Original Research

Supination-adduction (SAD)-type fractures of the ankle comprise approximately 5% to 20% of ankle fractures.^1-3 As the name describes, this fracture is caused by forceful adduction of the supinated foot. There are 2 stages of the fracture pattern: the injury usually occurs first on the lateral side of the ankle with injury to the soft tissues or a low transverse fracture of the distal fibula. With continued force, in the second stage, the talus causes a shearing of the medial malleolus, creating the vertical shear fracture pattern.^4-7 The vertical shear medial malleolus fracture pattern is the subject of this investigation.

Several techniques have been traditionally recommended for fixation of SAD-type ankle fracture, including: a 2-screw construct without plate fixation, oriented perpendicular to the fracture; and an AG plate construct with variable positioning and numbers of screws for fixation. There have been, however, only 2 published articles about the biomechanical properties of fixation of vertical shear medial malleolar fractures, which reported conflicting results.^8,9 The most recent of these studies argued that one-third tubular plate fixation offers significant mechanical advantage over screw-only fixation, supporting the use of AG plates for fixation of SAD ankle fractures.⁸

An additional design for fixation of medial malleolus fractures has been introduced, consisting of a hook plate (HP) contoured for the medial malleolus. To our knowledge, no studies have investigated HP’s biomechanical properties. Thus, the objective of this study was to investigate and compare the biomechanical properties of 3 constructs for fixation of SAD-ankle fractures: an antiglide (AG) plate, an AG plate with an additional lag-screw across the fracture, and a precontoured HP.

Materials and Methods

Thirty 4th-generation–composite polyurethane models of the left tibia were obtained (Sawbones, Pacific Research Laboratories, Inc.). Largely, our methods accorded with the precedent set by other studies on these fracture types.^8,9

Prior to creation of the fractures, each model was individually evaluated for pretest stiffness by using the slope of the linear portion of the load-displacement curve during offset-axial loading. This demonstrated the baseline elasticity of the models during loading. Assessing pretest stiffness was performed to reduce potential variables in the stiffness of individual models in the analysis of the testing data.

The models were numbered 1 through 30 on the shaft and on the medial malleolus. A custom jig was constructed with a table saw to create identical vertical shear medial malleolar fracture patterns in each model. The jig created the vertical shear SAD fracture described by Lauge-Hansen.⁷ All models were randomly assigned to 1 of 3 groups; each group consisted of 10 models (Figures 1A, 1B).

The 10 specimens in group 1 were fixed with a 5-hole, 3.5-mm, one-third tubular plate (Smith & Nephew) in a traditional AG fashion. The plates were placed at the same location on all tibiae. The proximal hole and the hole closest to the fracture line were filled with 3.5-mm cortical screws, which were long enough to achieve bicortical fixation. No lag screws were placed in this specimen group. In group 2, specimens were fixed with the same plate used in group 1 (Smith & Nephew). In this modified AG (MAG) construct, specimens were fixed identically to group 1 for plate placement and fixation of the 2 proximal screws. In this group, an additional screw was placed perpendicular to the fracture and parallel to the distal tibial articular surface. In both groups (AG and MAG), the plates were not bent before application.

Group 3 consisted of specimens fixed with a 5-hole, precontoured medial malleolar HP (Arthrex). This HP construct was fixed with two 3.5-mm cortical screws long enough to achieve bicortical fixation. The plate also engaged the bone at the tip of the medial malleolus by using 2 sharp prongs. The screws were placed in the most proximal hole and the hole just proximal to the fracture line. No lag screws were placed in the HP construct.

All models were tested in offset-axial loading to replicate a SAD moment similar to previous studies. To test offset-axial loading, a vice held each model identically with a 17º angle from the longitudinal axis (Figure 2). Loading was performed with a material testing system; a material testing system plunger was directed at the inferior articulating cartilage surface of the medial malleolus. The specimens were loaded at a rate of 1 mm/sec until 2 mm of displacement was reached (Figure 3) or catastrophic failure occurred. The raw data analyzed consisted of the initial stiffness of the construct and the overall load-to-failure. The slope of the linear portion of the load-displacement curve of stiffness determined stiffness of the construct.

One-way analysis of variance with post hoc Tukey HSD data analysis was performed to determine if there were statistical differences among the different fixation constructs during load-to-failure. To prevent skewing of results by different values of model elasticity, pretest stiffness was accounted for by calculating a ratio of construct stiffness as a function of pretest model stiffness. Total force-to-failure was the recorded maximum force (in N) to cause failure. A P value of < .05 was set for significance. All data were analyzed using SPSS software (SPSS Version 15.0; SPSS Inc.).

Results

Analysis of pretest stiffness showed no significant difference among models (P = .490). All models failed by a gap of 2 mm at the distal fracture site except for 3 models in the MAG group. These 3 models failed at a much higher load than the remainder of the models and failed by fracture of the models.

The MAG group demonstrated significantly superior stiffness to the 2 other models tested (Figure 4). On average, this group required 753.5 N of force before failure. This was 530 N higher than the HP (P < .05) and 638 N higher than the AG constructs, respectively (P < .05). The HP and AG groups required forces of 223.2 N and 115.5 N for failure, respectively. These numbers were not significant (P= .063).

The absolute construct stiffness and construct stiffness as a function of pretest stiffness of the MAG group was the highest of all groups, 271.7 N/mm and 57.2%, respectively (Figure 5). These numbers showed significance when compared with the values of the HP group (P < .05 for both) and the AG group (P < .05 for both). The average stiffness of the HP group was 159.7 N/mm, which was 36.8% of pretest stiffness.

The AG group had the lowest construct stiffness and percent of pretest stiffness (128.1 N/mm and 29.6%). The HP and AG groups were not statistically different in these comparisons, P = .350 for construct stiffness and P = .395 for percent of pretest stiffness.

Discussion

These results support the use of a one-third tubular plate and lag-screw construct for fixation of vertical shear medial malleolus fractures. This is clinically important because one-third tubular plates with 3.5-mm screws are readily available and cost significantly less than a precountoured anatomic-specific type of fixation. These results are based on the biomechanical properties of the constructs tested in this study.

The previous 2 studies^8,9showed conflicting results about the most biomechanically sound fixation for SAD medial malleolar fractures. The study by Toolan and colleagues⁹ reported that 2 screws placed perpendicular to the fracture demonstrated the strongest overall construct. This study compared 3 separate types of 2-screw–only fixations and 2 plate-and-screw fixations. One construct was similar to the AG group in our study, and the other construct had a lag screw at the apex of the fracture. This previous study,⁹ however, did not investigate a similar construct to the MAG group that was tested in our study.

According to Dumigan and associates,⁸ a construct that consisted of a 4-hole plate with 2 screws proximal to the fracture and 2 lag screws showed the strongest fixation. This study, however, did not include a group like our study’s AG group, which is the traditional AG form of fixation.

In our study, we examined the biomechanic properties of a traditional fixation (AG construct), a commonly used fixation (MAG construct), and a newer construct (HP construct). The HP group is unique to this study and, to our knowledge, there is no literature on its use as fixation for this fracture. We did not include a 2-screw–only group, which is a limitation, because this fixation type is not common for the SAD fracture. This study also did not include an HP construct with an additional lag screw, which is an available option as well.

The current investigation used synthetic bone models constructed for biomechanical testing. The models were thought to provide a consistent model for fixation as opposed to using potentially osteopenic cadaveric bone. Each model was the same size and laterality. The stiffness as determined by pretest stiffness was not significantly different among models. Because all models were similar in composition and size, this allowed for more consistent osteotomies and similarly sized malleolar fragments. Theoretically, this allowed a more uniform comparison of all specimens and constructs.

Using models, however, is a limit of this study. While the models were of similar biomechanical quality, it is possible that a model may not reproduce the biology of a cavaderic specimen or the physiology of a construct in vivo. Of the 2 studies that investigated SAD fractures, the Dumigan study8 used cadaveric specimens. The fact that these models were all mildly osteoporotic and were embalmed specimens were study limits. The Toolan study⁹ used synthetic models. Although these models were consistent, they were models of bones and not intended for biomechanical studies, thereby increasing the potential for skewed results.

Our study investigated loading only in the offset-axial direction, a difference when compared to the Dumigan and colleagues8 and Toolan and colleagues⁹ studies. The offest transverse loading previously investigated would most likely represent an external rotation moment. While fixation in vivo could experience an external rotation moment, the specific fracture pattern of interest fails in offset-axial loading. In the original discription of the SAD fracture, Lauge-Hanson7 stated that the talus causes the vertically oriented medial malleolar fracture in the extreme of ankle supination with an adduction moment. Considering this, we investigated failure with a force in the direction that causes this type of fracture.

There are some additional limitations. This study demonstrated superiority of a one-third tubular plate with 2 screws proximally and 1 lag screw. While this was shown in the laboratory under pure offset-axial loading conditions, this may not reproduce daily forces experienced by the constructs. Additionally, this study examined load-to-failure of the constructs and did not investigate cyclic loading that a construct would experience in vivo. Because the testing is not recognizably consistent with day-to-day stresses of these constructs in vivo, this confounds the clinical application of our study.

The stiffness required for clinical healing is undetermined and, therefore, all 3 types of fixation could be adequate clinically. Patients are typically instructed to adhere to weight-bearing limitations on the affected extremity, and casts or splints are applied postoperatively for extended periods of time. Clinical studies would have significant benefit in the evaluation of fixation of vertical shear medial malleolar fractures.

Conclusion

AG plating technique with lag-screw placement is biomechanically superior to the other 2 constructs investigated. The clinical applications of these results are not known, and clinical trials are suggested to determine the best type of fixation for SAD-type medial malleolar fractures.

Supination-adduction (SAD)-type fractures of the ankle comprise approximately 5% to 20% of ankle fractures.^1-3 As the name describes, this fracture is caused by forceful adduction of the supinated foot. There are 2 stages of the fracture pattern: the injury usually occurs first on the lateral side of the ankle with injury to the soft tissues or a low transverse fracture of the distal fibula. With continued force, in the second stage, the talus causes a shearing of the medial malleolus, creating the vertical shear fracture pattern.^4-7 The vertical shear medial malleolus fracture pattern is the subject of this investigation.

Several techniques have been traditionally recommended for fixation of SAD-type ankle fracture, including: a 2-screw construct without plate fixation, oriented perpendicular to the fracture; and an AG plate construct with variable positioning and numbers of screws for fixation. There have been, however, only 2 published articles about the biomechanical properties of fixation of vertical shear medial malleolar fractures, which reported conflicting results.^8,9 The most recent of these studies argued that one-third tubular plate fixation offers significant mechanical advantage over screw-only fixation, supporting the use of AG plates for fixation of SAD ankle fractures.⁸

An additional design for fixation of medial malleolus fractures has been introduced, consisting of a hook plate (HP) contoured for the medial malleolus. To our knowledge, no studies have investigated HP’s biomechanical properties. Thus, the objective of this study was to investigate and compare the biomechanical properties of 3 constructs for fixation of SAD-ankle fractures: an antiglide (AG) plate, an AG plate with an additional lag-screw across the fracture, and a precontoured HP.

Materials and Methods

Thirty 4th-generation–composite polyurethane models of the left tibia were obtained (Sawbones, Pacific Research Laboratories, Inc.). Largely, our methods accorded with the precedent set by other studies on these fracture types.^8,9

Prior to creation of the fractures, each model was individually evaluated for pretest stiffness by using the slope of the linear portion of the load-displacement curve during offset-axial loading. This demonstrated the baseline elasticity of the models during loading. Assessing pretest stiffness was performed to reduce potential variables in the stiffness of individual models in the analysis of the testing data.

The models were numbered 1 through 30 on the shaft and on the medial malleolus. A custom jig was constructed with a table saw to create identical vertical shear medial malleolar fracture patterns in each model. The jig created the vertical shear SAD fracture described by Lauge-Hansen.⁷ All models were randomly assigned to 1 of 3 groups; each group consisted of 10 models (Figures 1A, 1B).

The 10 specimens in group 1 were fixed with a 5-hole, 3.5-mm, one-third tubular plate (Smith & Nephew) in a traditional AG fashion. The plates were placed at the same location on all tibiae. The proximal hole and the hole closest to the fracture line were filled with 3.5-mm cortical screws, which were long enough to achieve bicortical fixation. No lag screws were placed in this specimen group. In group 2, specimens were fixed with the same plate used in group 1 (Smith & Nephew). In this modified AG (MAG) construct, specimens were fixed identically to group 1 for plate placement and fixation of the 2 proximal screws. In this group, an additional screw was placed perpendicular to the fracture and parallel to the distal tibial articular surface. In both groups (AG and MAG), the plates were not bent before application.

Group 3 consisted of specimens fixed with a 5-hole, precontoured medial malleolar HP (Arthrex). This HP construct was fixed with two 3.5-mm cortical screws long enough to achieve bicortical fixation. The plate also engaged the bone at the tip of the medial malleolus by using 2 sharp prongs. The screws were placed in the most proximal hole and the hole just proximal to the fracture line. No lag screws were placed in the HP construct.

All models were tested in offset-axial loading to replicate a SAD moment similar to previous studies. To test offset-axial loading, a vice held each model identically with a 17º angle from the longitudinal axis (Figure 2). Loading was performed with a material testing system; a material testing system plunger was directed at the inferior articulating cartilage surface of the medial malleolus. The specimens were loaded at a rate of 1 mm/sec until 2 mm of displacement was reached (Figure 3) or catastrophic failure occurred. The raw data analyzed consisted of the initial stiffness of the construct and the overall load-to-failure. The slope of the linear portion of the load-displacement curve of stiffness determined stiffness of the construct.

One-way analysis of variance with post hoc Tukey HSD data analysis was performed to determine if there were statistical differences among the different fixation constructs during load-to-failure. To prevent skewing of results by different values of model elasticity, pretest stiffness was accounted for by calculating a ratio of construct stiffness as a function of pretest model stiffness. Total force-to-failure was the recorded maximum force (in N) to cause failure. A P value of < .05 was set for significance. All data were analyzed using SPSS software (SPSS Version 15.0; SPSS Inc.).

Results

Analysis of pretest stiffness showed no significant difference among models (P = .490). All models failed by a gap of 2 mm at the distal fracture site except for 3 models in the MAG group. These 3 models failed at a much higher load than the remainder of the models and failed by fracture of the models.

The MAG group demonstrated significantly superior stiffness to the 2 other models tested (Figure 4). On average, this group required 753.5 N of force before failure. This was 530 N higher than the HP (P < .05) and 638 N higher than the AG constructs, respectively (P < .05). The HP and AG groups required forces of 223.2 N and 115.5 N for failure, respectively. These numbers were not significant (P= .063).

The absolute construct stiffness and construct stiffness as a function of pretest stiffness of the MAG group was the highest of all groups, 271.7 N/mm and 57.2%, respectively (Figure 5). These numbers showed significance when compared with the values of the HP group (P < .05 for both) and the AG group (P < .05 for both). The average stiffness of the HP group was 159.7 N/mm, which was 36.8% of pretest stiffness.

The AG group had the lowest construct stiffness and percent of pretest stiffness (128.1 N/mm and 29.6%). The HP and AG groups were not statistically different in these comparisons, P = .350 for construct stiffness and P = .395 for percent of pretest stiffness.

Discussion

These results support the use of a one-third tubular plate and lag-screw construct for fixation of vertical shear medial malleolus fractures. This is clinically important because one-third tubular plates with 3.5-mm screws are readily available and cost significantly less than a precountoured anatomic-specific type of fixation. These results are based on the biomechanical properties of the constructs tested in this study.

The previous 2 studies^8,9showed conflicting results about the most biomechanically sound fixation for SAD medial malleolar fractures. The study by Toolan and colleagues⁹ reported that 2 screws placed perpendicular to the fracture demonstrated the strongest overall construct. This study compared 3 separate types of 2-screw–only fixations and 2 plate-and-screw fixations. One construct was similar to the AG group in our study, and the other construct had a lag screw at the apex of the fracture. This previous study,⁹ however, did not investigate a similar construct to the MAG group that was tested in our study.

According to Dumigan and associates,⁸ a construct that consisted of a 4-hole plate with 2 screws proximal to the fracture and 2 lag screws showed the strongest fixation. This study, however, did not include a group like our study’s AG group, which is the traditional AG form of fixation.

In our study, we examined the biomechanic properties of a traditional fixation (AG construct), a commonly used fixation (MAG construct), and a newer construct (HP construct). The HP group is unique to this study and, to our knowledge, there is no literature on its use as fixation for this fracture. We did not include a 2-screw–only group, which is a limitation, because this fixation type is not common for the SAD fracture. This study also did not include an HP construct with an additional lag screw, which is an available option as well.

The current investigation used synthetic bone models constructed for biomechanical testing. The models were thought to provide a consistent model for fixation as opposed to using potentially osteopenic cadaveric bone. Each model was the same size and laterality. The stiffness as determined by pretest stiffness was not significantly different among models. Because all models were similar in composition and size, this allowed for more consistent osteotomies and similarly sized malleolar fragments. Theoretically, this allowed a more uniform comparison of all specimens and constructs.

Using models, however, is a limit of this study. While the models were of similar biomechanical quality, it is possible that a model may not reproduce the biology of a cavaderic specimen or the physiology of a construct in vivo. Of the 2 studies that investigated SAD fractures, the Dumigan study8 used cadaveric specimens. The fact that these models were all mildly osteoporotic and were embalmed specimens were study limits. The Toolan study⁹ used synthetic models. Although these models were consistent, they were models of bones and not intended for biomechanical studies, thereby increasing the potential for skewed results.

Our study investigated loading only in the offset-axial direction, a difference when compared to the Dumigan and colleagues8 and Toolan and colleagues⁹ studies. The offest transverse loading previously investigated would most likely represent an external rotation moment. While fixation in vivo could experience an external rotation moment, the specific fracture pattern of interest fails in offset-axial loading. In the original discription of the SAD fracture, Lauge-Hanson7 stated that the talus causes the vertically oriented medial malleolar fracture in the extreme of ankle supination with an adduction moment. Considering this, we investigated failure with a force in the direction that causes this type of fracture.

There are some additional limitations. This study demonstrated superiority of a one-third tubular plate with 2 screws proximally and 1 lag screw. While this was shown in the laboratory under pure offset-axial loading conditions, this may not reproduce daily forces experienced by the constructs. Additionally, this study examined load-to-failure of the constructs and did not investigate cyclic loading that a construct would experience in vivo. Because the testing is not recognizably consistent with day-to-day stresses of these constructs in vivo, this confounds the clinical application of our study.

The stiffness required for clinical healing is undetermined and, therefore, all 3 types of fixation could be adequate clinically. Patients are typically instructed to adhere to weight-bearing limitations on the affected extremity, and casts or splints are applied postoperatively for extended periods of time. Clinical studies would have significant benefit in the evaluation of fixation of vertical shear medial malleolar fractures.

Conclusion

AG plating technique with lag-screw placement is biomechanically superior to the other 2 constructs investigated. The clinical applications of these results are not known, and clinical trials are suggested to determine the best type of fixation for SAD-type medial malleolar fractures.

Supination-adduction (SAD)-type fractures of the ankle comprise approximately 5% to 20% of ankle fractures.^1-3 As the name describes, this fracture is caused by forceful adduction of the supinated foot. There are 2 stages of the fracture pattern: the injury usually occurs first on the lateral side of the ankle with injury to the soft tissues or a low transverse fracture of the distal fibula. With continued force, in the second stage, the talus causes a shearing of the medial malleolus, creating the vertical shear fracture pattern.^4-7 The vertical shear medial malleolus fracture pattern is the subject of this investigation.

Several techniques have been traditionally recommended for fixation of SAD-type ankle fracture, including: a 2-screw construct without plate fixation, oriented perpendicular to the fracture; and an AG plate construct with variable positioning and numbers of screws for fixation. There have been, however, only 2 published articles about the biomechanical properties of fixation of vertical shear medial malleolar fractures, which reported conflicting results.^8,9 The most recent of these studies argued that one-third tubular plate fixation offers significant mechanical advantage over screw-only fixation, supporting the use of AG plates for fixation of SAD ankle fractures.⁸

An additional design for fixation of medial malleolus fractures has been introduced, consisting of a hook plate (HP) contoured for the medial malleolus. To our knowledge, no studies have investigated HP’s biomechanical properties. Thus, the objective of this study was to investigate and compare the biomechanical properties of 3 constructs for fixation of SAD-ankle fractures: an antiglide (AG) plate, an AG plate with an additional lag-screw across the fracture, and a precontoured HP.

Materials and Methods

Thirty 4th-generation–composite polyurethane models of the left tibia were obtained (Sawbones, Pacific Research Laboratories, Inc.). Largely, our methods accorded with the precedent set by other studies on these fracture types.^8,9

Prior to creation of the fractures, each model was individually evaluated for pretest stiffness by using the slope of the linear portion of the load-displacement curve during offset-axial loading. This demonstrated the baseline elasticity of the models during loading. Assessing pretest stiffness was performed to reduce potential variables in the stiffness of individual models in the analysis of the testing data.

The models were numbered 1 through 30 on the shaft and on the medial malleolus. A custom jig was constructed with a table saw to create identical vertical shear medial malleolar fracture patterns in each model. The jig created the vertical shear SAD fracture described by Lauge-Hansen.⁷ All models were randomly assigned to 1 of 3 groups; each group consisted of 10 models (Figures 1A, 1B).

The 10 specimens in group 1 were fixed with a 5-hole, 3.5-mm, one-third tubular plate (Smith & Nephew) in a traditional AG fashion. The plates were placed at the same location on all tibiae. The proximal hole and the hole closest to the fracture line were filled with 3.5-mm cortical screws, which were long enough to achieve bicortical fixation. No lag screws were placed in this specimen group. In group 2, specimens were fixed with the same plate used in group 1 (Smith & Nephew). In this modified AG (MAG) construct, specimens were fixed identically to group 1 for plate placement and fixation of the 2 proximal screws. In this group, an additional screw was placed perpendicular to the fracture and parallel to the distal tibial articular surface. In both groups (AG and MAG), the plates were not bent before application.

Group 3 consisted of specimens fixed with a 5-hole, precontoured medial malleolar HP (Arthrex). This HP construct was fixed with two 3.5-mm cortical screws long enough to achieve bicortical fixation. The plate also engaged the bone at the tip of the medial malleolus by using 2 sharp prongs. The screws were placed in the most proximal hole and the hole just proximal to the fracture line. No lag screws were placed in the HP construct.

All models were tested in offset-axial loading to replicate a SAD moment similar to previous studies. To test offset-axial loading, a vice held each model identically with a 17º angle from the longitudinal axis (Figure 2). Loading was performed with a material testing system; a material testing system plunger was directed at the inferior articulating cartilage surface of the medial malleolus. The specimens were loaded at a rate of 1 mm/sec until 2 mm of displacement was reached (Figure 3) or catastrophic failure occurred. The raw data analyzed consisted of the initial stiffness of the construct and the overall load-to-failure. The slope of the linear portion of the load-displacement curve of stiffness determined stiffness of the construct.

One-way analysis of variance with post hoc Tukey HSD data analysis was performed to determine if there were statistical differences among the different fixation constructs during load-to-failure. To prevent skewing of results by different values of model elasticity, pretest stiffness was accounted for by calculating a ratio of construct stiffness as a function of pretest model stiffness. Total force-to-failure was the recorded maximum force (in N) to cause failure. A P value of < .05 was set for significance. All data were analyzed using SPSS software (SPSS Version 15.0; SPSS Inc.).

Results

Analysis of pretest stiffness showed no significant difference among models (P = .490). All models failed by a gap of 2 mm at the distal fracture site except for 3 models in the MAG group. These 3 models failed at a much higher load than the remainder of the models and failed by fracture of the models.

The MAG group demonstrated significantly superior stiffness to the 2 other models tested (Figure 4). On average, this group required 753.5 N of force before failure. This was 530 N higher than the HP (P < .05) and 638 N higher than the AG constructs, respectively (P < .05). The HP and AG groups required forces of 223.2 N and 115.5 N for failure, respectively. These numbers were not significant (P= .063).

The absolute construct stiffness and construct stiffness as a function of pretest stiffness of the MAG group was the highest of all groups, 271.7 N/mm and 57.2%, respectively (Figure 5). These numbers showed significance when compared with the values of the HP group (P < .05 for both) and the AG group (P < .05 for both). The average stiffness of the HP group was 159.7 N/mm, which was 36.8% of pretest stiffness.

The AG group had the lowest construct stiffness and percent of pretest stiffness (128.1 N/mm and 29.6%). The HP and AG groups were not statistically different in these comparisons, P = .350 for construct stiffness and P = .395 for percent of pretest stiffness.

Discussion

These results support the use of a one-third tubular plate and lag-screw construct for fixation of vertical shear medial malleolus fractures. This is clinically important because one-third tubular plates with 3.5-mm screws are readily available and cost significantly less than a precountoured anatomic-specific type of fixation. These results are based on the biomechanical properties of the constructs tested in this study.

The previous 2 studies^8,9showed conflicting results about the most biomechanically sound fixation for SAD medial malleolar fractures. The study by Toolan and colleagues⁹ reported that 2 screws placed perpendicular to the fracture demonstrated the strongest overall construct. This study compared 3 separate types of 2-screw–only fixations and 2 plate-and-screw fixations. One construct was similar to the AG group in our study, and the other construct had a lag screw at the apex of the fracture. This previous study,⁹ however, did not investigate a similar construct to the MAG group that was tested in our study.

According to Dumigan and associates,⁸ a construct that consisted of a 4-hole plate with 2 screws proximal to the fracture and 2 lag screws showed the strongest fixation. This study, however, did not include a group like our study’s AG group, which is the traditional AG form of fixation.

In our study, we examined the biomechanic properties of a traditional fixation (AG construct), a commonly used fixation (MAG construct), and a newer construct (HP construct). The HP group is unique to this study and, to our knowledge, there is no literature on its use as fixation for this fracture. We did not include a 2-screw–only group, which is a limitation, because this fixation type is not common for the SAD fracture. This study also did not include an HP construct with an additional lag screw, which is an available option as well.

The current investigation used synthetic bone models constructed for biomechanical testing. The models were thought to provide a consistent model for fixation as opposed to using potentially osteopenic cadaveric bone. Each model was the same size and laterality. The stiffness as determined by pretest stiffness was not significantly different among models. Because all models were similar in composition and size, this allowed for more consistent osteotomies and similarly sized malleolar fragments. Theoretically, this allowed a more uniform comparison of all specimens and constructs.

Using models, however, is a limit of this study. While the models were of similar biomechanical quality, it is possible that a model may not reproduce the biology of a cavaderic specimen or the physiology of a construct in vivo. Of the 2 studies that investigated SAD fractures, the Dumigan study8 used cadaveric specimens. The fact that these models were all mildly osteoporotic and were embalmed specimens were study limits. The Toolan study⁹ used synthetic models. Although these models were consistent, they were models of bones and not intended for biomechanical studies, thereby increasing the potential for skewed results.

Our study investigated loading only in the offset-axial direction, a difference when compared to the Dumigan and colleagues8 and Toolan and colleagues⁹ studies. The offest transverse loading previously investigated would most likely represent an external rotation moment. While fixation in vivo could experience an external rotation moment, the specific fracture pattern of interest fails in offset-axial loading. In the original discription of the SAD fracture, Lauge-Hanson7 stated that the talus causes the vertically oriented medial malleolar fracture in the extreme of ankle supination with an adduction moment. Considering this, we investigated failure with a force in the direction that causes this type of fracture.

There are some additional limitations. This study demonstrated superiority of a one-third tubular plate with 2 screws proximally and 1 lag screw. While this was shown in the laboratory under pure offset-axial loading conditions, this may not reproduce daily forces experienced by the constructs. Additionally, this study examined load-to-failure of the constructs and did not investigate cyclic loading that a construct would experience in vivo. Because the testing is not recognizably consistent with day-to-day stresses of these constructs in vivo, this confounds the clinical application of our study.

The stiffness required for clinical healing is undetermined and, therefore, all 3 types of fixation could be adequate clinically. Patients are typically instructed to adhere to weight-bearing limitations on the affected extremity, and casts or splints are applied postoperatively for extended periods of time. Clinical studies would have significant benefit in the evaluation of fixation of vertical shear medial malleolar fractures.

Conclusion

AG plating technique with lag-screw placement is biomechanically superior to the other 2 constructs investigated. The clinical applications of these results are not known, and clinical trials are suggested to determine the best type of fixation for SAD-type medial malleolar fractures.

Despite an abundance of peer-reviewed resources, there is wide variation in the surgical management of shoulder instability.^1,2 Current American Academy of Orthopaedic Surgeons (AAOS) clinical practice guidelines regarding the shoulder address only generalized shoulder pain, glenohumeral osteoarthritis, and rotator cuff injuries,^3,4 and treatment algorithms focus on conservative treatment, rather than surgical recommendations.^4-7

Shoulder instability most commonly results from 1 or more of 4 common lesions (capsular laxity, glenoid bone loss, humeral bone loss, and capsulolabral insufficiency).⁸ While it is a relatively common condition that represents 1% to 2% of all athletic injuries,^9,10 little consensus exists about surgical indications, ideal treatment algorithms, or optimal operative technique. This is a critical issue because more than 50% of patients with glenohumeral instability will undergo surgical intervention.¹¹ Chahal and associates⁶ surveyed 44 shoulder experts and reported strong consensus about diagnosis, but little agreement regarding surgical management. Owens and colleagues¹ have also evaluated current trends for surgical treatment of this pathology. Randelli and associates⁵ attempted to categorize operative management based upon case-specific shoulder scenarios through online surveys. Their survey, however, covered a broad range of shoulder injuries rather than instability in particular. In this study, we assess trends for surgical management of glenohumeral instability in a case-based survey of shoulder experts.

Materials and Methods

Survey Information

An online survey (Survey Monkey) of 417 active members of the American Shoulder and Elbow Surgeons (ASES) was administered on May 1, 2014. Respondents were blinded to the institution and co-investigators conducting the survey. The survey link was distributed via email because it has been shown to be a more efficacious conduit than standard postal mail.¹² The case-based, 25-question survey (Appendix) was designed to assess respondents’ selection of surgical intervention. Section 1 determined member demographics, including fellowship training, arthroscopy experience, and years of practice. Section 2 involved the presentation of 5 case scenarios. For each case, respondents were asked to identify the optimal surgical procedure in both primary and revision settings. Section 3 posed several general questions regarding shoulder-instability management.

Statistical Analysis

Data were stored using Microsoft Excel (Microsoft) and analyzed using SAS Software version 9.3 (SAS Institute, Inc.). Demographic survey responses were reported using descriptive statistics. Responses to clinical survey questions were reported using frequencies and percentages. To identify when a majority consensus was achieved for a given question, responses were flagged as reaching consensus when more than 50% of participants gave the same response.¹³In the event that only 2 response options were available, reaching consensus required 67% of respondents to choose a single answer (since, by default, a consensus would be reached with only 2 response options). Because this was an analysis of all respondents, an a priori power calculation was not performed. Associations between training and practice demographics and responses to clinical questions were investigated using chi-square analyses. All comparative analyses were two-tailed and used P = .05 as the threshold for statistical significance.

Results

Demographics

One hundred and twenty-five (29.9%) ASES members responded to the survey. Of the respondents, 71.2% reported at least 15 years of experience, and 71% performed more than 150 shoulder cases annually. Surgeons came from academic institutions (41.6%), private practice (24.8%), or mixed (33.6%). The majority of respondents were fellowship-trained in shoulder/elbow surgery (52.8%), while fewer had completed a sports-medicine fellowship (24.0%). For arthroscopic procedures, responses were nearly divided between those who preferred beach-chair positioning (47.2%) and those who preferred the lateral decubitus position (46.4%). The majority (70.4%) of respondents practiced in the United States and with a relatively even distribution among states and region. The remaining 29.6% of those surveyed practiced abroad.

Degree of Consensus Responses and Cases

Of the 25 survey questions, 6 questions were omitted from consensus calculations because these were designed for demographic categorization rather than professional opinion (questions 1-5, 8). Thirteen of the remaining 19 questions (68%) reached consensus response. All clinical case scenarios (5 of 5) reached consensus for selection of technique for the primary procedure; however, only 40% (2 of 5) of cases had a consensus in the revision setting.

In case 1, a young soccer player (noncontact athlete) with negligible bone loss, arthroscopic Bankart repair was recommended by 81.6% of respondents. In the event of revision surgery, only 22.4% recommended arthroscopic Bankart repair, and the remainder split between open Bankart repair with possible capsular shift (36%) or Latarjet procedure (32.8%).

In case 2, a college American football player (contact athlete) with negligible bone loss, arthroscopic Bankart repair was recommended by 56.8%. In the event of revision surgery, a majority of members (51.2%) suggested a Latarjet procedure.

In case 3, the weekend warrior with significant bone loss, most respondents recommended a Latarjet procedure for both primary (72.8%) and revision surgery (79.0%).

In case 4, a weekend warrior with multidirectional instability, 60% of respondents suggested arthroscopic Bankart repair, 21.6% recommended rotator interval closure, and 10.4% chose a capsular shift. As a revision procedure, there was less agreement, with a split between open Bankart repair (39.2%) and capsular shift only (39.2%).

In case 5, the weekend warrior with large engaging Hill-Sachs lesions, 60% of respondent selected a remplissage procedure. If revision was required, a Latarjet procedure was the choice of 48.8% of respondents (Table).

General Questions

For contact athletes, most respondents (87.2%) would allow return to play in the same season and recommended surgery after the end of the season. After surgical intervention, 56.8% prescribed 4 weeks of immobilization. When counseling a return to contact sports, 51.2% recommended waiting for 4 to 6 months.

The ASES members were divided on conservative management of instability injuries. Responses included immobilization in internal rotation (39.2%), no immobilization (39.2%), and external-rotation bracing (21.6%).

Finally, members thought the most important factor in choosing surgical technique was the patient’s pathology, then age; the least influential criteria was the patient’s sports participation.

