Original Research

Prolonged intensive care unit (ICU) stays, variably defined as > 48 h to > 14 days, are a known complication of cardiac surgery.^1-8 Prolonged stays are associated with higher resource utilization and higher mortality.^2,3,9-12 Although there are several cardiac surgery risk models that can be used preoperatively to identify patients at risk for prolonged ICU stay, factors that influence outcomes for patients who experience prolonged ICU stays are poorly understood.^2,13-19 Little information is available to inform discussions between health care practitioners (HCPs) and patients throughout a prolonged ICU stay, especially those ≥ 7 days.

As cardiac surgical complexity, patient age, and preexisting comorbidities have increased over time, so has the need to provide patients and HCPs with data to inform decision making, enhance prognostication, and set realistic expectations at varying time intervals during prolonged ICU stay. The purpose of this study was to evaluate short- and long-term outcomes in cardiac surgery patients after prolonged ICU stays at relevant time intervals (7, 14, 21, and 28 days) and to determine factors that may predict a patient’s outcome after a prolonged ICU stay.

Methods

The University of Michigan Health System Institutional Review Board approved this study and waived informed consent. We merged the University of Michigan Medical Center Society of Thoracic Surgeons (STS) database, which is updated periodically with late mortality, with elements of the electronic health record (EHR). Adult patients were included if they had cardiac surgery at the University of Michigan between January 2, 2001, and December 31, 2011. Late mortality was updated through December 1, 2014. Data are presented as frequency (%), mean (SD), and median (IQR) as appropriate. Bivariate comparisons between survivors and nonsurvivors were done with χ² or Fisher exact test for categorical data, Student t test for continuous normally distributed data, and Wilcoxon rank sum test for continuous not normally distributed data. To determine factors associated with operative mortality (death within 30 days of surgery or hospital discharge, whichever occurred later), we used logistic regression with forward selection. All available factors were initially entered in the models.

Separate logistic models were created based on all data available at days 7, 14, 21, and 28. Final models consisted of factors with statistically significant P values (< .05) and adjusted odds ratios (AORs) with 95% CIs that excluded 1. To determine factors associated with late mortality, we used a Cox proportional hazard model, which used data available at discharge and STS complications. As these complications did not include their timing, they could only be used in models created at discharge and not for days 7, 14, 21, and 28 models. Final models consisted of factors with P values < .05 and 95% CIs of the AORs or the hazard ratios (HRs) that excluded 1. As the EHR did not start recording data until January 2, 2004, and its capture of data remained incomplete for several years, rather than imputing these missing data or excluding these patients, we chose to create an extra categorical level for each factor to represent missing data. For continuous factors with missing data, we first converted the continuous data to terciles and the missing data became the fourth level.^20,21

The discrimination of the logistic models were determined by the c-statistic and for the Cox proportional hazards model with the Harrell concordance index (C index). Time trends were assessed with the Cochran-Armitage trend test. P < .05 was deemed statistically significant. Statistics were calculated with SPSS versions 21-23 or SAS 9.4.

Results

Of 8309 admissions to the ICU after cardiac surgery, 1174 (14%) had ICU stays ≥ 7 days, 386 (5%) ≥ 14 days, 201 (2%) ≥ 21 days, and 80 (0.9%) ≥ 28 days. The prolonged ICU study population was mostly male, White race, with a mean (SD) age of 62 (14) years. Patients had a variety of comorbidities, most notably 61% had hypertension and half had heart failure. Valve surgery (55%) was the most common procedure (n = 651). Twenty-nine percent required > 1 procedure (eAppendix 1).

