thoughts is coping self-instruction, which centers on identifying when negative automatic thoughts are focused on others’ behavior, accepting that their behavior cannot be changed, and using positive coping behaviors to minimize negative automatic thoughts.²³

The link between history of comorbid PTSD and TBI and combat driving, current driving anxiety, and coping strategies has not yet been extensively studied in veterans. Thus, the aim of the current study is to determine whether veterans with comorbid PTSD and TBI utilize coping self-instruction behind the wheel. Driving-specific coping self-instruction involves generating thoughts that are adaptive and accepting of others’ driving behaviors (eg, “Just turn up the radio and tune them out”). It was hypothesized that veterans with comorbid PTSD and TBI would endorse fewer coping self-instruction thoughts than would veterans without either PTSD or TBI.

Methods

The current project is part of a larger study that examines driving behaviors of post-9/11 combat veterans at the Michael J. Crescenz Veterans Affairs Medical Center in Philadelphia, Pennsylvania. Thirty-two male veterans aged between 22 and 48 years (M = 31.6, SD = 6.9) were included in the sample. Twenty-three were diagnosed with comorbid PTSD and TBI and 9 veterans with no major psychiatric or physical

Assessment

All participants completed a battery of questionnaires, including the Driver’s Angry Thoughts Questionnaire (DATQ).²³ The DATQ was used to investigate the specific thoughts that veterans experienced while driving.²³ Participants indicated on a Likert scale from 1 (not at all) to 5 (all the time) how often they experienced any of 65 thoughts while driving. Each item was categorized into 1 of 5 distinct subscales (Table 2). A frequency score was generated for each of the 5 subscales. Each subscale had good internal consistency and convergent, divergent, and predictive validity. The Coping Self-Instruction subscale, which is defined as engaging in relaxing thoughts to accept others’ driving behaviors, was of primary interest. It is a 9-item scale (frequency score can range from 9 to 45) with good reliability (α = .83).²³

Given the small and unequal sample sizes, nonparametric independent samples Mann-Whitney U-tests were selected to compare frequency of driving-related thoughts across veterans with comorbid PTSD and TBI and those of veterans without either PTSD or TBI.

Results

Descriptive statistics and results for each DATQ subscale are reported in Table 3. Group comparisons revealed that veterans with comorbid PTSD and TBI endorsed statistically significantly fewer coping self-instruction thoughts while driving (M = 11.5, SD = 7.2) than did combat veterans without either PTSD or TBI (M = 18.1, SD = 6.9; U = 56.0, P = .05). Conversely, frequency of angry thoughts were statistically significant in their difference as a function of PTSD or TBI diagnostic status.

Discussion

While driving, veterans with PTSD or TBI endorsed statistically significantly fewer coping self-instruction thoughts than did veterans without either PTSD or TBI. Prior research suggests that veterans with PTSD or TBI experience greater anxiety than do veterans without either condition while driving.^2,3 Taken together, this suggests that veterans with PTSD or TBI may lack efficient cognitive coping strategies related to the anxiety they experience while driving. Furthermore, the groups did not significantly differ in frequency of angry thoughts behind the wheel. This result was expected based on prior analyses that suggested that veterans with and without PTSD or TBI endorsed feelings of aggression, impatience, and frustration while driving at similar frequencies.³

Because all veterans in the current sample were exposed to combat, these results help to parse out the unique contribution of PTSD and TBI diagnoses on driving in civilian environments. Exposure to combat plus diagnoses of PTSD or TBI may be related to veterans’ ability to cope with typical driving situations at home. In the context of prior literature, results suggest that veterans with PTSD or TBI automatically may perceive neutral roadside stimuli as threatening, feel anxious in response to this perceived threat, and be ill-equipped to cope with this anxiety.^3,5,17,18 According to CBT models, negative automatic thoughts play a critical role in maintaining anxiety.²⁴ Particular cognitive distortions associated with PTSD symptomatology and combat driving experiences, such as misperceiving ambiguous stimuli as threatening because of an inability to suppress trauma-related schema and associations, may therefore maintain driving anxiety following military separation.

Research on CBT interventions suggests that cognitive restructuring, including coping self-instruction, are effective treatments to reduce anxiety.^22,24 The current findings suggest that combat veterans with PTSD and TBI who experience driving anxiety endorse significantly fewer coping self-instruction thoughts than do controls in response to anxiety-provoking driving situations. In fact, prior research suggests that a majority of veterans experiencing driving-related anxiety do not seek help for their symptoms, and many of those who do prefer to reach out to friends rather than mental health professionals.² However, due to their high levels of anxiety, these veterans likely would benefit from CBT interventions specifically targeted to coping strategies for civilian driving. These coping strategies should focus on recognizing that common roadside stimuli are not necessarily threatening in civilian environments. This type of cognitive restructuring may help veterans better manage anxiety while driving.

Limitations

The current study is limited by its small and unequal sample sizes and lack of a noncombat exposure comparison group. Additionally, while this study highlights a potential relationship between reduced cognitive coping strategies and behind-the-wheel anxiety in veterans with PTSD or TBI, causal inferences cannot be made. It is possible that individuals without coping strategies who are deployed to combat are more likely to develop PTSD or TBI. Being equipped with few coping strategies may then lead these veterans to experience greater anxiety while driving. Conversely, PTSD and TBI symptoms may prevent veterans from developing coping strategies over time.

Furthermore, the comorbid PTSD and TBI group was separated from the military for significantly longer than was the control group. Future studies using a longitudinal design could better examine the potential causal relationship between comorbid PTSD and TBI and coping and determine whether endorsement of coping self-instruction changes as a function of time since military separation.

Veterans in the current study report a variety of deployment experiences and locations. Methods of combat, type of vehicle, driving terrain, and prevalence of IEDs changed over the multiple post-9/11 military campaigns. Veterans who were deployed to Iraq in the mid-2000s were instructed to drive quickly and erratically to avoid IEDs and mortars, whereas veterans deployed in later years were taught to drive slowly and carefully to hunt for IEDs in heavily armored vehicles.³ Seventy-five percent of the veterans with PTSD or TBI in the current sample were deployed to Iraq in the early to mid-2000s, compared with 33% of the veterans without PTSD or TBI. Thus, the 2 groups in the current sample may have experienced different combat environments, which could impact how they perceived roadside stimuli. Future studies should recruit a larger and more balanced sample to better determine whether specific combat experiences impact coping strategies while driving.

Conclusion

To the best of the authors’ knowledge, the current study is the first to examine specific types of thoughts that veterans with and without PTSD or TBI experience while driving on civilian roads. Veterans with PTSD or TBI are not engaging in as many coping self-instruction thoughts behind the wheel, despite experiencing greater anxiety than that of veterans without either PTSD or TBI. Cognitive behavioral therapy interventions for anxiety include engaging in coping self-instruction during anxiety-provoking situations.²² Therefore, veterans with PTSD or TBI may benefit from learning targeted coping self-instruction thoughts that they can utilize when anxiety-provoking situations arise behind the wheel. Results suggest that clinicians should work with veterans with comorbid PTSD and TBI to develop specific coping self-instruction statements that they can utilize internally when faced with anxiety-provoking driving situations.

Acknowledgments
This study is the result of work supported by the Council on Brain Injury (grant #260472). The authors thank Dr. Rosette Biester for her guidance.

Dermatology departments in the United States have been facing challenges in recruiting and retaining dermatologists for academic positions. Accordingly, a survey study reported that academic dermatologists were more likely than those in private practice to state that their institutions were recruiting new associates.¹ Several factors could explain this phenomenon. Salary differences between jobs in academic and nonacademic settings may contribute to difficulty in recruiting dermatologists into academia, which is exacerbated by a theoretical shortage of dermatologists, leading to graduates who receive and accept private practice job offers.^1,2 Furthermore, a large survey study reported that challenges unique to academic dermatologists include longer patient wait times in addition to responsibilities such as research, hospital consultations, medical writing, and teaching. These patterns raise concerns for the future of teaching institutions because academic dermatologists not only train future physicians but also conduct clinical and basic science research necessary to advance the field and improve patient care.² Thus, it is important to evaluate the factors that affect career decisions in dermatology and to determine if these factors can be addressed. We hypothesized that student loan burden influences career plans in dermatology and that physicians are not fully educated on loan repayment options. The aims of this preliminary study were to explore the influence of student loan burden on career plans in dermatology and to determine if the Public Service Loan Forgiveness (PSLF) program could potentially encourage more dermatologists to consider academic careers.

Methods

The study aimed to investigate the factors that influence career decisions in dermatology and to assess attitudes toward the PSLF program as an option for student loan repayment. The target population included dermatology residents and attending physicians in the United States. Survey questions were adapted from a previously published study³ and were modified based on feedback from reviewers in the University of California (UC) Irvine department of dermatology. The survey was voluntary and did not collect identifying information. This study was granted exemption from oversight by the UC Irvine institutional review board.

Recruitment materials informed potential participants of the nature of the study and provided a hyperlink to the electronic survey. The UC Irvine department of dermatology emailed US dermatology residency program coordinators, requesting that they forward this study to residents and attending physicians in their programs. The survey also was distributed by one of the investigators (E.S.) on a social networking website for dermatologists. Data were collected from June 2015 to October 2015. A χ² analysis was conducted using SAS statistical software, and P<.05 was considered statistically significant.

Results

Demographics
The survey had 70 respondents including residents (56 [80.0%]) and attending physicians (14 [20.0%]). The mean age (SD) of the respondents was 32.4 (6.1) years, with 31 (44.3%) men and 39 (55.7%) women. The majority were married (38 [54.3%]) and did not have children (48 [68.6%]). Most respondents reported an annual household income of $200,000 or less (55 [78.6%]) and perceived a comfortable annual household income as greater than $200,000 (59 [84.3%])(Table 1).

Financing Medical Education
Most respondents currently had $200,000 or less in student loan debt (40 [57.1%]) and financed more than half of their medical education with student loans (53 [75.7%]). A large majority (61 [87.1%]) indicated that some portion of their medical education was funded by student loans. Other means of financing education included family assistance (29 [40.8%]), scholarships (26 [36.6%]), personal savings (18 [25.3%]), and military benefits (1 [1.4%]); 18.3% (n=13) of respondents did not accrue any educational debt. Respondents endorsed several plans for loan repayment including government-sponsored programs (34 [47.9%]), refinancing with a private company (13 [18.3%]), and the PSLF (1 [1.4%]); 12.7% (n=9) of respondents were uncertain how they were going to repay their student loans.

Career Goals in Dermatology and the Influence of Student Loans
Respondents were asked to specify their career plans before versus after starting dermatology residency training (ie, current career plan). Prior to starting residency, 36 (51.4%) and 34 (48.6%) respondents indicated they were interested in private practice and academia, respectively. After starting residency, the number of respondents interested in private practice increased to 45 (64.3%), and the number of respondents interested in academia decreased to 25 (35.7%). Fifteen (21.4%) respondents changed career trajectories from academia to private practice, 6 (8.6%) changed from private practice to academia, and 49 (70.0%) did not change career goals.

The majority of respondents (39 [55.7%]) indicated that the amount of their student loan debt did not influence their career goals (Table 2); however, those with more than $200,000 in debt were more likely to state that student loans impacted their career goals compared to those with $200,000 or less in debt (70.0% [21/30] vs 25.0% [10/40]; P<.001).

Comparison of Respondents Interested in Careers in Academia vs Private Practice
There were differences in financial circumstances between respondents interested in academia versus those interested in private practice. Compared to respondents interested in academia, those interested in private practice were more likely to have more than $200,000 in student loan debt (24 [53.3%] vs 6 [24.0%]; P<.05), have more than half of their education paid with student loans (38 [84.4%] vs 15 [60.0%]; P<.05), and state that student debt affected their career goals (28 [62.2%] vs 3 [12.0%]; P<.001)(Table 3). Demographic characteristics including gender, marital status, parental status, and current annual household income were not associated with a specific career goal.

Subgroup analysis was performed on respondents who were initially interested in academic careers but subsequently decided to pursue private practice (n=15). Compared to those who remained interested in academia, those who changed career plans from academia to private practice were more likely to have more than $200,000 in student debt (76.9% [10/13] vs 23.1% [3/13]; P<.001) and were more likely to state that loans affected their career goals (86.7% [13/15] vs 13.3% [2/15]; P<.001).

Residency program experience also may influence career trajectory. The majority (n=41 [58.6%]) of respondents indicated that their residency program experience affected their dermatology career goals. Of those, 41.7% and 58.5% stated that their residency program experiences dissuaded and persuaded them into academic positions, respectively. Those interested in academic dermatology were more likely to state that their residency program experience influenced their career goals (80.0% [20/25] vs 46.7% [21/45]; P<.05). Furthermore, those interested in academic positions responded with higher overall residency program satisfaction ratings on a scale of 1 to 10 (1 indicated the lowest satisfaction) than those interested in private practice, but the difference was not significant (mean [SD] score, 8.2 [2.3] vs 7.2 [1.9]; P=.07).

Respondents were asked to rate their interest in the following dermatology-related professional interests on a scale of 1 to 5 (1 indicated the least enjoyment): medical dermatology, dermatologic surgery, dermatopathology, cosmetics, and lasers. Those interested in private practice versus those interested in academic dermatology found more enjoyment in dermatologic surgery (mean [SD] score, 4.0 [0.8] vs 3.4 [1.3]; P<.05), cosmetics (3.4 [1.2] vs 2.6 [1.4]; P<.05), and lasers (3.7 [1.0] vs 2.8 [1.2]; P<.05)(Table 4).