Analysis of Training Demographics and Surgical Technique Preferences

Chi-square analyses demonstrated that respondents who completed a sports fellowship were more likely to do at least 50% of cases arthroscopically (odds ratio [OR], 15.3; P < .001) and were more likely to use the lateral decubitus position (OR, 2.8; P < .021). Furthermore, American respondents had a higher likelihood of having completed either a sports fellowship (OR, 12.8; P < .001) or a shoulder/elbow fellowship (OR, 4.6; P = .002) when compared with foreign respondents.

Discussion

In the absence of formal clinical practice guidelines, most surgeons formulate treatment strategy based upon a combination of experience and peer-reviewed evidence. The cohort analyzed in the current study was highly experienced, with more than 70% performing 150 shoulder cases annually and having more than 15 years of experience. We found a consensus response in 68% of questions and all primary surgical techniques for our shoulder instability scenarios. While expert consensus reported here is not equivalent to evidence-based clinical practice guidelines, it does provide important information to consider when treating anterior shoulder instability.

Specific responses to our case scenarios invite further reflection. Considering young (both noncontact and contact) athletes without bony pathology (cases 1 and 2, respectively), the ASES surgeons recommended arthroscopic Bankart repair for both. Randelli and associates5 found 71% of survey respondents recommended arthroscopic Bankart repair in a similar setting. It is interesting to note that consensus persisted regardless of the sport in which they engaged. Contact athletes have the highest rates of dislocation (up to 7 times higher incidence) compared with the general population.¹⁴ In addition, they have a higher recurrence rate after surgery.¹⁵ It should be noted, however, that although both cases reached consensus, the percentage of experts who recommended an arthroscopic procedure fell from 82% in the noncontact athlete to 57% in the contact athlete. This concurs with a recent review by Harris and Romeo,¹⁶ who recommended similar treatments for athletes without bony defects. In an older patient population with recurrent instability (case 3), responses varied more widely but still reached a consensus on primary surgical techniques. Respondents agreed that, even for patients with multidirectional instability, initial management should consist of arthroscopic capsulolabral repair. Overall, the agreement for arthroscopy for cases 1 through 3 mimics recent US practice patterns, showing 90% of stabilizations are being performed arthroscopically.¹⁷ Additionally, a recent meta-analysis by Harris and associates18 favored arthroscopic Bankart repair, showing no significant difference vs open stabilization even on long-term follow-up.

Glenoid bone loss is a difficult clinical scenario and that is reflected in this study’s findings. The literature suggests that arthroscopic Bankart repair, in this setting, is usually not sufficient and may result in a recurrence rate up to 75%, if bone loss greater than 20% is unaddressed.¹⁹ Our study supports this trend because ASES members recommended a Latarjet procedure when there is substantial bone loss.

While open Latarjet procedure was the consensus for dealing with glenoid bone loss, arthroscopic techniques were strongly favored for humeral head defects. This change in practice patterns results from the introduction of the arthroscopic remplissage technique.²⁰ Two recent systemic reviews have supported this technique, reporting good functional outcomes for engaging Hill-Sachs lesions.^21,22 Our study had similar agreement, with most respondents recommending remplissage for these patients.

This study found the lowest rates of expert consensus in the setting of revision surgery, likely caused, in part, by the paucity of available large cohort studies. This is a major void in the literature, and more studies are needed to help guide surgeons on the best techniques to deal with this difficult patient population.

Conservative bracing technique was 1 of the survey questions lacking a consensus response. Interestingly, 39% of members recommended no immobilization after an instability event. This contrasts with recent literature concerning the best position for bracing. We also found twice as many surgeons recommended internal rotation immobilization over external rotation. This is a subject of debate, with some studies stating improvement with external rotation immobilization,²³ while other studies reported no difference.²⁴ Overall, recommendations regarding type of immobilization are unclear, which will likely continue until larger studies can be performed.

The literature describing surgical trends in the treatment of shoulder instability is sparse and variable. With regard to other shoulder etiologies, only rotator cuff pathology has used expert consensus. Acevedo and colleagues¹³ reported agreement of ASES members surveyed regarding rotator cuff management. There was no consensus among surgeons in more than 50% of questions, despite AAOS published guidelines for rotator cuff treatment.²⁵ Despite the lack of guidelines for our topic, we found a consensus among respondents with 68% of survey questions.

To date, only 2 studies of shoulder instability management have elicited the opinion of experts in shoulder surgery. Chahal and associates⁶ surveyed 42 members of ASES and JOINTS (Joined Orthopaedic Initiatives for National Trials of the Shoulder) Canada on shoulder instability cases and found substantial agreement on diagnosis but little consensus regarding surgical technique. This lack of agreement on procedures differs from our findings and may be related to their complicated case scenarios that generated a wide array of treatment recommendations. Randelli and colleagues⁵ surveyed more than 1000 European Society of Sports Traumatology, Knee Surgery, and Arthroscopy members and reported similar agreement on arthroscopic Bankart repair in young male shoulder-dislocation patients, although no other instability scenarios were investigated. Our study is the first to report responses from expert shoulder surgeons on surgical-treatment strategies for an array of common shoulder instability pathologies.

This study had several limitations. First, while our study suffered from a low response rate (29.9%), it was similar to other published studies.^5,13 Second, because the case series included in the survey attempted to capture the most common instability scenarios, they were limited in their scope and failed to address additional etiologies or pathologic permutations. We believe, however, that a more comprehensive survey would have resulted in respondent fatigue and lowered the response rate. It is unlikely that any survey could capture all variables that come into play during clinical decision-making, and we sought to evaluate the most common shoulder instability scenarios. Third, 30% of respondents were from outside the United States, where the Latarjet procedure is much more popular. While this was not a majority, Latarjet’s regional preference may have decreased the consensus response in some scenarios if only the United States was included. Finally, there is inherent bias in a respondent pool that is heavily weighted to shoulder-surgery experts (ASES members) and does not consider the responses of the general orthopedic surgery community as have other studies.⁷

Conclusion

This study demonstrates that expert shoulder surgeons often agreed on shoulder-treatment principles for anterior shoulder instability. In the setting of primary repair, arthroscopic Bankart repair was favored in the absence of bony pathology, regardless of age (20 to 35 years) or nature of sport (contact versus noncontact). Latarjet procedures were favored in the setting of glenoid bone loss, and remplissage for an engaging Hill-Sachs lesion. Less agreement was observed for revision stabilization. It should be noted that, while consensus was often reached for our cases, there was a wide distribution of technical considerations and surgical preferences even among those who are fellowship-trained and high-volume surgeons, and who can be considered experts in the field of shoulder surgery.

Despite an abundance of peer-reviewed resources, there is wide variation in the surgical management of shoulder instability.^1,2 Current American Academy of Orthopaedic Surgeons (AAOS) clinical practice guidelines regarding the shoulder address only generalized shoulder pain, glenohumeral osteoarthritis, and rotator cuff injuries,^3,4 and treatment algorithms focus on conservative treatment, rather than surgical recommendations.^4-7

Shoulder instability most commonly results from 1 or more of 4 common lesions (capsular laxity, glenoid bone loss, humeral bone loss, and capsulolabral insufficiency).⁸ While it is a relatively common condition that represents 1% to 2% of all athletic injuries,^9,10 little consensus exists about surgical indications, ideal treatment algorithms, or optimal operative technique. This is a critical issue because more than 50% of patients with glenohumeral instability will undergo surgical intervention.¹¹ Chahal and associates⁶ surveyed 44 shoulder experts and reported strong consensus about diagnosis, but little agreement regarding surgical management. Owens and colleagues¹ have also evaluated current trends for surgical treatment of this pathology. Randelli and associates⁵ attempted to categorize operative management based upon case-specific shoulder scenarios through online surveys. Their survey, however, covered a broad range of shoulder injuries rather than instability in particular. In this study, we assess trends for surgical management of glenohumeral instability in a case-based survey of shoulder experts.

Materials and Methods

Survey Information

An online survey (Survey Monkey) of 417 active members of the American Shoulder and Elbow Surgeons (ASES) was administered on May 1, 2014. Respondents were blinded to the institution and co-investigators conducting the survey. The survey link was distributed via email because it has been shown to be a more efficacious conduit than standard postal mail.¹² The case-based, 25-question survey (Appendix) was designed to assess respondents’ selection of surgical intervention. Section 1 determined member demographics, including fellowship training, arthroscopy experience, and years of practice. Section 2 involved the presentation of 5 case scenarios. For each case, respondents were asked to identify the optimal surgical procedure in both primary and revision settings. Section 3 posed several general questions regarding shoulder-instability management.

Statistical Analysis

Data were stored using Microsoft Excel (Microsoft) and analyzed using SAS Software version 9.3 (SAS Institute, Inc.). Demographic survey responses were reported using descriptive statistics. Responses to clinical survey questions were reported using frequencies and percentages. To identify when a majority consensus was achieved for a given question, responses were flagged as reaching consensus when more than 50% of participants gave the same response.¹³In the event that only 2 response options were available, reaching consensus required 67% of respondents to choose a single answer (since, by default, a consensus would be reached with only 2 response options). Because this was an analysis of all respondents, an a priori power calculation was not performed. Associations between training and practice demographics and responses to clinical questions were investigated using chi-square analyses. All comparative analyses were two-tailed and used P = .05 as the threshold for statistical significance.

Results

Demographics

One hundred and twenty-five (29.9%) ASES members responded to the survey. Of the respondents, 71.2% reported at least 15 years of experience, and 71% performed more than 150 shoulder cases annually. Surgeons came from academic institutions (41.6%), private practice (24.8%), or mixed (33.6%). The majority of respondents were fellowship-trained in shoulder/elbow surgery (52.8%), while fewer had completed a sports-medicine fellowship (24.0%). For arthroscopic procedures, responses were nearly divided between those who preferred beach-chair positioning (47.2%) and those who preferred the lateral decubitus position (46.4%). The majority (70.4%) of respondents practiced in the United States and with a relatively even distribution among states and region. The remaining 29.6% of those surveyed practiced abroad.

Degree of Consensus Responses and Cases

Of the 25 survey questions, 6 questions were omitted from consensus calculations because these were designed for demographic categorization rather than professional opinion (questions 1-5, 8). Thirteen of the remaining 19 questions (68%) reached consensus response. All clinical case scenarios (5 of 5) reached consensus for selection of technique for the primary procedure; however, only 40% (2 of 5) of cases had a consensus in the revision setting.

In case 1, a young soccer player (noncontact athlete) with negligible bone loss, arthroscopic Bankart repair was recommended by 81.6% of respondents. In the event of revision surgery, only 22.4% recommended arthroscopic Bankart repair, and the remainder split between open Bankart repair with possible capsular shift (36%) or Latarjet procedure (32.8%).

In case 2, a college American football player (contact athlete) with negligible bone loss, arthroscopic Bankart repair was recommended by 56.8%. In the event of revision surgery, a majority of members (51.2%) suggested a Latarjet procedure.

In case 3, the weekend warrior with significant bone loss, most respondents recommended a Latarjet procedure for both primary (72.8%) and revision surgery (79.0%).

In case 4, a weekend warrior with multidirectional instability, 60% of respondents suggested arthroscopic Bankart repair, 21.6% recommended rotator interval closure, and 10.4% chose a capsular shift. As a revision procedure, there was less agreement, with a split between open Bankart repair (39.2%) and capsular shift only (39.2%).

In case 5, the weekend warrior with large engaging Hill-Sachs lesions, 60% of respondent selected a remplissage procedure. If revision was required, a Latarjet procedure was the choice of 48.8% of respondents (Table).

General Questions

For contact athletes, most respondents (87.2%) would allow return to play in the same season and recommended surgery after the end of the season. After surgical intervention, 56.8% prescribed 4 weeks of immobilization. When counseling a return to contact sports, 51.2% recommended waiting for 4 to 6 months.

The ASES members were divided on conservative management of instability injuries. Responses included immobilization in internal rotation (39.2%), no immobilization (39.2%), and external-rotation bracing (21.6%).

Finally, members thought the most important factor in choosing surgical technique was the patient’s pathology, then age; the least influential criteria was the patient’s sports participation.

Analysis of Training Demographics and Surgical Technique Preferences

Chi-square analyses demonstrated that respondents who completed a sports fellowship were more likely to do at least 50% of cases arthroscopically (odds ratio [OR], 15.3; P < .001) and were more likely to use the lateral decubitus position (OR, 2.8; P < .021). Furthermore, American respondents had a higher likelihood of having completed either a sports fellowship (OR, 12.8; P < .001) or a shoulder/elbow fellowship (OR, 4.6; P = .002) when compared with foreign respondents.

Discussion

In the absence of formal clinical practice guidelines, most surgeons formulate treatment strategy based upon a combination of experience and peer-reviewed evidence. The cohort analyzed in the current study was highly experienced, with more than 70% performing 150 shoulder cases annually and having more than 15 years of experience. We found a consensus response in 68% of questions and all primary surgical techniques for our shoulder instability scenarios. While expert consensus reported here is not equivalent to evidence-based clinical practice guidelines, it does provide important information to consider when treating anterior shoulder instability.

Specific responses to our case scenarios invite further reflection. Considering young (both noncontact and contact) athletes without bony pathology (cases 1 and 2, respectively), the ASES surgeons recommended arthroscopic Bankart repair for both. Randelli and associates5 found 71% of survey respondents recommended arthroscopic Bankart repair in a similar setting. It is interesting to note that consensus persisted regardless of the sport in which they engaged. Contact athletes have the highest rates of dislocation (up to 7 times higher incidence) compared with the general population.¹⁴ In addition, they have a higher recurrence rate after surgery.¹⁵ It should be noted, however, that although both cases reached consensus, the percentage of experts who recommended an arthroscopic procedure fell from 82% in the noncontact athlete to 57% in the contact athlete. This concurs with a recent review by Harris and Romeo,¹⁶ who recommended similar treatments for athletes without bony defects. In an older patient population with recurrent instability (case 3), responses varied more widely but still reached a consensus on primary surgical techniques. Respondents agreed that, even for patients with multidirectional instability, initial management should consist of arthroscopic capsulolabral repair. Overall, the agreement for arthroscopy for cases 1 through 3 mimics recent US practice patterns, showing 90% of stabilizations are being performed arthroscopically.¹⁷ Additionally, a recent meta-analysis by Harris and associates18 favored arthroscopic Bankart repair, showing no significant difference vs open stabilization even on long-term follow-up.

Glenoid bone loss is a difficult clinical scenario and that is reflected in this study’s findings. The literature suggests that arthroscopic Bankart repair, in this setting, is usually not sufficient and may result in a recurrence rate up to 75%, if bone loss greater than 20% is unaddressed.¹⁹ Our study supports this trend because ASES members recommended a Latarjet procedure when there is substantial bone loss.

While open Latarjet procedure was the consensus for dealing with glenoid bone loss, arthroscopic techniques were strongly favored for humeral head defects. This change in practice patterns results from the introduction of the arthroscopic remplissage technique.²⁰ Two recent systemic reviews have supported this technique, reporting good functional outcomes for engaging Hill-Sachs lesions.^21,22 Our study had similar agreement, with most respondents recommending remplissage for these patients.

This study found the lowest rates of expert consensus in the setting of revision surgery, likely caused, in part, by the paucity of available large cohort studies. This is a major void in the literature, and more studies are needed to help guide surgeons on the best techniques to deal with this difficult patient population.

Conservative bracing technique was 1 of the survey questions lacking a consensus response. Interestingly, 39% of members recommended no immobilization after an instability event. This contrasts with recent literature concerning the best position for bracing. We also found twice as many surgeons recommended internal rotation immobilization over external rotation. This is a subject of debate, with some studies stating improvement with external rotation immobilization,²³ while other studies reported no difference.²⁴ Overall, recommendations regarding type of immobilization are unclear, which will likely continue until larger studies can be performed.

The literature describing surgical trends in the treatment of shoulder instability is sparse and variable. With regard to other shoulder etiologies, only rotator cuff pathology has used expert consensus. Acevedo and colleagues¹³ reported agreement of ASES members surveyed regarding rotator cuff management. There was no consensus among surgeons in more than 50% of questions, despite AAOS published guidelines for rotator cuff treatment.²⁵ Despite the lack of guidelines for our topic, we found a consensus among respondents with 68% of survey questions.

To date, only 2 studies of shoulder instability management have elicited the opinion of experts in shoulder surgery. Chahal and associates⁶ surveyed 42 members of ASES and JOINTS (Joined Orthopaedic Initiatives for National Trials of the Shoulder) Canada on shoulder instability cases and found substantial agreement on diagnosis but little consensus regarding surgical technique. This lack of agreement on procedures differs from our findings and may be related to their complicated case scenarios that generated a wide array of treatment recommendations. Randelli and colleagues⁵ surveyed more than 1000 European Society of Sports Traumatology, Knee Surgery, and Arthroscopy members and reported similar agreement on arthroscopic Bankart repair in young male shoulder-dislocation patients, although no other instability scenarios were investigated. Our study is the first to report responses from expert shoulder surgeons on surgical-treatment strategies for an array of common shoulder instability pathologies.

This study had several limitations. First, while our study suffered from a low response rate (29.9%), it was similar to other published studies.^5,13 Second, because the case series included in the survey attempted to capture the most common instability scenarios, they were limited in their scope and failed to address additional etiologies or pathologic permutations. We believe, however, that a more comprehensive survey would have resulted in respondent fatigue and lowered the response rate. It is unlikely that any survey could capture all variables that come into play during clinical decision-making, and we sought to evaluate the most common shoulder instability scenarios. Third, 30% of respondents were from outside the United States, where the Latarjet procedure is much more popular. While this was not a majority, Latarjet’s regional preference may have decreased the consensus response in some scenarios if only the United States was included. Finally, there is inherent bias in a respondent pool that is heavily weighted to shoulder-surgery experts (ASES members) and does not consider the responses of the general orthopedic surgery community as have other studies.⁷

Conclusion

This study demonstrates that expert shoulder surgeons often agreed on shoulder-treatment principles for anterior shoulder instability. In the setting of primary repair, arthroscopic Bankart repair was favored in the absence of bony pathology, regardless of age (20 to 35 years) or nature of sport (contact versus noncontact). Latarjet procedures were favored in the setting of glenoid bone loss, and remplissage for an engaging Hill-Sachs lesion. Less agreement was observed for revision stabilization. It should be noted that, while consensus was often reached for our cases, there was a wide distribution of technical considerations and surgical preferences even among those who are fellowship-trained and high-volume surgeons, and who can be considered experts in the field of shoulder surgery.

Despite an abundance of peer-reviewed resources, there is wide variation in the surgical management of shoulder instability.^1,2 Current American Academy of Orthopaedic Surgeons (AAOS) clinical practice guidelines regarding the shoulder address only generalized shoulder pain, glenohumeral osteoarthritis, and rotator cuff injuries,^3,4 and treatment algorithms focus on conservative treatment, rather than surgical recommendations.^4-7

Shoulder instability most commonly results from 1 or more of 4 common lesions (capsular laxity, glenoid bone loss, humeral bone loss, and capsulolabral insufficiency).⁸ While it is a relatively common condition that represents 1% to 2% of all athletic injuries,^9,10 little consensus exists about surgical indications, ideal treatment algorithms, or optimal operative technique. This is a critical issue because more than 50% of patients with glenohumeral instability will undergo surgical intervention.¹¹ Chahal and associates⁶ surveyed 44 shoulder experts and reported strong consensus about diagnosis, but little agreement regarding surgical management. Owens and colleagues¹ have also evaluated current trends for surgical treatment of this pathology. Randelli and associates⁵ attempted to categorize operative management based upon case-specific shoulder scenarios through online surveys. Their survey, however, covered a broad range of shoulder injuries rather than instability in particular. In this study, we assess trends for surgical management of glenohumeral instability in a case-based survey of shoulder experts.

Materials and Methods

Survey Information

An online survey (Survey Monkey) of 417 active members of the American Shoulder and Elbow Surgeons (ASES) was administered on May 1, 2014. Respondents were blinded to the institution and co-investigators conducting the survey. The survey link was distributed via email because it has been shown to be a more efficacious conduit than standard postal mail.¹² The case-based, 25-question survey (Appendix) was designed to assess respondents’ selection of surgical intervention. Section 1 determined member demographics, including fellowship training, arthroscopy experience, and years of practice. Section 2 involved the presentation of 5 case scenarios. For each case, respondents were asked to identify the optimal surgical procedure in both primary and revision settings. Section 3 posed several general questions regarding shoulder-instability management.

Statistical Analysis

Data were stored using Microsoft Excel (Microsoft) and analyzed using SAS Software version 9.3 (SAS Institute, Inc.). Demographic survey responses were reported using descriptive statistics. Responses to clinical survey questions were reported using frequencies and percentages. To identify when a majority consensus was achieved for a given question, responses were flagged as reaching consensus when more than 50% of participants gave the same response.¹³In the event that only 2 response options were available, reaching consensus required 67% of respondents to choose a single answer (since, by default, a consensus would be reached with only 2 response options). Because this was an analysis of all respondents, an a priori power calculation was not performed. Associations between training and practice demographics and responses to clinical questions were investigated using chi-square analyses. All comparative analyses were two-tailed and used P = .05 as the threshold for statistical significance.

Results

Demographics

One hundred and twenty-five (29.9%) ASES members responded to the survey. Of the respondents, 71.2% reported at least 15 years of experience, and 71% performed more than 150 shoulder cases annually. Surgeons came from academic institutions (41.6%), private practice (24.8%), or mixed (33.6%). The majority of respondents were fellowship-trained in shoulder/elbow surgery (52.8%), while fewer had completed a sports-medicine fellowship (24.0%). For arthroscopic procedures, responses were nearly divided between those who preferred beach-chair positioning (47.2%) and those who preferred the lateral decubitus position (46.4%). The majority (70.4%) of respondents practiced in the United States and with a relatively even distribution among states and region. The remaining 29.6% of those surveyed practiced abroad.

Degree of Consensus Responses and Cases

Of the 25 survey questions, 6 questions were omitted from consensus calculations because these were designed for demographic categorization rather than professional opinion (questions 1-5, 8). Thirteen of the remaining 19 questions (68%) reached consensus response. All clinical case scenarios (5 of 5) reached consensus for selection of technique for the primary procedure; however, only 40% (2 of 5) of cases had a consensus in the revision setting.

In case 1, a young soccer player (noncontact athlete) with negligible bone loss, arthroscopic Bankart repair was recommended by 81.6% of respondents. In the event of revision surgery, only 22.4% recommended arthroscopic Bankart repair, and the remainder split between open Bankart repair with possible capsular shift (36%) or Latarjet procedure (32.8%).

In case 2, a college American football player (contact athlete) with negligible bone loss, arthroscopic Bankart repair was recommended by 56.8%. In the event of revision surgery, a majority of members (51.2%) suggested a Latarjet procedure.

In case 3, the weekend warrior with significant bone loss, most respondents recommended a Latarjet procedure for both primary (72.8%) and revision surgery (79.0%).

In case 4, a weekend warrior with multidirectional instability, 60% of respondents suggested arthroscopic Bankart repair, 21.6% recommended rotator interval closure, and 10.4% chose a capsular shift. As a revision procedure, there was less agreement, with a split between open Bankart repair (39.2%) and capsular shift only (39.2%).

In case 5, the weekend warrior with large engaging Hill-Sachs lesions, 60% of respondent selected a remplissage procedure. If revision was required, a Latarjet procedure was the choice of 48.8% of respondents (Table).

General Questions

For contact athletes, most respondents (87.2%) would allow return to play in the same season and recommended surgery after the end of the season. After surgical intervention, 56.8% prescribed 4 weeks of immobilization. When counseling a return to contact sports, 51.2% recommended waiting for 4 to 6 months.

The ASES members were divided on conservative management of instability injuries. Responses included immobilization in internal rotation (39.2%), no immobilization (39.2%), and external-rotation bracing (21.6%).

Finally, members thought the most important factor in choosing surgical technique was the patient’s pathology, then age; the least influential criteria was the patient’s sports participation.

Analysis of Training Demographics and Surgical Technique Preferences

Chi-square analyses demonstrated that respondents who completed a sports fellowship were more likely to do at least 50% of cases arthroscopically (odds ratio [OR], 15.3; P < .001) and were more likely to use the lateral decubitus position (OR, 2.8; P < .021). Furthermore, American respondents had a higher likelihood of having completed either a sports fellowship (OR, 12.8; P < .001) or a shoulder/elbow fellowship (OR, 4.6; P = .002) when compared with foreign respondents.

Discussion

In the absence of formal clinical practice guidelines, most surgeons formulate treatment strategy based upon a combination of experience and peer-reviewed evidence. The cohort analyzed in the current study was highly experienced, with more than 70% performing 150 shoulder cases annually and having more than 15 years of experience. We found a consensus response in 68% of questions and all primary surgical techniques for our shoulder instability scenarios. While expert consensus reported here is not equivalent to evidence-based clinical practice guidelines, it does provide important information to consider when treating anterior shoulder instability.

Specific responses to our case scenarios invite further reflection. Considering young (both noncontact and contact) athletes without bony pathology (cases 1 and 2, respectively), the ASES surgeons recommended arthroscopic Bankart repair for both. Randelli and associates5 found 71% of survey respondents recommended arthroscopic Bankart repair in a similar setting. It is interesting to note that consensus persisted regardless of the sport in which they engaged. Contact athletes have the highest rates of dislocation (up to 7 times higher incidence) compared with the general population.¹⁴ In addition, they have a higher recurrence rate after surgery.¹⁵ It should be noted, however, that although both cases reached consensus, the percentage of experts who recommended an arthroscopic procedure fell from 82% in the noncontact athlete to 57% in the contact athlete. This concurs with a recent review by Harris and Romeo,¹⁶ who recommended similar treatments for athletes without bony defects. In an older patient population with recurrent instability (case 3), responses varied more widely but still reached a consensus on primary surgical techniques. Respondents agreed that, even for patients with multidirectional instability, initial management should consist of arthroscopic capsulolabral repair. Overall, the agreement for arthroscopy for cases 1 through 3 mimics recent US practice patterns, showing 90% of stabilizations are being performed arthroscopically.¹⁷ Additionally, a recent meta-analysis by Harris and associates18 favored arthroscopic Bankart repair, showing no significant difference vs open stabilization even on long-term follow-up.

Glenoid bone loss is a difficult clinical scenario and that is reflected in this study’s findings. The literature suggests that arthroscopic Bankart repair, in this setting, is usually not sufficient and may result in a recurrence rate up to 75%, if bone loss greater than 20% is unaddressed.¹⁹ Our study supports this trend because ASES members recommended a Latarjet procedure when there is substantial bone loss.

While open Latarjet procedure was the consensus for dealing with glenoid bone loss, arthroscopic techniques were strongly favored for humeral head defects. This change in practice patterns results from the introduction of the arthroscopic remplissage technique.²⁰ Two recent systemic reviews have supported this technique, reporting good functional outcomes for engaging Hill-Sachs lesions.^21,22 Our study had similar agreement, with most respondents recommending remplissage for these patients.

This study found the lowest rates of expert consensus in the setting of revision surgery, likely caused, in part, by the paucity of available large cohort studies. This is a major void in the literature, and more studies are needed to help guide surgeons on the best techniques to deal with this difficult patient population.

Conservative bracing technique was 1 of the survey questions lacking a consensus response. Interestingly, 39% of members recommended no immobilization after an instability event. This contrasts with recent literature concerning the best position for bracing. We also found twice as many surgeons recommended internal rotation immobilization over external rotation. This is a subject of debate, with some studies stating improvement with external rotation immobilization,²³ while other studies reported no difference.²⁴ Overall, recommendations regarding type of immobilization are unclear, which will likely continue until larger studies can be performed.

The literature describing surgical trends in the treatment of shoulder instability is sparse and variable. With regard to other shoulder etiologies, only rotator cuff pathology has used expert consensus. Acevedo and colleagues¹³ reported agreement of ASES members surveyed regarding rotator cuff management. There was no consensus among surgeons in more than 50% of questions, despite AAOS published guidelines for rotator cuff treatment.²⁵ Despite the lack of guidelines for our topic, we found a consensus among respondents with 68% of survey questions.

To date, only 2 studies of shoulder instability management have elicited the opinion of experts in shoulder surgery. Chahal and associates⁶ surveyed 42 members of ASES and JOINTS (Joined Orthopaedic Initiatives for National Trials of the Shoulder) Canada on shoulder instability cases and found substantial agreement on diagnosis but little consensus regarding surgical technique. This lack of agreement on procedures differs from our findings and may be related to their complicated case scenarios that generated a wide array of treatment recommendations. Randelli and colleagues⁵ surveyed more than 1000 European Society of Sports Traumatology, Knee Surgery, and Arthroscopy members and reported similar agreement on arthroscopic Bankart repair in young male shoulder-dislocation patients, although no other instability scenarios were investigated. Our study is the first to report responses from expert shoulder surgeons on surgical-treatment strategies for an array of common shoulder instability pathologies.

This study had several limitations. First, while our study suffered from a low response rate (29.9%), it was similar to other published studies.^5,13 Second, because the case series included in the survey attempted to capture the most common instability scenarios, they were limited in their scope and failed to address additional etiologies or pathologic permutations. We believe, however, that a more comprehensive survey would have resulted in respondent fatigue and lowered the response rate. It is unlikely that any survey could capture all variables that come into play during clinical decision-making, and we sought to evaluate the most common shoulder instability scenarios. Third, 30% of respondents were from outside the United States, where the Latarjet procedure is much more popular. While this was not a majority, Latarjet’s regional preference may have decreased the consensus response in some scenarios if only the United States was included. Finally, there is inherent bias in a respondent pool that is heavily weighted to shoulder-surgery experts (ASES members) and does not consider the responses of the general orthopedic surgery community as have other studies.⁷

Conclusion

This study demonstrates that expert shoulder surgeons often agreed on shoulder-treatment principles for anterior shoulder instability. In the setting of primary repair, arthroscopic Bankart repair was favored in the absence of bony pathology, regardless of age (20 to 35 years) or nature of sport (contact versus noncontact). Latarjet procedures were favored in the setting of glenoid bone loss, and remplissage for an engaging Hill-Sachs lesion. Less agreement was observed for revision stabilization. It should be noted that, while consensus was often reached for our cases, there was a wide distribution of technical considerations and surgical preferences even among those who are fellowship-trained and high-volume surgeons, and who can be considered experts in the field of shoulder surgery.

Although recent clinical results of arthroscopic rotator cuff repair (RCR) have been encouraging, achieving anatomical healing of full-thickness rotator cuff tears remains a challenge.^1-4 Several factors influence rotator cuff healing after repair.^1,3-8 Patient-related factors include advanced patient age, tear size, tear chronicity, and amount of fatty infiltration.^1,3,5,6,8-10 Tension applied to the repair construct is a significant factor as well.^11,12

In the literature, limited consideration has been given to repair tension.¹³ The majority of studies have focused on other factors, mainly repair technique. Some surgeons advocate use of a double-row repair construct in which the rotator cuff tendon is pulled to the lateral margin of the footprint.^14-19 Double-row techniques, which include the transosseous-equivalent (TOE) construct, are biomechanically superior to other repairs.^20-26 Another purported benefit of double-row repair is more complete restoration and pressurization of the rotator cuff footprint.^21,24,27,28

Rotator cuff tears typically occur near the dysvascular region of the diseased musculotendinous unit, often leaving a stump of tissue attached to the tuberosity and ultimately a shortened tendon.²⁹ In addition, full-thickness tears often retract over time. Meyer and colleagues²⁹ recently demonstrated that this shortening is irreversible. Snyder³⁰ and Sostak and colleagues³¹ suggested that pulling a shortened, degenerative rotator cuff tendon to the lateral margin of the footprint results in increased tissue tension compared with that produced with a more medially based repair just off the articular margin. In our opinion, the possible increase in tension during a laterally based repair, whether single- or double-row, may place excessive strain on the diseased tissue as well as the surgical construct, potentially contributing to repair failure.

We conducted a study to evaluate the difference, if any, in tension applied to the rotator cuff tendon positioned at the medial versus lateral margin of the footprint during arthroscopic RCR. We hypothesized significantly more tension would be placed on the rotator cuff tendon when positioned at the lateral versus medial footprint.

Methods

After obtaining Institutional Review Board approval for this study, we collected data on a consecutive series of patients who underwent arthroscopic RCR performed by Dr. Getelman at a single institution. Only patients with primary full-thickness tears of the supraspinatus and/or infraspinatus were included. Exclusion criteria included revision rotator cuff surgeries, partial-thickness tears, concurrent subscapularis tears requiring anchor fixation, and any tears that could not be mobilized to the lateral footprint without interval slides or margin convergence. The 20 identified patients constituted the study group.

Demographic factors, including age and preoperative length of symptoms, were recorded after chart review. Magnetic resonance imaging (MRI) was performed for all patients before surgery and was retrospectively reviewed. Dr. Getelman assigned each patient a modified Goutallier score, based on MRI, to assess for fatty infiltration/atrophy.³² Each patient was placed in the lateral decubitus position with the operative arm in balanced suspension at 70° of abduction. Standard glenohumeral and subacromial diagnostic arthroscopy was performed. The rotator cuff tear was gently debrided back to a healthy-appearing margin in preparation for repair. The tear was then measured in the anterior-posterior (A-P) and medial-lateral (M-L) planes using a premeasured, marked suture, as previously described.33 Complete bursal and articular-sided releases were performed to allow for appropriate mobilization of the tendon. The tear was classified as crescent-shaped, U-shaped, or L-shaped.