The operative mortality

Using multivariable logistic regression to adjust for factors associated with mortality, we found that receiving mechanical ventilation on the day of analysis was associated with increased operative mortality with AOR increasing from 3.35 (95% CI, 2.82-3.98) for

Discussion

We found that operative mortality increased the longer the patient stayed in the ICU, ranging from 11% for ≥ 7 days to 35% for ≥ 28 days. We further found that in ICU survivors, median (IQR) survival was 10.7 (0.7) years. While previous studies have evaluated prolonged ICU stays, they have been limited by studying limited subpopulations, such as patients who are dependent on dialysis or octogenarians, or used a single cutoff to define prolonged ICU stays, variably defined from > 48 hours to > 14 days.^2-7,9-12,22 Our study is similar to others that used ≥ 2 cutoffs.^1,8 However, our study was novel by providing 4 cutoffs to improve temporal prediction of hospital outcomes. Unlike a study by Ryan and colleagues, which found no increase in mortality with longer stay (43.5% for ≥ 14 days and 45% for ≥ 28 days), our study findings are similar to those of Yu and colleagues (11.1% mortality for prolonged ICU stays of 1 to 2 weeks, 26.6% for 2 to 4 weeks, and 31% for > 4 weeks) and others (8%, 3 to 14 days; 40%, >14 days; 10%, 1 to 2 weeks; 25.7% > 2 weeks) in finding a progressively increased hospital mortality with longer ICU stays.^1,4,5,8 These differences may be related to different ICU populations or to improvements in care since Ryan and colleagues study was conducted.

Fewer studies have evaluated factors associated with mortality in cardiac surgery prolonged ICU stay patients. Our study is similar to other studies that evaluated risk factors by finding associations between a variety of comorbidities and process of care associated with both operative and long-term mortality; however, comparison between these studies is limited by the varying factors analyzed.^{1,3,5,6,8,9,11} We found that mechanical ventilation on days 7, 14, 21, and 28 was strongly associated with operative mortality, similar to noncardiac surgery patients and cardiac surgery patients.^6,23,24 While we found several processes of care, such as catecholamine use and transfusions to be associated with mortality, which is similar to other studies, notably, we did not find an association between renal replacement therapy and mortality.^1,25 While there is an association between renal replacement therapy and mortality in ICU patients, its status in cardiac surgery patients with prolonged ICU stays is less clear.²⁶ While Ryan and colleagues found an association between renal replacement therapy and hospital mortality in patients staying ≥ 14 days, they did not find it in patients staying ≥.

28 days.¹ Other studies of prolonged ICU stays for cardiac surgery patients have also failed to find an association between renal replacement therapy and mortality.^5,6,9 Importantly, practice that expedites liberation from mechanical ventilation, such as fast tracking, daily spontaneous breathing trials, extubation to noninvasive respiratory support, and pulmonary rehabilitation may all have potential to limit mechanical ventilation duration and improve hospital survival and deserve further study.^27-29Median (IQR) survival in hospital survivors was 10.7 (0.7) years, which is generally better than previously reported, but similar to that reported by Silberman and colleagues.^{2,4,6,8,11,12} Differences between these studies may relate to different patient populations within the cardiac surgery ICUs, definitions of prolonged ICU stays, or eras of care. Further study is needed to clarify these discrepancies. We found that cardiac transplantation and obesity were associated with the least risk of dying, while smoking, lung disease, and postoperative pneumonia were independently associated with increased hazard of dying. The obesity paradox, where obesity is protective, has been previously observed in cardiac surgery patients.³⁰

Strengths and Limitations

There are several limitations of this study. This is a single center study, and our patient population and processes of care may differ from other centers, limiting its generalizability. Notably, we do fewer coronary bypass operations and more aortic reconstructions and ventricular assist device insertions than do many other centers. Second, we did not have laboratory values for about one-third of patients (preceded EHR implementation). However, we were able to compensate for this by binning values and including missing data as an extra bin.^20,21

The main strength of this study is that we were able to combine disparate records to assess a large number of potential factors associated with both operative and long-term mortality. This produced models that had good to very good discrimination. By producing models at 7, 14, 21, and 28 days to predict operative mortality and a model at discharge, it may help to provide objective data to facilitate conversations with patients and their families. However, further studies to externally validate these models should be conducted.

Conclusions

We found that longer prolonged ICU stays are associated with both operative and late mortality. Receiving mechanical ventilation on days 7, 14, 21, or 28 was strongly associated with operative mortality.