Respondents also were asked to select primary motivating factors for their career goals (ie, academia or private practice) and to indicate reasons for not choosing the alternative. The majority of those pursuing academia were motivated by opportunities to collaborate with colleagues (23 [92.0%]), teach and mentor (20 [80%]), and manage complex cases (17 [68.0%]). Most of the respondents who were pursing private practice were motivated by focus on patient care and clinician duties (36 [80.0%]), flexible work hours (35 [77.8%]), higher income (33 [73.3%]), and location flexibility (29 [64.4%]). Among those interested in academic dermatology, the top factors for disinterest in private practice were running a business (9 [36.0%]) and high patient turnover (9 [36.0%]). Most of those interested in private practice indicated that they were not interested in an academic position because of lower income (31 [68.9%]) and research duties (31 [68.9%])(Table 5).

Awareness of and Attitudes Toward the PSLF Program
The majority of respondents were aware of PSLF (53 [75.7%]); however, only 1 respondent endorsed current plans to use PSLF for loan repayment. Respondents were asked how likely they would be to pursue an academic position if given the option to have their student loans forgiven by the PSLF program. Overall, 44.6% (n=25) of respondents indicated that this option would have no effect or would unlikely convince them to pursue an academic position, and 55.4% (n=31) of respondents indicated that they were somewhat likely, likely, or very likely to pursue academia if PSLF was an option. Of those who stated that they would consider enrolling in PSLF, 64.5% (20/31) of individuals were pursuing careers in private practice. Neither current student loan burden nor career goal was associated with likelihood of enrolling in the PSLF.

Comment

In 2015, 76% of medical school graduates in the United States accrued educational debt, with an average of $189,165, a number that has continued to increase over the years.⁴ In addition to the increasing cost of medical education, higher interest rates on federal student loans contribute to debt burden. Over the last 2 decades, some research has posited that debt may influence medical specialty selection, with most studies focusing on primary care.^5-9 However, there is limited information on the effect of student loan debt on career decisions within dermatology.

The results of our study suggest that financial factors including income and amount of educational debt may influence career decisions in dermatology. There is a known income gap between academic and nonacademic settings. According to the Association of American Medical Colleges, the median salaries for an assistant professor and associate/full professor in dermatology are $266,000 and $331,000, respectively.¹⁰ In contrast, for dermatologists in private clinical practice, the overall median salary is more than $450,000. These salary differences may impact career plans, as 73.3% (33/45) of survey respondents in our study who were interested in private practice stated that higher income was one of the primary motivating factors behind their career goals. Student loan burden also may impact career decisions in dermatology. Compared to those with lower student loan debt (≤$200,000), respondents with higher student loan debt (>$200,000) showed greater interest in pursuing positions in private practice and were more likely to change career goals from academia to private practice after starting residency. Thus, academic institutions are potentially losing physicians who are interested in academic careers but ultimately decide against it, at least partially due to high student loan debt.

The PSLF can potentially address this issue and be used as a recruiting tool for dermatology positions in academia. Under PSLF, borrowers can have the remainder of their loan balances forgiven after making 120 monthly payments while employed full time by public service employers, including some academic medical institutions. In our study, a large majority of respondents indicated that they are aware of the PSLF, and more than half said they would consider pursuing positions in academia if their loans could be forgiven through the program; however, when asked about plans for loan repayment, only 1 respondent endorsed current plans to enroll in PSLF. Thus, despite high interest in PSLF among the survey respondents, few had actual plans to use the service, suggesting that perhaps dermatologists are not provided enough information about PSLF to motivate enrollment. In the same way, almost a quarter of respondents were not familiar with the PSLF as a repayment option, further signifying that distribution of information about financial planning may be inadequate. If student loan burden is a notable factor in career decisions in dermatology, it is important that academic institutions provide sufficient information about repayment to encourage informed decisions. As such, it is possible that educating physicians about options such as PSLF can potentially recruit more dermatologists to academic positions.

Aside from financial reasons, residency program experience and differences in practices in academic and nonacademic settings may impact career trajectories. The majority of respondents stated their residency program experience influenced their career decisions; however, the majority of respondents did not change their minds about career goals since starting residency, suggesting that residency program experience may reinforce but not necessarily alter these choices. Interests in specific focuses within dermatology also may influence career decisions. This study suggests that those pursuing private practice positions are more interested in dermatologic surgery, lasers, and cosmetics. Roles in research, mentoring, and patient care differ between academic and nonacademic settings and also may be influential in career decisions in dermatology, mirroring the implications of other studies that posited that career decisions within medicine are complex and involve multiple factors.¹¹

In this study, we did not find an association between gender and career plans in dermatology. In 2013, more than 60% of dermatology resident physicians were female.¹² However, a recent study suggested that women face challenges in academic dermatology, including a downtrend in the number of female investigators with grants from the National Institutes of Health.¹³Taken together, female dermatologists, who currently represent the majority of resident physicians training in dermatology, may face unique challenges in academia, further highlighting the potential difficulties with recruitment of dermatologists by academic institutions in the future.

This preliminary study has several limitations. First, the small sample size limited generalizability to all dermatologists. Second, responder bias was possible, as those who have stronger opinions about this topic may have been more inclined to participate in this voluntary survey. Future studies with larger sample sizes are needed to further explore the factors that influence career decisions within dermatology and to determine if there are additional means to increase recruitment into academia.

Conclusion

It is recognized that there are challenges in recruiting dermatologists into academic positions. This study suggests that student loan burden influences career decisions in dermatology. Dermatologists may not be fully educated on options for student loan repayment. With increased awareness, the PSLF can potentially be used as a recruitment tool for positions in academic dermatology.

Dermatology departments in the United States have been facing challenges in recruiting and retaining dermatologists for academic positions. Accordingly, a survey study reported that academic dermatologists were more likely than those in private practice to state that their institutions were recruiting new associates.¹ Several factors could explain this phenomenon. Salary differences between jobs in academic and nonacademic settings may contribute to difficulty in recruiting dermatologists into academia, which is exacerbated by a theoretical shortage of dermatologists, leading to graduates who receive and accept private practice job offers.^1,2 Furthermore, a large survey study reported that challenges unique to academic dermatologists include longer patient wait times in addition to responsibilities such as research, hospital consultations, medical writing, and teaching. These patterns raise concerns for the future of teaching institutions because academic dermatologists not only train future physicians but also conduct clinical and basic science research necessary to advance the field and improve patient care.² Thus, it is important to evaluate the factors that affect career decisions in dermatology and to determine if these factors can be addressed. We hypothesized that student loan burden influences career plans in dermatology and that physicians are not fully educated on loan repayment options. The aims of this preliminary study were to explore the influence of student loan burden on career plans in dermatology and to determine if the Public Service Loan Forgiveness (PSLF) program could potentially encourage more dermatologists to consider academic careers.

Methods

The study aimed to investigate the factors that influence career decisions in dermatology and to assess attitudes toward the PSLF program as an option for student loan repayment. The target population included dermatology residents and attending physicians in the United States. Survey questions were adapted from a previously published study³ and were modified based on feedback from reviewers in the University of California (UC) Irvine department of dermatology. The survey was voluntary and did not collect identifying information. This study was granted exemption from oversight by the UC Irvine institutional review board.

Recruitment materials informed potential participants of the nature of the study and provided a hyperlink to the electronic survey. The UC Irvine department of dermatology emailed US dermatology residency program coordinators, requesting that they forward this study to residents and attending physicians in their programs. The survey also was distributed by one of the investigators (E.S.) on a social networking website for dermatologists. Data were collected from June 2015 to October 2015. A χ² analysis was conducted using SAS statistical software, and P<.05 was considered statistically significant.

Results

Demographics
The survey had 70 respondents including residents (56 [80.0%]) and attending physicians (14 [20.0%]). The mean age (SD) of the respondents was 32.4 (6.1) years, with 31 (44.3%) men and 39 (55.7%) women. The majority were married (38 [54.3%]) and did not have children (48 [68.6%]). Most respondents reported an annual household income of $200,000 or less (55 [78.6%]) and perceived a comfortable annual household income as greater than $200,000 (59 [84.3%])(Table 1).

Financing Medical Education
Most respondents currently had $200,000 or less in student loan debt (40 [57.1%]) and financed more than half of their medical education with student loans (53 [75.7%]). A large majority (61 [87.1%]) indicated that some portion of their medical education was funded by student loans. Other means of financing education included family assistance (29 [40.8%]), scholarships (26 [36.6%]), personal savings (18 [25.3%]), and military benefits (1 [1.4%]); 18.3% (n=13) of respondents did not accrue any educational debt. Respondents endorsed several plans for loan repayment including government-sponsored programs (34 [47.9%]), refinancing with a private company (13 [18.3%]), and the PSLF (1 [1.4%]); 12.7% (n=9) of respondents were uncertain how they were going to repay their student loans.

Career Goals in Dermatology and the Influence of Student Loans
Respondents were asked to specify their career plans before versus after starting dermatology residency training (ie, current career plan). Prior to starting residency, 36 (51.4%) and 34 (48.6%) respondents indicated they were interested in private practice and academia, respectively. After starting residency, the number of respondents interested in private practice increased to 45 (64.3%), and the number of respondents interested in academia decreased to 25 (35.7%). Fifteen (21.4%) respondents changed career trajectories from academia to private practice, 6 (8.6%) changed from private practice to academia, and 49 (70.0%) did not change career goals.

The majority of respondents (39 [55.7%]) indicated that the amount of their student loan debt did not influence their career goals (Table 2); however, those with more than $200,000 in debt were more likely to state that student loans impacted their career goals compared to those with $200,000 or less in debt (70.0% [21/30] vs 25.0% [10/40]; P<.001).

Comparison of Respondents Interested in Careers in Academia vs Private Practice
There were differences in financial circumstances between respondents interested in academia versus those interested in private practice. Compared to respondents interested in academia, those interested in private practice were more likely to have more than $200,000 in student loan debt (24 [53.3%] vs 6 [24.0%]; P<.05), have more than half of their education paid with student loans (38 [84.4%] vs 15 [60.0%]; P<.05), and state that student debt affected their career goals (28 [62.2%] vs 3 [12.0%]; P<.001)(Table 3). Demographic characteristics including gender, marital status, parental status, and current annual household income were not associated with a specific career goal.

Subgroup analysis was performed on respondents who were initially interested in academic careers but subsequently decided to pursue private practice (n=15). Compared to those who remained interested in academia, those who changed career plans from academia to private practice were more likely to have more than $200,000 in student debt (76.9% [10/13] vs 23.1% [3/13]; P<.001) and were more likely to state that loans affected their career goals (86.7% [13/15] vs 13.3% [2/15]; P<.001).

Residency program experience also may influence career trajectory. The majority (n=41 [58.6%]) of respondents indicated that their residency program experience affected their dermatology career goals. Of those, 41.7% and 58.5% stated that their residency program experiences dissuaded and persuaded them into academic positions, respectively. Those interested in academic dermatology were more likely to state that their residency program experience influenced their career goals (80.0% [20/25] vs 46.7% [21/45]; P<.05). Furthermore, those interested in academic positions responded with higher overall residency program satisfaction ratings on a scale of 1 to 10 (1 indicated the lowest satisfaction) than those interested in private practice, but the difference was not significant (mean [SD] score, 8.2 [2.3] vs 7.2 [1.9]; P=.07).

Respondents were asked to rate their interest in the following dermatology-related professional interests on a scale of 1 to 5 (1 indicated the least enjoyment): medical dermatology, dermatologic surgery, dermatopathology, cosmetics, and lasers. Those interested in private practice versus those interested in academic dermatology found more enjoyment in dermatologic surgery (mean [SD] score, 4.0 [0.8] vs 3.4 [1.3]; P<.05), cosmetics (3.4 [1.2] vs 2.6 [1.4]; P<.05), and lasers (3.7 [1.0] vs 2.8 [1.2]; P<.05)(Table 4).

Respondents also were asked to select primary motivating factors for their career goals (ie, academia or private practice) and to indicate reasons for not choosing the alternative. The majority of those pursuing academia were motivated by opportunities to collaborate with colleagues (23 [92.0%]), teach and mentor (20 [80%]), and manage complex cases (17 [68.0%]). Most of the respondents who were pursing private practice were motivated by focus on patient care and clinician duties (36 [80.0%]), flexible work hours (35 [77.8%]), higher income (33 [73.3%]), and location flexibility (29 [64.4%]). Among those interested in academic dermatology, the top factors for disinterest in private practice were running a business (9 [36.0%]) and high patient turnover (9 [36.0%]). Most of those interested in private practice indicated that they were not interested in an academic position because of lower income (31 [68.9%]) and research duties (31 [68.9%])(Table 5).

Awareness of and Attitudes Toward the PSLF Program
The majority of respondents were aware of PSLF (53 [75.7%]); however, only 1 respondent endorsed current plans to use PSLF for loan repayment. Respondents were asked how likely they would be to pursue an academic position if given the option to have their student loans forgiven by the PSLF program. Overall, 44.6% (n=25) of respondents indicated that this option would have no effect or would unlikely convince them to pursue an academic position, and 55.4% (n=31) of respondents indicated that they were somewhat likely, likely, or very likely to pursue academia if PSLF was an option. Of those who stated that they would consider enrolling in PSLF, 64.5% (20/31) of individuals were pursuing careers in private practice. Neither current student loan burden nor career goal was associated with likelihood of enrolling in the PSLF.

Comment

In 2015, 76% of medical school graduates in the United States accrued educational debt, with an average of $189,165, a number that has continued to increase over the years.⁴ In addition to the increasing cost of medical education, higher interest rates on federal student loans contribute to debt burden. Over the last 2 decades, some research has posited that debt may influence medical specialty selection, with most studies focusing on primary care.^5-9 However, there is limited information on the effect of student loan debt on career decisions within dermatology.