Viewing from the posterior portal, the surgeon inserted a tissue grasper through the lateral portal. The tendon was grasped at multiple points along its edge, anterior to posterior, and was translated laterally to assess its reducibility; the apex of the tear correlated with the point of maximal excursion and coverage of the footprint. Once confirmed, the rotator cuff tear apex was clamped with a tissue grasper. After placement in a sterile arthroscopic camera sleeve (DeRoyal camera drape with perforated tip), a calibrated digital weigh scale (American Weigh Scales model H22 portable electronic hanging scale, with accuracy of 0.01 lb) was attached to the tissue grasper with an S-hook (Figure 1). The tendon edge was first translated about 3 mm lateral to the articular margin (the medial footprint position), and tension was recorded (Figures 2A, 2B). After a 1-minute relaxation period, the tendon edge was translated to the lateral edge of the rotator cuff footprint (the lateral footprint position), and tension was recorded again (Figures 2C, 2D). A medially based single-row RCR with triple-loaded sutures and bone marrow vents placed in the lateral tuberosity was then completed, regardless of tension, tear size, or tear morphology.³¹ Typically, 1 anchor was used for every 10 to 15 mm of A-P tear length.

SAS software was used for statistical analysis, the Wilcoxon signed rank test for continuous or ordinal data comparisons between paired groups, and the Mann-Whitney test for continuous or ordinal data comparisons between independent, unmatched groups. One-way analysis of variance (ANOVA) was used to compare means among the 3 groups of morphology subtypes. Linear regression was performed to assess the simultaneous relationship between potential predictors (age, sex, length of symptoms, Goutallier score, tear size) and medial or lateral tension, where medial tension was included as an additional potential predictor for lateral tension. Restricted cubic splines were fit to assess linearity. Predictors were retained in multivariate regression using backward variable retention. Because of inadequate sample size, additivity was assumed except for sex. Statistical significance was set at P < .05.

Results

Of the 20 rotator cuff tears evaluated (Table 1), 13 were crescent-shaped, 5 were U-shaped, and 2 were L-shaped. Mean (SD) A-P tear size was 17.7 (5.8) mm, and mean (SD) M-L tear size was 19.1 (8.6) mm. Mean age of the 20 patients (15 men, 5 women) was 57.9 years (range, 44-72 years). Mean (SD) length of symptoms was 12.9 (12.4) months (range, 3-48 months). Mean (SD) modified Goutallier score was 1.4 (0.7; range, 0-3).

Mean (SD) rotator cuff tension for all tears approximated to the medial footprint was 0.41 (0.33) pound, and mean (SD) cuff tension for all tears approximated to the lateral footprint was 2.21 (1.20) pounds—representing a 5.4-fold difference (P < .0001).

No statistically significant differences were detected in the ANOVA comparing tensions at medial and lateral positions among tear morphologic subtypes (all Ps >.05).

Subgroup analysis (Table 2) was performed for smaller (≤20 mm A-P) and larger (>20 mm A-P) tears. For smaller tears, mean (SD) tension was 0.27 (0.24) pound applied with the cuff tendon pulled to the medial footprint and 2.06 (1.06) pounds applied with the tendon pulled to the lateral footprint—a 7.6-fold difference (P < .0018). For larger tears, mean (SD) tension was 0.58 (0.37) pound applied with the tendon pulled to the medial footprint and 2.38 (1.4) pounds applied with the tendon pulled to the lateral footprint—a 4.1-fold difference (P < .005).

A statistically significant difference in tensions was found between small and large cuff tears positioned at the medial footprint (0.27 vs 0.58 lb; P = .0367); no difference was found between groups with the tendon at the lateral footprint (2.06 vs 2.38 lb; P = .284).

Univariate and multivariate analyses were performed using linear regression analysis (Table 3). During univariate analysis for medial footprint position, A-P tear size and Goutallier score both positively correlated with increasing tension; for lateral footprint position, no factors statistically correlated with lateral tension, though there was a positive trend for medial tension and female sex. During multivariate analysis for medial footprint position, only A-P tear size positively correlated with increasing tension; for lateral footprint position, both age (in nonlinear fashion as function of age + age²) and medial tension positively correlated with increasing tension.

Discussion

Our results indicated that significantly more tension is placed on the torn rotator cuff tendon when it is reduced across the footprint from a medial to a more lateral position in vivo. More tension was required for all tears to be reduced to the lateral footprint compared with the medial footprint. As expected, compared with smaller tears, larger tears required significantly more tension in order to be reduced to the medial footprint. Interestingly, no statistical difference was found between tensions required to reduce either small or large tears to the lateral footprint, which suggests that, regardless of tear size, more force must be applied to reduce the torn tendon to the lateral footprint compared with the medial footprint.

Hersche and Gerber³⁴ were the first to report rotator cuff tension measurements in vivo. Although their study did not specifically compare cuff tensions reducing the tear to the medial versus lateral footprint, it did examine tension at displacement of 10 and 20 mm. Tension increased from 27 N to 60 N, correlating with a 2.2-fold difference between the 2 distances. Domb and colleagues35 also compared in vivo rotator cuff tension differences between the medial footprint and the lateral footprint in 4 patients. Mean tension applied to the cuff during reduction to the articular margin was 27 N, or 6 pounds. Mean tension needed to reduce the cuff to the lateral tuberosity was 76 N, or 17 pounds, for a 2.8-fold difference. Tears were not measured but were described as massive and retracted.

Although repair tension has long been recognized as a crucial factor in RCR healing, little clinical research has focused on the effects of excess tension. Davidson and Rivenburgh11 prospectively followed the clinical outcomes of 67 consecutive cuff repairs after intraoperative tension measurement and found that high-tension repairs (>8 lb) had significantly lower clinical outcome measures. However, the authors did not report on correlations with radiologic healing and stated, “Functional outcome is inversely proportional to rotator cuff repair tension.” Further study of the in vivo effects of increased tension on clinical and radiologic outcomes is needed.

Several animal studies have been conducted on the effects of tension on RCRs. Gerber and colleagues36 reported that the force needed to produce 1 cm of sheep supraspinatus tendon excursion increased 7-fold, from 6.8 N to 47.8 N, after 40 weeks of tendon tear. Coleman and colleagues³⁷ compared the modulus of elasticity in sheep supraspinatus tendon after 6 weeks and 18 weeks of detachment and reported increases of 60% and 70%, respectively. Gimbel and colleagues³⁸ showed that, in a rat model, “repair tension rapidly increased initially after injury followed by a progressive, but less dramatic, increase with additional time.” Of note, we did not identify any correlation between chronicity of symptoms and the tension needed to reduce the tendon medially or to a more lateral position on the footprint.

In acute tears, the cuff tissue is more compliant and mobile and can be pulled laterally across its anatomical footprint with minimal tension.³⁹ In contrast, cuff tissue in the more commonly encountered chronic tear is less compliant and is not mobile enough to be pulled to the lateral margin of the footprint without added stress.^30,34,35 In large, acute tears in which there are minimal tissue degeneration and retraction, a laterally based footprint-restoring technique may be performed with minimal tension. This technique may have advantages over a medially based repair. In the literature, more attention needs to be directed toward the biomechanics and biology of chronic rotator cuff tears, as these are more commonly encountered.

Almost all of the prospective studies that have compared single- and double-row RCR have found no significant differences in MRI healing rates or clinical results at follow-up up to 2 years.^14,16,40-45 Detailed analysis of the surgical techniques used in all these studies revealed that the rotator cuff tendons were repaired back to the lateral footprint in both the single- and double-row constructs.^14,16,40-45 Although no clinical studies have compared medially and laterally based single-row repairs, our data suggest that medially based repairs have lower tensions and therefore should not be considered equivalent. Sostak and colleagues³¹ and Murray and colleagues⁴⁶ have shown that a medially based single-row RCR can achieve excellent clinical and anatomical results, likely partly because of the lower tension applied to the torn cuff tissue.^31,46 Studies are needed to compare medially and laterally based repairs, including single- and double-row repairs.

The vast majority of recent research has aimed to counteract construct tension with stronger biomechanical constructs.^20-26 Surgeons have also aimed to improve biological healing by pulling the tendon laterally across the footprint to achieve complete footprint coverage, ultimately increasing the surface area for tendon–bone healing. This has led to the development of various double-row repair techniques, in which the cuff tendon is pulled to the lateral margin of its footprint. One row of anchors is placed in the medial aspect of the footprint, while a second is placed in the lateral aspect; the cuff is reduced and compressed to the tuberosity with various suture configurations. The TOE technique was developed to improve pressurization of the cuff tendon across the footprint by linking the 2 rows with bridging sutures. In doing so, however, the potentially deleterious effects of increased tension introduced by pulling the tendon laterally may have been overlooked. Nevertheless, the biomechanics and stress distribution likely differ between single-row repair and TOE repairs, and direct comparisons cannot be made at this time. The medial row of a double-row or TOE construct may stress-shield or “unload” the more lateral tissue. Studies are needed in order to better understand the tension differential and stress distribution of various double-row constructs.

Recognizing tear morphology is crucial in maximizing chances of healing after cuff repair. For example, a crescent-shaped tear is reduced to the tuberosity with direct lateral translation of the apex of the tear, which is also the deepest or most displaced part of the tear. On the other hand, reducing an L- or reverse L-shaped tear to the tuberosity is not as direct; reducing the deepest or most displaced part of the tear would lead to overreduction and overtensioning of the tendon. However, often the exact “elbow” of the tear is not obvious and appears more rounded; therefore, it is crucial for the surgeon to examine the mobility of the torn tendon along its entire length to minimize tension. Study is needed to assess tension along the entire length of the tear for different tear morphologies and sizes.

Although our results showed that increased tension was needed to reduce a torn tendon to its lateral footprint, no study has indicated exactly how much is “too much” tension. As stated earlier, use of stronger biomechanical constructs, including TOE constructs, may overcome the increased tension associated with laterally based repairs. In addition, laterally based repairs, either single- or double-row, may be best suited for tears with lower tension, whereas medially based repairs may be best suited for higher tension tears. It is also possible that the difference in tensions noted in this study is not significant enough to have a clinical impact on choice of construct or on anatomical healing. We need studies that correlate anatomical healing rates with repair tension in order to better guide surgeons on when to use a medially or laterally based repair.

Other possible effects of increased tension associated with laterally based repairs, including beneficial effects, must be considered as well. Viscoelastic properties of human rotator cuff tendon may dissipate increased tension over time through a variety of mechanisms. Stress relaxation, gap formation, creep, and the hysteresis effect, all associated with cyclical loading in the early healing period, may lead to dissipation of force over time.^47,48 These more complex biomechanical properties of RCR constructs are yet to be clearly defined.

This study had several weaknesses. Its data represent a static measurement of time-zero rotator cuff tension, which greatly simplifies the biomechanics of the torn rotator cuff and repair construct as well as changes that occur with healing. During cuff repair, forces typically are distributed through several fixation points in stepwise process and are not focused on a single point of tissue with a grasper. Therefore, the findings of this study may not directly correlate with medially versus laterally based repairs in vivo. Furthermore, as this is a time-zero measurement, we could not determine whether the tension differential between the 2 repair positions remained static over time. Current literature suggests that muscle atrophy, fatty infiltration, and loss of elasticity of the musculotendinous unit are relatively irreversible.^35,37,49 In addition, determining the precise apex of a cuff tear can be difficult, so error may have been introduced during this process. Last, although placement of the cuff tissue at the medial or lateral footprint position was based on visual estimation by an experienced and skilled arthroscopist, error may have been introduced based on this imprecise technique.

Conclusion

This study demonstrated a significant, 5.4-fold increase in in vivo time-zero rotator cuff tension with the tendon edge reduced to the lateral footprint rather than the medial footprint.

Although recent clinical results of arthroscopic rotator cuff repair (RCR) have been encouraging, achieving anatomical healing of full-thickness rotator cuff tears remains a challenge.^1-4 Several factors influence rotator cuff healing after repair.^1,3-8 Patient-related factors include advanced patient age, tear size, tear chronicity, and amount of fatty infiltration.^1,3,5,6,8-10 Tension applied to the repair construct is a significant factor as well.^11,12

In the literature, limited consideration has been given to repair tension.¹³ The majority of studies have focused on other factors, mainly repair technique. Some surgeons advocate use of a double-row repair construct in which the rotator cuff tendon is pulled to the lateral margin of the footprint.^14-19 Double-row techniques, which include the transosseous-equivalent (TOE) construct, are biomechanically superior to other repairs.^20-26 Another purported benefit of double-row repair is more complete restoration and pressurization of the rotator cuff footprint.^21,24,27,28

Rotator cuff tears typically occur near the dysvascular region of the diseased musculotendinous unit, often leaving a stump of tissue attached to the tuberosity and ultimately a shortened tendon.²⁹ In addition, full-thickness tears often retract over time. Meyer and colleagues²⁹ recently demonstrated that this shortening is irreversible. Snyder³⁰ and Sostak and colleagues³¹ suggested that pulling a shortened, degenerative rotator cuff tendon to the lateral margin of the footprint results in increased tissue tension compared with that produced with a more medially based repair just off the articular margin. In our opinion, the possible increase in tension during a laterally based repair, whether single- or double-row, may place excessive strain on the diseased tissue as well as the surgical construct, potentially contributing to repair failure.

We conducted a study to evaluate the difference, if any, in tension applied to the rotator cuff tendon positioned at the medial versus lateral margin of the footprint during arthroscopic RCR. We hypothesized significantly more tension would be placed on the rotator cuff tendon when positioned at the lateral versus medial footprint.

Methods

After obtaining Institutional Review Board approval for this study, we collected data on a consecutive series of patients who underwent arthroscopic RCR performed by Dr. Getelman at a single institution. Only patients with primary full-thickness tears of the supraspinatus and/or infraspinatus were included. Exclusion criteria included revision rotator cuff surgeries, partial-thickness tears, concurrent subscapularis tears requiring anchor fixation, and any tears that could not be mobilized to the lateral footprint without interval slides or margin convergence. The 20 identified patients constituted the study group.

Demographic factors, including age and preoperative length of symptoms, were recorded after chart review. Magnetic resonance imaging (MRI) was performed for all patients before surgery and was retrospectively reviewed. Dr. Getelman assigned each patient a modified Goutallier score, based on MRI, to assess for fatty infiltration/atrophy.³² Each patient was placed in the lateral decubitus position with the operative arm in balanced suspension at 70° of abduction. Standard glenohumeral and subacromial diagnostic arthroscopy was performed. The rotator cuff tear was gently debrided back to a healthy-appearing margin in preparation for repair. The tear was then measured in the anterior-posterior (A-P) and medial-lateral (M-L) planes using a premeasured, marked suture, as previously described.33 Complete bursal and articular-sided releases were performed to allow for appropriate mobilization of the tendon. The tear was classified as crescent-shaped, U-shaped, or L-shaped.

Viewing from the posterior portal, the surgeon inserted a tissue grasper through the lateral portal. The tendon was grasped at multiple points along its edge, anterior to posterior, and was translated laterally to assess its reducibility; the apex of the tear correlated with the point of maximal excursion and coverage of the footprint. Once confirmed, the rotator cuff tear apex was clamped with a tissue grasper. After placement in a sterile arthroscopic camera sleeve (DeRoyal camera drape with perforated tip), a calibrated digital weigh scale (American Weigh Scales model H22 portable electronic hanging scale, with accuracy of 0.01 lb) was attached to the tissue grasper with an S-hook (Figure 1). The tendon edge was first translated about 3 mm lateral to the articular margin (the medial footprint position), and tension was recorded (Figures 2A, 2B). After a 1-minute relaxation period, the tendon edge was translated to the lateral edge of the rotator cuff footprint (the lateral footprint position), and tension was recorded again (Figures 2C, 2D). A medially based single-row RCR with triple-loaded sutures and bone marrow vents placed in the lateral tuberosity was then completed, regardless of tension, tear size, or tear morphology.³¹ Typically, 1 anchor was used for every 10 to 15 mm of A-P tear length.

SAS software was used for statistical analysis, the Wilcoxon signed rank test for continuous or ordinal data comparisons between paired groups, and the Mann-Whitney test for continuous or ordinal data comparisons between independent, unmatched groups. One-way analysis of variance (ANOVA) was used to compare means among the 3 groups of morphology subtypes. Linear regression was performed to assess the simultaneous relationship between potential predictors (age, sex, length of symptoms, Goutallier score, tear size) and medial or lateral tension, where medial tension was included as an additional potential predictor for lateral tension. Restricted cubic splines were fit to assess linearity. Predictors were retained in multivariate regression using backward variable retention. Because of inadequate sample size, additivity was assumed except for sex. Statistical significance was set at P < .05.

Results

Of the 20 rotator cuff tears evaluated (Table 1), 13 were crescent-shaped, 5 were U-shaped, and 2 were L-shaped. Mean (SD) A-P tear size was 17.7 (5.8) mm, and mean (SD) M-L tear size was 19.1 (8.6) mm. Mean age of the 20 patients (15 men, 5 women) was 57.9 years (range, 44-72 years). Mean (SD) length of symptoms was 12.9 (12.4) months (range, 3-48 months). Mean (SD) modified Goutallier score was 1.4 (0.7; range, 0-3).

Mean (SD) rotator cuff tension for all tears approximated to the medial footprint was 0.41 (0.33) pound, and mean (SD) cuff tension for all tears approximated to the lateral footprint was 2.21 (1.20) pounds—representing a 5.4-fold difference (P < .0001).

No statistically significant differences were detected in the ANOVA comparing tensions at medial and lateral positions among tear morphologic subtypes (all Ps >.05).

Subgroup analysis (Table 2) was performed for smaller (≤20 mm A-P) and larger (>20 mm A-P) tears. For smaller tears, mean (SD) tension was 0.27 (0.24) pound applied with the cuff tendon pulled to the medial footprint and 2.06 (1.06) pounds applied with the tendon pulled to the lateral footprint—a 7.6-fold difference (P < .0018). For larger tears, mean (SD) tension was 0.58 (0.37) pound applied with the tendon pulled to the medial footprint and 2.38 (1.4) pounds applied with the tendon pulled to the lateral footprint—a 4.1-fold difference (P < .005).

A statistically significant difference in tensions was found between small and large cuff tears positioned at the medial footprint (0.27 vs 0.58 lb; P = .0367); no difference was found between groups with the tendon at the lateral footprint (2.06 vs 2.38 lb; P = .284).

Univariate and multivariate analyses were performed using linear regression analysis (Table 3). During univariate analysis for medial footprint position, A-P tear size and Goutallier score both positively correlated with increasing tension; for lateral footprint position, no factors statistically correlated with lateral tension, though there was a positive trend for medial tension and female sex. During multivariate analysis for medial footprint position, only A-P tear size positively correlated with increasing tension; for lateral footprint position, both age (in nonlinear fashion as function of age + age²) and medial tension positively correlated with increasing tension.

Discussion

Our results indicated that significantly more tension is placed on the torn rotator cuff tendon when it is reduced across the footprint from a medial to a more lateral position in vivo. More tension was required for all tears to be reduced to the lateral footprint compared with the medial footprint. As expected, compared with smaller tears, larger tears required significantly more tension in order to be reduced to the medial footprint. Interestingly, no statistical difference was found between tensions required to reduce either small or large tears to the lateral footprint, which suggests that, regardless of tear size, more force must be applied to reduce the torn tendon to the lateral footprint compared with the medial footprint.

Hersche and Gerber³⁴ were the first to report rotator cuff tension measurements in vivo. Although their study did not specifically compare cuff tensions reducing the tear to the medial versus lateral footprint, it did examine tension at displacement of 10 and 20 mm. Tension increased from 27 N to 60 N, correlating with a 2.2-fold difference between the 2 distances. Domb and colleagues35 also compared in vivo rotator cuff tension differences between the medial footprint and the lateral footprint in 4 patients. Mean tension applied to the cuff during reduction to the articular margin was 27 N, or 6 pounds. Mean tension needed to reduce the cuff to the lateral tuberosity was 76 N, or 17 pounds, for a 2.8-fold difference. Tears were not measured but were described as massive and retracted.

Although repair tension has long been recognized as a crucial factor in RCR healing, little clinical research has focused on the effects of excess tension. Davidson and Rivenburgh11 prospectively followed the clinical outcomes of 67 consecutive cuff repairs after intraoperative tension measurement and found that high-tension repairs (>8 lb) had significantly lower clinical outcome measures. However, the authors did not report on correlations with radiologic healing and stated, “Functional outcome is inversely proportional to rotator cuff repair tension.” Further study of the in vivo effects of increased tension on clinical and radiologic outcomes is needed.

Several animal studies have been conducted on the effects of tension on RCRs. Gerber and colleagues36 reported that the force needed to produce 1 cm of sheep supraspinatus tendon excursion increased 7-fold, from 6.8 N to 47.8 N, after 40 weeks of tendon tear. Coleman and colleagues³⁷ compared the modulus of elasticity in sheep supraspinatus tendon after 6 weeks and 18 weeks of detachment and reported increases of 60% and 70%, respectively. Gimbel and colleagues³⁸ showed that, in a rat model, “repair tension rapidly increased initially after injury followed by a progressive, but less dramatic, increase with additional time.” Of note, we did not identify any correlation between chronicity of symptoms and the tension needed to reduce the tendon medially or to a more lateral position on the footprint.

In acute tears, the cuff tissue is more compliant and mobile and can be pulled laterally across its anatomical footprint with minimal tension.³⁹ In contrast, cuff tissue in the more commonly encountered chronic tear is less compliant and is not mobile enough to be pulled to the lateral margin of the footprint without added stress.^30,34,35 In large, acute tears in which there are minimal tissue degeneration and retraction, a laterally based footprint-restoring technique may be performed with minimal tension. This technique may have advantages over a medially based repair. In the literature, more attention needs to be directed toward the biomechanics and biology of chronic rotator cuff tears, as these are more commonly encountered.

Almost all of the prospective studies that have compared single- and double-row RCR have found no significant differences in MRI healing rates or clinical results at follow-up up to 2 years.^14,16,40-45 Detailed analysis of the surgical techniques used in all these studies revealed that the rotator cuff tendons were repaired back to the lateral footprint in both the single- and double-row constructs.^14,16,40-45 Although no clinical studies have compared medially and laterally based single-row repairs, our data suggest that medially based repairs have lower tensions and therefore should not be considered equivalent. Sostak and colleagues³¹ and Murray and colleagues⁴⁶ have shown that a medially based single-row RCR can achieve excellent clinical and anatomical results, likely partly because of the lower tension applied to the torn cuff tissue.^31,46 Studies are needed to compare medially and laterally based repairs, including single- and double-row repairs.

The vast majority of recent research has aimed to counteract construct tension with stronger biomechanical constructs.^20-26 Surgeons have also aimed to improve biological healing by pulling the tendon laterally across the footprint to achieve complete footprint coverage, ultimately increasing the surface area for tendon–bone healing. This has led to the development of various double-row repair techniques, in which the cuff tendon is pulled to the lateral margin of its footprint. One row of anchors is placed in the medial aspect of the footprint, while a second is placed in the lateral aspect; the cuff is reduced and compressed to the tuberosity with various suture configurations. The TOE technique was developed to improve pressurization of the cuff tendon across the footprint by linking the 2 rows with bridging sutures. In doing so, however, the potentially deleterious effects of increased tension introduced by pulling the tendon laterally may have been overlooked. Nevertheless, the biomechanics and stress distribution likely differ between single-row repair and TOE repairs, and direct comparisons cannot be made at this time. The medial row of a double-row or TOE construct may stress-shield or “unload” the more lateral tissue. Studies are needed in order to better understand the tension differential and stress distribution of various double-row constructs.

Recognizing tear morphology is crucial in maximizing chances of healing after cuff repair. For example, a crescent-shaped tear is reduced to the tuberosity with direct lateral translation of the apex of the tear, which is also the deepest or most displaced part of the tear. On the other hand, reducing an L- or reverse L-shaped tear to the tuberosity is not as direct; reducing the deepest or most displaced part of the tear would lead to overreduction and overtensioning of the tendon. However, often the exact “elbow” of the tear is not obvious and appears more rounded; therefore, it is crucial for the surgeon to examine the mobility of the torn tendon along its entire length to minimize tension. Study is needed to assess tension along the entire length of the tear for different tear morphologies and sizes.

Although our results showed that increased tension was needed to reduce a torn tendon to its lateral footprint, no study has indicated exactly how much is “too much” tension. As stated earlier, use of stronger biomechanical constructs, including TOE constructs, may overcome the increased tension associated with laterally based repairs. In addition, laterally based repairs, either single- or double-row, may be best suited for tears with lower tension, whereas medially based repairs may be best suited for higher tension tears. It is also possible that the difference in tensions noted in this study is not significant enough to have a clinical impact on choice of construct or on anatomical healing. We need studies that correlate anatomical healing rates with repair tension in order to better guide surgeons on when to use a medially or laterally based repair.

Other possible effects of increased tension associated with laterally based repairs, including beneficial effects, must be considered as well. Viscoelastic properties of human rotator cuff tendon may dissipate increased tension over time through a variety of mechanisms. Stress relaxation, gap formation, creep, and the hysteresis effect, all associated with cyclical loading in the early healing period, may lead to dissipation of force over time.^47,48 These more complex biomechanical properties of RCR constructs are yet to be clearly defined.

This study had several weaknesses. Its data represent a static measurement of time-zero rotator cuff tension, which greatly simplifies the biomechanics of the torn rotator cuff and repair construct as well as changes that occur with healing. During cuff repair, forces typically are distributed through several fixation points in stepwise process and are not focused on a single point of tissue with a grasper. Therefore, the findings of this study may not directly correlate with medially versus laterally based repairs in vivo. Furthermore, as this is a time-zero measurement, we could not determine whether the tension differential between the 2 repair positions remained static over time. Current literature suggests that muscle atrophy, fatty infiltration, and loss of elasticity of the musculotendinous unit are relatively irreversible.^35,37,49 In addition, determining the precise apex of a cuff tear can be difficult, so error may have been introduced during this process. Last, although placement of the cuff tissue at the medial or lateral footprint position was based on visual estimation by an experienced and skilled arthroscopist, error may have been introduced based on this imprecise technique.

Conclusion

This study demonstrated a significant, 5.4-fold increase in in vivo time-zero rotator cuff tension with the tendon edge reduced to the lateral footprint rather than the medial footprint.

Although recent clinical results of arthroscopic rotator cuff repair (RCR) have been encouraging, achieving anatomical healing of full-thickness rotator cuff tears remains a challenge.^1-4 Several factors influence rotator cuff healing after repair.^1,3-8 Patient-related factors include advanced patient age, tear size, tear chronicity, and amount of fatty infiltration.^1,3,5,6,8-10 Tension applied to the repair construct is a significant factor as well.^11,12

In the literature, limited consideration has been given to repair tension.¹³ The majority of studies have focused on other factors, mainly repair technique. Some surgeons advocate use of a double-row repair construct in which the rotator cuff tendon is pulled to the lateral margin of the footprint.^14-19 Double-row techniques, which include the transosseous-equivalent (TOE) construct, are biomechanically superior to other repairs.^20-26 Another purported benefit of double-row repair is more complete restoration and pressurization of the rotator cuff footprint.^21,24,27,28

Rotator cuff tears typically occur near the dysvascular region of the diseased musculotendinous unit, often leaving a stump of tissue attached to the tuberosity and ultimately a shortened tendon.²⁹ In addition, full-thickness tears often retract over time. Meyer and colleagues²⁹ recently demonstrated that this shortening is irreversible. Snyder³⁰ and Sostak and colleagues³¹ suggested that pulling a shortened, degenerative rotator cuff tendon to the lateral margin of the footprint results in increased tissue tension compared with that produced with a more medially based repair just off the articular margin. In our opinion, the possible increase in tension during a laterally based repair, whether single- or double-row, may place excessive strain on the diseased tissue as well as the surgical construct, potentially contributing to repair failure.

We conducted a study to evaluate the difference, if any, in tension applied to the rotator cuff tendon positioned at the medial versus lateral margin of the footprint during arthroscopic RCR. We hypothesized significantly more tension would be placed on the rotator cuff tendon when positioned at the lateral versus medial footprint.

Methods

After obtaining Institutional Review Board approval for this study, we collected data on a consecutive series of patients who underwent arthroscopic RCR performed by Dr. Getelman at a single institution. Only patients with primary full-thickness tears of the supraspinatus and/or infraspinatus were included. Exclusion criteria included revision rotator cuff surgeries, partial-thickness tears, concurrent subscapularis tears requiring anchor fixation, and any tears that could not be mobilized to the lateral footprint without interval slides or margin convergence. The 20 identified patients constituted the study group.

Demographic factors, including age and preoperative length of symptoms, were recorded after chart review. Magnetic resonance imaging (MRI) was performed for all patients before surgery and was retrospectively reviewed. Dr. Getelman assigned each patient a modified Goutallier score, based on MRI, to assess for fatty infiltration/atrophy.³² Each patient was placed in the lateral decubitus position with the operative arm in balanced suspension at 70° of abduction. Standard glenohumeral and subacromial diagnostic arthroscopy was performed. The rotator cuff tear was gently debrided back to a healthy-appearing margin in preparation for repair. The tear was then measured in the anterior-posterior (A-P) and medial-lateral (M-L) planes using a premeasured, marked suture, as previously described.33 Complete bursal and articular-sided releases were performed to allow for appropriate mobilization of the tendon. The tear was classified as crescent-shaped, U-shaped, or L-shaped.

Viewing from the posterior portal, the surgeon inserted a tissue grasper through the lateral portal. The tendon was grasped at multiple points along its edge, anterior to posterior, and was translated laterally to assess its reducibility; the apex of the tear correlated with the point of maximal excursion and coverage of the footprint. Once confirmed, the rotator cuff tear apex was clamped with a tissue grasper. After placement in a sterile arthroscopic camera sleeve (DeRoyal camera drape with perforated tip), a calibrated digital weigh scale (American Weigh Scales model H22 portable electronic hanging scale, with accuracy of 0.01 lb) was attached to the tissue grasper with an S-hook (Figure 1). The tendon edge was first translated about 3 mm lateral to the articular margin (the medial footprint position), and tension was recorded (Figures 2A, 2B). After a 1-minute relaxation period, the tendon edge was translated to the lateral edge of the rotator cuff footprint (the lateral footprint position), and tension was recorded again (Figures 2C, 2D). A medially based single-row RCR with triple-loaded sutures and bone marrow vents placed in the lateral tuberosity was then completed, regardless of tension, tear size, or tear morphology.³¹ Typically, 1 anchor was used for every 10 to 15 mm of A-P tear length.

SAS software was used for statistical analysis, the Wilcoxon signed rank test for continuous or ordinal data comparisons between paired groups, and the Mann-Whitney test for continuous or ordinal data comparisons between independent, unmatched groups. One-way analysis of variance (ANOVA) was used to compare means among the 3 groups of morphology subtypes. Linear regression was performed to assess the simultaneous relationship between potential predictors (age, sex, length of symptoms, Goutallier score, tear size) and medial or lateral tension, where medial tension was included as an additional potential predictor for lateral tension. Restricted cubic splines were fit to assess linearity. Predictors were retained in multivariate regression using backward variable retention. Because of inadequate sample size, additivity was assumed except for sex. Statistical significance was set at P < .05.

Results

Of the 20 rotator cuff tears evaluated (Table 1), 13 were crescent-shaped, 5 were U-shaped, and 2 were L-shaped. Mean (SD) A-P tear size was 17.7 (5.8) mm, and mean (SD) M-L tear size was 19.1 (8.6) mm. Mean age of the 20 patients (15 men, 5 women) was 57.9 years (range, 44-72 years). Mean (SD) length of symptoms was 12.9 (12.4) months (range, 3-48 months). Mean (SD) modified Goutallier score was 1.4 (0.7; range, 0-3).

Mean (SD) rotator cuff tension for all tears approximated to the medial footprint was 0.41 (0.33) pound, and mean (SD) cuff tension for all tears approximated to the lateral footprint was 2.21 (1.20) pounds—representing a 5.4-fold difference (P < .0001).

No statistically significant differences were detected in the ANOVA comparing tensions at medial and lateral positions among tear morphologic subtypes (all Ps >.05).

Subgroup analysis (Table 2) was performed for smaller (≤20 mm A-P) and larger (>20 mm A-P) tears. For smaller tears, mean (SD) tension was 0.27 (0.24) pound applied with the cuff tendon pulled to the medial footprint and 2.06 (1.06) pounds applied with the tendon pulled to the lateral footprint—a 7.6-fold difference (P < .0018). For larger tears, mean (SD) tension was 0.58 (0.37) pound applied with the tendon pulled to the medial footprint and 2.38 (1.4) pounds applied with the tendon pulled to the lateral footprint—a 4.1-fold difference (P < .005).

A statistically significant difference in tensions was found between small and large cuff tears positioned at the medial footprint (0.27 vs 0.58 lb; P = .0367); no difference was found between groups with the tendon at the lateral footprint (2.06 vs 2.38 lb; P = .284).

Univariate and multivariate analyses were performed using linear regression analysis (Table 3). During univariate analysis for medial footprint position, A-P tear size and Goutallier score both positively correlated with increasing tension; for lateral footprint position, no factors statistically correlated with lateral tension, though there was a positive trend for medial tension and female sex. During multivariate analysis for medial footprint position, only A-P tear size positively correlated with increasing tension; for lateral footprint position, both age (in nonlinear fashion as function of age + age²) and medial tension positively correlated with increasing tension.

Discussion

Our results indicated that significantly more tension is placed on the torn rotator cuff tendon when it is reduced across the footprint from a medial to a more lateral position in vivo. More tension was required for all tears to be reduced to the lateral footprint compared with the medial footprint. As expected, compared with smaller tears, larger tears required significantly more tension in order to be reduced to the medial footprint. Interestingly, no statistical difference was found between tensions required to reduce either small or large tears to the lateral footprint, which suggests that, regardless of tear size, more force must be applied to reduce the torn tendon to the lateral footprint compared with the medial footprint.

Hersche and Gerber³⁴ were the first to report rotator cuff tension measurements in vivo. Although their study did not specifically compare cuff tensions reducing the tear to the medial versus lateral footprint, it did examine tension at displacement of 10 and 20 mm. Tension increased from 27 N to 60 N, correlating with a 2.2-fold difference between the 2 distances. Domb and colleagues35 also compared in vivo rotator cuff tension differences between the medial footprint and the lateral footprint in 4 patients. Mean tension applied to the cuff during reduction to the articular margin was 27 N, or 6 pounds. Mean tension needed to reduce the cuff to the lateral tuberosity was 76 N, or 17 pounds, for a 2.8-fold difference. Tears were not measured but were described as massive and retracted.