Prolonged intensive care unit (ICU) stays, variably defined as > 48 h to > 14 days, are a known complication of cardiac surgery.^1-8 Prolonged stays are associated with higher resource utilization and higher mortality.^2,3,9-12 Although there are several cardiac surgery risk models that can be used preoperatively to identify patients at risk for prolonged ICU stay, factors that influence outcomes for patients who experience prolonged ICU stays are poorly understood.^2,13-19 Little information is available to inform discussions between health care practitioners (HCPs) and patients throughout a prolonged ICU stay, especially those ≥ 7 days.

As cardiac surgical complexity, patient age, and preexisting comorbidities have increased over time, so has the need to provide patients and HCPs with data to inform decision making, enhance prognostication, and set realistic expectations at varying time intervals during prolonged ICU stay. The purpose of this study was to evaluate short- and long-term outcomes in cardiac surgery patients after prolonged ICU stays at relevant time intervals (7, 14, 21, and 28 days) and to determine factors that may predict a patient’s outcome after a prolonged ICU stay.

Methods

The University of Michigan Health System Institutional Review Board approved this study and waived informed consent. We merged the University of Michigan Medical Center Society of Thoracic Surgeons (STS) database, which is updated periodically with late mortality, with elements of the electronic health record (EHR). Adult patients were included if they had cardiac surgery at the University of Michigan between January 2, 2001, and December 31, 2011. Late mortality was updated through December 1, 2014. Data are presented as frequency (%), mean (SD), and median (IQR) as appropriate. Bivariate comparisons between survivors and nonsurvivors were done with χ² or Fisher exact test for categorical data, Student t test for continuous normally distributed data, and Wilcoxon rank sum test for continuous not normally distributed data. To determine factors associated with operative mortality (death within 30 days of surgery or hospital discharge, whichever occurred later), we used logistic regression with forward selection. All available factors were initially entered in the models.

Separate logistic models were created based on all data available at days 7, 14, 21, and 28. Final models consisted of factors with statistically significant P values (< .05) and adjusted odds ratios (AORs) with 95% CIs that excluded 1. To determine factors associated with late mortality, we used a Cox proportional hazard model, which used data available at discharge and STS complications. As these complications did not include their timing, they could only be used in models created at discharge and not for days 7, 14, 21, and 28 models. Final models consisted of factors with P values < .05 and 95% CIs of the AORs or the hazard ratios (HRs) that excluded 1. As the EHR did not start recording data until January 2, 2004, and its capture of data remained incomplete for several years, rather than imputing these missing data or excluding these patients, we chose to create an extra categorical level for each factor to represent missing data. For continuous factors with missing data, we first converted the continuous data to terciles and the missing data became the fourth level.^20,21

The discrimination of the logistic models were determined by the c-statistic and for the Cox proportional hazards model with the Harrell concordance index (C index). Time trends were assessed with the Cochran-Armitage trend test. P < .05 was deemed statistically significant. Statistics were calculated with SPSS versions 21-23 or SAS 9.4.

Results

Of 8309 admissions to the ICU after cardiac surgery, 1174 (14%) had ICU stays ≥ 7 days, 386 (5%) ≥ 14 days, 201 (2%) ≥ 21 days, and 80 (0.9%) ≥ 28 days. The prolonged ICU study population was mostly male, White race, with a mean (SD) age of 62 (14) years. Patients had a variety of comorbidities, most notably 61% had hypertension and half had heart failure. Valve surgery (55%) was the most common procedure (n = 651). Twenty-nine percent required > 1 procedure (eAppendix 1).

The operative mortality

Using multivariable logistic regression to adjust for factors associated with mortality, we found that receiving mechanical ventilation on the day of analysis was associated with increased operative mortality with AOR increasing from 3.35 (95% CI, 2.82-3.98) for