The results of our study suggest that financial factors including income and amount of educational debt may influence career decisions in dermatology. There is a known income gap between academic and nonacademic settings. According to the Association of American Medical Colleges, the median salaries for an assistant professor and associate/full professor in dermatology are $266,000 and $331,000, respectively.¹⁰ In contrast, for dermatologists in private clinical practice, the overall median salary is more than $450,000. These salary differences may impact career plans, as 73.3% (33/45) of survey respondents in our study who were interested in private practice stated that higher income was one of the primary motivating factors behind their career goals. Student loan burden also may impact career decisions in dermatology. Compared to those with lower student loan debt (≤$200,000), respondents with higher student loan debt (>$200,000) showed greater interest in pursuing positions in private practice and were more likely to change career goals from academia to private practice after starting residency. Thus, academic institutions are potentially losing physicians who are interested in academic careers but ultimately decide against it, at least partially due to high student loan debt.

The PSLF can potentially address this issue and be used as a recruiting tool for dermatology positions in academia. Under PSLF, borrowers can have the remainder of their loan balances forgiven after making 120 monthly payments while employed full time by public service employers, including some academic medical institutions. In our study, a large majority of respondents indicated that they are aware of the PSLF, and more than half said they would consider pursuing positions in academia if their loans could be forgiven through the program; however, when asked about plans for loan repayment, only 1 respondent endorsed current plans to enroll in PSLF. Thus, despite high interest in PSLF among the survey respondents, few had actual plans to use the service, suggesting that perhaps dermatologists are not provided enough information about PSLF to motivate enrollment. In the same way, almost a quarter of respondents were not familiar with the PSLF as a repayment option, further signifying that distribution of information about financial planning may be inadequate. If student loan burden is a notable factor in career decisions in dermatology, it is important that academic institutions provide sufficient information about repayment to encourage informed decisions. As such, it is possible that educating physicians about options such as PSLF can potentially recruit more dermatologists to academic positions.

Aside from financial reasons, residency program experience and differences in practices in academic and nonacademic settings may impact career trajectories. The majority of respondents stated their residency program experience influenced their career decisions; however, the majority of respondents did not change their minds about career goals since starting residency, suggesting that residency program experience may reinforce but not necessarily alter these choices. Interests in specific focuses within dermatology also may influence career decisions. This study suggests that those pursuing private practice positions are more interested in dermatologic surgery, lasers, and cosmetics. Roles in research, mentoring, and patient care differ between academic and nonacademic settings and also may be influential in career decisions in dermatology, mirroring the implications of other studies that posited that career decisions within medicine are complex and involve multiple factors.¹¹

In this study, we did not find an association between gender and career plans in dermatology. In 2013, more than 60% of dermatology resident physicians were female.¹² However, a recent study suggested that women face challenges in academic dermatology, including a downtrend in the number of female investigators with grants from the National Institutes of Health.¹³Taken together, female dermatologists, who currently represent the majority of resident physicians training in dermatology, may face unique challenges in academia, further highlighting the potential difficulties with recruitment of dermatologists by academic institutions in the future.

This preliminary study has several limitations. First, the small sample size limited generalizability to all dermatologists. Second, responder bias was possible, as those who have stronger opinions about this topic may have been more inclined to participate in this voluntary survey. Future studies with larger sample sizes are needed to further explore the factors that influence career decisions within dermatology and to determine if there are additional means to increase recruitment into academia.

Conclusion

It is recognized that there are challenges in recruiting dermatologists into academic positions. This study suggests that student loan burden influences career decisions in dermatology. Dermatologists may not be fully educated on options for student loan repayment. With increased awareness, the PSLF can potentially be used as a recruitment tool for positions in academic dermatology.

Dermatology departments in the United States have been facing challenges in recruiting and retaining dermatologists for academic positions. Accordingly, a survey study reported that academic dermatologists were more likely than those in private practice to state that their institutions were recruiting new associates.¹ Several factors could explain this phenomenon. Salary differences between jobs in academic and nonacademic settings may contribute to difficulty in recruiting dermatologists into academia, which is exacerbated by a theoretical shortage of dermatologists, leading to graduates who receive and accept private practice job offers.^1,2 Furthermore, a large survey study reported that challenges unique to academic dermatologists include longer patient wait times in addition to responsibilities such as research, hospital consultations, medical writing, and teaching. These patterns raise concerns for the future of teaching institutions because academic dermatologists not only train future physicians but also conduct clinical and basic science research necessary to advance the field and improve patient care.² Thus, it is important to evaluate the factors that affect career decisions in dermatology and to determine if these factors can be addressed. We hypothesized that student loan burden influences career plans in dermatology and that physicians are not fully educated on loan repayment options. The aims of this preliminary study were to explore the influence of student loan burden on career plans in dermatology and to determine if the Public Service Loan Forgiveness (PSLF) program could potentially encourage more dermatologists to consider academic careers.

Methods

The study aimed to investigate the factors that influence career decisions in dermatology and to assess attitudes toward the PSLF program as an option for student loan repayment. The target population included dermatology residents and attending physicians in the United States. Survey questions were adapted from a previously published study³ and were modified based on feedback from reviewers in the University of California (UC) Irvine department of dermatology. The survey was voluntary and did not collect identifying information. This study was granted exemption from oversight by the UC Irvine institutional review board.

Recruitment materials informed potential participants of the nature of the study and provided a hyperlink to the electronic survey. The UC Irvine department of dermatology emailed US dermatology residency program coordinators, requesting that they forward this study to residents and attending physicians in their programs. The survey also was distributed by one of the investigators (E.S.) on a social networking website for dermatologists. Data were collected from June 2015 to October 2015. A χ² analysis was conducted using SAS statistical software, and P<.05 was considered statistically significant.

Results

Demographics
The survey had 70 respondents including residents (56 [80.0%]) and attending physicians (14 [20.0%]). The mean age (SD) of the respondents was 32.4 (6.1) years, with 31 (44.3%) men and 39 (55.7%) women. The majority were married (38 [54.3%]) and did not have children (48 [68.6%]). Most respondents reported an annual household income of $200,000 or less (55 [78.6%]) and perceived a comfortable annual household income as greater than $200,000 (59 [84.3%])(Table 1).

Financing Medical Education
Most respondents currently had $200,000 or less in student loan debt (40 [57.1%]) and financed more than half of their medical education with student loans (53 [75.7%]). A large majority (61 [87.1%]) indicated that some portion of their medical education was funded by student loans. Other means of financing education included family assistance (29 [40.8%]), scholarships (26 [36.6%]), personal savings (18 [25.3%]), and military benefits (1 [1.4%]); 18.3% (n=13) of respondents did not accrue any educational debt. Respondents endorsed several plans for loan repayment including government-sponsored programs (34 [47.9%]), refinancing with a private company (13 [18.3%]), and the PSLF (1 [1.4%]); 12.7% (n=9) of respondents were uncertain how they were going to repay their student loans.

Career Goals in Dermatology and the Influence of Student Loans
Respondents were asked to specify their career plans before versus after starting dermatology residency training (ie, current career plan). Prior to starting residency, 36 (51.4%) and 34 (48.6%) respondents indicated they were interested in private practice and academia, respectively. After starting residency, the number of respondents interested in private practice increased to 45 (64.3%), and the number of respondents interested in academia decreased to 25 (35.7%). Fifteen (21.4%) respondents changed career trajectories from academia to private practice, 6 (8.6%) changed from private practice to academia, and 49 (70.0%) did not change career goals.

The majority of respondents (39 [55.7%]) indicated that the amount of their student loan debt did not influence their career goals (Table 2); however, those with more than $200,000 in debt were more likely to state that student loans impacted their career goals compared to those with $200,000 or less in debt (70.0% [21/30] vs 25.0% [10/40]; P<.001).

Comparison of Respondents Interested in Careers in Academia vs Private Practice
There were differences in financial circumstances between respondents interested in academia versus those interested in private practice. Compared to respondents interested in academia, those interested in private practice were more likely to have more than $200,000 in student loan debt (24 [53.3%] vs 6 [24.0%]; P<.05), have more than half of their education paid with student loans (38 [84.4%] vs 15 [60.0%]; P<.05), and state that student debt affected their career goals (28 [62.2%] vs 3 [12.0%]; P<.001)(Table 3). Demographic characteristics including gender, marital status, parental status, and current annual household income were not associated with a specific career goal.

Subgroup analysis was performed on respondents who were initially interested in academic careers but subsequently decided to pursue private practice (n=15). Compared to those who remained interested in academia, those who changed career plans from academia to private practice were more likely to have more than $200,000 in student debt (76.9% [10/13] vs 23.1% [3/13]; P<.001) and were more likely to state that loans affected their career goals (86.7% [13/15] vs 13.3% [2/15]; P<.001).

Residency program experience also may influence career trajectory. The majority (n=41 [58.6%]) of respondents indicated that their residency program experience affected their dermatology career goals. Of those, 41.7% and 58.5% stated that their residency program experiences dissuaded and persuaded them into academic positions, respectively. Those interested in academic dermatology were more likely to state that their residency program experience influenced their career goals (80.0% [20/25] vs 46.7% [21/45]; P<.05). Furthermore, those interested in academic positions responded with higher overall residency program satisfaction ratings on a scale of 1 to 10 (1 indicated the lowest satisfaction) than those interested in private practice, but the difference was not significant (mean [SD] score, 8.2 [2.3] vs 7.2 [1.9]; P=.07).

Respondents were asked to rate their interest in the following dermatology-related professional interests on a scale of 1 to 5 (1 indicated the least enjoyment): medical dermatology, dermatologic surgery, dermatopathology, cosmetics, and lasers. Those interested in private practice versus those interested in academic dermatology found more enjoyment in dermatologic surgery (mean [SD] score, 4.0 [0.8] vs 3.4 [1.3]; P<.05), cosmetics (3.4 [1.2] vs 2.6 [1.4]; P<.05), and lasers (3.7 [1.0] vs 2.8 [1.2]; P<.05)(Table 4).

Respondents also were asked to select primary motivating factors for their career goals (ie, academia or private practice) and to indicate reasons for not choosing the alternative. The majority of those pursuing academia were motivated by opportunities to collaborate with colleagues (23 [92.0%]), teach and mentor (20 [80%]), and manage complex cases (17 [68.0%]). Most of the respondents who were pursing private practice were motivated by focus on patient care and clinician duties (36 [80.0%]), flexible work hours (35 [77.8%]), higher income (33 [73.3%]), and location flexibility (29 [64.4%]). Among those interested in academic dermatology, the top factors for disinterest in private practice were running a business (9 [36.0%]) and high patient turnover (9 [36.0%]). Most of those interested in private practice indicated that they were not interested in an academic position because of lower income (31 [68.9%]) and research duties (31 [68.9%])(Table 5).

Awareness of and Attitudes Toward the PSLF Program
The majority of respondents were aware of PSLF (53 [75.7%]); however, only 1 respondent endorsed current plans to use PSLF for loan repayment. Respondents were asked how likely they would be to pursue an academic position if given the option to have their student loans forgiven by the PSLF program. Overall, 44.6% (n=25) of respondents indicated that this option would have no effect or would unlikely convince them to pursue an academic position, and 55.4% (n=31) of respondents indicated that they were somewhat likely, likely, or very likely to pursue academia if PSLF was an option. Of those who stated that they would consider enrolling in PSLF, 64.5% (20/31) of individuals were pursuing careers in private practice. Neither current student loan burden nor career goal was associated with likelihood of enrolling in the PSLF.

Comment

In 2015, 76% of medical school graduates in the United States accrued educational debt, with an average of $189,165, a number that has continued to increase over the years.⁴ In addition to the increasing cost of medical education, higher interest rates on federal student loans contribute to debt burden. Over the last 2 decades, some research has posited that debt may influence medical specialty selection, with most studies focusing on primary care.^5-9 However, there is limited information on the effect of student loan debt on career decisions within dermatology.

The results of our study suggest that financial factors including income and amount of educational debt may influence career decisions in dermatology. There is a known income gap between academic and nonacademic settings. According to the Association of American Medical Colleges, the median salaries for an assistant professor and associate/full professor in dermatology are $266,000 and $331,000, respectively.¹⁰ In contrast, for dermatologists in private clinical practice, the overall median salary is more than $450,000. These salary differences may impact career plans, as 73.3% (33/45) of survey respondents in our study who were interested in private practice stated that higher income was one of the primary motivating factors behind their career goals. Student loan burden also may impact career decisions in dermatology. Compared to those with lower student loan debt (≤$200,000), respondents with higher student loan debt (>$200,000) showed greater interest in pursuing positions in private practice and were more likely to change career goals from academia to private practice after starting residency. Thus, academic institutions are potentially losing physicians who are interested in academic careers but ultimately decide against it, at least partially due to high student loan debt.

The PSLF can potentially address this issue and be used as a recruiting tool for dermatology positions in academia. Under PSLF, borrowers can have the remainder of their loan balances forgiven after making 120 monthly payments while employed full time by public service employers, including some academic medical institutions. In our study, a large majority of respondents indicated that they are aware of the PSLF, and more than half said they would consider pursuing positions in academia if their loans could be forgiven through the program; however, when asked about plans for loan repayment, only 1 respondent endorsed current plans to enroll in PSLF. Thus, despite high interest in PSLF among the survey respondents, few had actual plans to use the service, suggesting that perhaps dermatologists are not provided enough information about PSLF to motivate enrollment. In the same way, almost a quarter of respondents were not familiar with the PSLF as a repayment option, further signifying that distribution of information about financial planning may be inadequate. If student loan burden is a notable factor in career decisions in dermatology, it is important that academic institutions provide sufficient information about repayment to encourage informed decisions. As such, it is possible that educating physicians about options such as PSLF can potentially recruit more dermatologists to academic positions.