Although repair tension has long been recognized as a crucial factor in RCR healing, little clinical research has focused on the effects of excess tension. Davidson and Rivenburgh11 prospectively followed the clinical outcomes of 67 consecutive cuff repairs after intraoperative tension measurement and found that high-tension repairs (>8 lb) had significantly lower clinical outcome measures. However, the authors did not report on correlations with radiologic healing and stated, “Functional outcome is inversely proportional to rotator cuff repair tension.” Further study of the in vivo effects of increased tension on clinical and radiologic outcomes is needed.

Several animal studies have been conducted on the effects of tension on RCRs. Gerber and colleagues36 reported that the force needed to produce 1 cm of sheep supraspinatus tendon excursion increased 7-fold, from 6.8 N to 47.8 N, after 40 weeks of tendon tear. Coleman and colleagues³⁷ compared the modulus of elasticity in sheep supraspinatus tendon after 6 weeks and 18 weeks of detachment and reported increases of 60% and 70%, respectively. Gimbel and colleagues³⁸ showed that, in a rat model, “repair tension rapidly increased initially after injury followed by a progressive, but less dramatic, increase with additional time.” Of note, we did not identify any correlation between chronicity of symptoms and the tension needed to reduce the tendon medially or to a more lateral position on the footprint.

In acute tears, the cuff tissue is more compliant and mobile and can be pulled laterally across its anatomical footprint with minimal tension.³⁹ In contrast, cuff tissue in the more commonly encountered chronic tear is less compliant and is not mobile enough to be pulled to the lateral margin of the footprint without added stress.^30,34,35 In large, acute tears in which there are minimal tissue degeneration and retraction, a laterally based footprint-restoring technique may be performed with minimal tension. This technique may have advantages over a medially based repair. In the literature, more attention needs to be directed toward the biomechanics and biology of chronic rotator cuff tears, as these are more commonly encountered.

Almost all of the prospective studies that have compared single- and double-row RCR have found no significant differences in MRI healing rates or clinical results at follow-up up to 2 years.^14,16,40-45 Detailed analysis of the surgical techniques used in all these studies revealed that the rotator cuff tendons were repaired back to the lateral footprint in both the single- and double-row constructs.^14,16,40-45 Although no clinical studies have compared medially and laterally based single-row repairs, our data suggest that medially based repairs have lower tensions and therefore should not be considered equivalent. Sostak and colleagues³¹ and Murray and colleagues⁴⁶ have shown that a medially based single-row RCR can achieve excellent clinical and anatomical results, likely partly because of the lower tension applied to the torn cuff tissue.^31,46 Studies are needed to compare medially and laterally based repairs, including single- and double-row repairs.

The vast majority of recent research has aimed to counteract construct tension with stronger biomechanical constructs.^20-26 Surgeons have also aimed to improve biological healing by pulling the tendon laterally across the footprint to achieve complete footprint coverage, ultimately increasing the surface area for tendon–bone healing. This has led to the development of various double-row repair techniques, in which the cuff tendon is pulled to the lateral margin of its footprint. One row of anchors is placed in the medial aspect of the footprint, while a second is placed in the lateral aspect; the cuff is reduced and compressed to the tuberosity with various suture configurations. The TOE technique was developed to improve pressurization of the cuff tendon across the footprint by linking the 2 rows with bridging sutures. In doing so, however, the potentially deleterious effects of increased tension introduced by pulling the tendon laterally may have been overlooked. Nevertheless, the biomechanics and stress distribution likely differ between single-row repair and TOE repairs, and direct comparisons cannot be made at this time. The medial row of a double-row or TOE construct may stress-shield or “unload” the more lateral tissue. Studies are needed in order to better understand the tension differential and stress distribution of various double-row constructs.

Recognizing tear morphology is crucial in maximizing chances of healing after cuff repair. For example, a crescent-shaped tear is reduced to the tuberosity with direct lateral translation of the apex of the tear, which is also the deepest or most displaced part of the tear. On the other hand, reducing an L- or reverse L-shaped tear to the tuberosity is not as direct; reducing the deepest or most displaced part of the tear would lead to overreduction and overtensioning of the tendon. However, often the exact “elbow” of the tear is not obvious and appears more rounded; therefore, it is crucial for the surgeon to examine the mobility of the torn tendon along its entire length to minimize tension. Study is needed to assess tension along the entire length of the tear for different tear morphologies and sizes.

Although our results showed that increased tension was needed to reduce a torn tendon to its lateral footprint, no study has indicated exactly how much is “too much” tension. As stated earlier, use of stronger biomechanical constructs, including TOE constructs, may overcome the increased tension associated with laterally based repairs. In addition, laterally based repairs, either single- or double-row, may be best suited for tears with lower tension, whereas medially based repairs may be best suited for higher tension tears. It is also possible that the difference in tensions noted in this study is not significant enough to have a clinical impact on choice of construct or on anatomical healing. We need studies that correlate anatomical healing rates with repair tension in order to better guide surgeons on when to use a medially or laterally based repair.

Other possible effects of increased tension associated with laterally based repairs, including beneficial effects, must be considered as well. Viscoelastic properties of human rotator cuff tendon may dissipate increased tension over time through a variety of mechanisms. Stress relaxation, gap formation, creep, and the hysteresis effect, all associated with cyclical loading in the early healing period, may lead to dissipation of force over time.^47,48 These more complex biomechanical properties of RCR constructs are yet to be clearly defined.

This study had several weaknesses. Its data represent a static measurement of time-zero rotator cuff tension, which greatly simplifies the biomechanics of the torn rotator cuff and repair construct as well as changes that occur with healing. During cuff repair, forces typically are distributed through several fixation points in stepwise process and are not focused on a single point of tissue with a grasper. Therefore, the findings of this study may not directly correlate with medially versus laterally based repairs in vivo. Furthermore, as this is a time-zero measurement, we could not determine whether the tension differential between the 2 repair positions remained static over time. Current literature suggests that muscle atrophy, fatty infiltration, and loss of elasticity of the musculotendinous unit are relatively irreversible.^35,37,49 In addition, determining the precise apex of a cuff tear can be difficult, so error may have been introduced during this process. Last, although placement of the cuff tissue at the medial or lateral footprint position was based on visual estimation by an experienced and skilled arthroscopist, error may have been introduced based on this imprecise technique.

Conclusion

This study demonstrated a significant, 5.4-fold increase in in vivo time-zero rotator cuff tension with the tendon edge reduced to the lateral footprint rather than the medial footprint.

Transforaminal lumbar interbody fusion (TLIF) has become an increasingly popular method of lumbar fusion, since its introduction by Harms and Rolinger in 1982.¹ The procedure allows for a circumferential fusion through a posterior-only approach, with improved sagittal alignment² and minimal risk for iatrogenic nerve injury. In the past decade, a minimally invasive surgical method of TLIF (MIS TLIF) has been introduced^3-5 and involves neural decompression and interbody fusion through a tubular retractor, and percutaneous placement of pedicle-screw instrumentation. This technique uses muscle dilation rather than large-scale detachment of muscle. Proponents of the MIS technique have postulated that decreased muscle damage would lead to better short-term, and possibly long-term, clinical outcomes, because of less iatrogenic soft-tissue damage.

Studies that have compared results of MIS TLIF with open TLIF have shown improved perioperative outcomes, but most have shown similar intermediate-term clinical outcomes.6 In the short term, multiple studies demonstrate that MIS TLIF is associated with decreased blood loss, less postoperative pain and narcotic requirements, and shorter hospital length of stay.^7-13 However, changes in pain score and disease-specific and generic health-related quality of life measures have been similar for the 2 procedures, beyond 6 months postoperatively.^10,13-15 These studies have generally involved retrospective reviews of unmatched patient groups, with small sample sizes and significant heterogeneity in surgical indications and case complexity. In our study, we compared intermediate-term clinical outcomes of MIS TLIF with open TLIF, using propensity matching to optimize baseline similarity of the groups.

Methods

This retrospective study was conducted after receiving approval from the Institutional Review Board. Surgical and clinical databases of 2 centers from 2008 to 2012 were reviewed for eligible subjects. Cases in 2007 were excluded because this was the year that MIS was introduced as a new technique in the practice. Inclusion criteria consisted of patients who underwent 1- to 2-level MIS TLIF and had complete baseline, 1- and 2-year postoperative outcome measures. Patients who had surgery for trauma, tumor, or osteomyelitis were excluded. Outcome measures collected and reviewed in this study included the Oswestry Disability Index (ODI),^16,17 the Medical Outcomes Study Short-Form 36 (SF-36),¹⁸ and numeric rating scales for back and leg pain (0-100 scale).¹⁹ The Physical Composite Summary (PCS) and Mental Composite Summary of the SF-36 were reviewed separately. We recorded the following patient demographic data: age, gender, American Society of Anesthesiologists (ASA) grade, body mass index, indication for surgery, workers’ compensation, and smoking status. Surgical data included number of levels fused, operative time, estimated blood loss, and length of hospital stay.

Propensity-scoring technique^20,21 was used to match the MIS TLIF patients to a control group of patients who underwent TLIF using an open approach (open TLIF), matching for multiple characteristics to produce 2 similar comparison groups. Propensity matching was performed to control for bias. In controlling for known confounders or biases, propensity matching, in theory, should also control for unknown confounders. Gender, age, body mass index, smoking status, indication for fusion, as well as preoperative ODI, SF-36 PCS, SF-36 Mental Composite Summary, and pain scores were used to generate a control open TLIF group.

MIS TLIF Surgical Technique

Patients in the MIS TLIF group underwent neural decompression and interbody fusion through a tubular retractor system (METRx, Medtronic Inc.), followed by percutaneous pedicle-screw fixation under fluoroscopic guidance (Sextant, Medtronic Inc.). After successful induction of general endotracheal anesthesia, patients were positioned prone on a radiolucent table. Posteroanterior (PA) and lateral fluoroscopic images were used to localize 2 paramedian incisions, approximately 3-cm to 5-cm lateral to midline, over the pedicles of interest. Modified Jamshidi needles (Medtronic Inc.) were used to cannulate the pedicles under PA, posterior-oblique, PA, and lateral fluoroscopic guidance. The pedicles were tapped with a cannulated tap. Pedicle screws and rods were introduced on the side contralateral to the TLIF and were used as needed to maintain intradiscal distraction during the TLIF portion of the procedure.

Decompression and TLIF were carried out on the side of the patient’s radicular pain or bilaterally, according to the surgeon’s discretion. A K-wire was advanced to the facet joint complex, after which sequential dilators were used to dilate through the muscles to establish an intramuscular corridor to the facet. A 26-mm fixed tubular retractor was docked over the facet and locked in place, using a post attached to the operating room table. Neural decompression was obtained by removal of the entire facet-joint complex and lamina to the base of the spinous process, using a combination of high-speed drills and Kerrison rongeurs. The ligamentum flavum was completely resected. The superior articular process of the caudal vertebra was removed all the way to the pedicle below. Ball-tipped probes were used to confirm that traversing and exiting nerve roots were completely free. An annulotomy was performed, and all disc material was removed from the disc through a combination of rotating shavers, serrated curettes, endplate scrapers, and rasps. Bone graft was placed anterior and contralateral to the interbody cage. (Bone grafts included autogenous iliac crest, local bone obtained from the decompression, recombinant human bone morphogenetic protein 2, or allograft demineralized bone matrix at the surgeon’s discretion.) After placement of the interbody cage, the ipsilateral pedicle-screw instrumentation was put over the remaining guide wires and compression applied across the construct to lock the interbody cage and restore lordosis. Wounds were closed without drains.

Open TLIF Surgical Technique

In patients undergoing open TLIF, a midline incision was made over the vertebrae of interest, and paraspinal muscles were subperiosteally dissected to the tips of the transverse processes. The appropriate level was confirmed with intraoperative radiograph. Pedicle screws were placed free-hand using anatomic landmarks, and appropriate placement was confirmed with intraoperative radiograph and evoked electromyography stimulation. Laminectomy and facetectomy were performed, and the disc was entered on the side of the facetectomy. After thorough disc-space preparation, bone graft and an interbody cage were placed, rods inserted, and compression carried out. A supplemental posterolateral fusion was also performed after decortication of the transverse processes and cartilaginous surface of the contralateral facet. Layered wound closure was performed over drains.

Analysis

Statistical analysis was carried out using SPSS Statistics version 17.0 (IBM) with significance set at the P < .01 level. A small, conservative P-value threshold was used to minimize type II error that resulted from the multiple comparisons performed. Student t test was used to determine any significant differences between continuous demographic variables, and to compare preoperative and postoperative outcome measure scores within and between study groups. Fisher’s exact test was used to compare categorical variables between the 2 groups.

Results

The MIS TLIF group consisted of 64 patients (average age, 52 years), and included 22 patients with degenerative spondylolisthesis, 33 with disc pathology, 8 with postdecompression, and 1 non-union patient. The open TLIF group consisted of 64 patients (average age, 54 years), and included 39 degenerative spondylolisthesis, 15 disc pathology, 7 postdecompression, and 3 nonunion patients (Table 1). All 64 open and 19 MIS cases were from a spine practice with 6 surgeons, and 45 MIS cases came from a spine practice with 2 surgeons. There was also an unequal distribution of the specific levels fused between the open and MIS groups.

Although the operative time was similar in both groups, the MIS TLIF group had a statistically significantly lower blood loss compared with the open TLIF group (Table 2). Both MIS TLIF and open TLIF lead to significant improvements in pain, ODI, and SF-36 PCS (P < .01) (Table 3). At 1 year, both groups had similar improvements in pain (36.9 vs 30.8, P = .178) and SF-36 PCS (9.9 vs 7.5, P = .231), but the MIS TLIF group had a statistically significantly greater improvement in ODI compared with the open TLIF group (30.4 vs 15.1, P < .000). At 2 years, both groups had similar improvements in SF-36 PCS (12.1 vs 7.5, P = .033), but the MIS TLIF group had a statistically significantly greater improvement in pain (40.2 vs 27.0, P = .005) and ODI (33.1 vs 15.4, P < .000) compared with the open TLIF group (Table 4).

Discussion

The current study compared intermediate-term clinical outcomes of MIS TLIF to open TLIF. We used propensity matching to identify a control group of open TLIFs that were comparable to the MIS TLIF group across a variety of covariates that are known to influence the results of lumbar fusion. This created comparison groups that were as closely matched at baseline as possible. We found that, at 2-year follow-up, MIS TLIF patients had less pain and less low-back pain–related disability as measured by ODI. There was also a trend toward better generic health-related quality of life in the MIS TLIF group.

These data suggest that the decreased soft-tissue trauma of the minimally invasive surgical technique, which leads to improved perioperative parameters in the short term, may also lead to some advantages that translate to improved intermediate-term clinical outcomes. Traditional lumbar fusion procedures have shown excellent clinical results when used for accepted clinical indications.²² However, the procedure requires extensive dissection of the paraspinal muscles, which causes significant muscle damage as evidenced by muscle breakdown products that can be detected in the bloodstream postoperatively.^23,24 The lateral dissection also transects the dorsal ramus of the segmental nerves, which innervate the paraspinal muscles, leading to significant scarring and atrophy on postoperative imaging studies.²³ Some authors have used the term “fusion disease” to describe the constellation of soft-tissue degradation seen after open lumbar fusion.⁵

An MIS version of the TLIF procedure that was described in 2003³ avoids much of this iatrogenic soft-tissue trauma. It involves intramuscular dilation to approach the spine and to carry out neural decompression and interbody fusion, in conjunction with percutaneous pedicle-screw instrumentation. Proponents of this technique point to diminished iatrogenic soft-tissue and muscle damage as an advantage. Multiple studies have, in fact, confirmed improved short-term perioperative parameters, such as less blood loss, lower narcotic requirements, and decreased length-of-hospital stay.²⁵ Economic analyses have also shown lower direct and indirect costs with the MIS technique.²⁶

Several studies have compared patient-reported outcome measures of MIS and open TLIF, and the results have been mixed. Most of these studies have shown similar improvement in clinical outcomes between the 2 procedures, but the MIS technique demonstrated short-term perioperative advantages, such as lower blood loss, less narcotic requirements, and shorter length of stay.^7-15 The authors of these studies conclude that the MIS technique can provide similar long-term results with lower short-term morbidity when compared with open TLIF. In contrast, some studies have shown better short- and intermediate-term clinical outcomes with the MIS technique.^23,27-29 As a whole, the literature comparing the 2 procedures consists of mostly small retrospective studies with nonrandomized patient samples, heterogeneous surgical indications, and differing surgical techniques, making it difficult to draw conclusions.

The current study suggests that MIS TLIF may lead to improved clinical results at 2-year follow-up, compared with open TLIF. Our study used propensity-score matching to minimize the effects of nonrandom assignment of subjects to MIS TLIF or open TLIF. A limitation of observational studies is that bias in assignment of subjects to treatment groups can lead to overestimation or underestimation of the effect of the treatment itself. Propensity-score matching attempts to reduce this bias by accounting for several covariates that predict whether a subject will receive a certain treatment. These covariates are used in a logistic regression to produce a propensity score, which can be used to match subjects to controls across multiple dimensions, thus ensuring groups are as comparable as possible at baseline.

Our study still has several limitations. Sample size is relatively small, and follow-up is still only intermediate, at 2 years. There was unequal distribution of specific levels of surgery. Because patients were not blinded to the treatment they received, it is possible that patient perception of receiving a newer, less-invasive treatment method may influence their subjective improvement. The study sample was drawn from 2 different centers, with one center providing mostly MIS cases and the other providing mostly open cases. Because of this, undetected differences in how patients were selected for surgery could also affect outcomes. Any latent confounding variables, which are not identified a priori, will not be accounted for in the matching process. Only a prospective, randomized study with large numbers can control for observed and unobserved confounding patient characteristics.

In summary, our study shows that MIS TLIF is associated with improved low back pain and low back–related disability at 2 years compared with open TLIF. Other studies comparing the 2 techniques have come to different conclusions regarding whether the short-term benefits of MIS TLIF translate into long-term differences in clinical outcome. This study adds to this evidence and suggests there may be longer term advantages to the MIS approach, but prospective randomized trials are needed to confirm this finding and determine the true magnitude of these differences.

Transforaminal lumbar interbody fusion (TLIF) has become an increasingly popular method of lumbar fusion, since its introduction by Harms and Rolinger in 1982.¹ The procedure allows for a circumferential fusion through a posterior-only approach, with improved sagittal alignment² and minimal risk for iatrogenic nerve injury. In the past decade, a minimally invasive surgical method of TLIF (MIS TLIF) has been introduced^3-5 and involves neural decompression and interbody fusion through a tubular retractor, and percutaneous placement of pedicle-screw instrumentation. This technique uses muscle dilation rather than large-scale detachment of muscle. Proponents of the MIS technique have postulated that decreased muscle damage would lead to better short-term, and possibly long-term, clinical outcomes, because of less iatrogenic soft-tissue damage.

Studies that have compared results of MIS TLIF with open TLIF have shown improved perioperative outcomes, but most have shown similar intermediate-term clinical outcomes.6 In the short term, multiple studies demonstrate that MIS TLIF is associated with decreased blood loss, less postoperative pain and narcotic requirements, and shorter hospital length of stay.^7-13 However, changes in pain score and disease-specific and generic health-related quality of life measures have been similar for the 2 procedures, beyond 6 months postoperatively.^10,13-15 These studies have generally involved retrospective reviews of unmatched patient groups, with small sample sizes and significant heterogeneity in surgical indications and case complexity. In our study, we compared intermediate-term clinical outcomes of MIS TLIF with open TLIF, using propensity matching to optimize baseline similarity of the groups.

Methods

This retrospective study was conducted after receiving approval from the Institutional Review Board. Surgical and clinical databases of 2 centers from 2008 to 2012 were reviewed for eligible subjects. Cases in 2007 were excluded because this was the year that MIS was introduced as a new technique in the practice. Inclusion criteria consisted of patients who underwent 1- to 2-level MIS TLIF and had complete baseline, 1- and 2-year postoperative outcome measures. Patients who had surgery for trauma, tumor, or osteomyelitis were excluded. Outcome measures collected and reviewed in this study included the Oswestry Disability Index (ODI),^16,17 the Medical Outcomes Study Short-Form 36 (SF-36),¹⁸ and numeric rating scales for back and leg pain (0-100 scale).¹⁹ The Physical Composite Summary (PCS) and Mental Composite Summary of the SF-36 were reviewed separately. We recorded the following patient demographic data: age, gender, American Society of Anesthesiologists (ASA) grade, body mass index, indication for surgery, workers’ compensation, and smoking status. Surgical data included number of levels fused, operative time, estimated blood loss, and length of hospital stay.

Propensity-scoring technique^20,21 was used to match the MIS TLIF patients to a control group of patients who underwent TLIF using an open approach (open TLIF), matching for multiple characteristics to produce 2 similar comparison groups. Propensity matching was performed to control for bias. In controlling for known confounders or biases, propensity matching, in theory, should also control for unknown confounders. Gender, age, body mass index, smoking status, indication for fusion, as well as preoperative ODI, SF-36 PCS, SF-36 Mental Composite Summary, and pain scores were used to generate a control open TLIF group.

MIS TLIF Surgical Technique

Patients in the MIS TLIF group underwent neural decompression and interbody fusion through a tubular retractor system (METRx, Medtronic Inc.), followed by percutaneous pedicle-screw fixation under fluoroscopic guidance (Sextant, Medtronic Inc.). After successful induction of general endotracheal anesthesia, patients were positioned prone on a radiolucent table. Posteroanterior (PA) and lateral fluoroscopic images were used to localize 2 paramedian incisions, approximately 3-cm to 5-cm lateral to midline, over the pedicles of interest. Modified Jamshidi needles (Medtronic Inc.) were used to cannulate the pedicles under PA, posterior-oblique, PA, and lateral fluoroscopic guidance. The pedicles were tapped with a cannulated tap. Pedicle screws and rods were introduced on the side contralateral to the TLIF and were used as needed to maintain intradiscal distraction during the TLIF portion of the procedure.

Decompression and TLIF were carried out on the side of the patient’s radicular pain or bilaterally, according to the surgeon’s discretion. A K-wire was advanced to the facet joint complex, after which sequential dilators were used to dilate through the muscles to establish an intramuscular corridor to the facet. A 26-mm fixed tubular retractor was docked over the facet and locked in place, using a post attached to the operating room table. Neural decompression was obtained by removal of the entire facet-joint complex and lamina to the base of the spinous process, using a combination of high-speed drills and Kerrison rongeurs. The ligamentum flavum was completely resected. The superior articular process of the caudal vertebra was removed all the way to the pedicle below. Ball-tipped probes were used to confirm that traversing and exiting nerve roots were completely free. An annulotomy was performed, and all disc material was removed from the disc through a combination of rotating shavers, serrated curettes, endplate scrapers, and rasps. Bone graft was placed anterior and contralateral to the interbody cage. (Bone grafts included autogenous iliac crest, local bone obtained from the decompression, recombinant human bone morphogenetic protein 2, or allograft demineralized bone matrix at the surgeon’s discretion.) After placement of the interbody cage, the ipsilateral pedicle-screw instrumentation was put over the remaining guide wires and compression applied across the construct to lock the interbody cage and restore lordosis. Wounds were closed without drains.

Open TLIF Surgical Technique

In patients undergoing open TLIF, a midline incision was made over the vertebrae of interest, and paraspinal muscles were subperiosteally dissected to the tips of the transverse processes. The appropriate level was confirmed with intraoperative radiograph. Pedicle screws were placed free-hand using anatomic landmarks, and appropriate placement was confirmed with intraoperative radiograph and evoked electromyography stimulation. Laminectomy and facetectomy were performed, and the disc was entered on the side of the facetectomy. After thorough disc-space preparation, bone graft and an interbody cage were placed, rods inserted, and compression carried out. A supplemental posterolateral fusion was also performed after decortication of the transverse processes and cartilaginous surface of the contralateral facet. Layered wound closure was performed over drains.

Analysis

Statistical analysis was carried out using SPSS Statistics version 17.0 (IBM) with significance set at the P < .01 level. A small, conservative P-value threshold was used to minimize type II error that resulted from the multiple comparisons performed. Student t test was used to determine any significant differences between continuous demographic variables, and to compare preoperative and postoperative outcome measure scores within and between study groups. Fisher’s exact test was used to compare categorical variables between the 2 groups.

Results

The MIS TLIF group consisted of 64 patients (average age, 52 years), and included 22 patients with degenerative spondylolisthesis, 33 with disc pathology, 8 with postdecompression, and 1 non-union patient. The open TLIF group consisted of 64 patients (average age, 54 years), and included 39 degenerative spondylolisthesis, 15 disc pathology, 7 postdecompression, and 3 nonunion patients (Table 1). All 64 open and 19 MIS cases were from a spine practice with 6 surgeons, and 45 MIS cases came from a spine practice with 2 surgeons. There was also an unequal distribution of the specific levels fused between the open and MIS groups.

Although the operative time was similar in both groups, the MIS TLIF group had a statistically significantly lower blood loss compared with the open TLIF group (Table 2). Both MIS TLIF and open TLIF lead to significant improvements in pain, ODI, and SF-36 PCS (P < .01) (Table 3). At 1 year, both groups had similar improvements in pain (36.9 vs 30.8, P = .178) and SF-36 PCS (9.9 vs 7.5, P = .231), but the MIS TLIF group had a statistically significantly greater improvement in ODI compared with the open TLIF group (30.4 vs 15.1, P < .000). At 2 years, both groups had similar improvements in SF-36 PCS (12.1 vs 7.5, P = .033), but the MIS TLIF group had a statistically significantly greater improvement in pain (40.2 vs 27.0, P = .005) and ODI (33.1 vs 15.4, P < .000) compared with the open TLIF group (Table 4).

Discussion

The current study compared intermediate-term clinical outcomes of MIS TLIF to open TLIF. We used propensity matching to identify a control group of open TLIFs that were comparable to the MIS TLIF group across a variety of covariates that are known to influence the results of lumbar fusion. This created comparison groups that were as closely matched at baseline as possible. We found that, at 2-year follow-up, MIS TLIF patients had less pain and less low-back pain–related disability as measured by ODI. There was also a trend toward better generic health-related quality of life in the MIS TLIF group.

These data suggest that the decreased soft-tissue trauma of the minimally invasive surgical technique, which leads to improved perioperative parameters in the short term, may also lead to some advantages that translate to improved intermediate-term clinical outcomes. Traditional lumbar fusion procedures have shown excellent clinical results when used for accepted clinical indications.²² However, the procedure requires extensive dissection of the paraspinal muscles, which causes significant muscle damage as evidenced by muscle breakdown products that can be detected in the bloodstream postoperatively.^23,24 The lateral dissection also transects the dorsal ramus of the segmental nerves, which innervate the paraspinal muscles, leading to significant scarring and atrophy on postoperative imaging studies.²³ Some authors have used the term “fusion disease” to describe the constellation of soft-tissue degradation seen after open lumbar fusion.⁵

An MIS version of the TLIF procedure that was described in 2003³ avoids much of this iatrogenic soft-tissue trauma. It involves intramuscular dilation to approach the spine and to carry out neural decompression and interbody fusion, in conjunction with percutaneous pedicle-screw instrumentation. Proponents of this technique point to diminished iatrogenic soft-tissue and muscle damage as an advantage. Multiple studies have, in fact, confirmed improved short-term perioperative parameters, such as less blood loss, lower narcotic requirements, and decreased length-of-hospital stay.²⁵ Economic analyses have also shown lower direct and indirect costs with the MIS technique.²⁶

Several studies have compared patient-reported outcome measures of MIS and open TLIF, and the results have been mixed. Most of these studies have shown similar improvement in clinical outcomes between the 2 procedures, but the MIS technique demonstrated short-term perioperative advantages, such as lower blood loss, less narcotic requirements, and shorter length of stay.^7-15 The authors of these studies conclude that the MIS technique can provide similar long-term results with lower short-term morbidity when compared with open TLIF. In contrast, some studies have shown better short- and intermediate-term clinical outcomes with the MIS technique.^23,27-29 As a whole, the literature comparing the 2 procedures consists of mostly small retrospective studies with nonrandomized patient samples, heterogeneous surgical indications, and differing surgical techniques, making it difficult to draw conclusions.

The current study suggests that MIS TLIF may lead to improved clinical results at 2-year follow-up, compared with open TLIF. Our study used propensity-score matching to minimize the effects of nonrandom assignment of subjects to MIS TLIF or open TLIF. A limitation of observational studies is that bias in assignment of subjects to treatment groups can lead to overestimation or underestimation of the effect of the treatment itself. Propensity-score matching attempts to reduce this bias by accounting for several covariates that predict whether a subject will receive a certain treatment. These covariates are used in a logistic regression to produce a propensity score, which can be used to match subjects to controls across multiple dimensions, thus ensuring groups are as comparable as possible at baseline.

Our study still has several limitations. Sample size is relatively small, and follow-up is still only intermediate, at 2 years. There was unequal distribution of specific levels of surgery. Because patients were not blinded to the treatment they received, it is possible that patient perception of receiving a newer, less-invasive treatment method may influence their subjective improvement. The study sample was drawn from 2 different centers, with one center providing mostly MIS cases and the other providing mostly open cases. Because of this, undetected differences in how patients were selected for surgery could also affect outcomes. Any latent confounding variables, which are not identified a priori, will not be accounted for in the matching process. Only a prospective, randomized study with large numbers can control for observed and unobserved confounding patient characteristics.

In summary, our study shows that MIS TLIF is associated with improved low back pain and low back–related disability at 2 years compared with open TLIF. Other studies comparing the 2 techniques have come to different conclusions regarding whether the short-term benefits of MIS TLIF translate into long-term differences in clinical outcome. This study adds to this evidence and suggests there may be longer term advantages to the MIS approach, but prospective randomized trials are needed to confirm this finding and determine the true magnitude of these differences.

Transforaminal lumbar interbody fusion (TLIF) has become an increasingly popular method of lumbar fusion, since its introduction by Harms and Rolinger in 1982.¹ The procedure allows for a circumferential fusion through a posterior-only approach, with improved sagittal alignment² and minimal risk for iatrogenic nerve injury. In the past decade, a minimally invasive surgical method of TLIF (MIS TLIF) has been introduced^3-5 and involves neural decompression and interbody fusion through a tubular retractor, and percutaneous placement of pedicle-screw instrumentation. This technique uses muscle dilation rather than large-scale detachment of muscle. Proponents of the MIS technique have postulated that decreased muscle damage would lead to better short-term, and possibly long-term, clinical outcomes, because of less iatrogenic soft-tissue damage.

Studies that have compared results of MIS TLIF with open TLIF have shown improved perioperative outcomes, but most have shown similar intermediate-term clinical outcomes.6 In the short term, multiple studies demonstrate that MIS TLIF is associated with decreased blood loss, less postoperative pain and narcotic requirements, and shorter hospital length of stay.^7-13 However, changes in pain score and disease-specific and generic health-related quality of life measures have been similar for the 2 procedures, beyond 6 months postoperatively.^10,13-15 These studies have generally involved retrospective reviews of unmatched patient groups, with small sample sizes and significant heterogeneity in surgical indications and case complexity. In our study, we compared intermediate-term clinical outcomes of MIS TLIF with open TLIF, using propensity matching to optimize baseline similarity of the groups.

Methods

This retrospective study was conducted after receiving approval from the Institutional Review Board. Surgical and clinical databases of 2 centers from 2008 to 2012 were reviewed for eligible subjects. Cases in 2007 were excluded because this was the year that MIS was introduced as a new technique in the practice. Inclusion criteria consisted of patients who underwent 1- to 2-level MIS TLIF and had complete baseline, 1- and 2-year postoperative outcome measures. Patients who had surgery for trauma, tumor, or osteomyelitis were excluded. Outcome measures collected and reviewed in this study included the Oswestry Disability Index (ODI),^16,17 the Medical Outcomes Study Short-Form 36 (SF-36),¹⁸ and numeric rating scales for back and leg pain (0-100 scale).¹⁹ The Physical Composite Summary (PCS) and Mental Composite Summary of the SF-36 were reviewed separately. We recorded the following patient demographic data: age, gender, American Society of Anesthesiologists (ASA) grade, body mass index, indication for surgery, workers’ compensation, and smoking status. Surgical data included number of levels fused, operative time, estimated blood loss, and length of hospital stay.

Propensity-scoring technique^20,21 was used to match the MIS TLIF patients to a control group of patients who underwent TLIF using an open approach (open TLIF), matching for multiple characteristics to produce 2 similar comparison groups. Propensity matching was performed to control for bias. In controlling for known confounders or biases, propensity matching, in theory, should also control for unknown confounders. Gender, age, body mass index, smoking status, indication for fusion, as well as preoperative ODI, SF-36 PCS, SF-36 Mental Composite Summary, and pain scores were used to generate a control open TLIF group.

MIS TLIF Surgical Technique

Patients in the MIS TLIF group underwent neural decompression and interbody fusion through a tubular retractor system (METRx, Medtronic Inc.), followed by percutaneous pedicle-screw fixation under fluoroscopic guidance (Sextant, Medtronic Inc.). After successful induction of general endotracheal anesthesia, patients were positioned prone on a radiolucent table. Posteroanterior (PA) and lateral fluoroscopic images were used to localize 2 paramedian incisions, approximately 3-cm to 5-cm lateral to midline, over the pedicles of interest. Modified Jamshidi needles (Medtronic Inc.) were used to cannulate the pedicles under PA, posterior-oblique, PA, and lateral fluoroscopic guidance. The pedicles were tapped with a cannulated tap. Pedicle screws and rods were introduced on the side contralateral to the TLIF and were used as needed to maintain intradiscal distraction during the TLIF portion of the procedure.