Discussion

We found that operative mortality increased the longer the patient stayed in the ICU, ranging from 11% for ≥ 7 days to 35% for ≥ 28 days. We further found that in ICU survivors, median (IQR) survival was 10.7 (0.7) years. While previous studies have evaluated prolonged ICU stays, they have been limited by studying limited subpopulations, such as patients who are dependent on dialysis or octogenarians, or used a single cutoff to define prolonged ICU stays, variably defined from > 48 hours to > 14 days.^2-7,9-12,22 Our study is similar to others that used ≥ 2 cutoffs.^1,8 However, our study was novel by providing 4 cutoffs to improve temporal prediction of hospital outcomes. Unlike a study by Ryan and colleagues, which found no increase in mortality with longer stay (43.5% for ≥ 14 days and 45% for ≥ 28 days), our study findings are similar to those of Yu and colleagues (11.1% mortality for prolonged ICU stays of 1 to 2 weeks, 26.6% for 2 to 4 weeks, and 31% for > 4 weeks) and others (8%, 3 to 14 days; 40%, >14 days; 10%, 1 to 2 weeks; 25.7% > 2 weeks) in finding a progressively increased hospital mortality with longer ICU stays.^1,4,5,8 These differences may be related to different ICU populations or to improvements in care since Ryan and colleagues study was conducted.

Fewer studies have evaluated factors associated with mortality in cardiac surgery prolonged ICU stay patients. Our study is similar to other studies that evaluated risk factors by finding associations between a variety of comorbidities and process of care associated with both operative and long-term mortality; however, comparison between these studies is limited by the varying factors analyzed.^{1,3,5,6,8,9,11} We found that mechanical ventilation on days 7, 14, 21, and 28 was strongly associated with operative mortality, similar to noncardiac surgery patients and cardiac surgery patients.^6,23,24 While we found several processes of care, such as catecholamine use and transfusions to be associated with mortality, which is similar to other studies, notably, we did not find an association between renal replacement therapy and mortality.^1,25 While there is an association between renal replacement therapy and mortality in ICU patients, its status in cardiac surgery patients with prolonged ICU stays is less clear.²⁶ While Ryan and colleagues found an association between renal replacement therapy and hospital mortality in patients staying ≥ 14 days, they did not find it in patients staying ≥.

28 days.¹ Other studies of prolonged ICU stays for cardiac surgery patients have also failed to find an association between renal replacement therapy and mortality.^5,6,9 Importantly, practice that expedites liberation from mechanical ventilation, such as fast tracking, daily spontaneous breathing trials, extubation to noninvasive respiratory support, and pulmonary rehabilitation may all have potential to limit mechanical ventilation duration and improve hospital survival and deserve further study.^27-29Median (IQR) survival in hospital survivors was 10.7 (0.7) years, which is generally better than previously reported, but similar to that reported by Silberman and colleagues.^{2,4,6,8,11,12} Differences between these studies may relate to different patient populations within the cardiac surgery ICUs, definitions of prolonged ICU stays, or eras of care. Further study is needed to clarify these discrepancies. We found that cardiac transplantation and obesity were associated with the least risk of dying, while smoking, lung disease, and postoperative pneumonia were independently associated with increased hazard of dying. The obesity paradox, where obesity is protective, has been previously observed in cardiac surgery patients.³⁰

Strengths and Limitations

There are several limitations of this study. This is a single center study, and our patient population and processes of care may differ from other centers, limiting its generalizability. Notably, we do fewer coronary bypass operations and more aortic reconstructions and ventricular assist device insertions than do many other centers. Second, we did not have laboratory values for about one-third of patients (preceded EHR implementation). However, we were able to compensate for this by binning values and including missing data as an extra bin.^20,21

The main strength of this study is that we were able to combine disparate records to assess a large number of potential factors associated with both operative and long-term mortality. This produced models that had good to very good discrimination. By producing models at 7, 14, 21, and 28 days to predict operative mortality and a model at discharge, it may help to provide objective data to facilitate conversations with patients and their families. However, further studies to externally validate these models should be conducted.

Conclusions

We found that longer prolonged ICU stays are associated with both operative and late mortality. Receiving mechanical ventilation on days 7, 14, 21, or 28 was strongly associated with operative mortality.

Prolonged intensive care unit (ICU) stays, variably defined as > 48 h to > 14 days, are a known complication of cardiac surgery.^1-8 Prolonged stays are associated with higher resource utilization and higher mortality.^2,3,9-12 Although there are several cardiac surgery risk models that can be used preoperatively to identify patients at risk for prolonged ICU stay, factors that influence outcomes for patients who experience prolonged ICU stays are poorly understood.^2,13-19 Little information is available to inform discussions between health care practitioners (HCPs) and patients throughout a prolonged ICU stay, especially those ≥ 7 days.