Aside from financial reasons, residency program experience and differences in practices in academic and nonacademic settings may impact career trajectories. The majority of respondents stated their residency program experience influenced their career decisions; however, the majority of respondents did not change their minds about career goals since starting residency, suggesting that residency program experience may reinforce but not necessarily alter these choices. Interests in specific focuses within dermatology also may influence career decisions. This study suggests that those pursuing private practice positions are more interested in dermatologic surgery, lasers, and cosmetics. Roles in research, mentoring, and patient care differ between academic and nonacademic settings and also may be influential in career decisions in dermatology, mirroring the implications of other studies that posited that career decisions within medicine are complex and involve multiple factors.¹¹

In this study, we did not find an association between gender and career plans in dermatology. In 2013, more than 60% of dermatology resident physicians were female.¹² However, a recent study suggested that women face challenges in academic dermatology, including a downtrend in the number of female investigators with grants from the National Institutes of Health.¹³Taken together, female dermatologists, who currently represent the majority of resident physicians training in dermatology, may face unique challenges in academia, further highlighting the potential difficulties with recruitment of dermatologists by academic institutions in the future.

This preliminary study has several limitations. First, the small sample size limited generalizability to all dermatologists. Second, responder bias was possible, as those who have stronger opinions about this topic may have been more inclined to participate in this voluntary survey. Future studies with larger sample sizes are needed to further explore the factors that influence career decisions within dermatology and to determine if there are additional means to increase recruitment into academia.

Conclusion

It is recognized that there are challenges in recruiting dermatologists into academic positions. This study suggests that student loan burden influences career decisions in dermatology. Dermatologists may not be fully educated on options for student loan repayment. With increased awareness, the PSLF can potentially be used as a recruitment tool for positions in academic dermatology.

Access to outpatient specialty care is notably limited due to time and out-of-pocket costs to patients, leading to patient dissatisfaction and worsened clinical outcomes. Lost time and earnings pose considerable opportunity costs for patients, with the total opportunity cost for all physician visits per year estimated at $52 billion in 2010 in the United States.¹

The field of dermatology exemplifies the access issues patients may face when seeking specialty care given the ongoing national shortage of dermatologists and notably long wait times exceeding 60 days in major cities.^2-4 With the high demand and limited number of providers, patients may have longer wait times to see dermatologists in their communities or have to travel further to see dermatologists in distant locations who have available appointments; therefore, patients may be subject to higher associated time, travel, and monetary costs. According to the 2013 Medical Expenditure Panel Survey, dermatology visits in the United States cost an average of $221 per visit compared to $166 for primary care. Dermatology visits had the highest median cost per office visit ($124) and were more often associated with out-of-pocket expenses (60.7%) compared to other specialties.⁵ Despite these high costs, the number of dermatology visits is increasing each year, with more than 38 million dermatology visits in 2012.⁶

In light of these factors that limit patient access to dermatologists compared to other specialists, we performed an evaluation of the direct and indirect costs to patients visiting an outpatient dermatology clinic in Boston, Massachusetts, to better understand obstacles to receiving dermatologic care. The impact that time and money have on how patients prefer to receive their care also was evaluated. Conducting this study in Boston may best reflect patient barriers to obtaining dermatologic treatment, as nationwide surveys have found that Boston has the highest cumulative average wait times for physician appointments compared to other US metropolitan cities, with an average wait time of 72 days to see a dermatologist.⁴ New studies of patient costs associated with dermatology clinic visits are lacking, and existing economic analyses rarely include time costs. Understanding time burden and opportunity costs from the patient perspective may motivate patients and physicians to alter how they receive and provide health care, respectively, to minimize these expenses. Advances in health care technology such as telecommunication may facilitate these changes.

Methods

Study Design
This survey study took place from October 1, 2015, to March 4, 2016, at the department of dermatology outpatient clinic of Tufts Medical Center, an academic university hospital located in downtown Boston, Massachusetts, with no satellite clinics. Five general dermatologists, 2 dermatologic surgeons, and 9 dermatology residents comprised the dermatology department. The study protocol and questionnaire received exemption status from the Tufts University Health Science’s institutional review board.

All adult patients (aged ≥18 years) attending a scheduled dermatology clinic visit within the designated time frame were invited to complete a questionnaire available in English, Spanish, or Chinese. Patients completed the questionnaire on paper or electronically using handheld tablets. Data were then compiled into the REDCap (Research Electronic Data Capture) online database. The questionnaire surveyed patient age; gender; ethnicity; language spoken; highest level of education; employment status; reason for visit (ie, skin condition); duration, cost, and mode of transportation; duration of visit including wait time; companion accompaniment; profession; hourly wage; and number of work hours requested off to attend the visit. Lastly, patients were surveyed on whether they prefer to receive dermatologic care at Tufts, to receive in-person care elsewhere, to use teledermatology, or none of the above.

Statistical Analysis
Total time attributed to the visit was the sum of time for round-trip travel to and from the clinic, wait time, and face-to-face time with care providers. Out-of-pocket patient expenses included round-trip travel expenses, child care expenses, and direct payments such as deductibles and co-pays. Opportunity cost for employed patients was calculated as the patient’s average hourly wage multiplied by either the number of hours taken off from work or the number of hours the patient attributed to the visit, whichever value was higher at the individual level. For the purpose of calculating opportunity costs, travel time, wait time, and face-to-face time were imputed using average values for these variables when not reported. Patients could provide exact hourly wage and annual income or select the closest approximation from 10 wage ranges. For patients who selected a wage range, the midpoint of the range was used as the hourly wage. Total costs were the sum of reported out-of-pocket expenses and calculated opportunity costs. For unemployed patients and those who did not report employment status, hourly wage was assumed to be $0, resulting in opportunity costs of $0. Costs are tabulated for individual patients and analyzed in aggregate.

Differences in patient characteristics between those who preferred their current care provider versus those who preferred to seek care elsewhere or via teledermatology were compared using the χ² and Student t test. A multivariate logistic regression was then performed to identify predictors of patient preference for their current provider. Potential predictors for regression model were time and cost variables as well as factors selected based on results from bivariate analysis. Data analysis was performed using statistical software.

Results

Demographics
Demographic data for respondents are outlined in Table 1. Of 145 patients who completed the survey, the majority had already seen a dermatologist for their presenting condition (87.4%), and were English speaking (96.5%), white (76.6%), employed (59.4%), and male (50.3%), with a mean age (SD) of 52.3 (18.1) years and education level of 4-year university or higher (64.7%). The most common reasons for dermatology clinic attendance were general skin checks (30.8%) and psoriasis (26.6%). A smaller proportion of patients (16.1%) presented for surgical visits. Other less common conditions that brought patients into the clinic included acne (6.3%), eczema (4.9%), and skin rash (2.8%).

The mean (SD) reported hourly wage of employed patients was $36.60 (15.8). The most common reasons for unemployment were retirement (65.5% [38/58]), disability (10.3% [6/58]), and schooling (10.3% [6/58]).

Time Attributed to Attending Dermatology Clinic Visits
Time costs are reported in Table 2. Patients traveled to the clinic mainly by car (56.5% [78/138]) or train/subway (25.3% [35/138]). One in approximately 5 patients (21.3%) spent more than 1 hour traveling one-way to the clinic. Most patients waited less than 20 minutes to see their care providers. Face-to-face time with providers (ie, residents and attending physicians) ranged from less than 21 minutes to more than 1 hour, with a mean (SD) time of 36.8 (18.9) minutes.

Of the employed respondents, 76.5% (65/85) took off time from work for the appointment. Patients took a mean (SD) of 4.1 (2.4) hours off from work, which was considered sick pay (35%), paid time off (36.6%), or unpaid time (28.3%). The total mean (SD) time dedicated to attending the clinic appointment averaged 144.8 (60.47) minutes. On average, the time spent traveling for the clinic visit was double the amount of time spent with the care provider (77.4 vs 36.8 minutes).

Monetary and Opportunity Costs
The mean (SD) monetary cost associated with clinic attendance for employed patients who reported their wages was $187.50 (103.2)(range, $37.50–$489), most of which was opportunity cost from loss of potential work income (mean [SD], $144.30 [93.6]; range, $27–$432)(Table 3). Similar total and opportunity costs were found for employed patients using the imputed average wage. The mean (SD) total cost per visit for unemployed patients or those who did not report employment status was $38.65 (103.6)(range, $0–$800), which was 4-times less than the cost per visit for employed patients. Mean (SD) and median one-way travel expenses were $16.60 (40.5) and $10, respectively. Mean (SD) and median reported costs for deductibles/co-pays were $44.20 (66.1) and $25, respectively. Only 2 patients reported child care costs, which were valued at $65 and $75.

Patient Provider Preference
The majority (59.3% [67/113]) of patients preferred their current care providers, whereas 33.6% (38/113) preferred providers closer to work, home, or in a different unspecified setting. Only 7.0% (8/113) of patients who answered this survey question would choose teledermatology over their current providers.

On multivariate logistic regression (Table 4), patients who had additional out-of-pocket costs were significantly less likely to prefer their current care provider compared to patients with no out-of-pocket costs (odds ratio [OR], 0.27; 95% confidence interval [CI], 0.10-0.71; P<.05). Opportunity costs were not a significant predictor of provider preference. For every minute the travel time increased, the likelihood of preference for the current care provider decreased by 2% (OR, 0.98; 95% CI, 0.95–0.99), and patients who traveled 60 minutes or more round-trip were 71% less likely to choose current provider care than those who traveled less than 60 minutes (OR, 0.29; 95% CI, 0.09-0.96; P<.05). Patients with higher education (≥4 years of college) were 3.29-times more likely to stay with their current care provider than those with lower education (≤2 years of college). Those presenting for skin checks also preferred the current provider more than those with noninflammatory skin conditions such as alopecia and warts (OR, 9.01; 95% CI, 2.28-35.59). Age and gender were not statistically significant predictors of patient provider preference.

Comment

Our study revealed that patients spend a substantial amount of time and money attending dermatology clinic appointments. Round-trip travel time exceeded 2 hours for 20% of patients and accounted for the majority of the total time attributed to the visit. Patients who were employed typically requested an average of 4 hours off from work, resulting in a mean (SD) opportunity cost of $144.30 (93.6) due to lost wages. Direct costs such as co-pays, deductibles, travel expenses, and child care accounted for a smaller proportion of total costs. The study assumed a wage of $0 for unemployed patients, thus underestimating the true costs of the visit for these patients whose time may otherwise have been spent on leisure, education, volunteerism, or other activities that contribute to individual and societal productivity. The total costs for unemployed patients reflected only direct costs, and thus were notably lower than those for employed patients.

Direct out-of-pocket costs and travel time negatively impacted provider preference. Patients with out-of-pocket costs were much less likely to stay with their current care provider (OR, 0.27; 95% CI, 0.10-0.71), preferring to seek care closer to home/work or teledermatology services. Similarly, for each minute that travel time increased, preference for current care provider decreased by 2%. Those who traveled 60 minutes or more were 71% less likely than those who traveled less than 60 minutes to stay with their current provider when given other options for care. Opportunity costs did not affect provider preference, even though they far exceeded direct costs for employed patients. Perhaps opportunity costs are not as immediately apparent to patients as out-of-pocket costs and travel time, and thus they do not factor as heavily in provider preference.

Despite high time and monetary costs, the majority of patients (60%) still preferred their current care provider, especially those with 4-year university degrees or higher education level (OR, 3.29; 95% CI, 1.23-5.26) and those presenting for skin checks (OR, 9.01; 95% CI, 2.28-35.59). Patients with higher levels of education likely have higher incomes and thus may not be as adversely affected by direct and/or indirect visit costs. Patients presenting for skin checks may value continuity and prefer providers with whom they already have an established therapeutic relationship. Future studies are needed to analyze the impact of these nonmonetary factors on provider preference.

Seeking Alternative Care
Tufts Medical Center does not have satellite dermatology clinics, making it the only option for patients who wish to receive care within the Tufts hospital network. However, patients do have the option of visiting non–Tufts-affiliated dermatology clinics outside of the city. To our knowledge, no formal studies have been performed comparing wait times for dermatology appointments in suburban versus urban Boston areas; however, it has been reported that rural practitioners have longer wait times than urban dermatologists, possibly due to the fact that physicians tend to aggregate in metropolitan areas.² Thus, the potential for shorter wait times in the Boston metropolitan area may make it a more desirable location to receive care compared to more suburban or even rural areas of Massachusetts, but additional data are needed to substantiate this hypothesis. Additionally, health insurance restrictions, refractory or complex dermatologic conditions, and referring providers’ preference may affect patients’ decisions to seek care at a particular clinic. However, these factors do not alter our finding that those who travel long distances to our dermatology clinic are less likely to stay with their current provider if given the choice to seek care closer to home/work or utilize teledermatology services.

Prior studies have demonstrated patient preference and willingness to accept alternative modes of care delivery to reduce time and monetary costs associated with in-person medical visits.^7,8 Dermatology patients at a clinic in Ontario, Canada, considered the time they spent attending the clinic to be even more burdensome than the monetary cost.⁷ Patients with nondermatologic chronic diseases and high out-of-pocket costs would prefer email rather than a clinic visit as the first method of contact with care providers.⁸ The explosive growth of direct-to-consumer (DTC) teledermatology services in the last 10 years speaks to patient demand for alternative care delivery that saves time and money. Although telemedicine has been implemented in various specialties, including ophthalmology and neurology, one of the most common applications is teledermatology. With DTC teledermatology, patients can take photographs or videos using personal smartphones and communicate directly with care providers using mobile or online applications. More recent review articles have identified 22 to 29 DTC mobile and web-based teledermatology services, with costs varying from $0 to $250.^9-11 The median consultation fee of $59 for DTC teledermatology services is substantially less than total visit costs for employed patients in our study.⁹ Teledermatology has become an accessible and affordable modality of care, though perhaps not yet fully optimized for quality of care.