Decompression and TLIF were carried out on the side of the patient’s radicular pain or bilaterally, according to the surgeon’s discretion. A K-wire was advanced to the facet joint complex, after which sequential dilators were used to dilate through the muscles to establish an intramuscular corridor to the facet. A 26-mm fixed tubular retractor was docked over the facet and locked in place, using a post attached to the operating room table. Neural decompression was obtained by removal of the entire facet-joint complex and lamina to the base of the spinous process, using a combination of high-speed drills and Kerrison rongeurs. The ligamentum flavum was completely resected. The superior articular process of the caudal vertebra was removed all the way to the pedicle below. Ball-tipped probes were used to confirm that traversing and exiting nerve roots were completely free. An annulotomy was performed, and all disc material was removed from the disc through a combination of rotating shavers, serrated curettes, endplate scrapers, and rasps. Bone graft was placed anterior and contralateral to the interbody cage. (Bone grafts included autogenous iliac crest, local bone obtained from the decompression, recombinant human bone morphogenetic protein 2, or allograft demineralized bone matrix at the surgeon’s discretion.) After placement of the interbody cage, the ipsilateral pedicle-screw instrumentation was put over the remaining guide wires and compression applied across the construct to lock the interbody cage and restore lordosis. Wounds were closed without drains.

Open TLIF Surgical Technique

In patients undergoing open TLIF, a midline incision was made over the vertebrae of interest, and paraspinal muscles were subperiosteally dissected to the tips of the transverse processes. The appropriate level was confirmed with intraoperative radiograph. Pedicle screws were placed free-hand using anatomic landmarks, and appropriate placement was confirmed with intraoperative radiograph and evoked electromyography stimulation. Laminectomy and facetectomy were performed, and the disc was entered on the side of the facetectomy. After thorough disc-space preparation, bone graft and an interbody cage were placed, rods inserted, and compression carried out. A supplemental posterolateral fusion was also performed after decortication of the transverse processes and cartilaginous surface of the contralateral facet. Layered wound closure was performed over drains.

Analysis

Statistical analysis was carried out using SPSS Statistics version 17.0 (IBM) with significance set at the P < .01 level. A small, conservative P-value threshold was used to minimize type II error that resulted from the multiple comparisons performed. Student t test was used to determine any significant differences between continuous demographic variables, and to compare preoperative and postoperative outcome measure scores within and between study groups. Fisher’s exact test was used to compare categorical variables between the 2 groups.

Results

The MIS TLIF group consisted of 64 patients (average age, 52 years), and included 22 patients with degenerative spondylolisthesis, 33 with disc pathology, 8 with postdecompression, and 1 non-union patient. The open TLIF group consisted of 64 patients (average age, 54 years), and included 39 degenerative spondylolisthesis, 15 disc pathology, 7 postdecompression, and 3 nonunion patients (Table 1). All 64 open and 19 MIS cases were from a spine practice with 6 surgeons, and 45 MIS cases came from a spine practice with 2 surgeons. There was also an unequal distribution of the specific levels fused between the open and MIS groups.

Although the operative time was similar in both groups, the MIS TLIF group had a statistically significantly lower blood loss compared with the open TLIF group (Table 2). Both MIS TLIF and open TLIF lead to significant improvements in pain, ODI, and SF-36 PCS (P < .01) (Table 3). At 1 year, both groups had similar improvements in pain (36.9 vs 30.8, P = .178) and SF-36 PCS (9.9 vs 7.5, P = .231), but the MIS TLIF group had a statistically significantly greater improvement in ODI compared with the open TLIF group (30.4 vs 15.1, P < .000). At 2 years, both groups had similar improvements in SF-36 PCS (12.1 vs 7.5, P = .033), but the MIS TLIF group had a statistically significantly greater improvement in pain (40.2 vs 27.0, P = .005) and ODI (33.1 vs 15.4, P < .000) compared with the open TLIF group (Table 4).

Discussion

The current study compared intermediate-term clinical outcomes of MIS TLIF to open TLIF. We used propensity matching to identify a control group of open TLIFs that were comparable to the MIS TLIF group across a variety of covariates that are known to influence the results of lumbar fusion. This created comparison groups that were as closely matched at baseline as possible. We found that, at 2-year follow-up, MIS TLIF patients had less pain and less low-back pain–related disability as measured by ODI. There was also a trend toward better generic health-related quality of life in the MIS TLIF group.

These data suggest that the decreased soft-tissue trauma of the minimally invasive surgical technique, which leads to improved perioperative parameters in the short term, may also lead to some advantages that translate to improved intermediate-term clinical outcomes. Traditional lumbar fusion procedures have shown excellent clinical results when used for accepted clinical indications.²² However, the procedure requires extensive dissection of the paraspinal muscles, which causes significant muscle damage as evidenced by muscle breakdown products that can be detected in the bloodstream postoperatively.^23,24 The lateral dissection also transects the dorsal ramus of the segmental nerves, which innervate the paraspinal muscles, leading to significant scarring and atrophy on postoperative imaging studies.²³ Some authors have used the term “fusion disease” to describe the constellation of soft-tissue degradation seen after open lumbar fusion.⁵

An MIS version of the TLIF procedure that was described in 2003³ avoids much of this iatrogenic soft-tissue trauma. It involves intramuscular dilation to approach the spine and to carry out neural decompression and interbody fusion, in conjunction with percutaneous pedicle-screw instrumentation. Proponents of this technique point to diminished iatrogenic soft-tissue and muscle damage as an advantage. Multiple studies have, in fact, confirmed improved short-term perioperative parameters, such as less blood loss, lower narcotic requirements, and decreased length-of-hospital stay.²⁵ Economic analyses have also shown lower direct and indirect costs with the MIS technique.²⁶

Several studies have compared patient-reported outcome measures of MIS and open TLIF, and the results have been mixed. Most of these studies have shown similar improvement in clinical outcomes between the 2 procedures, but the MIS technique demonstrated short-term perioperative advantages, such as lower blood loss, less narcotic requirements, and shorter length of stay.^7-15 The authors of these studies conclude that the MIS technique can provide similar long-term results with lower short-term morbidity when compared with open TLIF. In contrast, some studies have shown better short- and intermediate-term clinical outcomes with the MIS technique.^23,27-29 As a whole, the literature comparing the 2 procedures consists of mostly small retrospective studies with nonrandomized patient samples, heterogeneous surgical indications, and differing surgical techniques, making it difficult to draw conclusions.

The current study suggests that MIS TLIF may lead to improved clinical results at 2-year follow-up, compared with open TLIF. Our study used propensity-score matching to minimize the effects of nonrandom assignment of subjects to MIS TLIF or open TLIF. A limitation of observational studies is that bias in assignment of subjects to treatment groups can lead to overestimation or underestimation of the effect of the treatment itself. Propensity-score matching attempts to reduce this bias by accounting for several covariates that predict whether a subject will receive a certain treatment. These covariates are used in a logistic regression to produce a propensity score, which can be used to match subjects to controls across multiple dimensions, thus ensuring groups are as comparable as possible at baseline.

Our study still has several limitations. Sample size is relatively small, and follow-up is still only intermediate, at 2 years. There was unequal distribution of specific levels of surgery. Because patients were not blinded to the treatment they received, it is possible that patient perception of receiving a newer, less-invasive treatment method may influence their subjective improvement. The study sample was drawn from 2 different centers, with one center providing mostly MIS cases and the other providing mostly open cases. Because of this, undetected differences in how patients were selected for surgery could also affect outcomes. Any latent confounding variables, which are not identified a priori, will not be accounted for in the matching process. Only a prospective, randomized study with large numbers can control for observed and unobserved confounding patient characteristics.

In summary, our study shows that MIS TLIF is associated with improved low back pain and low back–related disability at 2 years compared with open TLIF. Other studies comparing the 2 techniques have come to different conclusions regarding whether the short-term benefits of MIS TLIF translate into long-term differences in clinical outcome. This study adds to this evidence and suggests there may be longer term advantages to the MIS approach, but prospective randomized trials are needed to confirm this finding and determine the true magnitude of these differences.

The optimal management strategy for acute fractures of the metadiaphyseal fifth metatarsal (Jones fractures) is controversial. Patients can be successfully treated nonoperatively with non-weight-bearing and immobilization in a short leg cast^1-7 or operatively with placement of an intramedullary screw.^8-10 The primary advantage of nonoperative treatment is avoiding the risks and discomfort of surgery; disadvantages include the need for prolonged immobilization and protected weight-bearing as well as a decreased union rate.^8,9 Advantages of operative treatment include accelerated functional recovery and an improved union rate; disadvantages include exposure to the risks, inconvenience, and discomfort of surgery. Clear, definitive evidence for guiding treatment is not available in the orthopedic literature, and treatment strategies vary substantially according to surgeon and patient preference.

Expected-value decision analysis, a research tool that helps guide decision-making in situations of uncertainty, has been effectively applied to other areas of uncertainty in the orthopedic literature.^11-14 Borrowed from gaming theory, the technique involves creating a decision tree to define the clinical problem, determining outcome probabilities and utilities, performing a fold-back analysis to determine the optimal decision-making strategy, and performing a sensitivity analysis to model the effect of varying outcome probabilities and utilities on decision-making. Decision analysis may therefore allow the clinician and the patient to optimize decision-making based on best available evidence and patient preferences. It also helps determine the most important factors affecting management strategies and the decision-making process, which may not always be intuitive.

In the present study, we used expected-value decision analysis to determine the optimal management strategy, operative or nonoperative, for acute Jones fracture. We also explored factors with the most influence on the model and identified important questions for future research.

Materials and Methods

Institutional review board approval was obtained for this study. Analysis was performed with Treeage Pro statistical software (Treeage Software).

Outcome Probabilities

Outcome probabilities were determined by reviewing the literature for articles on Jones fractures. This body of literature was summarized in a comprehensive review by Dean and colleagues¹⁵, who extracted data from 19 studies: 1 randomized controlled trial, 1 prospective case series, and 17 retrospective case series.¹⁵ We used data from these studies to determine outcome probabilities (Table).

Outcome Utilities

Utilities represent patient preferences for various disease states. Outcome utility values were obtained from 32 adults (25 women, 7 men) with no history of foot injury. Mean age was 32.4 years (range, 20-69 years). The questionnaire presented scenarios for the different outcomes and asked patients to rate these outcomes on a scale ranging from 0 (worst possible outcome) to 10 (best possible outcome). The Sports subscale of the Foot and Ankle Ability Measure (FAAM) 16 was used to quantify patient activity level.

Decision Tree and Fold-Back Analysis

A decision tree was constructed with 1 decision node, 4 chance nodes, and 7 terminal nodes (Figure 1). The decision tree demonstrates 2 different strategies for managing a Jones fracture. The decision node divides the tree into 2 branches: initial operative or nonoperative treatment. Both branches are followed by various chance nodes, each terminating in a discrete clinical outcome. Per convention, utility data were placed to the right of the terminal nodes, and probability data were placed under the terminal nodes.

Fold-back analysis was performed to identify the optimal strategy. Fold-back analysis involves multiplying each outcome utility by its associated probability, thereby providing an “expected value” for each clinical endpoint. Then, the expected values for each endpoint can be summed for a given management strategy, and the ultimate expected values of the different strategies can be compared. The management strategy associated with the highest expected value is optimal for the given outcome utilities and probabilities.

Sensitivity Analysis

One-way sensitivity analysis was performed to model the effect on decision-making of changing the values for utility for uncomplicated surgery, utility for healing with nonoperative treatment, utility for uncomplicated treatment of nonunion, likelihood of healing with nonoperative treatment, likelihood of healing with surgery, and likelihood of minor complication with surgery. These were the variables found to affect the decision-making strategy within their clinically plausible ranges.

Results

Outcome Probabilities and Utilities

Outcome probabilities and utilities are illustrated in Figure 1. By convention, probabilities appear below the corresponding branches of the decision tree, and utilities appear at the end of each branch. Mean (SD) FAAM Sports subscale score was 84.6 (27.4). This subscale is scored as a percentage from 0% to 100%, with higher scores indicating a higher level of physical function.

Decision Analysis

The expected value for nonoperative treatment was 7.74, and the expected value for intramedullary screw fixation was 7.88 (Figure 1). Therefore, operative treatment was identified as the optimal treatment strategy.

Sensitivity analyses revealed that the optimal decision making strategy was very sensitive to small changes in several variables. Nonoperative treatment becomes the preferred strategy when the utility value for uncomplicated surgery falls below 8.04 (Figure 2), when the utility for healing with nonoperative treatment rises above 8.49 (Figure 3), when the likelihood of healing with nonoperative treatment rises above 82% (Figure 4), or when the probability of healing after surgery falls below 92% (Figure 5). Nonoperative treatment is also favored when the probability of minor complication with surgery is above 17% (Figure 6) and when utility for a successfully treated nonunion is higher than 6.9 (Figure 7).

Discussion

Optimal management of a metadiaphyseal fracture of the fifth metatarsal (Jones fracture) remains controversial. The decision between initial operative or nonoperative treatment lends itself to expected-value decision analysis because of well-defined treatment options and relatively discrete outcomes. The principal advantages of nonoperative treatment are that it allows the patient to avoid the risks and discomfort of surgery, and the principal advantages of operative treatment are that it maximizes the chance of fracture union and may accelerate functional recovery.

Our decision analysis determined that operative fixation is the optimal decision path, given the outcome probabilities derived from the literature and the utilities obtained from surveys. This finding is in accordance with several expert opinions in foot and ankle fracture surgery.^17,18 However, the expected values of the operative and nonoperative treatment strategies differed by only 0.3 on a 10-point scale. Such similar expected values in our model are not surprising given the controversy surrounding clinical decision making in the treatment of these fractures.¹⁹

In addition, our analysis identified the important variables in the decision-making process. Patients averse to surgery, patients not averse to successful nonoperative treatment, and patients who view successful nonunion surgery after initial nonoperative treatment as a relatively positive outcome may be best treated nonoperatively. These findings emphasize the importance of patient preferences and shared decision-making. Higher rates of healing with nonoperative treatment, lower rates of healing with surgery, and higher complication rates with surgery also favor nonoperative management. It would therefore be valuable to identify risk factors for nonunion with nonoperative treatment and to identify the technical details of surgery that maximize rates of healing and minimize the risk of complications.

The limitations of decision analysis involve the methods by which probabilities and utilities are obtained. In general, the most accurate, stable, and robust estimates of outcome probabilities are derived from a meta-analytic synthesis of randomized clinical trials, the highest level of clinical evidence. In our model, data were extracted primarily from level IV studies; only 1 level III study²⁰ and 1 level II study²¹ were available for analysis. Thus, as is the case with many foot and ankle disorders²², the information on treatment of Jones fractures is very limited in its level of clinical evidence.

Determination of outcome utility also has limitations. Utility is a subjective value that an individual places on a specific outcome. This can be very difficult to quantify. In general, the most robust estimates of patient-derived utilities are derived from complex qualitative methods, such as the standard reference gamble or time trade-offs, in which patients are asked to gamble or choose between health states usually referenced to death. In this study, we determined patient-derived utility values from a direct scaling method using a Likert scale because of the complexity of the standard reference gamble and the difficulty of referencing to death for metatarsal fracture. Although use of a direct scale to determine utility values is less rigorous than the standard reference gamble, this technique has been corroborated methodologically,²³ is advantageous in terms of feasibility and reliability,²⁴ and has been successfully used in other orthopedic decision analysis models.^12,25,26 In our estimation, generally active patients without a history of foot pathology constituted a sample of convenience but also were representative of individuals at risk for Jones fracture. Although specific scenarios were presented, the patients who completed the questionnaire may not have had deep insights into the subtleties and implications of the various disease states and treatments. Regardless of how outcome probabilities and utilities are determined, they are considered point estimates in decision analysis, and sensitivity analyses are therefore performed to assess how decision making changes over a range of values.

Conclusion

The results of this study may help optimize the process of deciding between operative and nonoperative treatment for Jones fracture. For a given patient, the optimal strategy depends not only on the probabilities of the various outcomes but also on personal preference. Thus, there may not be one right answer for all patients. Patients who value a higher chance of fracture healing with initial treatment or an earlier return to sports are best treated operatively, whereas patients who are risk-averse and place a high value on fracture healing without surgery should be managed nonoperatively. We therefore advocate a model of shared medical decision-making in which the physician and the patient are jointly involved, considering both outcome probabilities and patient preferences. Ongoing research efforts should focus on predictors of nonunion with nonoperative treatment.

The optimal management strategy for acute fractures of the metadiaphyseal fifth metatarsal (Jones fractures) is controversial. Patients can be successfully treated nonoperatively with non-weight-bearing and immobilization in a short leg cast^1-7 or operatively with placement of an intramedullary screw.^8-10 The primary advantage of nonoperative treatment is avoiding the risks and discomfort of surgery; disadvantages include the need for prolonged immobilization and protected weight-bearing as well as a decreased union rate.^8,9 Advantages of operative treatment include accelerated functional recovery and an improved union rate; disadvantages include exposure to the risks, inconvenience, and discomfort of surgery. Clear, definitive evidence for guiding treatment is not available in the orthopedic literature, and treatment strategies vary substantially according to surgeon and patient preference.

Expected-value decision analysis, a research tool that helps guide decision-making in situations of uncertainty, has been effectively applied to other areas of uncertainty in the orthopedic literature.^11-14 Borrowed from gaming theory, the technique involves creating a decision tree to define the clinical problem, determining outcome probabilities and utilities, performing a fold-back analysis to determine the optimal decision-making strategy, and performing a sensitivity analysis to model the effect of varying outcome probabilities and utilities on decision-making. Decision analysis may therefore allow the clinician and the patient to optimize decision-making based on best available evidence and patient preferences. It also helps determine the most important factors affecting management strategies and the decision-making process, which may not always be intuitive.

In the present study, we used expected-value decision analysis to determine the optimal management strategy, operative or nonoperative, for acute Jones fracture. We also explored factors with the most influence on the model and identified important questions for future research.

Materials and Methods

Institutional review board approval was obtained for this study. Analysis was performed with Treeage Pro statistical software (Treeage Software).

Outcome Probabilities

Outcome probabilities were determined by reviewing the literature for articles on Jones fractures. This body of literature was summarized in a comprehensive review by Dean and colleagues¹⁵, who extracted data from 19 studies: 1 randomized controlled trial, 1 prospective case series, and 17 retrospective case series.¹⁵ We used data from these studies to determine outcome probabilities (Table).

Outcome Utilities

Utilities represent patient preferences for various disease states. Outcome utility values were obtained from 32 adults (25 women, 7 men) with no history of foot injury. Mean age was 32.4 years (range, 20-69 years). The questionnaire presented scenarios for the different outcomes and asked patients to rate these outcomes on a scale ranging from 0 (worst possible outcome) to 10 (best possible outcome). The Sports subscale of the Foot and Ankle Ability Measure (FAAM) 16 was used to quantify patient activity level.

Decision Tree and Fold-Back Analysis

A decision tree was constructed with 1 decision node, 4 chance nodes, and 7 terminal nodes (Figure 1). The decision tree demonstrates 2 different strategies for managing a Jones fracture. The decision node divides the tree into 2 branches: initial operative or nonoperative treatment. Both branches are followed by various chance nodes, each terminating in a discrete clinical outcome. Per convention, utility data were placed to the right of the terminal nodes, and probability data were placed under the terminal nodes.

Fold-back analysis was performed to identify the optimal strategy. Fold-back analysis involves multiplying each outcome utility by its associated probability, thereby providing an “expected value” for each clinical endpoint. Then, the expected values for each endpoint can be summed for a given management strategy, and the ultimate expected values of the different strategies can be compared. The management strategy associated with the highest expected value is optimal for the given outcome utilities and probabilities.

Sensitivity Analysis

One-way sensitivity analysis was performed to model the effect on decision-making of changing the values for utility for uncomplicated surgery, utility for healing with nonoperative treatment, utility for uncomplicated treatment of nonunion, likelihood of healing with nonoperative treatment, likelihood of healing with surgery, and likelihood of minor complication with surgery. These were the variables found to affect the decision-making strategy within their clinically plausible ranges.

Results

Outcome Probabilities and Utilities

Outcome probabilities and utilities are illustrated in Figure 1. By convention, probabilities appear below the corresponding branches of the decision tree, and utilities appear at the end of each branch. Mean (SD) FAAM Sports subscale score was 84.6 (27.4). This subscale is scored as a percentage from 0% to 100%, with higher scores indicating a higher level of physical function.

Decision Analysis

The expected value for nonoperative treatment was 7.74, and the expected value for intramedullary screw fixation was 7.88 (Figure 1). Therefore, operative treatment was identified as the optimal treatment strategy.

Sensitivity analyses revealed that the optimal decision making strategy was very sensitive to small changes in several variables. Nonoperative treatment becomes the preferred strategy when the utility value for uncomplicated surgery falls below 8.04 (Figure 2), when the utility for healing with nonoperative treatment rises above 8.49 (Figure 3), when the likelihood of healing with nonoperative treatment rises above 82% (Figure 4), or when the probability of healing after surgery falls below 92% (Figure 5). Nonoperative treatment is also favored when the probability of minor complication with surgery is above 17% (Figure 6) and when utility for a successfully treated nonunion is higher than 6.9 (Figure 7).

Discussion

Optimal management of a metadiaphyseal fracture of the fifth metatarsal (Jones fracture) remains controversial. The decision between initial operative or nonoperative treatment lends itself to expected-value decision analysis because of well-defined treatment options and relatively discrete outcomes. The principal advantages of nonoperative treatment are that it allows the patient to avoid the risks and discomfort of surgery, and the principal advantages of operative treatment are that it maximizes the chance of fracture union and may accelerate functional recovery.

Our decision analysis determined that operative fixation is the optimal decision path, given the outcome probabilities derived from the literature and the utilities obtained from surveys. This finding is in accordance with several expert opinions in foot and ankle fracture surgery.^17,18 However, the expected values of the operative and nonoperative treatment strategies differed by only 0.3 on a 10-point scale. Such similar expected values in our model are not surprising given the controversy surrounding clinical decision making in the treatment of these fractures.¹⁹

In addition, our analysis identified the important variables in the decision-making process. Patients averse to surgery, patients not averse to successful nonoperative treatment, and patients who view successful nonunion surgery after initial nonoperative treatment as a relatively positive outcome may be best treated nonoperatively. These findings emphasize the importance of patient preferences and shared decision-making. Higher rates of healing with nonoperative treatment, lower rates of healing with surgery, and higher complication rates with surgery also favor nonoperative management. It would therefore be valuable to identify risk factors for nonunion with nonoperative treatment and to identify the technical details of surgery that maximize rates of healing and minimize the risk of complications.

The limitations of decision analysis involve the methods by which probabilities and utilities are obtained. In general, the most accurate, stable, and robust estimates of outcome probabilities are derived from a meta-analytic synthesis of randomized clinical trials, the highest level of clinical evidence. In our model, data were extracted primarily from level IV studies; only 1 level III study²⁰ and 1 level II study²¹ were available for analysis. Thus, as is the case with many foot and ankle disorders²², the information on treatment of Jones fractures is very limited in its level of clinical evidence.

Determination of outcome utility also has limitations. Utility is a subjective value that an individual places on a specific outcome. This can be very difficult to quantify. In general, the most robust estimates of patient-derived utilities are derived from complex qualitative methods, such as the standard reference gamble or time trade-offs, in which patients are asked to gamble or choose between health states usually referenced to death. In this study, we determined patient-derived utility values from a direct scaling method using a Likert scale because of the complexity of the standard reference gamble and the difficulty of referencing to death for metatarsal fracture. Although use of a direct scale to determine utility values is less rigorous than the standard reference gamble, this technique has been corroborated methodologically,²³ is advantageous in terms of feasibility and reliability,²⁴ and has been successfully used in other orthopedic decision analysis models.^12,25,26 In our estimation, generally active patients without a history of foot pathology constituted a sample of convenience but also were representative of individuals at risk for Jones fracture. Although specific scenarios were presented, the patients who completed the questionnaire may not have had deep insights into the subtleties and implications of the various disease states and treatments. Regardless of how outcome probabilities and utilities are determined, they are considered point estimates in decision analysis, and sensitivity analyses are therefore performed to assess how decision making changes over a range of values.

Conclusion

The results of this study may help optimize the process of deciding between operative and nonoperative treatment for Jones fracture. For a given patient, the optimal strategy depends not only on the probabilities of the various outcomes but also on personal preference. Thus, there may not be one right answer for all patients. Patients who value a higher chance of fracture healing with initial treatment or an earlier return to sports are best treated operatively, whereas patients who are risk-averse and place a high value on fracture healing without surgery should be managed nonoperatively. We therefore advocate a model of shared medical decision-making in which the physician and the patient are jointly involved, considering both outcome probabilities and patient preferences. Ongoing research efforts should focus on predictors of nonunion with nonoperative treatment.

The optimal management strategy for acute fractures of the metadiaphyseal fifth metatarsal (Jones fractures) is controversial. Patients can be successfully treated nonoperatively with non-weight-bearing and immobilization in a short leg cast^1-7 or operatively with placement of an intramedullary screw.^8-10 The primary advantage of nonoperative treatment is avoiding the risks and discomfort of surgery; disadvantages include the need for prolonged immobilization and protected weight-bearing as well as a decreased union rate.^8,9 Advantages of operative treatment include accelerated functional recovery and an improved union rate; disadvantages include exposure to the risks, inconvenience, and discomfort of surgery. Clear, definitive evidence for guiding treatment is not available in the orthopedic literature, and treatment strategies vary substantially according to surgeon and patient preference.

Expected-value decision analysis, a research tool that helps guide decision-making in situations of uncertainty, has been effectively applied to other areas of uncertainty in the orthopedic literature.^11-14 Borrowed from gaming theory, the technique involves creating a decision tree to define the clinical problem, determining outcome probabilities and utilities, performing a fold-back analysis to determine the optimal decision-making strategy, and performing a sensitivity analysis to model the effect of varying outcome probabilities and utilities on decision-making. Decision analysis may therefore allow the clinician and the patient to optimize decision-making based on best available evidence and patient preferences. It also helps determine the most important factors affecting management strategies and the decision-making process, which may not always be intuitive.

In the present study, we used expected-value decision analysis to determine the optimal management strategy, operative or nonoperative, for acute Jones fracture. We also explored factors with the most influence on the model and identified important questions for future research.

Materials and Methods

Institutional review board approval was obtained for this study. Analysis was performed with Treeage Pro statistical software (Treeage Software).

Outcome Probabilities

Outcome probabilities were determined by reviewing the literature for articles on Jones fractures. This body of literature was summarized in a comprehensive review by Dean and colleagues¹⁵, who extracted data from 19 studies: 1 randomized controlled trial, 1 prospective case series, and 17 retrospective case series.¹⁵ We used data from these studies to determine outcome probabilities (Table).

Outcome Utilities

Utilities represent patient preferences for various disease states. Outcome utility values were obtained from 32 adults (25 women, 7 men) with no history of foot injury. Mean age was 32.4 years (range, 20-69 years). The questionnaire presented scenarios for the different outcomes and asked patients to rate these outcomes on a scale ranging from 0 (worst possible outcome) to 10 (best possible outcome). The Sports subscale of the Foot and Ankle Ability Measure (FAAM) 16 was used to quantify patient activity level.

Decision Tree and Fold-Back Analysis

A decision tree was constructed with 1 decision node, 4 chance nodes, and 7 terminal nodes (Figure 1). The decision tree demonstrates 2 different strategies for managing a Jones fracture. The decision node divides the tree into 2 branches: initial operative or nonoperative treatment. Both branches are followed by various chance nodes, each terminating in a discrete clinical outcome. Per convention, utility data were placed to the right of the terminal nodes, and probability data were placed under the terminal nodes.

Fold-back analysis was performed to identify the optimal strategy. Fold-back analysis involves multiplying each outcome utility by its associated probability, thereby providing an “expected value” for each clinical endpoint. Then, the expected values for each endpoint can be summed for a given management strategy, and the ultimate expected values of the different strategies can be compared. The management strategy associated with the highest expected value is optimal for the given outcome utilities and probabilities.

Sensitivity Analysis

One-way sensitivity analysis was performed to model the effect on decision-making of changing the values for utility for uncomplicated surgery, utility for healing with nonoperative treatment, utility for uncomplicated treatment of nonunion, likelihood of healing with nonoperative treatment, likelihood of healing with surgery, and likelihood of minor complication with surgery. These were the variables found to affect the decision-making strategy within their clinically plausible ranges.

Results

Outcome Probabilities and Utilities

Outcome probabilities and utilities are illustrated in Figure 1. By convention, probabilities appear below the corresponding branches of the decision tree, and utilities appear at the end of each branch. Mean (SD) FAAM Sports subscale score was 84.6 (27.4). This subscale is scored as a percentage from 0% to 100%, with higher scores indicating a higher level of physical function.

Decision Analysis

The expected value for nonoperative treatment was 7.74, and the expected value for intramedullary screw fixation was 7.88 (Figure 1). Therefore, operative treatment was identified as the optimal treatment strategy.

Sensitivity analyses revealed that the optimal decision making strategy was very sensitive to small changes in several variables. Nonoperative treatment becomes the preferred strategy when the utility value for uncomplicated surgery falls below 8.04 (Figure 2), when the utility for healing with nonoperative treatment rises above 8.49 (Figure 3), when the likelihood of healing with nonoperative treatment rises above 82% (Figure 4), or when the probability of healing after surgery falls below 92% (Figure 5). Nonoperative treatment is also favored when the probability of minor complication with surgery is above 17% (Figure 6) and when utility for a successfully treated nonunion is higher than 6.9 (Figure 7).

Discussion

Optimal management of a metadiaphyseal fracture of the fifth metatarsal (Jones fracture) remains controversial. The decision between initial operative or nonoperative treatment lends itself to expected-value decision analysis because of well-defined treatment options and relatively discrete outcomes. The principal advantages of nonoperative treatment are that it allows the patient to avoid the risks and discomfort of surgery, and the principal advantages of operative treatment are that it maximizes the chance of fracture union and may accelerate functional recovery.

Our decision analysis determined that operative fixation is the optimal decision path, given the outcome probabilities derived from the literature and the utilities obtained from surveys. This finding is in accordance with several expert opinions in foot and ankle fracture surgery.^17,18 However, the expected values of the operative and nonoperative treatment strategies differed by only 0.3 on a 10-point scale. Such similar expected values in our model are not surprising given the controversy surrounding clinical decision making in the treatment of these fractures.¹⁹

In addition, our analysis identified the important variables in the decision-making process. Patients averse to surgery, patients not averse to successful nonoperative treatment, and patients who view successful nonunion surgery after initial nonoperative treatment as a relatively positive outcome may be best treated nonoperatively. These findings emphasize the importance of patient preferences and shared decision-making. Higher rates of healing with nonoperative treatment, lower rates of healing with surgery, and higher complication rates with surgery also favor nonoperative management. It would therefore be valuable to identify risk factors for nonunion with nonoperative treatment and to identify the technical details of surgery that maximize rates of healing and minimize the risk of complications.

The limitations of decision analysis involve the methods by which probabilities and utilities are obtained. In general, the most accurate, stable, and robust estimates of outcome probabilities are derived from a meta-analytic synthesis of randomized clinical trials, the highest level of clinical evidence. In our model, data were extracted primarily from level IV studies; only 1 level III study²⁰ and 1 level II study²¹ were available for analysis. Thus, as is the case with many foot and ankle disorders²², the information on treatment of Jones fractures is very limited in its level of clinical evidence.

Determination of outcome utility also has limitations. Utility is a subjective value that an individual places on a specific outcome. This can be very difficult to quantify. In general, the most robust estimates of patient-derived utilities are derived from complex qualitative methods, such as the standard reference gamble or time trade-offs, in which patients are asked to gamble or choose between health states usually referenced to death. In this study, we determined patient-derived utility values from a direct scaling method using a Likert scale because of the complexity of the standard reference gamble and the difficulty of referencing to death for metatarsal fracture. Although use of a direct scale to determine utility values is less rigorous than the standard reference gamble, this technique has been corroborated methodologically,²³ is advantageous in terms of feasibility and reliability,²⁴ and has been successfully used in other orthopedic decision analysis models.^12,25,26 In our estimation, generally active patients without a history of foot pathology constituted a sample of convenience but also were representative of individuals at risk for Jones fracture. Although specific scenarios were presented, the patients who completed the questionnaire may not have had deep insights into the subtleties and implications of the various disease states and treatments. Regardless of how outcome probabilities and utilities are determined, they are considered point estimates in decision analysis, and sensitivity analyses are therefore performed to assess how decision making changes over a range of values.

Conclusion

The results of this study may help optimize the process of deciding between operative and nonoperative treatment for Jones fracture. For a given patient, the optimal strategy depends not only on the probabilities of the various outcomes but also on personal preference. Thus, there may not be one right answer for all patients. Patients who value a higher chance of fracture healing with initial treatment or an earlier return to sports are best treated operatively, whereas patients who are risk-averse and place a high value on fracture healing without surgery should be managed nonoperatively. We therefore advocate a model of shared medical decision-making in which the physician and the patient are jointly involved, considering both outcome probabilities and patient preferences. Ongoing research efforts should focus on predictors of nonunion with nonoperative treatment.

Injuries among professional baseball players have been on the rise for several years.^1,2 From 1989 to 1999, the number of disabled list (DL) reports increased 38% (266 to 367 annual reports),¹ and a similar increase in injury rates was noted from the 2002 to the 2008 seasons (37%).² These injuries have important implications for future injury risk and time away from play. Identifying these injuries and determining correlates and risk factors is important for targeted prevention efforts.

Several studies have explored the prevalence of upper extremity injuries in professional and collegiate baseball players;^2-4 however, detailed epidemiology of knee injuries in Major League Baseball (MLB) and Minor League Baseball (MiLB) players is lacking. Much more is known about the prevalence, treatment, and outcomes of knee injuries in other professional sporting organizations, such as the National Basketball Association (NBA), National Football League (NFL), and National Hockey League (NHL).^4-12 A recent meta-analysis exploring injuries in professional athletes found that studies on lower extremity injuries comprised approximately 12% of the literature reporting injuries in MLB players.⁴ In other professional leagues, publications on lower extremity injuries comprise approximately 56% of the sports medicine literature in the NFL, 54% in the NBA, and 62% in the NHL.⁴ Since few studies have investigated lower extremity injuries among professional baseball players, there is an opportunity for additional research to guide evidence-based prevention strategies.