As cardiac surgical complexity, patient age, and preexisting comorbidities have increased over time, so has the need to provide patients and HCPs with data to inform decision making, enhance prognostication, and set realistic expectations at varying time intervals during prolonged ICU stay. The purpose of this study was to evaluate short- and long-term outcomes in cardiac surgery patients after prolonged ICU stays at relevant time intervals (7, 14, 21, and 28 days) and to determine factors that may predict a patient’s outcome after a prolonged ICU stay.

Methods

The University of Michigan Health System Institutional Review Board approved this study and waived informed consent. We merged the University of Michigan Medical Center Society of Thoracic Surgeons (STS) database, which is updated periodically with late mortality, with elements of the electronic health record (EHR). Adult patients were included if they had cardiac surgery at the University of Michigan between January 2, 2001, and December 31, 2011. Late mortality was updated through December 1, 2014. Data are presented as frequency (%), mean (SD), and median (IQR) as appropriate. Bivariate comparisons between survivors and nonsurvivors were done with χ² or Fisher exact test for categorical data, Student t test for continuous normally distributed data, and Wilcoxon rank sum test for continuous not normally distributed data. To determine factors associated with operative mortality (death within 30 days of surgery or hospital discharge, whichever occurred later), we used logistic regression with forward selection. All available factors were initially entered in the models.

Separate logistic models were created based on all data available at days 7, 14, 21, and 28. Final models consisted of factors with statistically significant P values (< .05) and adjusted odds ratios (AORs) with 95% CIs that excluded 1. To determine factors associated with late mortality, we used a Cox proportional hazard model, which used data available at discharge and STS complications. As these complications did not include their timing, they could only be used in models created at discharge and not for days 7, 14, 21, and 28 models. Final models consisted of factors with P values < .05 and 95% CIs of the AORs or the hazard ratios (HRs) that excluded 1. As the EHR did not start recording data until January 2, 2004, and its capture of data remained incomplete for several years, rather than imputing these missing data or excluding these patients, we chose to create an extra categorical level for each factor to represent missing data. For continuous factors with missing data, we first converted the continuous data to terciles and the missing data became the fourth level.^20,21

The discrimination of the logistic models were determined by the c-statistic and for the Cox proportional hazards model with the Harrell concordance index (C index). Time trends were assessed with the Cochran-Armitage trend test. P < .05 was deemed statistically significant. Statistics were calculated with SPSS versions 21-23 or SAS 9.4.

Results

Of 8309 admissions to the ICU after cardiac surgery, 1174 (14%) had ICU stays ≥ 7 days, 386 (5%) ≥ 14 days, 201 (2%) ≥ 21 days, and 80 (0.9%) ≥ 28 days. The prolonged ICU study population was mostly male, White race, with a mean (SD) age of 62 (14) years. Patients had a variety of comorbidities, most notably 61% had hypertension and half had heart failure. Valve surgery (55%) was the most common procedure (n = 651). Twenty-nine percent required > 1 procedure (eAppendix 1).

The operative mortality

Using multivariable logistic regression to adjust for factors associated with mortality, we found that receiving mechanical ventilation on the day of analysis was associated with increased operative mortality with AOR increasing from 3.35 (95% CI, 2.82-3.98) for

Discussion

We found that operative mortality increased the longer the patient stayed in the ICU, ranging from 11% for ≥ 7 days to 35% for ≥ 28 days. We further found that in ICU survivors, median (IQR) survival was 10.7 (0.7) years. While previous studies have evaluated prolonged ICU stays, they have been limited by studying limited subpopulations, such as patients who are dependent on dialysis or octogenarians, or used a single cutoff to define prolonged ICU stays, variably defined from > 48 hours to > 14 days.^2-7,9-12,22 Our study is similar to others that used ≥ 2 cutoffs.^1,8 However, our study was novel by providing 4 cutoffs to improve temporal prediction of hospital outcomes. Unlike a study by Ryan and colleagues, which found no increase in mortality with longer stay (43.5% for ≥ 14 days and 45% for ≥ 28 days), our study findings are similar to those of Yu and colleagues (11.1% mortality for prolonged ICU stays of 1 to 2 weeks, 26.6% for 2 to 4 weeks, and 31% for > 4 weeks) and others (8%, 3 to 14 days; 40%, >14 days; 10%, 1 to 2 weeks; 25.7% > 2 weeks) in finding a progressively increased hospital mortality with longer ICU stays.^1,4,5,8 These differences may be related to different ICU populations or to improvements in care since Ryan and colleagues study was conducted.