With increasingly higher co-pays and high-deductible insurance plans, time and monetary factors play increasingly important roles in patient preference for specialty care providers,¹² as demonstrated by our study. Dermatologists can work with patients to reduce the costs of medical visits. Perhaps monitoring of chronic but stable conditions can be accomplished through telecommunication to reduce the number of follow-up visits. For instance, psoriasis patients enrolled in telemonitoring perceived savings of time and expenses through reduction of clinic visits, resulting in high patient satisfaction levels.¹³ Telephone calls and secure email messaging are other feasible alternatives shown to aid in clinical management and decrease the need for in-person care.^8,14 Fewer unnecessary follow-up visits also means more availability for new patients and those with acute needs.

Barriers to obtaining care are not limited to dermatology and are pervasive across most medical specialties. Issues of patient time burden and out-of-pocket expenses are reflected in recent reports focused on quantifying these costs throughout ambulatory care visits and services such as colorectal, cervical, and breast cancer screenings.^1,15-18 Similar to our findings, many of these studies also show high time and opportunity costs from the patient perspective. Expansion of telemedicine to reduce patient costs is becoming a viable option for many specialists, though low reimbursement rates restrict its widespread application.^9,19 However, this obstacle is not impossible to surmount. One study found that offering teledermatology to Medicaid patients through their primary care providers significantly improved access, allowing for a 63.8% increase in the number of patients visiting a dermatologist (P<.01).²⁰ Currently, a total of 48 state Medicaid programs now cover telemedicine, and a growing number of states are requiring private insurers to cover telehealth services.²¹ As more dermatologists adopt telemedicine practices, it may allow for better access as well as expanded insurance coverage.

Limitations
The results of our study are limited by the single-institution survey design. Patients were asked to complete the survey while still at the clinic visit to minimize recall bias. Because these patients actually attended their appointments, they might perceive the time and monetary costs associated with the visit to be less problematic than those who canceled their appointments or transferred care elsewhere; however, we were still able to detect a significant impact of time and monetary costs on provider preference in this cohort (P<.05). Larger studies in different geographic settings and other specialty clinics are needed to confirm our findings and to determine if nonmonetary factors such as specific diagnoses, length of time with a certain care provider, or patient socioeconomic status can modulate the impact of time and monetary costs on provider preference.

Conclusion

This study showed that patients expend a substantial amount of time and monetary costs to attend dermatology clinic visits. Data from the current and prior studies suggest that these costs affect patient provider preference for dermatologic care and may pose barriers to necessary medical care. Recognizing direct and indirect patient costs may drive critical changes in health care delivery, such as increased telecommunication utilization, the more cost-saving alternative. Telemedicine, when integrated appropriately, can help minimize expenses for patients while continuing to maintain a high level of care.

Access to outpatient specialty care is notably limited due to time and out-of-pocket costs to patients, leading to patient dissatisfaction and worsened clinical outcomes. Lost time and earnings pose considerable opportunity costs for patients, with the total opportunity cost for all physician visits per year estimated at $52 billion in 2010 in the United States.¹

The field of dermatology exemplifies the access issues patients may face when seeking specialty care given the ongoing national shortage of dermatologists and notably long wait times exceeding 60 days in major cities.^2-4 With the high demand and limited number of providers, patients may have longer wait times to see dermatologists in their communities or have to travel further to see dermatologists in distant locations who have available appointments; therefore, patients may be subject to higher associated time, travel, and monetary costs. According to the 2013 Medical Expenditure Panel Survey, dermatology visits in the United States cost an average of $221 per visit compared to $166 for primary care. Dermatology visits had the highest median cost per office visit ($124) and were more often associated with out-of-pocket expenses (60.7%) compared to other specialties.⁵ Despite these high costs, the number of dermatology visits is increasing each year, with more than 38 million dermatology visits in 2012.⁶

In light of these factors that limit patient access to dermatologists compared to other specialists, we performed an evaluation of the direct and indirect costs to patients visiting an outpatient dermatology clinic in Boston, Massachusetts, to better understand obstacles to receiving dermatologic care. The impact that time and money have on how patients prefer to receive their care also was evaluated. Conducting this study in Boston may best reflect patient barriers to obtaining dermatologic treatment, as nationwide surveys have found that Boston has the highest cumulative average wait times for physician appointments compared to other US metropolitan cities, with an average wait time of 72 days to see a dermatologist.⁴ New studies of patient costs associated with dermatology clinic visits are lacking, and existing economic analyses rarely include time costs. Understanding time burden and opportunity costs from the patient perspective may motivate patients and physicians to alter how they receive and provide health care, respectively, to minimize these expenses. Advances in health care technology such as telecommunication may facilitate these changes.

Methods

Study Design
This survey study took place from October 1, 2015, to March 4, 2016, at the department of dermatology outpatient clinic of Tufts Medical Center, an academic university hospital located in downtown Boston, Massachusetts, with no satellite clinics. Five general dermatologists, 2 dermatologic surgeons, and 9 dermatology residents comprised the dermatology department. The study protocol and questionnaire received exemption status from the Tufts University Health Science’s institutional review board.

All adult patients (aged ≥18 years) attending a scheduled dermatology clinic visit within the designated time frame were invited to complete a questionnaire available in English, Spanish, or Chinese. Patients completed the questionnaire on paper or electronically using handheld tablets. Data were then compiled into the REDCap (Research Electronic Data Capture) online database. The questionnaire surveyed patient age; gender; ethnicity; language spoken; highest level of education; employment status; reason for visit (ie, skin condition); duration, cost, and mode of transportation; duration of visit including wait time; companion accompaniment; profession; hourly wage; and number of work hours requested off to attend the visit. Lastly, patients were surveyed on whether they prefer to receive dermatologic care at Tufts, to receive in-person care elsewhere, to use teledermatology, or none of the above.

Statistical Analysis
Total time attributed to the visit was the sum of time for round-trip travel to and from the clinic, wait time, and face-to-face time with care providers. Out-of-pocket patient expenses included round-trip travel expenses, child care expenses, and direct payments such as deductibles and co-pays. Opportunity cost for employed patients was calculated as the patient’s average hourly wage multiplied by either the number of hours taken off from work or the number of hours the patient attributed to the visit, whichever value was higher at the individual level. For the purpose of calculating opportunity costs, travel time, wait time, and face-to-face time were imputed using average values for these variables when not reported. Patients could provide exact hourly wage and annual income or select the closest approximation from 10 wage ranges. For patients who selected a wage range, the midpoint of the range was used as the hourly wage. Total costs were the sum of reported out-of-pocket expenses and calculated opportunity costs. For unemployed patients and those who did not report employment status, hourly wage was assumed to be $0, resulting in opportunity costs of $0. Costs are tabulated for individual patients and analyzed in aggregate.

Differences in patient characteristics between those who preferred their current care provider versus those who preferred to seek care elsewhere or via teledermatology were compared using the χ² and Student t test. A multivariate logistic regression was then performed to identify predictors of patient preference for their current provider. Potential predictors for regression model were time and cost variables as well as factors selected based on results from bivariate analysis. Data analysis was performed using statistical software.

Results

Demographics
Demographic data for respondents are outlined in Table 1. Of 145 patients who completed the survey, the majority had already seen a dermatologist for their presenting condition (87.4%), and were English speaking (96.5%), white (76.6%), employed (59.4%), and male (50.3%), with a mean age (SD) of 52.3 (18.1) years and education level of 4-year university or higher (64.7%). The most common reasons for dermatology clinic attendance were general skin checks (30.8%) and psoriasis (26.6%). A smaller proportion of patients (16.1%) presented for surgical visits. Other less common conditions that brought patients into the clinic included acne (6.3%), eczema (4.9%), and skin rash (2.8%).

The mean (SD) reported hourly wage of employed patients was $36.60 (15.8). The most common reasons for unemployment were retirement (65.5% [38/58]), disability (10.3% [6/58]), and schooling (10.3% [6/58]).

Time Attributed to Attending Dermatology Clinic Visits
Time costs are reported in Table 2. Patients traveled to the clinic mainly by car (56.5% [78/138]) or train/subway (25.3% [35/138]). One in approximately 5 patients (21.3%) spent more than 1 hour traveling one-way to the clinic. Most patients waited less than 20 minutes to see their care providers. Face-to-face time with providers (ie, residents and attending physicians) ranged from less than 21 minutes to more than 1 hour, with a mean (SD) time of 36.8 (18.9) minutes.

Of the employed respondents, 76.5% (65/85) took off time from work for the appointment. Patients took a mean (SD) of 4.1 (2.4) hours off from work, which was considered sick pay (35%), paid time off (36.6%), or unpaid time (28.3%). The total mean (SD) time dedicated to attending the clinic appointment averaged 144.8 (60.47) minutes. On average, the time spent traveling for the clinic visit was double the amount of time spent with the care provider (77.4 vs 36.8 minutes).

Monetary and Opportunity Costs
The mean (SD) monetary cost associated with clinic attendance for employed patients who reported their wages was $187.50 (103.2)(range, $37.50–$489), most of which was opportunity cost from loss of potential work income (mean [SD], $144.30 [93.6]; range, $27–$432)(Table 3). Similar total and opportunity costs were found for employed patients using the imputed average wage. The mean (SD) total cost per visit for unemployed patients or those who did not report employment status was $38.65 (103.6)(range, $0–$800), which was 4-times less than the cost per visit for employed patients. Mean (SD) and median one-way travel expenses were $16.60 (40.5) and $10, respectively. Mean (SD) and median reported costs for deductibles/co-pays were $44.20 (66.1) and $25, respectively. Only 2 patients reported child care costs, which were valued at $65 and $75.

Patient Provider Preference
The majority (59.3% [67/113]) of patients preferred their current care providers, whereas 33.6% (38/113) preferred providers closer to work, home, or in a different unspecified setting. Only 7.0% (8/113) of patients who answered this survey question would choose teledermatology over their current providers.

On multivariate logistic regression (Table 4), patients who had additional out-of-pocket costs were significantly less likely to prefer their current care provider compared to patients with no out-of-pocket costs (odds ratio [OR], 0.27; 95% confidence interval [CI], 0.10-0.71; P<.05). Opportunity costs were not a significant predictor of provider preference. For every minute the travel time increased, the likelihood of preference for the current care provider decreased by 2% (OR, 0.98; 95% CI, 0.95–0.99), and patients who traveled 60 minutes or more round-trip were 71% less likely to choose current provider care than those who traveled less than 60 minutes (OR, 0.29; 95% CI, 0.09-0.96; P<.05). Patients with higher education (≥4 years of college) were 3.29-times more likely to stay with their current care provider than those with lower education (≤2 years of college). Those presenting for skin checks also preferred the current provider more than those with noninflammatory skin conditions such as alopecia and warts (OR, 9.01; 95% CI, 2.28-35.59). Age and gender were not statistically significant predictors of patient provider preference.

Comment

Our study revealed that patients spend a substantial amount of time and money attending dermatology clinic appointments. Round-trip travel time exceeded 2 hours for 20% of patients and accounted for the majority of the total time attributed to the visit. Patients who were employed typically requested an average of 4 hours off from work, resulting in a mean (SD) opportunity cost of $144.30 (93.6) due to lost wages. Direct costs such as co-pays, deductibles, travel expenses, and child care accounted for a smaller proportion of total costs. The study assumed a wage of $0 for unemployed patients, thus underestimating the true costs of the visit for these patients whose time may otherwise have been spent on leisure, education, volunteerism, or other activities that contribute to individual and societal productivity. The total costs for unemployed patients reflected only direct costs, and thus were notably lower than those for employed patients.

Direct out-of-pocket costs and travel time negatively impacted provider preference. Patients with out-of-pocket costs were much less likely to stay with their current care provider (OR, 0.27; 95% CI, 0.10-0.71), preferring to seek care closer to home/work or teledermatology services. Similarly, for each minute that travel time increased, preference for current care provider decreased by 2%. Those who traveled 60 minutes or more were 71% less likely than those who traveled less than 60 minutes to stay with their current provider when given other options for care. Opportunity costs did not affect provider preference, even though they far exceeded direct costs for employed patients. Perhaps opportunity costs are not as immediately apparent to patients as out-of-pocket costs and travel time, and thus they do not factor as heavily in provider preference.

Despite high time and monetary costs, the majority of patients (60%) still preferred their current care provider, especially those with 4-year university degrees or higher education level (OR, 3.29; 95% CI, 1.23-5.26) and those presenting for skin checks (OR, 9.01; 95% CI, 2.28-35.59). Patients with higher levels of education likely have higher incomes and thus may not be as adversely affected by direct and/or indirect visit costs. Patients presenting for skin checks may value continuity and prefer providers with whom they already have an established therapeutic relationship. Future studies are needed to analyze the impact of these nonmonetary factors on provider preference.

Seeking Alternative Care
Tufts Medical Center does not have satellite dermatology clinics, making it the only option for patients who wish to receive care within the Tufts hospital network. However, patients do have the option of visiting non–Tufts-affiliated dermatology clinics outside of the city. To our knowledge, no formal studies have been performed comparing wait times for dermatology appointments in suburban versus urban Boston areas; however, it has been reported that rural practitioners have longer wait times than urban dermatologists, possibly due to the fact that physicians tend to aggregate in metropolitan areas.² Thus, the potential for shorter wait times in the Boston metropolitan area may make it a more desirable location to receive care compared to more suburban or even rural areas of Massachusetts, but additional data are needed to substantiate this hypothesis. Additionally, health insurance restrictions, refractory or complex dermatologic conditions, and referring providers’ preference may affect patients’ decisions to seek care at a particular clinic. However, these factors do not alter our finding that those who travel long distances to our dermatology clinic are less likely to stay with their current provider if given the choice to seek care closer to home/work or utilize teledermatology services.