A better understanding of the nature of these injuries is one of the first steps towards developing targeted injury prevention programs and treatment algorithms. The study of injury epidemiology among professional baseball players has been aided by the creation of an injury tracking system initiated by the MLB, its minor league affiliates, and the Major League Baseball Players Association.^5,13,14 This surveillance system allows for the tracking of medical histories and injuries to players as they move across major and minor league organizations. Similar systems have been utilized in the National Collegiate Athletic Association and other professional sports organizations.^3,15-17 A unique advantage of the MLB surveillance system is the required participation of all major and minor league teams, which allows for investigation of the entire population of players rather than simply a sample of players from select teams. This system has propelled an effort to identify injury patterns as a means of developing appropriate targets for potential preventative measures.⁵

The purpose of this descriptive epidemiologic study is to better understand the distribution and characteristics of knee injuries in these elite athletes by reporting on all knee injuries occurring over a span of 4 seasons (2011-2014). Additionally, this study seeks to characterize the impact of these injuries by analyzing the time required for return to play and the treatments rendered (surgical and nonsurgical).

Materials and Methods

After approval from the Johns Hopkins Bloomberg School of Public Health Institutional Review Board, detailed data regarding knee injuries in both MLB and MiLB baseball players were extracted from the de-identified MLB Health and Injury Tracking System (HITS). The HITS database is a centralized database that contains data on injuries from an electronic medical record (EMR). All players provided consent to have their data included in this EMR. HITS system captures injuries reported by the athletic trainers for all professional baseball players from 30 MLB clubs and their 230 minor league affiliates. Additional details on this population of professional baseball players have been published elsewhere.⁵ Only injuries that result in time out of play (≥1 day missed) are included in the database, and they are logged with basic information such as region of the body, diagnosis, date, player position, activity leading to injury, and general treatment. Any injury that affects participation in any aspect of baseball-related activity (eg, game, practice, warm-up, conditioning, weight training) is captured in HITS.

All baseball-related knee injuries occurring during the 2011-2014 seasons that resulted in time out of sport were included in the study. These injuries were identified based on the Sports Medicine Diagnostic Coding System (SMDCS) to capture injuries by diagnostic groups.¹⁸ Knee injuries were included if they occurred during spring training, regular season, or postseason play. Offseason injuries were not included. Injury events that were classified as “season-ending” were not included in the analysis of days missed because many of these players may not have been cleared to play until the beginning of the following season. To determine the proportion of knee injuries during the study period, all injuries were included for comparative purposes (subdivided based on 30 anatomic regions or types).

For each knee injury, a number of variables were analyzed, including diagnosis, level of play (MLB vs. MiLB), age, player position at the time of injury (pitcher, catcher, infield, outfield, base runner, or batter), field location where the injury occurred (home plate, pitcher’s mound, infield, outfield, foul territory or bullpen, or other), mechanism of injury, days missed, and treatment rendered (conservative vs surgical). The classification used to describe the mechanism of injury consisted of contact with ball, contact with ground, contact with another player, contact with another object, or noncontact.

Statistical Analysis
Epidemiologic data are presented with descriptive statistics such as mean, median, frequency, and percentage where appropriate. When comparing player age, days missed, and surgical vs nonsurgical treatment between MLB and MiLB players, t-tests and tests for difference in proportions were applied as appropriate. Statistical significance was established for P values < .05.

The distribution of days missed for the variables considered was often skewed to the right (ie, days missed mostly concentrated on the low to moderate number of days, with fewer values in the much higher days missed range), even after excluding the season-ending injuries; hence the mean (or average) days missed was often larger than the median days missed. Reporting the median would allow for a robust estimate of the expected number of days missed, but would down weight those instances when knee injuries result in much longer missed days, as reflected by the mean. Because of the importance of the days missed measure for professional baseball, both the mean and median are presented.

In order to estimate exposure, the average number of players per team per game was calculated based on analysis of regular season game participation via box scores. This average number over a season, multiplied by the number of team games at each professional level of baseball, was used as an estimate of athlete exposures in order to provide rates comparable to those of other injury surveillance systems. Injury rates were reported as injuries per 1000 athlete-exposures (AE) for those knee injuries that occurred during the regular season. It should be noted that the number of regular season knee injuries and the subsequent AE rates are based on injuries that were deemed work-related during the regular season. This does not necessarily only include injuries occurring during the course of a game, but injuries in game preparation as well. Due to the variations in spring training games and fluctuating rosters, an exposure rate could not be calculated for spring training knee injuries.

RESULTS

Overall Summary

Of the 30 general body regions/systems included in the HITS database, injuries to the knee were the fifth most common reason for days missed in all of professional baseball from 2011-2014 (Table 1). Injuries to the knee represented 6.5% of the nearly 34,000 injuries sustained during the study period. Knee injuries were the fifth most common reason for time out of play for players in both the MiLB and MLB.

A total of 2171 isolated knee injuries resulted in time out of sport for professional baseball players (Table 2). Of these, 410 (19%) occurred in MLB players and 1761 (81%) occurred in MiLB players. MLB players were older than MiLB players at the time of injury (29.5 vs 22.8 years, respectively). Overall mean number of days missed was 16.2 days per knee injury, with MLB players missing an approximately 7 days more per injury than MiLB athletes (21.8 vs. 14.9 days respectively; P = .001).Over the course of the 4 seasons, a total of 30,449 days were missed due to knee injuries in professional baseball, giving an average rate of 7612 days lost per season. Surgery was performed for 263 (12.1%) of the 2171 knee injuries, with a greater proportion of MLB players requiring surgery than MiLB players (17.3% vs 10.9%) (P < .001). With respect to number of days missed per injury, 26% of knee injuries in the minor leagues resulted in greater than 30 days missed, while this number rose to 32% for knee injuries in MLB players (Table 3).

For regular season games, it was estimated that there were 1,197,738 MiLB and 276,608 MLB AE, respectively, over the course of the 4 seasons (2011-2014). The overall knee injury rate across both the MiLB and MLB was 1.2 per 1000 AE, based on the subset of 308 and 1473 regular season knee injuries in MiLB and MLB, respectively. The rate of knee injury was similar and not significantly different between the MiLB and MLB (1.2 per 1000 AE in the MiLB and 1.1 per 1000 AE in the MLB).

Characteristics of Injuries

When considering the position of the player during injury, defensive players were most frequently injured (n = 742, 56.5%), with pitchers (n = 227, 17.3%), infielders (n =193, 14.7%), outfielders (n = 193, 14.7%), and catchers (n = 129, 9.8%) sustaining injuries in decreasing frequency. Injuries while on offense (n = 571, 43.5%) were most frequent in base runners (n = 320, 24.4%) followed by batters (n = 251, 19.1%) (Table 4). Injuries while on defense occurring in infielders and catchers resulted in the longest period of time away from play (average of 22.4 and 20.8 days missed, respectively), while those occurring in batters resulted in the least average days missed (8.9 days).

The most common field location for knee injuries to occur was the infield, which was responsible for n = 647 (29.8%) of the total knee injuries (Table 4). This was followed by home plate (n = 493, 22.7%), other locations outside those specified (n = 394, 18.1%), outfield (n = 320, 14.7%), pitcher’s mound (n = 210, 9.7%), and foul territory or the bullpen (n = 107, 4.9%). Of the knee injuries with a specified location, those occurring in foul territory or the bullpen resulted in the highest mean days missed (18.4), while those occurring at home plate resulted in the least mean days missed (13.4 days).

When analyzed by mechanism of injury, noncontact injuries (n = 953, 43.9%) were more common than being hit with the ball (n = 374, 17.2%), striking the ground (n = 409, 18.8%), other mechanisms not listed (n = 196, 9%), contact with another player (n = 176, 8.1%), or contact with other objects (n = 63, 2.9%) (Table 4). Noncontact injuries and player to player collisions resulted in the greatest number of missed days (21.6 and 17.1 days, respectively) while being struck by the ball resulted in the least mean days missed (5.1).

Of the n = 493 knee injuries occurring at home plate, n = 212 (43%) occurred to the batter, n = 100 (20%) to the catcher, n = 34 (6.9%) to base runners, and n = 7 (1.4%) to pitchers (Table 5). The majority of knee injuries in the infield occurred to base runners (n = 283, 43.7%). Player-to-player collisions at home plate were responsible for 51 (2.3%) knee injuries, while 163 (24%) were noncontact injuries and 376 (56%) were the result of a player being hit by the ball (Table 5).

Injury Diagnosis

By diagnosis, the most common knee injuries observed were contusions or hematomas (n = 662, 30.5%), other injuries (n = 415, 19.1%), sprains or ligament injuries (n = 380, 17.5%), tendinopathies or bursitis (n = 367, 16.9%), and meniscal or cartilage injury (n = 200, 9.2%) (Table 6). Injuries resulting in the greatest mean number of days missed included meniscal or cartilage injuries (44 days), sprains or ligament injuries (30 days), or dislocations (22 days).

Based on specific SMDCS descriptors, the most frequent knee injuries reported were contusion (n = 662, 30.5%), patella tendinopathy (n = 222, 10.2%), and meniscal tears (n = 200, 9.2%) (Table 6). Complete anterior cruciate ligament tears, although infrequent, were responsible for the greatest mean days missed (156.2 days). This was followed by lateral meniscus tears (47.5 days) and medial meniscus tears (41.2 days). Knee contusions, although very common, resulted in the least number of days missed (6.0 days).

Discussion

Although much is known about knee injuries in other professional athletic leagues, little is known about knee injuries in professional baseball players.^2-4 The majority of epidemiologic studies regarding baseball players at any level emphasizes the study of shoulder and elbow injuries.^3,4,19 Since the implementation of the electronic medical record and the HITS database in professional baseball, there has been increased effort to document injuries that have received less attention in the existing literature. Understanding the epidemiology of these injuries is important for the development of targeted prevention efforts.

Prior studies of injuries in professional baseball relied on data captured by the publicly available DL. Posner and colleagues² provide one of the most comprehensive reports on MLB injuries in a report utilizing DL assignment data over a period of 7 seasons.They demonstrated that knee injuries were responsible for 7.7% (12.5% for fielders and 3.7% for pitchers) of assignments to the DL. The current study utilized a comprehensive surveillance and builds on this existing knowledge. The present study found similar trends to Posner and colleagues² in that knee injuries were responsible for 6.5% of injuries in professional baseball players that resulted in missed games. From the 2002 season to the 2008 season, knee injuries were the fifth most common reason MLB players were placed on the DL,² and the current study indicates that they remain the fifth most common reason for missed time from play based on the HITS data. Since the prevalence of these injuries have remained constant since the 2002 season, efforts to better understand these injuries are warranted in order to identify strategies to prevent them. These analyses have generated important data towards achieving this understanding.

As with most injuries in professional sports, goals for treatment are aimed at maximizing patient function and performance while minimizing time out of play. For the 2011-2014 professional baseball seasons, a total of 2171 players sustained knee injuries and missed an average of 16.2 days per injury. Knee injuries were responsible for a total of 7612 days of missed work for MLB and MiLB players per season (30,449 days over the 4-season study period). This is equivalent to a total of 20.9 years of players’ time lost in professional baseball per season over the last 4 years. The implications of this amount of time away from sport are significant, and further study should be targeted at prevention of these injuries and optimizing return to play times.

When attempting to reduce the burden of knee injuries in professional baseball, it may prove beneficial to first understand how the injuries occur, where on the field, and who is at greatest risk. From 2011 to 2014, nearly 44% of knee injuries occurred by noncontact mechanisms. Among all locations on the field where knee injuries occurred, those occurring in the infield were responsible for the greatest mean days missed. The players who seem to be at greatest risk for knee injuries appear to be base runners. These data suggest the need for prevention efforts targeting base runners and infield players, as well as players in MiLB, where the largest number of injuries occurred.

Recently, playing rules implemented by MLB after consultation with players have focused on reducing the number of player-to-player collisions at home plate in an attempt to decrease the injury burden to catchers and base runners.²⁰ This present analysis suggests that this rule change may also reduce the occurrence of knee injuries, as player collisions at home plate were responsible for a total of 51 knee injuries during the study period. The impact of this rule change on injury rates should also be explored. Interestingly, of the 51 knees injuries occurring due to contact at home plate, 23 occurred in 2011, and only 2 occurred in 2014—the first year of the new rule. Additional areas that resulted in high numbers of knee injuries were player-to-player contact in the infield and player contact with the ground in the infield.

Attempting to reduce injury burden and time out of play related to knee injuries in professional baseball players will likely prove to be a difficult task. In order to generate meaningful improvement, a comprehensive approach that involves players, management, trainers, therapists, and physicians will likely be required. As the first report of the epidemiology of knee injuries in professional baseball players, this study is one important step in that process. The strengths of this study are its comprehensive nature that analyzes injuries from an entire population of players on more than 200 teams over a 3-year period. Also, this research is strengthened by its focus on one particular region of the body that has received limited attention in the empirical literature, but represents a significant source of lost time during the baseball season.

There are some limitations to this study. As with any injury surveillance system, there is the possibility that not all cases were captured. Additionally, since the surveillance system is based on data from multiple teams, data entry discrepancy is possible; however, the presence of dropdown boxes and systematic definitions for injuries reduces this risk. Finally, this study did not investigate the various treatments for knee injuries beyond whether or not the injury required surgery. Since this was the first comprehensive exploration of knee injuries in professional baseball, future studies are needed to explore additional facets including outcomes related to treatment, return to play, and performance.

Conclusion

Knee injuries represent 6.5% of all injuries in professional baseball, occurring at a rate of 1.3 per 1000 AE. The burden of these injuries is significant for professional baseball players. This study fills a critical gap in sports injury research by contributing to the knowledge about the effect of knee injuries in professional baseball. It also provides an important foundation for future epidemiologic inquiry to identify modifiable risk factors and interventions that may reduce the impact of these injuries in athletes.

Injuries among professional baseball players have been on the rise for several years.^1,2 From 1989 to 1999, the number of disabled list (DL) reports increased 38% (266 to 367 annual reports),¹ and a similar increase in injury rates was noted from the 2002 to the 2008 seasons (37%).² These injuries have important implications for future injury risk and time away from play. Identifying these injuries and determining correlates and risk factors is important for targeted prevention efforts.

Several studies have explored the prevalence of upper extremity injuries in professional and collegiate baseball players;^2-4 however, detailed epidemiology of knee injuries in Major League Baseball (MLB) and Minor League Baseball (MiLB) players is lacking. Much more is known about the prevalence, treatment, and outcomes of knee injuries in other professional sporting organizations, such as the National Basketball Association (NBA), National Football League (NFL), and National Hockey League (NHL).^4-12 A recent meta-analysis exploring injuries in professional athletes found that studies on lower extremity injuries comprised approximately 12% of the literature reporting injuries in MLB players.⁴ In other professional leagues, publications on lower extremity injuries comprise approximately 56% of the sports medicine literature in the NFL, 54% in the NBA, and 62% in the NHL.⁴ Since few studies have investigated lower extremity injuries among professional baseball players, there is an opportunity for additional research to guide evidence-based prevention strategies.

A better understanding of the nature of these injuries is one of the first steps towards developing targeted injury prevention programs and treatment algorithms. The study of injury epidemiology among professional baseball players has been aided by the creation of an injury tracking system initiated by the MLB, its minor league affiliates, and the Major League Baseball Players Association.^5,13,14 This surveillance system allows for the tracking of medical histories and injuries to players as they move across major and minor league organizations. Similar systems have been utilized in the National Collegiate Athletic Association and other professional sports organizations.^3,15-17 A unique advantage of the MLB surveillance system is the required participation of all major and minor league teams, which allows for investigation of the entire population of players rather than simply a sample of players from select teams. This system has propelled an effort to identify injury patterns as a means of developing appropriate targets for potential preventative measures.⁵

The purpose of this descriptive epidemiologic study is to better understand the distribution and characteristics of knee injuries in these elite athletes by reporting on all knee injuries occurring over a span of 4 seasons (2011-2014). Additionally, this study seeks to characterize the impact of these injuries by analyzing the time required for return to play and the treatments rendered (surgical and nonsurgical).

Materials and Methods

After approval from the Johns Hopkins Bloomberg School of Public Health Institutional Review Board, detailed data regarding knee injuries in both MLB and MiLB baseball players were extracted from the de-identified MLB Health and Injury Tracking System (HITS). The HITS database is a centralized database that contains data on injuries from an electronic medical record (EMR). All players provided consent to have their data included in this EMR. HITS system captures injuries reported by the athletic trainers for all professional baseball players from 30 MLB clubs and their 230 minor league affiliates. Additional details on this population of professional baseball players have been published elsewhere.⁵ Only injuries that result in time out of play (≥1 day missed) are included in the database, and they are logged with basic information such as region of the body, diagnosis, date, player position, activity leading to injury, and general treatment. Any injury that affects participation in any aspect of baseball-related activity (eg, game, practice, warm-up, conditioning, weight training) is captured in HITS.

All baseball-related knee injuries occurring during the 2011-2014 seasons that resulted in time out of sport were included in the study. These injuries were identified based on the Sports Medicine Diagnostic Coding System (SMDCS) to capture injuries by diagnostic groups.¹⁸ Knee injuries were included if they occurred during spring training, regular season, or postseason play. Offseason injuries were not included. Injury events that were classified as “season-ending” were not included in the analysis of days missed because many of these players may not have been cleared to play until the beginning of the following season. To determine the proportion of knee injuries during the study period, all injuries were included for comparative purposes (subdivided based on 30 anatomic regions or types).

For each knee injury, a number of variables were analyzed, including diagnosis, level of play (MLB vs. MiLB), age, player position at the time of injury (pitcher, catcher, infield, outfield, base runner, or batter), field location where the injury occurred (home plate, pitcher’s mound, infield, outfield, foul territory or bullpen, or other), mechanism of injury, days missed, and treatment rendered (conservative vs surgical). The classification used to describe the mechanism of injury consisted of contact with ball, contact with ground, contact with another player, contact with another object, or noncontact.

Statistical Analysis
Epidemiologic data are presented with descriptive statistics such as mean, median, frequency, and percentage where appropriate. When comparing player age, days missed, and surgical vs nonsurgical treatment between MLB and MiLB players, t-tests and tests for difference in proportions were applied as appropriate. Statistical significance was established for P values < .05.

The distribution of days missed for the variables considered was often skewed to the right (ie, days missed mostly concentrated on the low to moderate number of days, with fewer values in the much higher days missed range), even after excluding the season-ending injuries; hence the mean (or average) days missed was often larger than the median days missed. Reporting the median would allow for a robust estimate of the expected number of days missed, but would down weight those instances when knee injuries result in much longer missed days, as reflected by the mean. Because of the importance of the days missed measure for professional baseball, both the mean and median are presented.

In order to estimate exposure, the average number of players per team per game was calculated based on analysis of regular season game participation via box scores. This average number over a season, multiplied by the number of team games at each professional level of baseball, was used as an estimate of athlete exposures in order to provide rates comparable to those of other injury surveillance systems. Injury rates were reported as injuries per 1000 athlete-exposures (AE) for those knee injuries that occurred during the regular season. It should be noted that the number of regular season knee injuries and the subsequent AE rates are based on injuries that were deemed work-related during the regular season. This does not necessarily only include injuries occurring during the course of a game, but injuries in game preparation as well. Due to the variations in spring training games and fluctuating rosters, an exposure rate could not be calculated for spring training knee injuries.

RESULTS

Overall Summary

Of the 30 general body regions/systems included in the HITS database, injuries to the knee were the fifth most common reason for days missed in all of professional baseball from 2011-2014 (Table 1). Injuries to the knee represented 6.5% of the nearly 34,000 injuries sustained during the study period. Knee injuries were the fifth most common reason for time out of play for players in both the MiLB and MLB.

A total of 2171 isolated knee injuries resulted in time out of sport for professional baseball players (Table 2). Of these, 410 (19%) occurred in MLB players and 1761 (81%) occurred in MiLB players. MLB players were older than MiLB players at the time of injury (29.5 vs 22.8 years, respectively). Overall mean number of days missed was 16.2 days per knee injury, with MLB players missing an approximately 7 days more per injury than MiLB athletes (21.8 vs. 14.9 days respectively; P = .001).Over the course of the 4 seasons, a total of 30,449 days were missed due to knee injuries in professional baseball, giving an average rate of 7612 days lost per season. Surgery was performed for 263 (12.1%) of the 2171 knee injuries, with a greater proportion of MLB players requiring surgery than MiLB players (17.3% vs 10.9%) (P < .001). With respect to number of days missed per injury, 26% of knee injuries in the minor leagues resulted in greater than 30 days missed, while this number rose to 32% for knee injuries in MLB players (Table 3).

For regular season games, it was estimated that there were 1,197,738 MiLB and 276,608 MLB AE, respectively, over the course of the 4 seasons (2011-2014). The overall knee injury rate across both the MiLB and MLB was 1.2 per 1000 AE, based on the subset of 308 and 1473 regular season knee injuries in MiLB and MLB, respectively. The rate of knee injury was similar and not significantly different between the MiLB and MLB (1.2 per 1000 AE in the MiLB and 1.1 per 1000 AE in the MLB).

Characteristics of Injuries

When considering the position of the player during injury, defensive players were most frequently injured (n = 742, 56.5%), with pitchers (n = 227, 17.3%), infielders (n =193, 14.7%), outfielders (n = 193, 14.7%), and catchers (n = 129, 9.8%) sustaining injuries in decreasing frequency. Injuries while on offense (n = 571, 43.5%) were most frequent in base runners (n = 320, 24.4%) followed by batters (n = 251, 19.1%) (Table 4). Injuries while on defense occurring in infielders and catchers resulted in the longest period of time away from play (average of 22.4 and 20.8 days missed, respectively), while those occurring in batters resulted in the least average days missed (8.9 days).

The most common field location for knee injuries to occur was the infield, which was responsible for n = 647 (29.8%) of the total knee injuries (Table 4). This was followed by home plate (n = 493, 22.7%), other locations outside those specified (n = 394, 18.1%), outfield (n = 320, 14.7%), pitcher’s mound (n = 210, 9.7%), and foul territory or the bullpen (n = 107, 4.9%). Of the knee injuries with a specified location, those occurring in foul territory or the bullpen resulted in the highest mean days missed (18.4), while those occurring at home plate resulted in the least mean days missed (13.4 days).

When analyzed by mechanism of injury, noncontact injuries (n = 953, 43.9%) were more common than being hit with the ball (n = 374, 17.2%), striking the ground (n = 409, 18.8%), other mechanisms not listed (n = 196, 9%), contact with another player (n = 176, 8.1%), or contact with other objects (n = 63, 2.9%) (Table 4). Noncontact injuries and player to player collisions resulted in the greatest number of missed days (21.6 and 17.1 days, respectively) while being struck by the ball resulted in the least mean days missed (5.1).

Of the n = 493 knee injuries occurring at home plate, n = 212 (43%) occurred to the batter, n = 100 (20%) to the catcher, n = 34 (6.9%) to base runners, and n = 7 (1.4%) to pitchers (Table 5). The majority of knee injuries in the infield occurred to base runners (n = 283, 43.7%). Player-to-player collisions at home plate were responsible for 51 (2.3%) knee injuries, while 163 (24%) were noncontact injuries and 376 (56%) were the result of a player being hit by the ball (Table 5).

Injury Diagnosis

By diagnosis, the most common knee injuries observed were contusions or hematomas (n = 662, 30.5%), other injuries (n = 415, 19.1%), sprains or ligament injuries (n = 380, 17.5%), tendinopathies or bursitis (n = 367, 16.9%), and meniscal or cartilage injury (n = 200, 9.2%) (Table 6). Injuries resulting in the greatest mean number of days missed included meniscal or cartilage injuries (44 days), sprains or ligament injuries (30 days), or dislocations (22 days).

Based on specific SMDCS descriptors, the most frequent knee injuries reported were contusion (n = 662, 30.5%), patella tendinopathy (n = 222, 10.2%), and meniscal tears (n = 200, 9.2%) (Table 6). Complete anterior cruciate ligament tears, although infrequent, were responsible for the greatest mean days missed (156.2 days). This was followed by lateral meniscus tears (47.5 days) and medial meniscus tears (41.2 days). Knee contusions, although very common, resulted in the least number of days missed (6.0 days).

Discussion

Although much is known about knee injuries in other professional athletic leagues, little is known about knee injuries in professional baseball players.^2-4 The majority of epidemiologic studies regarding baseball players at any level emphasizes the study of shoulder and elbow injuries.^3,4,19 Since the implementation of the electronic medical record and the HITS database in professional baseball, there has been increased effort to document injuries that have received less attention in the existing literature. Understanding the epidemiology of these injuries is important for the development of targeted prevention efforts.

Prior studies of injuries in professional baseball relied on data captured by the publicly available DL. Posner and colleagues² provide one of the most comprehensive reports on MLB injuries in a report utilizing DL assignment data over a period of 7 seasons.They demonstrated that knee injuries were responsible for 7.7% (12.5% for fielders and 3.7% for pitchers) of assignments to the DL. The current study utilized a comprehensive surveillance and builds on this existing knowledge. The present study found similar trends to Posner and colleagues² in that knee injuries were responsible for 6.5% of injuries in professional baseball players that resulted in missed games. From the 2002 season to the 2008 season, knee injuries were the fifth most common reason MLB players were placed on the DL,² and the current study indicates that they remain the fifth most common reason for missed time from play based on the HITS data. Since the prevalence of these injuries have remained constant since the 2002 season, efforts to better understand these injuries are warranted in order to identify strategies to prevent them. These analyses have generated important data towards achieving this understanding.

As with most injuries in professional sports, goals for treatment are aimed at maximizing patient function and performance while minimizing time out of play. For the 2011-2014 professional baseball seasons, a total of 2171 players sustained knee injuries and missed an average of 16.2 days per injury. Knee injuries were responsible for a total of 7612 days of missed work for MLB and MiLB players per season (30,449 days over the 4-season study period). This is equivalent to a total of 20.9 years of players’ time lost in professional baseball per season over the last 4 years. The implications of this amount of time away from sport are significant, and further study should be targeted at prevention of these injuries and optimizing return to play times.

When attempting to reduce the burden of knee injuries in professional baseball, it may prove beneficial to first understand how the injuries occur, where on the field, and who is at greatest risk. From 2011 to 2014, nearly 44% of knee injuries occurred by noncontact mechanisms. Among all locations on the field where knee injuries occurred, those occurring in the infield were responsible for the greatest mean days missed. The players who seem to be at greatest risk for knee injuries appear to be base runners. These data suggest the need for prevention efforts targeting base runners and infield players, as well as players in MiLB, where the largest number of injuries occurred.

Recently, playing rules implemented by MLB after consultation with players have focused on reducing the number of player-to-player collisions at home plate in an attempt to decrease the injury burden to catchers and base runners.²⁰ This present analysis suggests that this rule change may also reduce the occurrence of knee injuries, as player collisions at home plate were responsible for a total of 51 knee injuries during the study period. The impact of this rule change on injury rates should also be explored. Interestingly, of the 51 knees injuries occurring due to contact at home plate, 23 occurred in 2011, and only 2 occurred in 2014—the first year of the new rule. Additional areas that resulted in high numbers of knee injuries were player-to-player contact in the infield and player contact with the ground in the infield.

Attempting to reduce injury burden and time out of play related to knee injuries in professional baseball players will likely prove to be a difficult task. In order to generate meaningful improvement, a comprehensive approach that involves players, management, trainers, therapists, and physicians will likely be required. As the first report of the epidemiology of knee injuries in professional baseball players, this study is one important step in that process. The strengths of this study are its comprehensive nature that analyzes injuries from an entire population of players on more than 200 teams over a 3-year period. Also, this research is strengthened by its focus on one particular region of the body that has received limited attention in the empirical literature, but represents a significant source of lost time during the baseball season.

There are some limitations to this study. As with any injury surveillance system, there is the possibility that not all cases were captured. Additionally, since the surveillance system is based on data from multiple teams, data entry discrepancy is possible; however, the presence of dropdown boxes and systematic definitions for injuries reduces this risk. Finally, this study did not investigate the various treatments for knee injuries beyond whether or not the injury required surgery. Since this was the first comprehensive exploration of knee injuries in professional baseball, future studies are needed to explore additional facets including outcomes related to treatment, return to play, and performance.

Conclusion

Knee injuries represent 6.5% of all injuries in professional baseball, occurring at a rate of 1.3 per 1000 AE. The burden of these injuries is significant for professional baseball players. This study fills a critical gap in sports injury research by contributing to the knowledge about the effect of knee injuries in professional baseball. It also provides an important foundation for future epidemiologic inquiry to identify modifiable risk factors and interventions that may reduce the impact of these injuries in athletes.

Injuries among professional baseball players have been on the rise for several years.^1,2 From 1989 to 1999, the number of disabled list (DL) reports increased 38% (266 to 367 annual reports),¹ and a similar increase in injury rates was noted from the 2002 to the 2008 seasons (37%).² These injuries have important implications for future injury risk and time away from play. Identifying these injuries and determining correlates and risk factors is important for targeted prevention efforts.

Several studies have explored the prevalence of upper extremity injuries in professional and collegiate baseball players;^2-4 however, detailed epidemiology of knee injuries in Major League Baseball (MLB) and Minor League Baseball (MiLB) players is lacking. Much more is known about the prevalence, treatment, and outcomes of knee injuries in other professional sporting organizations, such as the National Basketball Association (NBA), National Football League (NFL), and National Hockey League (NHL).^4-12 A recent meta-analysis exploring injuries in professional athletes found that studies on lower extremity injuries comprised approximately 12% of the literature reporting injuries in MLB players.⁴ In other professional leagues, publications on lower extremity injuries comprise approximately 56% of the sports medicine literature in the NFL, 54% in the NBA, and 62% in the NHL.⁴ Since few studies have investigated lower extremity injuries among professional baseball players, there is an opportunity for additional research to guide evidence-based prevention strategies.

A better understanding of the nature of these injuries is one of the first steps towards developing targeted injury prevention programs and treatment algorithms. The study of injury epidemiology among professional baseball players has been aided by the creation of an injury tracking system initiated by the MLB, its minor league affiliates, and the Major League Baseball Players Association.^5,13,14 This surveillance system allows for the tracking of medical histories and injuries to players as they move across major and minor league organizations. Similar systems have been utilized in the National Collegiate Athletic Association and other professional sports organizations.^3,15-17 A unique advantage of the MLB surveillance system is the required participation of all major and minor league teams, which allows for investigation of the entire population of players rather than simply a sample of players from select teams. This system has propelled an effort to identify injury patterns as a means of developing appropriate targets for potential preventative measures.⁵

The purpose of this descriptive epidemiologic study is to better understand the distribution and characteristics of knee injuries in these elite athletes by reporting on all knee injuries occurring over a span of 4 seasons (2011-2014). Additionally, this study seeks to characterize the impact of these injuries by analyzing the time required for return to play and the treatments rendered (surgical and nonsurgical).

Materials and Methods

After approval from the Johns Hopkins Bloomberg School of Public Health Institutional Review Board, detailed data regarding knee injuries in both MLB and MiLB baseball players were extracted from the de-identified MLB Health and Injury Tracking System (HITS). The HITS database is a centralized database that contains data on injuries from an electronic medical record (EMR). All players provided consent to have their data included in this EMR. HITS system captures injuries reported by the athletic trainers for all professional baseball players from 30 MLB clubs and their 230 minor league affiliates. Additional details on this population of professional baseball players have been published elsewhere.⁵ Only injuries that result in time out of play (≥1 day missed) are included in the database, and they are logged with basic information such as region of the body, diagnosis, date, player position, activity leading to injury, and general treatment. Any injury that affects participation in any aspect of baseball-related activity (eg, game, practice, warm-up, conditioning, weight training) is captured in HITS.

All baseball-related knee injuries occurring during the 2011-2014 seasons that resulted in time out of sport were included in the study. These injuries were identified based on the Sports Medicine Diagnostic Coding System (SMDCS) to capture injuries by diagnostic groups.¹⁸ Knee injuries were included if they occurred during spring training, regular season, or postseason play. Offseason injuries were not included. Injury events that were classified as “season-ending” were not included in the analysis of days missed because many of these players may not have been cleared to play until the beginning of the following season. To determine the proportion of knee injuries during the study period, all injuries were included for comparative purposes (subdivided based on 30 anatomic regions or types).

For each knee injury, a number of variables were analyzed, including diagnosis, level of play (MLB vs. MiLB), age, player position at the time of injury (pitcher, catcher, infield, outfield, base runner, or batter), field location where the injury occurred (home plate, pitcher’s mound, infield, outfield, foul territory or bullpen, or other), mechanism of injury, days missed, and treatment rendered (conservative vs surgical). The classification used to describe the mechanism of injury consisted of contact with ball, contact with ground, contact with another player, contact with another object, or noncontact.

Statistical Analysis
Epidemiologic data are presented with descriptive statistics such as mean, median, frequency, and percentage where appropriate. When comparing player age, days missed, and surgical vs nonsurgical treatment between MLB and MiLB players, t-tests and tests for difference in proportions were applied as appropriate. Statistical significance was established for P values < .05.

The distribution of days missed for the variables considered was often skewed to the right (ie, days missed mostly concentrated on the low to moderate number of days, with fewer values in the much higher days missed range), even after excluding the season-ending injuries; hence the mean (or average) days missed was often larger than the median days missed. Reporting the median would allow for a robust estimate of the expected number of days missed, but would down weight those instances when knee injuries result in much longer missed days, as reflected by the mean. Because of the importance of the days missed measure for professional baseball, both the mean and median are presented.