Fewer studies have evaluated factors associated with mortality in cardiac surgery prolonged ICU stay patients. Our study is similar to other studies that evaluated risk factors by finding associations between a variety of comorbidities and process of care associated with both operative and long-term mortality; however, comparison between these studies is limited by the varying factors analyzed.^{1,3,5,6,8,9,11} We found that mechanical ventilation on days 7, 14, 21, and 28 was strongly associated with operative mortality, similar to noncardiac surgery patients and cardiac surgery patients.^6,23,24 While we found several processes of care, such as catecholamine use and transfusions to be associated with mortality, which is similar to other studies, notably, we did not find an association between renal replacement therapy and mortality.^1,25 While there is an association between renal replacement therapy and mortality in ICU patients, its status in cardiac surgery patients with prolonged ICU stays is less clear.²⁶ While Ryan and colleagues found an association between renal replacement therapy and hospital mortality in patients staying ≥ 14 days, they did not find it in patients staying ≥.

28 days.¹ Other studies of prolonged ICU stays for cardiac surgery patients have also failed to find an association between renal replacement therapy and mortality.^5,6,9 Importantly, practice that expedites liberation from mechanical ventilation, such as fast tracking, daily spontaneous breathing trials, extubation to noninvasive respiratory support, and pulmonary rehabilitation may all have potential to limit mechanical ventilation duration and improve hospital survival and deserve further study.^27-29Median (IQR) survival in hospital survivors was 10.7 (0.7) years, which is generally better than previously reported, but similar to that reported by Silberman and colleagues.^{2,4,6,8,11,12} Differences between these studies may relate to different patient populations within the cardiac surgery ICUs, definitions of prolonged ICU stays, or eras of care. Further study is needed to clarify these discrepancies. We found that cardiac transplantation and obesity were associated with the least risk of dying, while smoking, lung disease, and postoperative pneumonia were independently associated with increased hazard of dying. The obesity paradox, where obesity is protective, has been previously observed in cardiac surgery patients.³⁰

Strengths and Limitations

There are several limitations of this study. This is a single center study, and our patient population and processes of care may differ from other centers, limiting its generalizability. Notably, we do fewer coronary bypass operations and more aortic reconstructions and ventricular assist device insertions than do many other centers. Second, we did not have laboratory values for about one-third of patients (preceded EHR implementation). However, we were able to compensate for this by binning values and including missing data as an extra bin.^20,21

The main strength of this study is that we were able to combine disparate records to assess a large number of potential factors associated with both operative and long-term mortality. This produced models that had good to very good discrimination. By producing models at 7, 14, 21, and 28 days to predict operative mortality and a model at discharge, it may help to provide objective data to facilitate conversations with patients and their families. However, further studies to externally validate these models should be conducted.

Conclusions

We found that longer prolonged ICU stays are associated with both operative and late mortality. Receiving mechanical ventilation on days 7, 14, 21, or 28 was strongly associated with operative mortality.

The US Department of Veterans Affairs (VA) is the largest integrated health care system in the United States, providing care for more than 9 million veterans.¹ With veterans experiencing mental health conditions like posttraumatic stress disorder (PTSD), substance use disorders, and other serious mental illnesses (SMI) at higher rates compared with the general population, the VA plays an important role in the provision of mental health services.^2-5 Since the implementation of its Mental Health Strategic Plan in 2004, the VA has overseen the development of a wide array of mental health programs geared toward the complex needs of veterans. Research has demonstrated VA care outperforming Medicaid-reimbursed services in terms of the percentage of veterans filling antidepressants for at least 12 weeks after initiation of treatment for major depressive disorder (MDD), as well as posthospitalization follow-up.⁶