Prior studies have demonstrated patient preference and willingness to accept alternative modes of care delivery to reduce time and monetary costs associated with in-person medical visits.^7,8 Dermatology patients at a clinic in Ontario, Canada, considered the time they spent attending the clinic to be even more burdensome than the monetary cost.⁷ Patients with nondermatologic chronic diseases and high out-of-pocket costs would prefer email rather than a clinic visit as the first method of contact with care providers.⁸ The explosive growth of direct-to-consumer (DTC) teledermatology services in the last 10 years speaks to patient demand for alternative care delivery that saves time and money. Although telemedicine has been implemented in various specialties, including ophthalmology and neurology, one of the most common applications is teledermatology. With DTC teledermatology, patients can take photographs or videos using personal smartphones and communicate directly with care providers using mobile or online applications. More recent review articles have identified 22 to 29 DTC mobile and web-based teledermatology services, with costs varying from $0 to $250.^9-11 The median consultation fee of $59 for DTC teledermatology services is substantially less than total visit costs for employed patients in our study.⁹ Teledermatology has become an accessible and affordable modality of care, though perhaps not yet fully optimized for quality of care.

With increasingly higher co-pays and high-deductible insurance plans, time and monetary factors play increasingly important roles in patient preference for specialty care providers,¹² as demonstrated by our study. Dermatologists can work with patients to reduce the costs of medical visits. Perhaps monitoring of chronic but stable conditions can be accomplished through telecommunication to reduce the number of follow-up visits. For instance, psoriasis patients enrolled in telemonitoring perceived savings of time and expenses through reduction of clinic visits, resulting in high patient satisfaction levels.¹³ Telephone calls and secure email messaging are other feasible alternatives shown to aid in clinical management and decrease the need for in-person care.^8,14 Fewer unnecessary follow-up visits also means more availability for new patients and those with acute needs.

Barriers to obtaining care are not limited to dermatology and are pervasive across most medical specialties. Issues of patient time burden and out-of-pocket expenses are reflected in recent reports focused on quantifying these costs throughout ambulatory care visits and services such as colorectal, cervical, and breast cancer screenings.^1,15-18 Similar to our findings, many of these studies also show high time and opportunity costs from the patient perspective. Expansion of telemedicine to reduce patient costs is becoming a viable option for many specialists, though low reimbursement rates restrict its widespread application.^9,19 However, this obstacle is not impossible to surmount. One study found that offering teledermatology to Medicaid patients through their primary care providers significantly improved access, allowing for a 63.8% increase in the number of patients visiting a dermatologist (P<.01).²⁰ Currently, a total of 48 state Medicaid programs now cover telemedicine, and a growing number of states are requiring private insurers to cover telehealth services.²¹ As more dermatologists adopt telemedicine practices, it may allow for better access as well as expanded insurance coverage.

Limitations
The results of our study are limited by the single-institution survey design. Patients were asked to complete the survey while still at the clinic visit to minimize recall bias. Because these patients actually attended their appointments, they might perceive the time and monetary costs associated with the visit to be less problematic than those who canceled their appointments or transferred care elsewhere; however, we were still able to detect a significant impact of time and monetary costs on provider preference in this cohort (P<.05). Larger studies in different geographic settings and other specialty clinics are needed to confirm our findings and to determine if nonmonetary factors such as specific diagnoses, length of time with a certain care provider, or patient socioeconomic status can modulate the impact of time and monetary costs on provider preference.

Conclusion

This study showed that patients expend a substantial amount of time and monetary costs to attend dermatology clinic visits. Data from the current and prior studies suggest that these costs affect patient provider preference for dermatologic care and may pose barriers to necessary medical care. Recognizing direct and indirect patient costs may drive critical changes in health care delivery, such as increased telecommunication utilization, the more cost-saving alternative. Telemedicine, when integrated appropriately, can help minimize expenses for patients while continuing to maintain a high level of care.

Access to outpatient specialty care is notably limited due to time and out-of-pocket costs to patients, leading to patient dissatisfaction and worsened clinical outcomes. Lost time and earnings pose considerable opportunity costs for patients, with the total opportunity cost for all physician visits per year estimated at $52 billion in 2010 in the United States.¹

The field of dermatology exemplifies the access issues patients may face when seeking specialty care given the ongoing national shortage of dermatologists and notably long wait times exceeding 60 days in major cities.^2-4 With the high demand and limited number of providers, patients may have longer wait times to see dermatologists in their communities or have to travel further to see dermatologists in distant locations who have available appointments; therefore, patients may be subject to higher associated time, travel, and monetary costs. According to the 2013 Medical Expenditure Panel Survey, dermatology visits in the United States cost an average of $221 per visit compared to $166 for primary care. Dermatology visits had the highest median cost per office visit ($124) and were more often associated with out-of-pocket expenses (60.7%) compared to other specialties.⁵ Despite these high costs, the number of dermatology visits is increasing each year, with more than 38 million dermatology visits in 2012.⁶

In light of these factors that limit patient access to dermatologists compared to other specialists, we performed an evaluation of the direct and indirect costs to patients visiting an outpatient dermatology clinic in Boston, Massachusetts, to better understand obstacles to receiving dermatologic care. The impact that time and money have on how patients prefer to receive their care also was evaluated. Conducting this study in Boston may best reflect patient barriers to obtaining dermatologic treatment, as nationwide surveys have found that Boston has the highest cumulative average wait times for physician appointments compared to other US metropolitan cities, with an average wait time of 72 days to see a dermatologist.⁴ New studies of patient costs associated with dermatology clinic visits are lacking, and existing economic analyses rarely include time costs. Understanding time burden and opportunity costs from the patient perspective may motivate patients and physicians to alter how they receive and provide health care, respectively, to minimize these expenses. Advances in health care technology such as telecommunication may facilitate these changes.

Methods

Study Design
This survey study took place from October 1, 2015, to March 4, 2016, at the department of dermatology outpatient clinic of Tufts Medical Center, an academic university hospital located in downtown Boston, Massachusetts, with no satellite clinics. Five general dermatologists, 2 dermatologic surgeons, and 9 dermatology residents comprised the dermatology department. The study protocol and questionnaire received exemption status from the Tufts University Health Science’s institutional review board.

All adult patients (aged ≥18 years) attending a scheduled dermatology clinic visit within the designated time frame were invited to complete a questionnaire available in English, Spanish, or Chinese. Patients completed the questionnaire on paper or electronically using handheld tablets. Data were then compiled into the REDCap (Research Electronic Data Capture) online database. The questionnaire surveyed patient age; gender; ethnicity; language spoken; highest level of education; employment status; reason for visit (ie, skin condition); duration, cost, and mode of transportation; duration of visit including wait time; companion accompaniment; profession; hourly wage; and number of work hours requested off to attend the visit. Lastly, patients were surveyed on whether they prefer to receive dermatologic care at Tufts, to receive in-person care elsewhere, to use teledermatology, or none of the above.

Statistical Analysis
Total time attributed to the visit was the sum of time for round-trip travel to and from the clinic, wait time, and face-to-face time with care providers. Out-of-pocket patient expenses included round-trip travel expenses, child care expenses, and direct payments such as deductibles and co-pays. Opportunity cost for employed patients was calculated as the patient’s average hourly wage multiplied by either the number of hours taken off from work or the number of hours the patient attributed to the visit, whichever value was higher at the individual level. For the purpose of calculating opportunity costs, travel time, wait time, and face-to-face time were imputed using average values for these variables when not reported. Patients could provide exact hourly wage and annual income or select the closest approximation from 10 wage ranges. For patients who selected a wage range, the midpoint of the range was used as the hourly wage. Total costs were the sum of reported out-of-pocket expenses and calculated opportunity costs. For unemployed patients and those who did not report employment status, hourly wage was assumed to be $0, resulting in opportunity costs of $0. Costs are tabulated for individual patients and analyzed in aggregate.

Differences in patient characteristics between those who preferred their current care provider versus those who preferred to seek care elsewhere or via teledermatology were compared using the χ² and Student t test. A multivariate logistic regression was then performed to identify predictors of patient preference for their current provider. Potential predictors for regression model were time and cost variables as well as factors selected based on results from bivariate analysis. Data analysis was performed using statistical software.

Results

Demographics
Demographic data for respondents are outlined in Table 1. Of 145 patients who completed the survey, the majority had already seen a dermatologist for their presenting condition (87.4%), and were English speaking (96.5%), white (76.6%), employed (59.4%), and male (50.3%), with a mean age (SD) of 52.3 (18.1) years and education level of 4-year university or higher (64.7%). The most common reasons for dermatology clinic attendance were general skin checks (30.8%) and psoriasis (26.6%). A smaller proportion of patients (16.1%) presented for surgical visits. Other less common conditions that brought patients into the clinic included acne (6.3%), eczema (4.9%), and skin rash (2.8%).

The mean (SD) reported hourly wage of employed patients was $36.60 (15.8). The most common reasons for unemployment were retirement (65.5% [38/58]), disability (10.3% [6/58]), and schooling (10.3% [6/58]).

Time Attributed to Attending Dermatology Clinic Visits
Time costs are reported in Table 2. Patients traveled to the clinic mainly by car (56.5% [78/138]) or train/subway (25.3% [35/138]). One in approximately 5 patients (21.3%) spent more than 1 hour traveling one-way to the clinic. Most patients waited less than 20 minutes to see their care providers. Face-to-face time with providers (ie, residents and attending physicians) ranged from less than 21 minutes to more than 1 hour, with a mean (SD) time of 36.8 (18.9) minutes.

Of the employed respondents, 76.5% (65/85) took off time from work for the appointment. Patients took a mean (SD) of 4.1 (2.4) hours off from work, which was considered sick pay (35%), paid time off (36.6%), or unpaid time (28.3%). The total mean (SD) time dedicated to attending the clinic appointment averaged 144.8 (60.47) minutes. On average, the time spent traveling for the clinic visit was double the amount of time spent with the care provider (77.4 vs 36.8 minutes).

Monetary and Opportunity Costs
The mean (SD) monetary cost associated with clinic attendance for employed patients who reported their wages was $187.50 (103.2)(range, $37.50–$489), most of which was opportunity cost from loss of potential work income (mean [SD], $144.30 [93.6]; range, $27–$432)(Table 3). Similar total and opportunity costs were found for employed patients using the imputed average wage. The mean (SD) total cost per visit for unemployed patients or those who did not report employment status was $38.65 (103.6)(range, $0–$800), which was 4-times less than the cost per visit for employed patients. Mean (SD) and median one-way travel expenses were $16.60 (40.5) and $10, respectively. Mean (SD) and median reported costs for deductibles/co-pays were $44.20 (66.1) and $25, respectively. Only 2 patients reported child care costs, which were valued at $65 and $75.

Patient Provider Preference
The majority (59.3% [67/113]) of patients preferred their current care providers, whereas 33.6% (38/113) preferred providers closer to work, home, or in a different unspecified setting. Only 7.0% (8/113) of patients who answered this survey question would choose teledermatology over their current providers.

On multivariate logistic regression (Table 4), patients who had additional out-of-pocket costs were significantly less likely to prefer their current care provider compared to patients with no out-of-pocket costs (odds ratio [OR], 0.27; 95% confidence interval [CI], 0.10-0.71; P<.05). Opportunity costs were not a significant predictor of provider preference. For every minute the travel time increased, the likelihood of preference for the current care provider decreased by 2% (OR, 0.98; 95% CI, 0.95–0.99), and patients who traveled 60 minutes or more round-trip were 71% less likely to choose current provider care than those who traveled less than 60 minutes (OR, 0.29; 95% CI, 0.09-0.96; P<.05). Patients with higher education (≥4 years of college) were 3.29-times more likely to stay with their current care provider than those with lower education (≤2 years of college). Those presenting for skin checks also preferred the current provider more than those with noninflammatory skin conditions such as alopecia and warts (OR, 9.01; 95% CI, 2.28-35.59). Age and gender were not statistically significant predictors of patient provider preference.

Comment

Our study revealed that patients spend a substantial amount of time and money attending dermatology clinic appointments. Round-trip travel time exceeded 2 hours for 20% of patients and accounted for the majority of the total time attributed to the visit. Patients who were employed typically requested an average of 4 hours off from work, resulting in a mean (SD) opportunity cost of $144.30 (93.6) due to lost wages. Direct costs such as co-pays, deductibles, travel expenses, and child care accounted for a smaller proportion of total costs. The study assumed a wage of $0 for unemployed patients, thus underestimating the true costs of the visit for these patients whose time may otherwise have been spent on leisure, education, volunteerism, or other activities that contribute to individual and societal productivity. The total costs for unemployed patients reflected only direct costs, and thus were notably lower than those for employed patients.

Direct out-of-pocket costs and travel time negatively impacted provider preference. Patients with out-of-pocket costs were much less likely to stay with their current care provider (OR, 0.27; 95% CI, 0.10-0.71), preferring to seek care closer to home/work or teledermatology services. Similarly, for each minute that travel time increased, preference for current care provider decreased by 2%. Those who traveled 60 minutes or more were 71% less likely than those who traveled less than 60 minutes to stay with their current provider when given other options for care. Opportunity costs did not affect provider preference, even though they far exceeded direct costs for employed patients. Perhaps opportunity costs are not as immediately apparent to patients as out-of-pocket costs and travel time, and thus they do not factor as heavily in provider preference.

Despite high time and monetary costs, the majority of patients (60%) still preferred their current care provider, especially those with 4-year university degrees or higher education level (OR, 3.29; 95% CI, 1.23-5.26) and those presenting for skin checks (OR, 9.01; 95% CI, 2.28-35.59). Patients with higher levels of education likely have higher incomes and thus may not be as adversely affected by direct and/or indirect visit costs. Patients presenting for skin checks may value continuity and prefer providers with whom they already have an established therapeutic relationship. Future studies are needed to analyze the impact of these nonmonetary factors on provider preference.