In order to estimate exposure, the average number of players per team per game was calculated based on analysis of regular season game participation via box scores. This average number over a season, multiplied by the number of team games at each professional level of baseball, was used as an estimate of athlete exposures in order to provide rates comparable to those of other injury surveillance systems. Injury rates were reported as injuries per 1000 athlete-exposures (AE) for those knee injuries that occurred during the regular season. It should be noted that the number of regular season knee injuries and the subsequent AE rates are based on injuries that were deemed work-related during the regular season. This does not necessarily only include injuries occurring during the course of a game, but injuries in game preparation as well. Due to the variations in spring training games and fluctuating rosters, an exposure rate could not be calculated for spring training knee injuries.

RESULTS

Overall Summary

Of the 30 general body regions/systems included in the HITS database, injuries to the knee were the fifth most common reason for days missed in all of professional baseball from 2011-2014 (Table 1). Injuries to the knee represented 6.5% of the nearly 34,000 injuries sustained during the study period. Knee injuries were the fifth most common reason for time out of play for players in both the MiLB and MLB.

A total of 2171 isolated knee injuries resulted in time out of sport for professional baseball players (Table 2). Of these, 410 (19%) occurred in MLB players and 1761 (81%) occurred in MiLB players. MLB players were older than MiLB players at the time of injury (29.5 vs 22.8 years, respectively). Overall mean number of days missed was 16.2 days per knee injury, with MLB players missing an approximately 7 days more per injury than MiLB athletes (21.8 vs. 14.9 days respectively; P = .001).Over the course of the 4 seasons, a total of 30,449 days were missed due to knee injuries in professional baseball, giving an average rate of 7612 days lost per season. Surgery was performed for 263 (12.1%) of the 2171 knee injuries, with a greater proportion of MLB players requiring surgery than MiLB players (17.3% vs 10.9%) (P < .001). With respect to number of days missed per injury, 26% of knee injuries in the minor leagues resulted in greater than 30 days missed, while this number rose to 32% for knee injuries in MLB players (Table 3).

For regular season games, it was estimated that there were 1,197,738 MiLB and 276,608 MLB AE, respectively, over the course of the 4 seasons (2011-2014). The overall knee injury rate across both the MiLB and MLB was 1.2 per 1000 AE, based on the subset of 308 and 1473 regular season knee injuries in MiLB and MLB, respectively. The rate of knee injury was similar and not significantly different between the MiLB and MLB (1.2 per 1000 AE in the MiLB and 1.1 per 1000 AE in the MLB).

Characteristics of Injuries

When considering the position of the player during injury, defensive players were most frequently injured (n = 742, 56.5%), with pitchers (n = 227, 17.3%), infielders (n =193, 14.7%), outfielders (n = 193, 14.7%), and catchers (n = 129, 9.8%) sustaining injuries in decreasing frequency. Injuries while on offense (n = 571, 43.5%) were most frequent in base runners (n = 320, 24.4%) followed by batters (n = 251, 19.1%) (Table 4). Injuries while on defense occurring in infielders and catchers resulted in the longest period of time away from play (average of 22.4 and 20.8 days missed, respectively), while those occurring in batters resulted in the least average days missed (8.9 days).

The most common field location for knee injuries to occur was the infield, which was responsible for n = 647 (29.8%) of the total knee injuries (Table 4). This was followed by home plate (n = 493, 22.7%), other locations outside those specified (n = 394, 18.1%), outfield (n = 320, 14.7%), pitcher’s mound (n = 210, 9.7%), and foul territory or the bullpen (n = 107, 4.9%). Of the knee injuries with a specified location, those occurring in foul territory or the bullpen resulted in the highest mean days missed (18.4), while those occurring at home plate resulted in the least mean days missed (13.4 days).

When analyzed by mechanism of injury, noncontact injuries (n = 953, 43.9%) were more common than being hit with the ball (n = 374, 17.2%), striking the ground (n = 409, 18.8%), other mechanisms not listed (n = 196, 9%), contact with another player (n = 176, 8.1%), or contact with other objects (n = 63, 2.9%) (Table 4). Noncontact injuries and player to player collisions resulted in the greatest number of missed days (21.6 and 17.1 days, respectively) while being struck by the ball resulted in the least mean days missed (5.1).

Of the n = 493 knee injuries occurring at home plate, n = 212 (43%) occurred to the batter, n = 100 (20%) to the catcher, n = 34 (6.9%) to base runners, and n = 7 (1.4%) to pitchers (Table 5). The majority of knee injuries in the infield occurred to base runners (n = 283, 43.7%). Player-to-player collisions at home plate were responsible for 51 (2.3%) knee injuries, while 163 (24%) were noncontact injuries and 376 (56%) were the result of a player being hit by the ball (Table 5).

Injury Diagnosis

By diagnosis, the most common knee injuries observed were contusions or hematomas (n = 662, 30.5%), other injuries (n = 415, 19.1%), sprains or ligament injuries (n = 380, 17.5%), tendinopathies or bursitis (n = 367, 16.9%), and meniscal or cartilage injury (n = 200, 9.2%) (Table 6). Injuries resulting in the greatest mean number of days missed included meniscal or cartilage injuries (44 days), sprains or ligament injuries (30 days), or dislocations (22 days).

Based on specific SMDCS descriptors, the most frequent knee injuries reported were contusion (n = 662, 30.5%), patella tendinopathy (n = 222, 10.2%), and meniscal tears (n = 200, 9.2%) (Table 6). Complete anterior cruciate ligament tears, although infrequent, were responsible for the greatest mean days missed (156.2 days). This was followed by lateral meniscus tears (47.5 days) and medial meniscus tears (41.2 days). Knee contusions, although very common, resulted in the least number of days missed (6.0 days).

Discussion

Although much is known about knee injuries in other professional athletic leagues, little is known about knee injuries in professional baseball players.^2-4 The majority of epidemiologic studies regarding baseball players at any level emphasizes the study of shoulder and elbow injuries.^3,4,19 Since the implementation of the electronic medical record and the HITS database in professional baseball, there has been increased effort to document injuries that have received less attention in the existing literature. Understanding the epidemiology of these injuries is important for the development of targeted prevention efforts.

Prior studies of injuries in professional baseball relied on data captured by the publicly available DL. Posner and colleagues² provide one of the most comprehensive reports on MLB injuries in a report utilizing DL assignment data over a period of 7 seasons.They demonstrated that knee injuries were responsible for 7.7% (12.5% for fielders and 3.7% for pitchers) of assignments to the DL. The current study utilized a comprehensive surveillance and builds on this existing knowledge. The present study found similar trends to Posner and colleagues² in that knee injuries were responsible for 6.5% of injuries in professional baseball players that resulted in missed games. From the 2002 season to the 2008 season, knee injuries were the fifth most common reason MLB players were placed on the DL,² and the current study indicates that they remain the fifth most common reason for missed time from play based on the HITS data. Since the prevalence of these injuries have remained constant since the 2002 season, efforts to better understand these injuries are warranted in order to identify strategies to prevent them. These analyses have generated important data towards achieving this understanding.

As with most injuries in professional sports, goals for treatment are aimed at maximizing patient function and performance while minimizing time out of play. For the 2011-2014 professional baseball seasons, a total of 2171 players sustained knee injuries and missed an average of 16.2 days per injury. Knee injuries were responsible for a total of 7612 days of missed work for MLB and MiLB players per season (30,449 days over the 4-season study period). This is equivalent to a total of 20.9 years of players’ time lost in professional baseball per season over the last 4 years. The implications of this amount of time away from sport are significant, and further study should be targeted at prevention of these injuries and optimizing return to play times.

When attempting to reduce the burden of knee injuries in professional baseball, it may prove beneficial to first understand how the injuries occur, where on the field, and who is at greatest risk. From 2011 to 2014, nearly 44% of knee injuries occurred by noncontact mechanisms. Among all locations on the field where knee injuries occurred, those occurring in the infield were responsible for the greatest mean days missed. The players who seem to be at greatest risk for knee injuries appear to be base runners. These data suggest the need for prevention efforts targeting base runners and infield players, as well as players in MiLB, where the largest number of injuries occurred.

Recently, playing rules implemented by MLB after consultation with players have focused on reducing the number of player-to-player collisions at home plate in an attempt to decrease the injury burden to catchers and base runners.²⁰ This present analysis suggests that this rule change may also reduce the occurrence of knee injuries, as player collisions at home plate were responsible for a total of 51 knee injuries during the study period. The impact of this rule change on injury rates should also be explored. Interestingly, of the 51 knees injuries occurring due to contact at home plate, 23 occurred in 2011, and only 2 occurred in 2014—the first year of the new rule. Additional areas that resulted in high numbers of knee injuries were player-to-player contact in the infield and player contact with the ground in the infield.

Attempting to reduce injury burden and time out of play related to knee injuries in professional baseball players will likely prove to be a difficult task. In order to generate meaningful improvement, a comprehensive approach that involves players, management, trainers, therapists, and physicians will likely be required. As the first report of the epidemiology of knee injuries in professional baseball players, this study is one important step in that process. The strengths of this study are its comprehensive nature that analyzes injuries from an entire population of players on more than 200 teams over a 3-year period. Also, this research is strengthened by its focus on one particular region of the body that has received limited attention in the empirical literature, but represents a significant source of lost time during the baseball season.

There are some limitations to this study. As with any injury surveillance system, there is the possibility that not all cases were captured. Additionally, since the surveillance system is based on data from multiple teams, data entry discrepancy is possible; however, the presence of dropdown boxes and systematic definitions for injuries reduces this risk. Finally, this study did not investigate the various treatments for knee injuries beyond whether or not the injury required surgery. Since this was the first comprehensive exploration of knee injuries in professional baseball, future studies are needed to explore additional facets including outcomes related to treatment, return to play, and performance.

Conclusion

Knee injuries represent 6.5% of all injuries in professional baseball, occurring at a rate of 1.3 per 1000 AE. The burden of these injuries is significant for professional baseball players. This study fills a critical gap in sports injury research by contributing to the knowledge about the effect of knee injuries in professional baseball. It also provides an important foundation for future epidemiologic inquiry to identify modifiable risk factors and interventions that may reduce the impact of these injuries in athletes.

Rotator cuff injuries can be a source of debilitating pain and dysfunction in athletes at all levels, occasionally precluding return to competitive sport. Overhead athletes place extraordinary physiologic demands on the shoulder, as humeral angular velocities of 7000° to 8000° per second and rotational torques higher than 70 Nm have been measured during the baseball pitch.¹ Repetitive supraphysiologic loading of the rotator cuff throughout the coordinated phases of throwing can result in a characteristic spectrum of shoulder pathology in overhead throwers. Several studies have demonstrated partial-thickness articular-sided rotator cuff tears (RCTs) in the area of the posterior supraspinatus and anterior infraspinatus tendons.^2-4 Although the precise mechanism remains unclear, plausible explanations for the pathogenesis of these injuries include eccentric tensile and shear forces that lead to tendon failure with repetitive throwing, as well as internal impingement (mechanical impingement of the aforementioned tendons against the posterosuperior glenoid at 90° of shoulder abduction and maximum external rotation).^5,6

Whereas partial-thickness articular-sided RCTs have been described in overhead athletes with rotator cuff pathology, full-thickness tears are encountered less often.^7,8 Accordingly, there is a paucity of literature on clinical outcomes in professional baseball players with these injuries. To our knowledge, only 2 studies have investigated functional outcomes of open surgical repair of full-thickness tears in this population, and the outcomes have been uniformly poor.^8,9

An anatomical description of rotator cuff anatomy has demonstrated a consistent pattern of supraspinatus and infraspinatus tendon insertion relative to the articular surface, biceps groove, and the bare area of the humerus.¹⁰ Using gross and microscopic analyses, the authors noted that the supraspinatus tendon inserted immediately adjacent to the articular margin, and the infraspinatus and teres minor tapered laterally away from the margin to form the bare area. Detailed knowledge of the insertional anatomy of the rotator cuff is important, as surgical repair should recreate the broad footprint to restore normal biomechanics and increase the surface area available for healing.^11,12 Medial advancement of the rotator cuff insertion during surgical repair can have deleterious biomechanical effects on glenohumeral motion.¹¹

Given the unfavorable results found after routine open repair of full-thickness tears, we altered our approach to these injuries and adopted an arthroscopic technique in which the tendon is repaired immediately lateral to the anatomical footprint. Research studies have demonstrated that chronic stress from repetitive throwing can lead to attenuation of soft-tissue restraints, and we think preservation of these adaptive changes after surgical repair may be important for these athletes to maintain extraordinary glenohumeral rotation and achieve high throwing velocities.¹³ We conducted a study to describe the lateralized repair technique for full-thickness RCTs and to report functional outcomes in Major League Baseball (MLB) pitchers treated with this procedure at minimum 2-year follow-up. We hypothesized that use of this novel technique would result in a higher rate of return to preinjury level of play in comparison with open rotator cuff repair in comparable cohorts, as reported in other studies.^8,9

Materials and Methods

After obtaining Institutional Review Board approval for this study, we performed a retrospective chart review of MLB players treated by Dr. Altchek. We identified all professional baseball players who received a diagnosis of full-thickness RCT after preoperative magnetic resonance imaging with subsequent confirmation during surgery. Any patient who underwent arthroscopic repair using the lateralized footprint technique was included in the study. Demographic and preoperative injury information was collected from the chart, and final follow-up data were collected at the last available clinic visit. From available team records, we also obtained return-to-play data and objective pitching statistics: seasons played, games played, innings pitched, strikeouts per 9 innings, walks per 9 innings, and earned run average.

Surgical Technique

We routinely perform arthroscopic rotator cuff repairs with the patient under regional anesthesia in the beach-chair position. The operative extremity is placed in a Spider Limb Positioner (Smith & Nephew) to facilitate easy manipulation of the arm throughout the procedure. A standard posterior portal is established, and then an anterior portal is placed in the superolateral aspect of the rotator interval directly anterior to the leading edge of the supraspinatus tendon. A lateral portal created 2 to 3 cm distal to the anterolateral margin of the acromion may be used as an additional working portal. A thorough diagnostic arthroscopy is performed to evaluate the glenohumeral joint for any concomitant intra-articular pathology. Particular attention is directed to inspection of the superior labrum, biceps tendon, and capsuloligamentous structures, as injuries to these structures are often associated with rotator cuff pathology in overhead athletes.

Once presence of an RCT is confirmed, a thorough subacromial bursectomy is performed to help with visualization and inspection of the injury. The tissue is provisionally grasped and mobilized to measure the amount of available tendon excursion. In this unique population, the vast majority of injuries are diagnosed in an expeditious manner, thereby precluding the presence of significant retraction, poor tissue quality, and inadequate mobilization of the tendons. The greater tuberosity is identified, and the area immediately adjacent to the articular margin is abraded with a mechanical shaver to enhance healing potential. For supraspinatus tears, an anchor is placed immediately lateral to the articular margin in the region of the anterior attachment of the rotator cable (Figure 1). The posterior anchor is placed about 10 to 15 mm lateral to the articular margin to reattach the infraspinatus tendon (Figure 2). When the medial row sutures are tied down, anatomical placement of these anchors effectively re-creates the bare area described by Curtis and colleagues¹⁰ (Figure 3). In most cases, the medial row sutures are left intact and fixed laterally with a knotless anchor to provide a transosseous equivalent (double-row) repair.

Results

We identified 6 MLB pitchers who underwent arthroscopic rotator cuff repair using the aforementioned technique over an 8-year period. Each patient presented with complaints of debilitating shoulder pain and decreased pitching performance, including loss of throwing accuracy and velocity. There were 4 right-hand–dominant pitchers and 2 left-hand–dominant pitchers; rotator cuff pathology was observed in the dominant pitching arm in each case. Three players were classified as starting pitchers; the other 3 pitched in a relief role. Mean age of all pitchers at time of surgery was 29.8 years (range, 25-37 years). According to records, 2 patients (33%) underwent previous rotator cuff débridement for partial-thickness RCTs before surgical intervention at our institution. Operative information on the depth of the partial-thickness tears observed during the previous procedures was not available for review. At time of rotator cuff repair, 3 patients (50%) underwent concomitant procedures, including superior labrum anterior-posterior (SLAP) lesion repair (1 patient) and posterior labrum débridement (2 patients). A double-row fixation construct was achieved in each case. Review of operative records revealed a mean tear size of 2.1 cm (range, 1.5-3.0 cm) measured anterior to posterior, and all tears involved the supraspinatus and/or infraspinatus tendons. Postoperative rehabilitation included immobilization in a sling for 4 weeks. Hand, wrist, and elbow range-of-motion (ROM) exercises were started immediately to help reduce inflammation. Passive ROM exercises in the plane of the scapula were begun 4 weeks after surgery. Isometric scapular stabilization exercises were also incorporated at that time. Active-assisted ROM exercises were started at about 6 weeks, and isometric strengthening exercises were started at week 8 with progression to eccentric strengthening and weight training at about 3 months. Most pitchers were allowed to begin an interval throwing program at 24 weeks. There were no significant differences in the therapy programs for pitchers who underwent concomitant labral procedures, but the patient who underwent SLAP repair was limited to 30° of external rotation and 90° of forward flexion, with avoidance of active biceps contractions, for the first 6 weeks of rehabilitation.

By mean follow-up of 66.7 months (range, 23.2-94.6 months), 5 pitchers (83%) returned to their preinjury level of competition for at least 1 full season. One player pitched at Minor League Class AA level for about 1 season but was forced to retire because of persistent symptoms related to the shoulder. This pitcher underwent simultaneous rotator cuff and SLAP lesion repair. Of the 5 pitchers who resumed MLB play, none returned to their preoperative pitching productivity; mean number of innings pitched decreased from 1806.5 to 183.7. Three (60%) of these 5 pitchers experienced a slight reduction in performance as measured by earned run average. Interestingly, both players over age 30 years at time of surgery, versus 3 of the 4 pitchers under age 30 years, returned to their preoperative level of competition for at least 1 season. The Table summarizes MLB player data and objective pitching statistics. There were no perioperative complications related to this arthroscopic technique, and there were no glenohumeral ROM deficits at final follow-up.

Discussion

Although the incidence of full-thickness RCTs in professional baseball players is presumably low, available studies suggest that it is a debilitating injury with a poor prognosis for return to high-level athletics. Mazoué and Andrews⁹ reviewed the outcomes of 16 professional baseball players (12 pitchers, 4 position players) who underwent mini-open repair of full-thickness RCTs that involved more than 90% of the rotator cuff. Fifteen patients underwent mini-open rotator cuff repair using suture anchors in the anatomical footprint along with bone tunnels established near the lateral margin of the greater tuberosity to create a 2-level anatomical repair. One patient was treated with a mini-open repair using suture anchors in the greater tuberosity with a side-side repair of a longitudinal split within the rotator cuff. In the evaluation of outcomes by player position, only 1 pitcher (8%) returned to a competitive level of pitching at a mean follow-up of 67 months. On review of 2 position players with a full-thickness RCT in the dominant shoulder, only 1 (50%) returned to Major League play at a mean follow-up of 62.5 months. The remaining 2 position players underwent surgical repair of the nondominant shoulder, and, not surprisingly, both returned to their previous level of athletic activity without any difficulty. These results should be examined carefully, as the associated pathology in this high-demand cohort should not be discounted. Eleven (almost 92%) of the 12 pitchers had undergone at least 1 previous procedure on the shoulder. Furthermore, at time of full-thickness rotator cuff repair, 9 (75%) of the 12 pitchers were treated for concomitant intra-articular pathology, including SLAP tears, capsular attenuation, and/or labral fraying. In our study, 50% of pitchers underwent an associated labral procedure. Although labral débridement did not have a significant effect on return to play, the 1 pitcher who underwent SLAP repair was not able to return to preinjury level of play.

Tibone and colleagues⁸ reviewed postoperative outcomes in 45 athletes with rotator cuff pathology. Within their series, 5 professional baseball pitchers with full-thickness tears were treated with open subacromial decompression and rotator cuff repair. Two baseball pitchers with RCTs larger than 2 cm underwent open transosseous footprint repair in which the cuff was reinserted using bone tunnels created within the greater tuberosity. At long-term follow-up, only 2 (40%) of the 5 pitchers returned to competitive pitching. Interestingly, both pitchers who underwent transosseous footprint fixation were unable to return to professional baseball.

Overhead athletes require a delicate balance of shoulder mobility and stability to meet the high functional demands of their sports. Significant debate continues as to whether innate alterations in glenohumeral mobility preselect individuals for overhead sports, or if these changes are acquired through adaptations in supporting soft-tissue and osseous structures. Sethi and colleagues¹⁴ used an instrumented manual laxity examination to compare anterior-posterior laxity in asymptomatic professional and Division I college baseball players. The authors noted asymmetric anterior-posterior translation (>3 mm) between the throwing shoulder and the nondominant shoulder in 12 (60%) of 20 professional pitchers and 10 (59%) of 17 college pitchers. Although the authors did not correlate translational differences with corresponding shoulder pathology, the observed asymmetry supported the idea that these athletes may experience adaptive glenohumeral changes with repetitive throwing. The association between adaptive changes and shoulder biomechanics has been studied. Burkhart and Lo¹⁵ used a cadaveric model to describe the cam effect of the proximal humerus and the biomechanical consequences of a relative reduction in this effect after pathologic changes within the glenohumeral joint (constriction of posteroinferior capsule). They noted that a posterosuperior shift in the glenohumeral contact point in the throwing position can result in anterior capsular redundancy that may contribute to microinstability of the shoulder. This relative laxity increases external rotation, resulting in increased torsional and shear forces at the rotator cuff insertion.¹⁶ Ultimately, these abnormal forces may predispose overhead athletes to rotator cuff injury.

Given the available literature, it is clear that full-thickness RCTs are potentially career-ending injuries for professional baseball players. The question arises as to why the results are so poor. Ultimately, the high incidence of concomitant intra-articular pathology associated with full-thickness RCTs underscores the severity of soft-tissue damage sustained with repetitive overhead throwing. Mazoué and Andrews⁹ proposed the presence of associated labral and capsular pathology as a potential explanation for poor outcomes of surgical repair. Given the myriad of additional pathology observed in each patient, it is difficult to ascertain the precise impact of these injuries on postoperative outcome. However, early diagnosis and aggressive surgical intervention are clearly necessary to prevent accumulative injury. Regarding surgical intervention, both Tibone and colleagues⁸ and Mazoué and Andrews⁹ reported use of an open surgical repair technique in which the tendon was repaired to the anatomical footprint. Certainly, the benefits of an all-arthroscopic technique include optimal visualization of the RCT, less perioperative morbidity, and minimal soft-tissue injury. With our arthroscopic technique, the rotator cuff was fixed immediately lateral to the anatomical footprint, thereby leaving the medial aspect of the footprint uncovered. Functionally, the goal of this procedure is to restore the integrity of the rotator cuff without compromising glenohumeral mobility acquired through soft-tissue adaptation. Investigation of the insertional anatomy of the rotator cuff has demonstrated that the supraspinatus tendon inserts about 0.9 mm from the edge of the articular surface, and the infraspinatus insertional footprint tapers away from the articular surface to form the bare area as it extends inferiorly on the greater tuberosity.¹⁰ We think preexisting adaptations in glenohumeral anatomy are important for peak performance in this unique population, and even small alterations in the repair location can have deleterious effects on throwing mechanics. Lateralized repair of the cuff precludes potential medialization of the cuff insertion and may facilitate preservation of soft-tissue adaptations that these athletes rely on to achieve extraordinary glenohumeral motion.

Interestingly, with this technique we noted a higher rate of return to MLB play in pitchers over age 30 years. Although several individual factors (eg, player talent level, work ethics, compliance with rehabilitation) may play a role in this finding, it is possible that older, more mature patients may be more willing to assume diminished roles to continue to play. Jones and colleagues¹⁷ recently reported similar findings in older MLB pitchers after revision ulnar collateral ligament reconstruction.

This study had several limitations. First, the patient cohort was small (a result of the nature and relatively infrequent incidence of the clinical problem). Second, clinical information was collected retrospectively, which limited our ability to determine precise differences between preoperative and postoperative glenohumeral ROM with this technique. Third, the cohort included patients who demonstrated additional intra-articular (labral) pathology. Although associated pathology is common in this high-demand athletic population, it is clear that advanced pathology (eg, SLAP tears) may affect clinical outcomes, as in our study. Despite these limitations, our study is the largest review of professional baseball players treated for full-thickness rotator cuff injuries with an arthroscopic technique. Overall, the results of this study are promising and call for further clinical and biomechanical evaluation.

Conclusion

Surgical management of rotator cuff injuries in professional baseball players remains an extremely difficult problem. Current studies of full-thickness RCTs highlight these athletes’ poor functional outcomes. These unfavorable results prompted us to alter our surgical technique. Initial outcomes have been encouraging, and extended follow-up in this cohort of patients will provide a more definitive assessment of the success of this technique.

Rotator cuff injuries can be a source of debilitating pain and dysfunction in athletes at all levels, occasionally precluding return to competitive sport. Overhead athletes place extraordinary physiologic demands on the shoulder, as humeral angular velocities of 7000° to 8000° per second and rotational torques higher than 70 Nm have been measured during the baseball pitch.¹ Repetitive supraphysiologic loading of the rotator cuff throughout the coordinated phases of throwing can result in a characteristic spectrum of shoulder pathology in overhead throwers. Several studies have demonstrated partial-thickness articular-sided rotator cuff tears (RCTs) in the area of the posterior supraspinatus and anterior infraspinatus tendons.^2-4 Although the precise mechanism remains unclear, plausible explanations for the pathogenesis of these injuries include eccentric tensile and shear forces that lead to tendon failure with repetitive throwing, as well as internal impingement (mechanical impingement of the aforementioned tendons against the posterosuperior glenoid at 90° of shoulder abduction and maximum external rotation).^5,6

Whereas partial-thickness articular-sided RCTs have been described in overhead athletes with rotator cuff pathology, full-thickness tears are encountered less often.^7,8 Accordingly, there is a paucity of literature on clinical outcomes in professional baseball players with these injuries. To our knowledge, only 2 studies have investigated functional outcomes of open surgical repair of full-thickness tears in this population, and the outcomes have been uniformly poor.^8,9

An anatomical description of rotator cuff anatomy has demonstrated a consistent pattern of supraspinatus and infraspinatus tendon insertion relative to the articular surface, biceps groove, and the bare area of the humerus.¹⁰ Using gross and microscopic analyses, the authors noted that the supraspinatus tendon inserted immediately adjacent to the articular margin, and the infraspinatus and teres minor tapered laterally away from the margin to form the bare area. Detailed knowledge of the insertional anatomy of the rotator cuff is important, as surgical repair should recreate the broad footprint to restore normal biomechanics and increase the surface area available for healing.^11,12 Medial advancement of the rotator cuff insertion during surgical repair can have deleterious biomechanical effects on glenohumeral motion.¹¹

Given the unfavorable results found after routine open repair of full-thickness tears, we altered our approach to these injuries and adopted an arthroscopic technique in which the tendon is repaired immediately lateral to the anatomical footprint. Research studies have demonstrated that chronic stress from repetitive throwing can lead to attenuation of soft-tissue restraints, and we think preservation of these adaptive changes after surgical repair may be important for these athletes to maintain extraordinary glenohumeral rotation and achieve high throwing velocities.¹³ We conducted a study to describe the lateralized repair technique for full-thickness RCTs and to report functional outcomes in Major League Baseball (MLB) pitchers treated with this procedure at minimum 2-year follow-up. We hypothesized that use of this novel technique would result in a higher rate of return to preinjury level of play in comparison with open rotator cuff repair in comparable cohorts, as reported in other studies.^8,9

Materials and Methods

After obtaining Institutional Review Board approval for this study, we performed a retrospective chart review of MLB players treated by Dr. Altchek. We identified all professional baseball players who received a diagnosis of full-thickness RCT after preoperative magnetic resonance imaging with subsequent confirmation during surgery. Any patient who underwent arthroscopic repair using the lateralized footprint technique was included in the study. Demographic and preoperative injury information was collected from the chart, and final follow-up data were collected at the last available clinic visit. From available team records, we also obtained return-to-play data and objective pitching statistics: seasons played, games played, innings pitched, strikeouts per 9 innings, walks per 9 innings, and earned run average.

Surgical Technique

We routinely perform arthroscopic rotator cuff repairs with the patient under regional anesthesia in the beach-chair position. The operative extremity is placed in a Spider Limb Positioner (Smith & Nephew) to facilitate easy manipulation of the arm throughout the procedure. A standard posterior portal is established, and then an anterior portal is placed in the superolateral aspect of the rotator interval directly anterior to the leading edge of the supraspinatus tendon. A lateral portal created 2 to 3 cm distal to the anterolateral margin of the acromion may be used as an additional working portal. A thorough diagnostic arthroscopy is performed to evaluate the glenohumeral joint for any concomitant intra-articular pathology. Particular attention is directed to inspection of the superior labrum, biceps tendon, and capsuloligamentous structures, as injuries to these structures are often associated with rotator cuff pathology in overhead athletes.

Once presence of an RCT is confirmed, a thorough subacromial bursectomy is performed to help with visualization and inspection of the injury. The tissue is provisionally grasped and mobilized to measure the amount of available tendon excursion. In this unique population, the vast majority of injuries are diagnosed in an expeditious manner, thereby precluding the presence of significant retraction, poor tissue quality, and inadequate mobilization of the tendons. The greater tuberosity is identified, and the area immediately adjacent to the articular margin is abraded with a mechanical shaver to enhance healing potential. For supraspinatus tears, an anchor is placed immediately lateral to the articular margin in the region of the anterior attachment of the rotator cable (Figure 1). The posterior anchor is placed about 10 to 15 mm lateral to the articular margin to reattach the infraspinatus tendon (Figure 2). When the medial row sutures are tied down, anatomical placement of these anchors effectively re-creates the bare area described by Curtis and colleagues¹⁰ (Figure 3). In most cases, the medial row sutures are left intact and fixed laterally with a knotless anchor to provide a transosseous equivalent (double-row) repair.

Results

We identified 6 MLB pitchers who underwent arthroscopic rotator cuff repair using the aforementioned technique over an 8-year period. Each patient presented with complaints of debilitating shoulder pain and decreased pitching performance, including loss of throwing accuracy and velocity. There were 4 right-hand–dominant pitchers and 2 left-hand–dominant pitchers; rotator cuff pathology was observed in the dominant pitching arm in each case. Three players were classified as starting pitchers; the other 3 pitched in a relief role. Mean age of all pitchers at time of surgery was 29.8 years (range, 25-37 years). According to records, 2 patients (33%) underwent previous rotator cuff débridement for partial-thickness RCTs before surgical intervention at our institution. Operative information on the depth of the partial-thickness tears observed during the previous procedures was not available for review. At time of rotator cuff repair, 3 patients (50%) underwent concomitant procedures, including superior labrum anterior-posterior (SLAP) lesion repair (1 patient) and posterior labrum débridement (2 patients). A double-row fixation construct was achieved in each case. Review of operative records revealed a mean tear size of 2.1 cm (range, 1.5-3.0 cm) measured anterior to posterior, and all tears involved the supraspinatus and/or infraspinatus tendons. Postoperative rehabilitation included immobilization in a sling for 4 weeks. Hand, wrist, and elbow range-of-motion (ROM) exercises were started immediately to help reduce inflammation. Passive ROM exercises in the plane of the scapula were begun 4 weeks after surgery. Isometric scapular stabilization exercises were also incorporated at that time. Active-assisted ROM exercises were started at about 6 weeks, and isometric strengthening exercises were started at week 8 with progression to eccentric strengthening and weight training at about 3 months. Most pitchers were allowed to begin an interval throwing program at 24 weeks. There were no significant differences in the therapy programs for pitchers who underwent concomitant labral procedures, but the patient who underwent SLAP repair was limited to 30° of external rotation and 90° of forward flexion, with avoidance of active biceps contractions, for the first 6 weeks of rehabilitation.

By mean follow-up of 66.7 months (range, 23.2-94.6 months), 5 pitchers (83%) returned to their preinjury level of competition for at least 1 full season. One player pitched at Minor League Class AA level for about 1 season but was forced to retire because of persistent symptoms related to the shoulder. This pitcher underwent simultaneous rotator cuff and SLAP lesion repair. Of the 5 pitchers who resumed MLB play, none returned to their preoperative pitching productivity; mean number of innings pitched decreased from 1806.5 to 183.7. Three (60%) of these 5 pitchers experienced a slight reduction in performance as measured by earned run average. Interestingly, both players over age 30 years at time of surgery, versus 3 of the 4 pitchers under age 30 years, returned to their preoperative level of competition for at least 1 season. The Table summarizes MLB player data and objective pitching statistics. There were no perioperative complications related to this arthroscopic technique, and there were no glenohumeral ROM deficits at final follow-up.

Discussion

Although the incidence of full-thickness RCTs in professional baseball players is presumably low, available studies suggest that it is a debilitating injury with a poor prognosis for return to high-level athletics. Mazoué and Andrews⁹ reviewed the outcomes of 16 professional baseball players (12 pitchers, 4 position players) who underwent mini-open repair of full-thickness RCTs that involved more than 90% of the rotator cuff. Fifteen patients underwent mini-open rotator cuff repair using suture anchors in the anatomical footprint along with bone tunnels established near the lateral margin of the greater tuberosity to create a 2-level anatomical repair. One patient was treated with a mini-open repair using suture anchors in the greater tuberosity with a side-side repair of a longitudinal split within the rotator cuff. In the evaluation of outcomes by player position, only 1 pitcher (8%) returned to a competitive level of pitching at a mean follow-up of 67 months. On review of 2 position players with a full-thickness RCT in the dominant shoulder, only 1 (50%) returned to Major League play at a mean follow-up of 62.5 months. The remaining 2 position players underwent surgical repair of the nondominant shoulder, and, not surprisingly, both returned to their previous level of athletic activity without any difficulty. These results should be examined carefully, as the associated pathology in this high-demand cohort should not be discounted. Eleven (almost 92%) of the 12 pitchers had undergone at least 1 previous procedure on the shoulder. Furthermore, at time of full-thickness rotator cuff repair, 9 (75%) of the 12 pitchers were treated for concomitant intra-articular pathology, including SLAP tears, capsular attenuation, and/or labral fraying. In our study, 50% of pitchers underwent an associated labral procedure. Although labral débridement did not have a significant effect on return to play, the 1 pitcher who underwent SLAP repair was not able to return to preinjury level of play.