Eligible veterans enrolled in the VA often also seek non-VA care. Medicaid covers nearly 10% of all nonelderly veterans, and of these veterans, 39% rely solely on Medicaid for health care access.⁷ Today, Medicaid is the largest payer for mental health services in the US, providing coverage for approximately 27% of Americans who have SMI and helping fulfill unmet mental health needs.^8,9 Understanding which of these systems veterans choose to use, and under which circumstances, is essential in guiding the allocation of limited health care resources.¹⁰

Beyond Medicaid, alternatives to VA care may include TRICARE, Medicare, Indian Health Services, and employer-based or self-purchased private insurance. While these options potentially increase convenience, choice, and access to health care practitioners (HCPs) and services not available at local VA systems, cross-system utilization with poor integration may cause care coordination and continuity problems, such as medication mismanagement and opioid overdose, unnecessary duplicate utilization, and possible increased mortality.^11-15 As recent national legislative changes, such as the Patient Protection and Affordable Care Act (ACA), Veterans Access, Choice and Accountability Act, and the VA MISSION Act, continue to shift the health care landscape for veterans, questions surrounding how veterans are changing their health care use become significant.^16,17

Here, we approach the impacts of Medicaid expansion on veterans’ reliance on the VA for mental health services with a unique lens. We leverage a difference-in-difference design to study 2 historical Medicaid expansions in Arizona (AZ) and New York (NY), which extended eligibility to childless adults in 2001. Prior Medicaid dual-eligible mental health research investigated reliance shifts during the immediate postenrollment year in a subset of veterans newly enrolled in Medicaid.¹⁸ However, this study took place in a period of relative policy stability. In contrast, we investigate the potential effects of a broad policy shift by analyzing state-level changes in veterans’ reliance over 6 years after a statewide Medicaid expansion. We match expansion states with demographically similar nonexpansion states to account for unobserved trends and confounding effects. Prior studies have used this method to evaluate post-Medicaid expansion mortality changes and changes in veteran dual enrollment and hospitalizations.^10,19 While a study of ACA Medicaid expansion states would be ideal, Medicaid data from most states were only available through 2014 at the time of this analysis. Our study offers a quasi-experimental framework leveraging longitudinal data that can be applied as more post-ACA data become available.

Given the rising incidence of suicide among veterans, understanding care-seeking behaviors for depression among veterans is important as it is the most common psychiatric condition found in those who died by suicide.^20,21 Furthermore, depression may be useful as a clinical proxy for mental health policy impacts, given that the Patient Health Questionnaire-9 (PHQ-9) screening tool is well validated and increasingly research accessible, and it is a chronic condition responsive to both well-managed pharmacologic treatment and psychotherapeutic interventions.^22,23

In this study, we quantify the change in care-seeking behavior for depression among veterans after Medicaid expansion, using a quasi-experimental design. We hypothesize that new access to Medicaid would be associated with a shift away from using VA services for depression. Given the income-dependent eligibility requirements of Medicaid, we also hypothesize that veterans who qualified for VA coverage due to low income, determined by a regional means test (Priority group 5, “income-eligible”), would be more likely to shift care compared with those whose serviced-connected conditions related to their military service (Priority groups 1-4, “service-connected”) provide VA access.

Methods

To investigate the relative changes in veterans’ reliance on the VA for depression care after the 2001 NY and AZ Medicaid expansions We used a retrospective, difference-in-difference analysis. Our comparison pairings, based on prior demographic analyses were as follows: NY with Pennsylvania(PA); AZ with New Mexico and Nevada (NM/NV).¹⁹ The time frame of our analysis was 1999 to 2006, with pre- and postexpansion periods defined as 1999 to 2000 and 2001 to 2006, respectively.

Data

We included veterans aged 18 to 64 years, seeking care for depression from 1999 to 2006, who were also VA-enrolled and residing in our states of interest. We counted veterans as enrolled in Medicaid if they were enrolled at least 1 month in a given year.

Using similar methods like those used in prior studies, we selected patients with encounters documenting depression as the primary outpatient or inpatient diagnosis using International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes: 296.2x for a single episode of major depressive disorder, 296.3x for a recurrent episode of MDD, 300.4 for dysthymia, and 311.0 for depression not otherwise specified.^18,24 We used data from the Medicaid Analytic eXtract files (MAX) for Medicaid data and the VA Corporate Data Warehouse (CDW) for VA data. We chose 1999 as the first study year because it was the earliest year MAX data were available.