Seeking Alternative Care
Tufts Medical Center does not have satellite dermatology clinics, making it the only option for patients who wish to receive care within the Tufts hospital network. However, patients do have the option of visiting non–Tufts-affiliated dermatology clinics outside of the city. To our knowledge, no formal studies have been performed comparing wait times for dermatology appointments in suburban versus urban Boston areas; however, it has been reported that rural practitioners have longer wait times than urban dermatologists, possibly due to the fact that physicians tend to aggregate in metropolitan areas.² Thus, the potential for shorter wait times in the Boston metropolitan area may make it a more desirable location to receive care compared to more suburban or even rural areas of Massachusetts, but additional data are needed to substantiate this hypothesis. Additionally, health insurance restrictions, refractory or complex dermatologic conditions, and referring providers’ preference may affect patients’ decisions to seek care at a particular clinic. However, these factors do not alter our finding that those who travel long distances to our dermatology clinic are less likely to stay with their current provider if given the choice to seek care closer to home/work or utilize teledermatology services.

Prior studies have demonstrated patient preference and willingness to accept alternative modes of care delivery to reduce time and monetary costs associated with in-person medical visits.^7,8 Dermatology patients at a clinic in Ontario, Canada, considered the time they spent attending the clinic to be even more burdensome than the monetary cost.⁷ Patients with nondermatologic chronic diseases and high out-of-pocket costs would prefer email rather than a clinic visit as the first method of contact with care providers.⁸ The explosive growth of direct-to-consumer (DTC) teledermatology services in the last 10 years speaks to patient demand for alternative care delivery that saves time and money. Although telemedicine has been implemented in various specialties, including ophthalmology and neurology, one of the most common applications is teledermatology. With DTC teledermatology, patients can take photographs or videos using personal smartphones and communicate directly with care providers using mobile or online applications. More recent review articles have identified 22 to 29 DTC mobile and web-based teledermatology services, with costs varying from $0 to $250.^9-11 The median consultation fee of $59 for DTC teledermatology services is substantially less than total visit costs for employed patients in our study.⁹ Teledermatology has become an accessible and affordable modality of care, though perhaps not yet fully optimized for quality of care.

With increasingly higher co-pays and high-deductible insurance plans, time and monetary factors play increasingly important roles in patient preference for specialty care providers,¹² as demonstrated by our study. Dermatologists can work with patients to reduce the costs of medical visits. Perhaps monitoring of chronic but stable conditions can be accomplished through telecommunication to reduce the number of follow-up visits. For instance, psoriasis patients enrolled in telemonitoring perceived savings of time and expenses through reduction of clinic visits, resulting in high patient satisfaction levels.¹³ Telephone calls and secure email messaging are other feasible alternatives shown to aid in clinical management and decrease the need for in-person care.^8,14 Fewer unnecessary follow-up visits also means more availability for new patients and those with acute needs.

Barriers to obtaining care are not limited to dermatology and are pervasive across most medical specialties. Issues of patient time burden and out-of-pocket expenses are reflected in recent reports focused on quantifying these costs throughout ambulatory care visits and services such as colorectal, cervical, and breast cancer screenings.^1,15-18 Similar to our findings, many of these studies also show high time and opportunity costs from the patient perspective. Expansion of telemedicine to reduce patient costs is becoming a viable option for many specialists, though low reimbursement rates restrict its widespread application.^9,19 However, this obstacle is not impossible to surmount. One study found that offering teledermatology to Medicaid patients through their primary care providers significantly improved access, allowing for a 63.8% increase in the number of patients visiting a dermatologist (P<.01).²⁰ Currently, a total of 48 state Medicaid programs now cover telemedicine, and a growing number of states are requiring private insurers to cover telehealth services.²¹ As more dermatologists adopt telemedicine practices, it may allow for better access as well as expanded insurance coverage.

Limitations
The results of our study are limited by the single-institution survey design. Patients were asked to complete the survey while still at the clinic visit to minimize recall bias. Because these patients actually attended their appointments, they might perceive the time and monetary costs associated with the visit to be less problematic than those who canceled their appointments or transferred care elsewhere; however, we were still able to detect a significant impact of time and monetary costs on provider preference in this cohort (P<.05). Larger studies in different geographic settings and other specialty clinics are needed to confirm our findings and to determine if nonmonetary factors such as specific diagnoses, length of time with a certain care provider, or patient socioeconomic status can modulate the impact of time and monetary costs on provider preference.

Conclusion

This study showed that patients expend a substantial amount of time and monetary costs to attend dermatology clinic visits. Data from the current and prior studies suggest that these costs affect patient provider preference for dermatologic care and may pose barriers to necessary medical care. Recognizing direct and indirect patient costs may drive critical changes in health care delivery, such as increased telecommunication utilization, the more cost-saving alternative. Telemedicine, when integrated appropriately, can help minimize expenses for patients while continuing to maintain a high level of care.

Academic promotion requires evidence of scholastic production. The number of publications by a scientist is the most frequently reported metric of scholastic production, but it does not account for the impact of publications. The h-index is a bibliometric measure that combines both volume and impact of scientific contributions. The physicist Jorge E. Hirsch introduced this metric in 2005.¹ He defined it as the number of publications (h) by an author that have been cited at least h times. For example, a scientist with 30 publications including 12 that have been cited at least 12 times each has an h-index of 12. h-Index is a superior predictor of future scientific achievement in physics compared with total citation count, total publication count, and citations per publication. Hirsch² proposed h-index thresholds of 12 and 18 for advancement to associate professor and full professor in physics, respectively.²

h-Index values are not comparable across academic disciplines because they are influenced by the number of journals and authors within the field. Scientists in disciplines with numerous scholars and publications will have higher h-indices. For example, the mean h-index for full professors of cardiothoracic anesthesiology is 12, but the mean h-index for full professors of urology is 22.^3,4 Hence, h-index thresholds for professional advancement cannot be generalized but must be calculated on a granular, specialty-specific basis.

In a prior study on h-index among academic dermatologists in the United States, John et al⁵ reported that fellowship-trained dermatologists had a significantly higher mean h-index than those without fellowship training (13.2 vs 11.7; P<.001). They further found the mean h-index increased with academic rank.⁵

In our study, we measured mean and median h-indices among associate and full professors of dermatology in academic training programs in the United States with the goal of describing h-index distributions in these 2 academic ranks. We further sought to measure regional differences in h-index between northeastern, southern, central, and western states as defined by the National Resident Matching Program.

Methods

Institutional review board approval was deferred because the study did not require patient information or participation. Using the Association of American Medical Colleges Electronic Residency Application Service website (https://www.aamc.org/services/eras/) we identified dermatology residency training programs accredited by the Accreditation Council for Graduate Medical Education and participating in the Electronic Residency Application Service for the National Resident Matching Program in the United States. We visited the official website of each residency program and identified all associate and full professors of dermatology for further study. We included all faculty members listed as professor, clinical professor, associate professor, or clinical associate professor, and excluded assistant professor, volunteer faculty, research professor, and research associate professor. All faculty held an MD degree or an equivalent degree, such as MBBS or MDCM.

We used the Thomson Reuters (now Clarivate Analytics) Web of Science to calculate h-index and publication counts. The initial search was basic using the professor’s last name and first initial. We then augmented this list by searching for all variations of each professor’s name, with or without middle initial. Each publication in the search results was confirmed as belonging to the author of interest by verifying coauthors, institution information, and subject material. For authors with common names, we additionally consulted their online university profiles for specific names used in their “Selected Publications” lists. In a minority of cases, we also limited Research Domain to “dermatology.” Referring to the verified publication list for each dermatology professor, we used the Web of Science Citation Report function to determine number of publications and h-index for the individual. We tabulated results for associate and full professors and subgrouped those results into 4 geographic regions—northeastern, southern, central, and western states—according to the map used by the National Resident Matching Program. Descriptive statistics were performed with Microsoft Excel.

Results

We identified 300 associate professors and 352 full professors from 81 academic institutions. The number of associate professors per institution ranged from 1 to 25; the number of full professors per institution ranged from 1 to 16. The median and mean h-indices for associate and full professors, including interquartile values, are shown in the Table. There was a broad range of h-index scores among both academic ranks; median and mean h-indices varied more than 5-fold between the bottom and upper quartiles in both associate and full professor cohorts. Median interquartile h-index values for upper-quartile associate professors overlapped with those of lower-quartile full professors (Figure 1). h-Index for associate and full professors was similar across the 4 regions defined by the National Resident Matching Program. Median h-index was highest for full professors in western states and lowest for associate professors in southern states (Figure 2).

Comment

Professional advancement in academic medicine requires scholastic production. The h-index, defined as the number of publications (h) that have been cited at least h times, is a bibliometric measure that accounts for both volume and impact of an individual’s scientific productivity. The h-index would be a useful tool for determining professional advancement in academic dermatology departments. In this project, we calculated h-index values for 300 associate professors and 352 full professors of dermatology in the United States. We found the median h-index for associate professors was 8 and the median h-index for full professors was 21. There was more than a 5-fold variation in median and mean h-indices between lower and upper quartiles within both the associate and full professor cohorts. The highest median and mean h-indices were found among full professors of dermatology in western states. These results provide the opportunity for academic dermatologists and institutions to compare their research contributions with peers across the United States.

Our results support those of John et al⁵ who also found academic rank in dermatology was correlated with h-index. Scopus, Web of Science, and Google Scholar can be used to calculate h-index, but they may return different scores for the same individual.⁶ John et al⁵ used the Scopus database to calculate h-index. We used Web of Science because Scopus only includes citations since 1996 and Web of Science was used in the original h-index studies by Hirsch.^1,2 Institutions that adopt h-index criteria for advancement and resource distribution decisions should be aware that database selection can affect h-index scores.

Caveats With the h-Index
Flaws in the h-index include inflationary effects of self-citation, time bias, and excessive coauthorship. Individuals can increase their h-index by routinely citing their own publications. However, Engqvist and Frommen⁷ found tripling self-citations increased the h-index by only 1.

Citations tend to increase with time, and authors who have been active for longer periods will have a higher h-index. It is more difficult for junior faculty to distinguish themselves with the h-index, as it takes time for even the most impactful publications to gain citations. Major scientific papers can take years from conception to publication, and an outstanding paper that is 1 year old would have fewer citations than an equally impactful paper that is 10 years old. To adjust for the effect of time bias, Hirsch² proposed the m-index, in which the h-index is divided by the years between the author’s first and last publication. He proposed that an m-index of 1 would indicate a successful scientist, 2 an outstanding scientist, and 3 a unique individual.²

The literature is increasingly dominated by teams of coauthors, and the number of coauthors within each team has increased over the last 5 decades.⁸ h-Indices will increase if this trend continues, making it difficult to compare h-indices between different eras. Prosperi et al⁹ found national differences in kinship-based coauthorship, suggesting nepotism may influence decisions in assigning authorship status. h-Index valuations do not require evidence of meaningful contribution to the work but simply rely on contributors’ self-governance in assigning authorship status.

The h-index also has a bias against highly cited papers. A scientist with a small number of highly influential papers may have a smaller h-index than a scientist with more papers of modest impact. Finally, an author who has changed names (eg, due to marriage) may have an artificially low h-index, as a standard database search would miss publications under a maiden name.

Limitations
This study is limited by possible operator error when compiling each author’s publication list through Web of Science. Our search and refinement methodology took into account that authors may publish with slight variations in name, in various subject areas and fields, and with different institutions and coauthors. Each publication populated through Web of Science was carefully verified by the principal investigator; however, overestimation or underestimation of the number of publications and citations was possible, as the publication lists were not verified by the studied associate and full professors themselves. Our results are consistent with the h-index bar charts published by John et al⁵ using an alternate citation index, Scopus, which tends to corroborate our findings. This study also is limited by possible time bias because we did not correct the h-index for years of active publication (m-index).

Conclusion

In summary, we found the median h-index for associate professors was 8 and the median h-index for full professors was 21. We found a broad range of h-index values within each academic rank. h-Index for upper-quartile associate professors overlapped with those of lower-quartile full professors. Our results suggest professional advancement occurs over a broad range of scholastic production. Adopting requirements for minimum h-index thresholds for application for promotion might reduce disparities between rank and scientific contributions. We encourage use of the h-index for tracking academic progression and as a parameter to consider in academic promotion.

Academic promotion requires evidence of scholastic production. The number of publications by a scientist is the most frequently reported metric of scholastic production, but it does not account for the impact of publications. The h-index is a bibliometric measure that combines both volume and impact of scientific contributions. The physicist Jorge E. Hirsch introduced this metric in 2005.¹ He defined it as the number of publications (h) by an author that have been cited at least h times. For example, a scientist with 30 publications including 12 that have been cited at least 12 times each has an h-index of 12. h-Index is a superior predictor of future scientific achievement in physics compared with total citation count, total publication count, and citations per publication. Hirsch² proposed h-index thresholds of 12 and 18 for advancement to associate professor and full professor in physics, respectively.²

h-Index values are not comparable across academic disciplines because they are influenced by the number of journals and authors within the field. Scientists in disciplines with numerous scholars and publications will have higher h-indices. For example, the mean h-index for full professors of cardiothoracic anesthesiology is 12, but the mean h-index for full professors of urology is 22.^3,4 Hence, h-index thresholds for professional advancement cannot be generalized but must be calculated on a granular, specialty-specific basis.