Tibone and colleagues⁸ reviewed postoperative outcomes in 45 athletes with rotator cuff pathology. Within their series, 5 professional baseball pitchers with full-thickness tears were treated with open subacromial decompression and rotator cuff repair. Two baseball pitchers with RCTs larger than 2 cm underwent open transosseous footprint repair in which the cuff was reinserted using bone tunnels created within the greater tuberosity. At long-term follow-up, only 2 (40%) of the 5 pitchers returned to competitive pitching. Interestingly, both pitchers who underwent transosseous footprint fixation were unable to return to professional baseball.

Overhead athletes require a delicate balance of shoulder mobility and stability to meet the high functional demands of their sports. Significant debate continues as to whether innate alterations in glenohumeral mobility preselect individuals for overhead sports, or if these changes are acquired through adaptations in supporting soft-tissue and osseous structures. Sethi and colleagues¹⁴ used an instrumented manual laxity examination to compare anterior-posterior laxity in asymptomatic professional and Division I college baseball players. The authors noted asymmetric anterior-posterior translation (>3 mm) between the throwing shoulder and the nondominant shoulder in 12 (60%) of 20 professional pitchers and 10 (59%) of 17 college pitchers. Although the authors did not correlate translational differences with corresponding shoulder pathology, the observed asymmetry supported the idea that these athletes may experience adaptive glenohumeral changes with repetitive throwing. The association between adaptive changes and shoulder biomechanics has been studied. Burkhart and Lo¹⁵ used a cadaveric model to describe the cam effect of the proximal humerus and the biomechanical consequences of a relative reduction in this effect after pathologic changes within the glenohumeral joint (constriction of posteroinferior capsule). They noted that a posterosuperior shift in the glenohumeral contact point in the throwing position can result in anterior capsular redundancy that may contribute to microinstability of the shoulder. This relative laxity increases external rotation, resulting in increased torsional and shear forces at the rotator cuff insertion.¹⁶ Ultimately, these abnormal forces may predispose overhead athletes to rotator cuff injury.

Given the available literature, it is clear that full-thickness RCTs are potentially career-ending injuries for professional baseball players. The question arises as to why the results are so poor. Ultimately, the high incidence of concomitant intra-articular pathology associated with full-thickness RCTs underscores the severity of soft-tissue damage sustained with repetitive overhead throwing. Mazoué and Andrews⁹ proposed the presence of associated labral and capsular pathology as a potential explanation for poor outcomes of surgical repair. Given the myriad of additional pathology observed in each patient, it is difficult to ascertain the precise impact of these injuries on postoperative outcome. However, early diagnosis and aggressive surgical intervention are clearly necessary to prevent accumulative injury. Regarding surgical intervention, both Tibone and colleagues⁸ and Mazoué and Andrews⁹ reported use of an open surgical repair technique in which the tendon was repaired to the anatomical footprint. Certainly, the benefits of an all-arthroscopic technique include optimal visualization of the RCT, less perioperative morbidity, and minimal soft-tissue injury. With our arthroscopic technique, the rotator cuff was fixed immediately lateral to the anatomical footprint, thereby leaving the medial aspect of the footprint uncovered. Functionally, the goal of this procedure is to restore the integrity of the rotator cuff without compromising glenohumeral mobility acquired through soft-tissue adaptation. Investigation of the insertional anatomy of the rotator cuff has demonstrated that the supraspinatus tendon inserts about 0.9 mm from the edge of the articular surface, and the infraspinatus insertional footprint tapers away from the articular surface to form the bare area as it extends inferiorly on the greater tuberosity.¹⁰ We think preexisting adaptations in glenohumeral anatomy are important for peak performance in this unique population, and even small alterations in the repair location can have deleterious effects on throwing mechanics. Lateralized repair of the cuff precludes potential medialization of the cuff insertion and may facilitate preservation of soft-tissue adaptations that these athletes rely on to achieve extraordinary glenohumeral motion.

Interestingly, with this technique we noted a higher rate of return to MLB play in pitchers over age 30 years. Although several individual factors (eg, player talent level, work ethics, compliance with rehabilitation) may play a role in this finding, it is possible that older, more mature patients may be more willing to assume diminished roles to continue to play. Jones and colleagues¹⁷ recently reported similar findings in older MLB pitchers after revision ulnar collateral ligament reconstruction.

This study had several limitations. First, the patient cohort was small (a result of the nature and relatively infrequent incidence of the clinical problem). Second, clinical information was collected retrospectively, which limited our ability to determine precise differences between preoperative and postoperative glenohumeral ROM with this technique. Third, the cohort included patients who demonstrated additional intra-articular (labral) pathology. Although associated pathology is common in this high-demand athletic population, it is clear that advanced pathology (eg, SLAP tears) may affect clinical outcomes, as in our study. Despite these limitations, our study is the largest review of professional baseball players treated for full-thickness rotator cuff injuries with an arthroscopic technique. Overall, the results of this study are promising and call for further clinical and biomechanical evaluation.

Conclusion

Surgical management of rotator cuff injuries in professional baseball players remains an extremely difficult problem. Current studies of full-thickness RCTs highlight these athletes’ poor functional outcomes. These unfavorable results prompted us to alter our surgical technique. Initial outcomes have been encouraging, and extended follow-up in this cohort of patients will provide a more definitive assessment of the success of this technique.

Rotator cuff injuries can be a source of debilitating pain and dysfunction in athletes at all levels, occasionally precluding return to competitive sport. Overhead athletes place extraordinary physiologic demands on the shoulder, as humeral angular velocities of 7000° to 8000° per second and rotational torques higher than 70 Nm have been measured during the baseball pitch.¹ Repetitive supraphysiologic loading of the rotator cuff throughout the coordinated phases of throwing can result in a characteristic spectrum of shoulder pathology in overhead throwers. Several studies have demonstrated partial-thickness articular-sided rotator cuff tears (RCTs) in the area of the posterior supraspinatus and anterior infraspinatus tendons.^2-4 Although the precise mechanism remains unclear, plausible explanations for the pathogenesis of these injuries include eccentric tensile and shear forces that lead to tendon failure with repetitive throwing, as well as internal impingement (mechanical impingement of the aforementioned tendons against the posterosuperior glenoid at 90° of shoulder abduction and maximum external rotation).^5,6

Whereas partial-thickness articular-sided RCTs have been described in overhead athletes with rotator cuff pathology, full-thickness tears are encountered less often.^7,8 Accordingly, there is a paucity of literature on clinical outcomes in professional baseball players with these injuries. To our knowledge, only 2 studies have investigated functional outcomes of open surgical repair of full-thickness tears in this population, and the outcomes have been uniformly poor.^8,9

An anatomical description of rotator cuff anatomy has demonstrated a consistent pattern of supraspinatus and infraspinatus tendon insertion relative to the articular surface, biceps groove, and the bare area of the humerus.¹⁰ Using gross and microscopic analyses, the authors noted that the supraspinatus tendon inserted immediately adjacent to the articular margin, and the infraspinatus and teres minor tapered laterally away from the margin to form the bare area. Detailed knowledge of the insertional anatomy of the rotator cuff is important, as surgical repair should recreate the broad footprint to restore normal biomechanics and increase the surface area available for healing.^11,12 Medial advancement of the rotator cuff insertion during surgical repair can have deleterious biomechanical effects on glenohumeral motion.¹¹

Given the unfavorable results found after routine open repair of full-thickness tears, we altered our approach to these injuries and adopted an arthroscopic technique in which the tendon is repaired immediately lateral to the anatomical footprint. Research studies have demonstrated that chronic stress from repetitive throwing can lead to attenuation of soft-tissue restraints, and we think preservation of these adaptive changes after surgical repair may be important for these athletes to maintain extraordinary glenohumeral rotation and achieve high throwing velocities.¹³ We conducted a study to describe the lateralized repair technique for full-thickness RCTs and to report functional outcomes in Major League Baseball (MLB) pitchers treated with this procedure at minimum 2-year follow-up. We hypothesized that use of this novel technique would result in a higher rate of return to preinjury level of play in comparison with open rotator cuff repair in comparable cohorts, as reported in other studies.^8,9

Materials and Methods

After obtaining Institutional Review Board approval for this study, we performed a retrospective chart review of MLB players treated by Dr. Altchek. We identified all professional baseball players who received a diagnosis of full-thickness RCT after preoperative magnetic resonance imaging with subsequent confirmation during surgery. Any patient who underwent arthroscopic repair using the lateralized footprint technique was included in the study. Demographic and preoperative injury information was collected from the chart, and final follow-up data were collected at the last available clinic visit. From available team records, we also obtained return-to-play data and objective pitching statistics: seasons played, games played, innings pitched, strikeouts per 9 innings, walks per 9 innings, and earned run average.

Surgical Technique

We routinely perform arthroscopic rotator cuff repairs with the patient under regional anesthesia in the beach-chair position. The operative extremity is placed in a Spider Limb Positioner (Smith & Nephew) to facilitate easy manipulation of the arm throughout the procedure. A standard posterior portal is established, and then an anterior portal is placed in the superolateral aspect of the rotator interval directly anterior to the leading edge of the supraspinatus tendon. A lateral portal created 2 to 3 cm distal to the anterolateral margin of the acromion may be used as an additional working portal. A thorough diagnostic arthroscopy is performed to evaluate the glenohumeral joint for any concomitant intra-articular pathology. Particular attention is directed to inspection of the superior labrum, biceps tendon, and capsuloligamentous structures, as injuries to these structures are often associated with rotator cuff pathology in overhead athletes.

Once presence of an RCT is confirmed, a thorough subacromial bursectomy is performed to help with visualization and inspection of the injury. The tissue is provisionally grasped and mobilized to measure the amount of available tendon excursion. In this unique population, the vast majority of injuries are diagnosed in an expeditious manner, thereby precluding the presence of significant retraction, poor tissue quality, and inadequate mobilization of the tendons. The greater tuberosity is identified, and the area immediately adjacent to the articular margin is abraded with a mechanical shaver to enhance healing potential. For supraspinatus tears, an anchor is placed immediately lateral to the articular margin in the region of the anterior attachment of the rotator cable (Figure 1). The posterior anchor is placed about 10 to 15 mm lateral to the articular margin to reattach the infraspinatus tendon (Figure 2). When the medial row sutures are tied down, anatomical placement of these anchors effectively re-creates the bare area described by Curtis and colleagues¹⁰ (Figure 3). In most cases, the medial row sutures are left intact and fixed laterally with a knotless anchor to provide a transosseous equivalent (double-row) repair.

Results

We identified 6 MLB pitchers who underwent arthroscopic rotator cuff repair using the aforementioned technique over an 8-year period. Each patient presented with complaints of debilitating shoulder pain and decreased pitching performance, including loss of throwing accuracy and velocity. There were 4 right-hand–dominant pitchers and 2 left-hand–dominant pitchers; rotator cuff pathology was observed in the dominant pitching arm in each case. Three players were classified as starting pitchers; the other 3 pitched in a relief role. Mean age of all pitchers at time of surgery was 29.8 years (range, 25-37 years). According to records, 2 patients (33%) underwent previous rotator cuff débridement for partial-thickness RCTs before surgical intervention at our institution. Operative information on the depth of the partial-thickness tears observed during the previous procedures was not available for review. At time of rotator cuff repair, 3 patients (50%) underwent concomitant procedures, including superior labrum anterior-posterior (SLAP) lesion repair (1 patient) and posterior labrum débridement (2 patients). A double-row fixation construct was achieved in each case. Review of operative records revealed a mean tear size of 2.1 cm (range, 1.5-3.0 cm) measured anterior to posterior, and all tears involved the supraspinatus and/or infraspinatus tendons. Postoperative rehabilitation included immobilization in a sling for 4 weeks. Hand, wrist, and elbow range-of-motion (ROM) exercises were started immediately to help reduce inflammation. Passive ROM exercises in the plane of the scapula were begun 4 weeks after surgery. Isometric scapular stabilization exercises were also incorporated at that time. Active-assisted ROM exercises were started at about 6 weeks, and isometric strengthening exercises were started at week 8 with progression to eccentric strengthening and weight training at about 3 months. Most pitchers were allowed to begin an interval throwing program at 24 weeks. There were no significant differences in the therapy programs for pitchers who underwent concomitant labral procedures, but the patient who underwent SLAP repair was limited to 30° of external rotation and 90° of forward flexion, with avoidance of active biceps contractions, for the first 6 weeks of rehabilitation.

By mean follow-up of 66.7 months (range, 23.2-94.6 months), 5 pitchers (83%) returned to their preinjury level of competition for at least 1 full season. One player pitched at Minor League Class AA level for about 1 season but was forced to retire because of persistent symptoms related to the shoulder. This pitcher underwent simultaneous rotator cuff and SLAP lesion repair. Of the 5 pitchers who resumed MLB play, none returned to their preoperative pitching productivity; mean number of innings pitched decreased from 1806.5 to 183.7. Three (60%) of these 5 pitchers experienced a slight reduction in performance as measured by earned run average. Interestingly, both players over age 30 years at time of surgery, versus 3 of the 4 pitchers under age 30 years, returned to their preoperative level of competition for at least 1 season. The Table summarizes MLB player data and objective pitching statistics. There were no perioperative complications related to this arthroscopic technique, and there were no glenohumeral ROM deficits at final follow-up.

Discussion

Although the incidence of full-thickness RCTs in professional baseball players is presumably low, available studies suggest that it is a debilitating injury with a poor prognosis for return to high-level athletics. Mazoué and Andrews⁹ reviewed the outcomes of 16 professional baseball players (12 pitchers, 4 position players) who underwent mini-open repair of full-thickness RCTs that involved more than 90% of the rotator cuff. Fifteen patients underwent mini-open rotator cuff repair using suture anchors in the anatomical footprint along with bone tunnels established near the lateral margin of the greater tuberosity to create a 2-level anatomical repair. One patient was treated with a mini-open repair using suture anchors in the greater tuberosity with a side-side repair of a longitudinal split within the rotator cuff. In the evaluation of outcomes by player position, only 1 pitcher (8%) returned to a competitive level of pitching at a mean follow-up of 67 months. On review of 2 position players with a full-thickness RCT in the dominant shoulder, only 1 (50%) returned to Major League play at a mean follow-up of 62.5 months. The remaining 2 position players underwent surgical repair of the nondominant shoulder, and, not surprisingly, both returned to their previous level of athletic activity without any difficulty. These results should be examined carefully, as the associated pathology in this high-demand cohort should not be discounted. Eleven (almost 92%) of the 12 pitchers had undergone at least 1 previous procedure on the shoulder. Furthermore, at time of full-thickness rotator cuff repair, 9 (75%) of the 12 pitchers were treated for concomitant intra-articular pathology, including SLAP tears, capsular attenuation, and/or labral fraying. In our study, 50% of pitchers underwent an associated labral procedure. Although labral débridement did not have a significant effect on return to play, the 1 pitcher who underwent SLAP repair was not able to return to preinjury level of play.

Tibone and colleagues⁸ reviewed postoperative outcomes in 45 athletes with rotator cuff pathology. Within their series, 5 professional baseball pitchers with full-thickness tears were treated with open subacromial decompression and rotator cuff repair. Two baseball pitchers with RCTs larger than 2 cm underwent open transosseous footprint repair in which the cuff was reinserted using bone tunnels created within the greater tuberosity. At long-term follow-up, only 2 (40%) of the 5 pitchers returned to competitive pitching. Interestingly, both pitchers who underwent transosseous footprint fixation were unable to return to professional baseball.

Overhead athletes require a delicate balance of shoulder mobility and stability to meet the high functional demands of their sports. Significant debate continues as to whether innate alterations in glenohumeral mobility preselect individuals for overhead sports, or if these changes are acquired through adaptations in supporting soft-tissue and osseous structures. Sethi and colleagues¹⁴ used an instrumented manual laxity examination to compare anterior-posterior laxity in asymptomatic professional and Division I college baseball players. The authors noted asymmetric anterior-posterior translation (>3 mm) between the throwing shoulder and the nondominant shoulder in 12 (60%) of 20 professional pitchers and 10 (59%) of 17 college pitchers. Although the authors did not correlate translational differences with corresponding shoulder pathology, the observed asymmetry supported the idea that these athletes may experience adaptive glenohumeral changes with repetitive throwing. The association between adaptive changes and shoulder biomechanics has been studied. Burkhart and Lo¹⁵ used a cadaveric model to describe the cam effect of the proximal humerus and the biomechanical consequences of a relative reduction in this effect after pathologic changes within the glenohumeral joint (constriction of posteroinferior capsule). They noted that a posterosuperior shift in the glenohumeral contact point in the throwing position can result in anterior capsular redundancy that may contribute to microinstability of the shoulder. This relative laxity increases external rotation, resulting in increased torsional and shear forces at the rotator cuff insertion.¹⁶ Ultimately, these abnormal forces may predispose overhead athletes to rotator cuff injury.

Given the available literature, it is clear that full-thickness RCTs are potentially career-ending injuries for professional baseball players. The question arises as to why the results are so poor. Ultimately, the high incidence of concomitant intra-articular pathology associated with full-thickness RCTs underscores the severity of soft-tissue damage sustained with repetitive overhead throwing. Mazoué and Andrews⁹ proposed the presence of associated labral and capsular pathology as a potential explanation for poor outcomes of surgical repair. Given the myriad of additional pathology observed in each patient, it is difficult to ascertain the precise impact of these injuries on postoperative outcome. However, early diagnosis and aggressive surgical intervention are clearly necessary to prevent accumulative injury. Regarding surgical intervention, both Tibone and colleagues⁸ and Mazoué and Andrews⁹ reported use of an open surgical repair technique in which the tendon was repaired to the anatomical footprint. Certainly, the benefits of an all-arthroscopic technique include optimal visualization of the RCT, less perioperative morbidity, and minimal soft-tissue injury. With our arthroscopic technique, the rotator cuff was fixed immediately lateral to the anatomical footprint, thereby leaving the medial aspect of the footprint uncovered. Functionally, the goal of this procedure is to restore the integrity of the rotator cuff without compromising glenohumeral mobility acquired through soft-tissue adaptation. Investigation of the insertional anatomy of the rotator cuff has demonstrated that the supraspinatus tendon inserts about 0.9 mm from the edge of the articular surface, and the infraspinatus insertional footprint tapers away from the articular surface to form the bare area as it extends inferiorly on the greater tuberosity.¹⁰ We think preexisting adaptations in glenohumeral anatomy are important for peak performance in this unique population, and even small alterations in the repair location can have deleterious effects on throwing mechanics. Lateralized repair of the cuff precludes potential medialization of the cuff insertion and may facilitate preservation of soft-tissue adaptations that these athletes rely on to achieve extraordinary glenohumeral motion.

Interestingly, with this technique we noted a higher rate of return to MLB play in pitchers over age 30 years. Although several individual factors (eg, player talent level, work ethics, compliance with rehabilitation) may play a role in this finding, it is possible that older, more mature patients may be more willing to assume diminished roles to continue to play. Jones and colleagues¹⁷ recently reported similar findings in older MLB pitchers after revision ulnar collateral ligament reconstruction.

This study had several limitations. First, the patient cohort was small (a result of the nature and relatively infrequent incidence of the clinical problem). Second, clinical information was collected retrospectively, which limited our ability to determine precise differences between preoperative and postoperative glenohumeral ROM with this technique. Third, the cohort included patients who demonstrated additional intra-articular (labral) pathology. Although associated pathology is common in this high-demand athletic population, it is clear that advanced pathology (eg, SLAP tears) may affect clinical outcomes, as in our study. Despite these limitations, our study is the largest review of professional baseball players treated for full-thickness rotator cuff injuries with an arthroscopic technique. Overall, the results of this study are promising and call for further clinical and biomechanical evaluation.

Conclusion

Surgical management of rotator cuff injuries in professional baseball players remains an extremely difficult problem. Current studies of full-thickness RCTs highlight these athletes’ poor functional outcomes. These unfavorable results prompted us to alter our surgical technique. Initial outcomes have been encouraging, and extended follow-up in this cohort of patients will provide a more definitive assessment of the success of this technique.

Repair of the ulnar collateral ligament (UCL) was first reported by Norwood and colleagues¹ in a group of athletes who sustained acute UCL ruptures. Of the 4 athletes in their cohort who underwent direct UCL repair, none were noted to have any residual instability 2 years after the surgery. However, none of these 4 were overhead throwing athletes. Jobe and colleagues² first published Jobe’s technique of UCL reconstruction in 1986, but it was Conway and colleagus’³ 1992 publication describing Jobe’s experience with UCL injury and surgical treatment in throwing athletes that set the early standard for management in that population. Since those landmark studies, there has been a tremendous increase in attention to this near-epidemic clinical problem.

Although these studies were the first to describe the surgical procedure that is now often referred to as “Tommy John surgery,” named after Jobe’s initial patient in 1974, Conway and colleagues³ also reported on Jobe’s early experience with UCL repair. In fact, of the 70 patients reported in the Conway and colleagues’³ article, 14 were treated with repair of the ligament. Only 7 of the 14 (50%) of those who underwent UCL repair were able to return to the same level of play, and only 2 of the 7 (29%) of Major League Baseball (MLB) players who underwent UCL repair were able to return to competition at the MLB level. This compared very poorly with the nearly 75% rate of return to competition in patients who underwent UCL reconstructions in the same cohort. In Azar and colleagues’⁴ 2000 report on Dr. James Andrews’ experience with UCL injury and treatment in male college and professional baseball players, UCL repair again did poorly when compared to UCL reconstruction, with only 5 of the 8 (63%) of UCL repair patients returning to the same level of play compared to 41 of the 51 (81%) of UCL reconstructions using a modification of Jobe’s original technique.

Since the mid-1990s, numerous new techniques have been described and shown to have acceptable and largely successful outcomes in treating UCL injuries.^5-9 All of them involve placing or anchoring a spanning piece of tendon graft from the native origin on the medial epicondyle of the humerus to the native insertion on the sublime tubercle of the ulna. These palpable and visible anatomic landmarks are important to the UCL surgeon due to the need to place the graft or repair the torn ligament tissue to its normal anatomic origin and/or insertion.¹⁰ Regardless of whether the graft is sewn, docked, tunneled, or anchored, these types of procedures have demonstrated rates of return to competition at the same or higher level of play in the 75% to 92% range.^3,4,7,11-13 In the largest published series of 1281 UCL reconstructions by Cain and colleagues⁷ at American Sports Medicine Institute (Birmingham, AL), the rate of return to play at the same or higher level was 84%, with the average time to return to play of 11.4 months. On the basis of these robust clinical studies and numerous basic science studies demonstrating essentially equivalent strength and function among reconstruction techniques, UCL reconstruction now enjoys an acceptance among clinicians, athletes, athletic trainers, coaches, and team management at all levels of overhead sports.

In comparison to UCL reconstruction, relatively little has been published on UCL repair since 2000. Certainly this is in part due to the success of its clinical descendant. UCL repair did not appear on the pages of peer-reviewed literature until 2006, when Argo and colleagues¹¹ published a report on the outcome of 17 UCL repairs in female athletes using a variety of techniques, including plication, anchor-to-bone, and drill holes. Although there was only 1 pitcher in the group, 16 of the 17 (94%) returned to the same or higher level of competition at an average of only 3 months after surgery.¹¹

Savoie and colleagues¹³ followed this in 2008 with a report on 60 UCL repairs in overhead athletes. Of the 51 patients in this study in which the ligament was repaired to bone using suture anchors, 93% returned to the same or higher level of play at an average of only 6 months after surgery. Including Jobe’s original group, there have been less than 100 patients ever reported to have had a UCL repair performed. In comparison to the thousands of UCL reconstructions that have been reported over the last 20 years, it is not surprising that UCL repair has not gained great popularity among surgeons and patients. It is also important to remember that suture and anchor technology has come a long way since the 1970s, and our overall knowledge of the injury and its treatments and rehabilitation have grown tremendously since that time.

A New Technique for UCL Repair

Since we began data collection in Birmingham, Alabama in the mid 1990s, our practice has successfully treated thousands of overhead athletes of all types with the modified Jobe technique of UCL reconstruction, using either a palmaris longus tendon or a gracilis tendon graft.⁷ Until August 2013, this technique was exclusively utilized regardless of the amount and location of pathology encountered at the time of surgery. The range of pathology, from partial undersurface tearing to complete disruption of the ligament tissue, was treated by placing a graft at the anatomic insertion points of the native ligament. While the success of this experience cannot be overlooked, we also realized that we were treating a broad spectrum of pathology and injury with the same operation.

Recognizing the valuable contributions of earlier authors who had attempted UCL repair previously, we asked whether we were doing too much of an operation for all of the various pathology we saw at the time of surgery, and whether the availability of modern anchor and suture technology, vast clinical experience with these injuries and their outcomes, and even biologic additives could be applied to some of these patients in order to achieve an equal or superior outcome in less time. In particular, could such a technique be applied to the ever-increasing number of younger athletes with less pathology, who more frequently suffer end-avulsions and partial tears of their UCL?

These thoughts, along with Savoie and colleagues’¹³ experience with UCL repair using suture anchors, led us to create a construct that could be used to not only repair the torn native UCL tissue to bone, but also span the anatomic native ligament from its origin to its insertion. The construct includes an ultra-strong collagen coated tape (FiberTape, Arthrex) attached at the anatomic insertions of the ligament using two 3.5-mm nonabsorbable PEEK corkscrew anchors (SwiveLock, Arthrex), and a suture through the eyelet of one of the anchors (Figure 1). Using this construct, the native ligament disruption can be repaired directly to bone using the suture through the eyelet of the anchor, and the remainder of the native ligament is augmented with the spanning biologic enhanced tape (Figures 2A-2C). The construct is created by placing one end of the tape through the eyelet of the first anchor, and then placing one end of a No. zero braided permanent suture through the same eyelet. Both ends of the tape are then placed through the eyelet of the second anchor. The first anchor is inserted into a hole drilled at the apex of the insertion of the torn end of the native ligament. This anchor is placed first in order to allow for direct repair of the native torn ligament using the free suture through the eyelet of the first anchor. The second hole is then drilled at the insertion of the native ligament on the uninjured end of the native ligament. In order to accommodate the volume of tape in the hole created for the second anchor, a slightly oversized drill and tap were created specifically for this technique (Arthrex).

Before attempting this in vivo, a cadaveric study was carried out in order to ensure that the time-zero function of the construct would be at least as good as the standard UCL reconstruction technique we have used for several decades.¹⁴ The time-zero gap formation under valgus load was less for the repair/augmentation than for the standard reconstruction with palmaris longus, and the ultimate failure strength of the repair was the same as in the reconstruction group, with all failures through bone in the cadaveric specimens. No anchors pulled out of bone, and the tape did not tear in any specimen.

This basic science study has given us confidence to proceed with the use of this technique in patients. The first patient was treated with this construct in August 2013. The outcomes of our first series of patients were presented on Saturday, March 5 at American Orthopaedic Society for Sports Medicine Specialty Day during the 2016 American Academy of Orthopaedic Surgeons annual meeting in Orlando, FL.

We do not feel that this technique is adequate for the treatment of the UCL that has sustained attritional injury and contains poor quality native ligament tissue. Before we do these procedures, we always discuss with the patient the possibility that full reconstruction may be required, and that the decision to proceed with UCL repair is contingent upon the quality and quantity of the native UCL tissue present at the time of surgery. If the quality of the native tissue is poor (chronic degenerative changes, etc), full reconstruction with autograft tendon is recommended. It is our hope that this technique will afford the UCL surgeon another option for treating end-avulsions and partial thickness injuries, with a more rapid and successful return to normal function and competition.

Repair of the ulnar collateral ligament (UCL) was first reported by Norwood and colleagues¹ in a group of athletes who sustained acute UCL ruptures. Of the 4 athletes in their cohort who underwent direct UCL repair, none were noted to have any residual instability 2 years after the surgery. However, none of these 4 were overhead throwing athletes. Jobe and colleagues² first published Jobe’s technique of UCL reconstruction in 1986, but it was Conway and colleagus’³ 1992 publication describing Jobe’s experience with UCL injury and surgical treatment in throwing athletes that set the early standard for management in that population. Since those landmark studies, there has been a tremendous increase in attention to this near-epidemic clinical problem.

Although these studies were the first to describe the surgical procedure that is now often referred to as “Tommy John surgery,” named after Jobe’s initial patient in 1974, Conway and colleagues³ also reported on Jobe’s early experience with UCL repair. In fact, of the 70 patients reported in the Conway and colleagues’³ article, 14 were treated with repair of the ligament. Only 7 of the 14 (50%) of those who underwent UCL repair were able to return to the same level of play, and only 2 of the 7 (29%) of Major League Baseball (MLB) players who underwent UCL repair were able to return to competition at the MLB level. This compared very poorly with the nearly 75% rate of return to competition in patients who underwent UCL reconstructions in the same cohort. In Azar and colleagues’⁴ 2000 report on Dr. James Andrews’ experience with UCL injury and treatment in male college and professional baseball players, UCL repair again did poorly when compared to UCL reconstruction, with only 5 of the 8 (63%) of UCL repair patients returning to the same level of play compared to 41 of the 51 (81%) of UCL reconstructions using a modification of Jobe’s original technique.

Since the mid-1990s, numerous new techniques have been described and shown to have acceptable and largely successful outcomes in treating UCL injuries.^5-9 All of them involve placing or anchoring a spanning piece of tendon graft from the native origin on the medial epicondyle of the humerus to the native insertion on the sublime tubercle of the ulna. These palpable and visible anatomic landmarks are important to the UCL surgeon due to the need to place the graft or repair the torn ligament tissue to its normal anatomic origin and/or insertion.¹⁰ Regardless of whether the graft is sewn, docked, tunneled, or anchored, these types of procedures have demonstrated rates of return to competition at the same or higher level of play in the 75% to 92% range.^3,4,7,11-13 In the largest published series of 1281 UCL reconstructions by Cain and colleagues⁷ at American Sports Medicine Institute (Birmingham, AL), the rate of return to play at the same or higher level was 84%, with the average time to return to play of 11.4 months. On the basis of these robust clinical studies and numerous basic science studies demonstrating essentially equivalent strength and function among reconstruction techniques, UCL reconstruction now enjoys an acceptance among clinicians, athletes, athletic trainers, coaches, and team management at all levels of overhead sports.

In comparison to UCL reconstruction, relatively little has been published on UCL repair since 2000. Certainly this is in part due to the success of its clinical descendant. UCL repair did not appear on the pages of peer-reviewed literature until 2006, when Argo and colleagues¹¹ published a report on the outcome of 17 UCL repairs in female athletes using a variety of techniques, including plication, anchor-to-bone, and drill holes. Although there was only 1 pitcher in the group, 16 of the 17 (94%) returned to the same or higher level of competition at an average of only 3 months after surgery.¹¹

Savoie and colleagues¹³ followed this in 2008 with a report on 60 UCL repairs in overhead athletes. Of the 51 patients in this study in which the ligament was repaired to bone using suture anchors, 93% returned to the same or higher level of play at an average of only 6 months after surgery. Including Jobe’s original group, there have been less than 100 patients ever reported to have had a UCL repair performed. In comparison to the thousands of UCL reconstructions that have been reported over the last 20 years, it is not surprising that UCL repair has not gained great popularity among surgeons and patients. It is also important to remember that suture and anchor technology has come a long way since the 1970s, and our overall knowledge of the injury and its treatments and rehabilitation have grown tremendously since that time.

A New Technique for UCL Repair

Since we began data collection in Birmingham, Alabama in the mid 1990s, our practice has successfully treated thousands of overhead athletes of all types with the modified Jobe technique of UCL reconstruction, using either a palmaris longus tendon or a gracilis tendon graft.⁷ Until August 2013, this technique was exclusively utilized regardless of the amount and location of pathology encountered at the time of surgery. The range of pathology, from partial undersurface tearing to complete disruption of the ligament tissue, was treated by placing a graft at the anatomic insertion points of the native ligament. While the success of this experience cannot be overlooked, we also realized that we were treating a broad spectrum of pathology and injury with the same operation.

Recognizing the valuable contributions of earlier authors who had attempted UCL repair previously, we asked whether we were doing too much of an operation for all of the various pathology we saw at the time of surgery, and whether the availability of modern anchor and suture technology, vast clinical experience with these injuries and their outcomes, and even biologic additives could be applied to some of these patients in order to achieve an equal or superior outcome in less time. In particular, could such a technique be applied to the ever-increasing number of younger athletes with less pathology, who more frequently suffer end-avulsions and partial tears of their UCL?

These thoughts, along with Savoie and colleagues’¹³ experience with UCL repair using suture anchors, led us to create a construct that could be used to not only repair the torn native UCL tissue to bone, but also span the anatomic native ligament from its origin to its insertion. The construct includes an ultra-strong collagen coated tape (FiberTape, Arthrex) attached at the anatomic insertions of the ligament using two 3.5-mm nonabsorbable PEEK corkscrew anchors (SwiveLock, Arthrex), and a suture through the eyelet of one of the anchors (Figure 1). Using this construct, the native ligament disruption can be repaired directly to bone using the suture through the eyelet of the anchor, and the remainder of the native ligament is augmented with the spanning biologic enhanced tape (Figures 2A-2C). The construct is created by placing one end of the tape through the eyelet of the first anchor, and then placing one end of a No. zero braided permanent suture through the same eyelet. Both ends of the tape are then placed through the eyelet of the second anchor. The first anchor is inserted into a hole drilled at the apex of the insertion of the torn end of the native ligament. This anchor is placed first in order to allow for direct repair of the native torn ligament using the free suture through the eyelet of the first anchor. The second hole is then drilled at the insertion of the native ligament on the uninjured end of the native ligament. In order to accommodate the volume of tape in the hole created for the second anchor, a slightly oversized drill and tap were created specifically for this technique (Arthrex).