In a prior study on h-index among academic dermatologists in the United States, John et al⁵ reported that fellowship-trained dermatologists had a significantly higher mean h-index than those without fellowship training (13.2 vs 11.7; P<.001). They further found the mean h-index increased with academic rank.⁵

In our study, we measured mean and median h-indices among associate and full professors of dermatology in academic training programs in the United States with the goal of describing h-index distributions in these 2 academic ranks. We further sought to measure regional differences in h-index between northeastern, southern, central, and western states as defined by the National Resident Matching Program.

Methods

Institutional review board approval was deferred because the study did not require patient information or participation. Using the Association of American Medical Colleges Electronic Residency Application Service website (https://www.aamc.org/services/eras/) we identified dermatology residency training programs accredited by the Accreditation Council for Graduate Medical Education and participating in the Electronic Residency Application Service for the National Resident Matching Program in the United States. We visited the official website of each residency program and identified all associate and full professors of dermatology for further study. We included all faculty members listed as professor, clinical professor, associate professor, or clinical associate professor, and excluded assistant professor, volunteer faculty, research professor, and research associate professor. All faculty held an MD degree or an equivalent degree, such as MBBS or MDCM.

We used the Thomson Reuters (now Clarivate Analytics) Web of Science to calculate h-index and publication counts. The initial search was basic using the professor’s last name and first initial. We then augmented this list by searching for all variations of each professor’s name, with or without middle initial. Each publication in the search results was confirmed as belonging to the author of interest by verifying coauthors, institution information, and subject material. For authors with common names, we additionally consulted their online university profiles for specific names used in their “Selected Publications” lists. In a minority of cases, we also limited Research Domain to “dermatology.” Referring to the verified publication list for each dermatology professor, we used the Web of Science Citation Report function to determine number of publications and h-index for the individual. We tabulated results for associate and full professors and subgrouped those results into 4 geographic regions—northeastern, southern, central, and western states—according to the map used by the National Resident Matching Program. Descriptive statistics were performed with Microsoft Excel.

Results

We identified 300 associate professors and 352 full professors from 81 academic institutions. The number of associate professors per institution ranged from 1 to 25; the number of full professors per institution ranged from 1 to 16. The median and mean h-indices for associate and full professors, including interquartile values, are shown in the Table. There was a broad range of h-index scores among both academic ranks; median and mean h-indices varied more than 5-fold between the bottom and upper quartiles in both associate and full professor cohorts. Median interquartile h-index values for upper-quartile associate professors overlapped with those of lower-quartile full professors (Figure 1). h-Index for associate and full professors was similar across the 4 regions defined by the National Resident Matching Program. Median h-index was highest for full professors in western states and lowest for associate professors in southern states (Figure 2).

Comment

Professional advancement in academic medicine requires scholastic production. The h-index, defined as the number of publications (h) that have been cited at least h times, is a bibliometric measure that accounts for both volume and impact of an individual’s scientific productivity. The h-index would be a useful tool for determining professional advancement in academic dermatology departments. In this project, we calculated h-index values for 300 associate professors and 352 full professors of dermatology in the United States. We found the median h-index for associate professors was 8 and the median h-index for full professors was 21. There was more than a 5-fold variation in median and mean h-indices between lower and upper quartiles within both the associate and full professor cohorts. The highest median and mean h-indices were found among full professors of dermatology in western states. These results provide the opportunity for academic dermatologists and institutions to compare their research contributions with peers across the United States.

Our results support those of John et al⁵ who also found academic rank in dermatology was correlated with h-index. Scopus, Web of Science, and Google Scholar can be used to calculate h-index, but they may return different scores for the same individual.⁶ John et al⁵ used the Scopus database to calculate h-index. We used Web of Science because Scopus only includes citations since 1996 and Web of Science was used in the original h-index studies by Hirsch.^1,2 Institutions that adopt h-index criteria for advancement and resource distribution decisions should be aware that database selection can affect h-index scores.

Caveats With the h-Index
Flaws in the h-index include inflationary effects of self-citation, time bias, and excessive coauthorship. Individuals can increase their h-index by routinely citing their own publications. However, Engqvist and Frommen⁷ found tripling self-citations increased the h-index by only 1.

Citations tend to increase with time, and authors who have been active for longer periods will have a higher h-index. It is more difficult for junior faculty to distinguish themselves with the h-index, as it takes time for even the most impactful publications to gain citations. Major scientific papers can take years from conception to publication, and an outstanding paper that is 1 year old would have fewer citations than an equally impactful paper that is 10 years old. To adjust for the effect of time bias, Hirsch² proposed the m-index, in which the h-index is divided by the years between the author’s first and last publication. He proposed that an m-index of 1 would indicate a successful scientist, 2 an outstanding scientist, and 3 a unique individual.²

The literature is increasingly dominated by teams of coauthors, and the number of coauthors within each team has increased over the last 5 decades.⁸ h-Indices will increase if this trend continues, making it difficult to compare h-indices between different eras. Prosperi et al⁹ found national differences in kinship-based coauthorship, suggesting nepotism may influence decisions in assigning authorship status. h-Index valuations do not require evidence of meaningful contribution to the work but simply rely on contributors’ self-governance in assigning authorship status.

The h-index also has a bias against highly cited papers. A scientist with a small number of highly influential papers may have a smaller h-index than a scientist with more papers of modest impact. Finally, an author who has changed names (eg, due to marriage) may have an artificially low h-index, as a standard database search would miss publications under a maiden name.

Limitations
This study is limited by possible operator error when compiling each author’s publication list through Web of Science. Our search and refinement methodology took into account that authors may publish with slight variations in name, in various subject areas and fields, and with different institutions and coauthors. Each publication populated through Web of Science was carefully verified by the principal investigator; however, overestimation or underestimation of the number of publications and citations was possible, as the publication lists were not verified by the studied associate and full professors themselves. Our results are consistent with the h-index bar charts published by John et al⁵ using an alternate citation index, Scopus, which tends to corroborate our findings. This study also is limited by possible time bias because we did not correct the h-index for years of active publication (m-index).

Conclusion

In summary, we found the median h-index for associate professors was 8 and the median h-index for full professors was 21. We found a broad range of h-index values within each academic rank. h-Index for upper-quartile associate professors overlapped with those of lower-quartile full professors. Our results suggest professional advancement occurs over a broad range of scholastic production. Adopting requirements for minimum h-index thresholds for application for promotion might reduce disparities between rank and scientific contributions. We encourage use of the h-index for tracking academic progression and as a parameter to consider in academic promotion.

Academic promotion requires evidence of scholastic production. The number of publications by a scientist is the most frequently reported metric of scholastic production, but it does not account for the impact of publications. The h-index is a bibliometric measure that combines both volume and impact of scientific contributions. The physicist Jorge E. Hirsch introduced this metric in 2005.¹ He defined it as the number of publications (h) by an author that have been cited at least h times. For example, a scientist with 30 publications including 12 that have been cited at least 12 times each has an h-index of 12. h-Index is a superior predictor of future scientific achievement in physics compared with total citation count, total publication count, and citations per publication. Hirsch² proposed h-index thresholds of 12 and 18 for advancement to associate professor and full professor in physics, respectively.²

h-Index values are not comparable across academic disciplines because they are influenced by the number of journals and authors within the field. Scientists in disciplines with numerous scholars and publications will have higher h-indices. For example, the mean h-index for full professors of cardiothoracic anesthesiology is 12, but the mean h-index for full professors of urology is 22.^3,4 Hence, h-index thresholds for professional advancement cannot be generalized but must be calculated on a granular, specialty-specific basis.

In a prior study on h-index among academic dermatologists in the United States, John et al⁵ reported that fellowship-trained dermatologists had a significantly higher mean h-index than those without fellowship training (13.2 vs 11.7; P<.001). They further found the mean h-index increased with academic rank.⁵

In our study, we measured mean and median h-indices among associate and full professors of dermatology in academic training programs in the United States with the goal of describing h-index distributions in these 2 academic ranks. We further sought to measure regional differences in h-index between northeastern, southern, central, and western states as defined by the National Resident Matching Program.

Methods

Institutional review board approval was deferred because the study did not require patient information or participation. Using the Association of American Medical Colleges Electronic Residency Application Service website (https://www.aamc.org/services/eras/) we identified dermatology residency training programs accredited by the Accreditation Council for Graduate Medical Education and participating in the Electronic Residency Application Service for the National Resident Matching Program in the United States. We visited the official website of each residency program and identified all associate and full professors of dermatology for further study. We included all faculty members listed as professor, clinical professor, associate professor, or clinical associate professor, and excluded assistant professor, volunteer faculty, research professor, and research associate professor. All faculty held an MD degree or an equivalent degree, such as MBBS or MDCM.

We used the Thomson Reuters (now Clarivate Analytics) Web of Science to calculate h-index and publication counts. The initial search was basic using the professor’s last name and first initial. We then augmented this list by searching for all variations of each professor’s name, with or without middle initial. Each publication in the search results was confirmed as belonging to the author of interest by verifying coauthors, institution information, and subject material. For authors with common names, we additionally consulted their online university profiles for specific names used in their “Selected Publications” lists. In a minority of cases, we also limited Research Domain to “dermatology.” Referring to the verified publication list for each dermatology professor, we used the Web of Science Citation Report function to determine number of publications and h-index for the individual. We tabulated results for associate and full professors and subgrouped those results into 4 geographic regions—northeastern, southern, central, and western states—according to the map used by the National Resident Matching Program. Descriptive statistics were performed with Microsoft Excel.

Results

We identified 300 associate professors and 352 full professors from 81 academic institutions. The number of associate professors per institution ranged from 1 to 25; the number of full professors per institution ranged from 1 to 16. The median and mean h-indices for associate and full professors, including interquartile values, are shown in the Table. There was a broad range of h-index scores among both academic ranks; median and mean h-indices varied more than 5-fold between the bottom and upper quartiles in both associate and full professor cohorts. Median interquartile h-index values for upper-quartile associate professors overlapped with those of lower-quartile full professors (Figure 1). h-Index for associate and full professors was similar across the 4 regions defined by the National Resident Matching Program. Median h-index was highest for full professors in western states and lowest for associate professors in southern states (Figure 2).

Comment

Professional advancement in academic medicine requires scholastic production. The h-index, defined as the number of publications (h) that have been cited at least h times, is a bibliometric measure that accounts for both volume and impact of an individual’s scientific productivity. The h-index would be a useful tool for determining professional advancement in academic dermatology departments. In this project, we calculated h-index values for 300 associate professors and 352 full professors of dermatology in the United States. We found the median h-index for associate professors was 8 and the median h-index for full professors was 21. There was more than a 5-fold variation in median and mean h-indices between lower and upper quartiles within both the associate and full professor cohorts. The highest median and mean h-indices were found among full professors of dermatology in western states. These results provide the opportunity for academic dermatologists and institutions to compare their research contributions with peers across the United States.

Our results support those of John et al⁵ who also found academic rank in dermatology was correlated with h-index. Scopus, Web of Science, and Google Scholar can be used to calculate h-index, but they may return different scores for the same individual.⁶ John et al⁵ used the Scopus database to calculate h-index. We used Web of Science because Scopus only includes citations since 1996 and Web of Science was used in the original h-index studies by Hirsch.^1,2 Institutions that adopt h-index criteria for advancement and resource distribution decisions should be aware that database selection can affect h-index scores.

Caveats With the h-Index
Flaws in the h-index include inflationary effects of self-citation, time bias, and excessive coauthorship. Individuals can increase their h-index by routinely citing their own publications. However, Engqvist and Frommen⁷ found tripling self-citations increased the h-index by only 1.

Citations tend to increase with time, and authors who have been active for longer periods will have a higher h-index. It is more difficult for junior faculty to distinguish themselves with the h-index, as it takes time for even the most impactful publications to gain citations. Major scientific papers can take years from conception to publication, and an outstanding paper that is 1 year old would have fewer citations than an equally impactful paper that is 10 years old. To adjust for the effect of time bias, Hirsch² proposed the m-index, in which the h-index is divided by the years between the author’s first and last publication. He proposed that an m-index of 1 would indicate a successful scientist, 2 an outstanding scientist, and 3 a unique individual.²

The literature is increasingly dominated by teams of coauthors, and the number of coauthors within each team has increased over the last 5 decades.⁸ h-Indices will increase if this trend continues, making it difficult to compare h-indices between different eras. Prosperi et al⁹ found national differences in kinship-based coauthorship, suggesting nepotism may influence decisions in assigning authorship status. h-Index valuations do not require evidence of meaningful contribution to the work but simply rely on contributors’ self-governance in assigning authorship status.

The h-index also has a bias against highly cited papers. A scientist with a small number of highly influential papers may have a smaller h-index than a scientist with more papers of modest impact. Finally, an author who has changed names (eg, due to marriage) may have an artificially low h-index, as a standard database search would miss publications under a maiden name.

Limitations
This study is limited by possible operator error when compiling each author’s publication list through Web of Science. Our search and refinement methodology took into account that authors may publish with slight variations in name, in various subject areas and fields, and with different institutions and coauthors. Each publication populated through Web of Science was carefully verified by the principal investigator; however, overestimation or underestimation of the number of publications and citations was possible, as the publication lists were not verified by the studied associate and full professors themselves. Our results are consistent with the h-index bar charts published by John et al⁵ using an alternate citation index, Scopus, which tends to corroborate our findings. This study also is limited by possible time bias because we did not correct the h-index for years of active publication (m-index).

Conclusion

In summary, we found the median h-index for associate professors was 8 and the median h-index for full professors was 21. We found a broad range of h-index values within each academic rank. h-Index for upper-quartile associate professors overlapped with those of lower-quartile full professors. Our results suggest professional advancement occurs over a broad range of scholastic production. Adopting requirements for minimum h-index thresholds for application for promotion might reduce disparities between rank and scientific contributions. We encourage use of the h-index for tracking academic progression and as a parameter to consider in academic promotion.