Our final sample included 1833 person-years pre-expansion and 7157 postexpansion in our inpatient analysis, as well as 31,767 person-years pre-expansion and 130,382 postexpansion in our outpatient analysis.

Outcomes and Variables

Our primary outcomes were comparative shifts in VA reliance between expansion and nonexpansion states after Medicaid expansion for both inpatient and outpatient depression care. For each year of study, we calculated a veteran’s VA reliance by aggregating the number of days with depression-related encounters at the VA and dividing by the total number of days with a VA or Medicaid depression-related encounters for the year. To provide context to these shifts in VA reliance, we further analyzed the changes in the proportion of annual VA-Medicaid dual users and annual per capita utilization of depression care across the VA and Medicaid. Changes in the proportion would indicate a relative shift in usage between the VA and Medicaid. Annual per capita changes demonstrate changes in the volume of usage. Understanding how proportion and volume interact is critical to understanding likely ramifications for resource management and cost. For example, a relative shift in the proportion of care toward Medicaid might be explained by a substitution effect of increased Medicaid usage and lower VA per capita usage, or an additive (or complementary) effect, with more Medicaid services coming on top of the current VA services.

We conducted subanalyses by income-eligible and service-connected veterans and adjusted our models for age, non-White race, sex, distances to the nearest inpatient and outpatient VA facilities, and VA Relative Risk Score, which is a measure of disease burden and clinical complexity validated specifically for veterans.²⁵

Statistical Analysis

We used fractional logistic regression to model the adjusted effect of Medicaid expansion on VA reliance for depression care. In parallel, we leveraged ordered logit regression and negative binomial regression models to examine the proportion of VA-Medicaid dual users and the per capita utilization of Medicaid and VA depression care, respectively. To estimate the difference-in-difference effects, we used the interaction term of 2 categorical variables—expansion vs nonexpansion states and pre- vs postexpansion status—as the independent variable. We then calculated the average marginal effects with 95% CIs to estimate the differences in outcomes between expansion and nonexpansion states from pre- to postexpansion periods, as well as year-by-year shifts as a robustness check. We conducted these analyses using Stata MP, version 15.

This project was approved by the Baylor College of Medicine Institutional Review Board (IRB # H-40441) and the Michael E. Debakey Veterans Affairs Medical Center Research and Development Committee.

Results

Baseline and postexpansion characteristics

VA Reliance

Overall, we observed postexpansion decreases in VA reliance for depression care

At the state level, reliance on the VA for inpatient depression care in NY decreased by 13.53 pp (95% CI, -22.58 to -4.49) for income-eligible veterans and 16.67 pp (95% CI, -24.53 to -8.80) for service-connected veterans. No relative differences were observed in the outpatient comparisons for both income-eligible (-0.58 pp; 95% CI, -2.13 to 0.98) and service-connected (0.05 pp; 95% CI, -1.00 to 1.10) veterans. In AZ, Medicaid expansion was associated with decreased VA reliance for outpatient depression care among income-eligible veterans (-8.60 pp; 95% CI, -10.60 to -6.61), greater than that for service-connected veterans (-2.89 pp; 95% CI, -4.02 to -1.77). This decrease in VA reliance was significant in the inpatient context only for service-connected veterans (-4.55 pp; 95% CI, -8.14 to -0.97), not income-eligible veterans (-8.38 pp; 95% CI, -17.91 to 1.16).

By applying the aggregate pp changes toward the postexpansion number of visits across both expansion and nonexpansion states, we found that expansion of Medicaid across all our study states would have resulted in 996 fewer hospitalizations and 10,109 fewer outpatient visits for depression at VA in the postexpansion period vs if no states had chosen to expand Medicaid.

Dual Use/Per Capita Utilization

Overall, Medicaid expansion was associated with greater dual use for inpatient depression care—a 0.97-pp (95% CI, 0.46 to 1.48) increase among service-connected veterans and a 0.64-pp (95% CI, 0.35 to 0.94) increase among income-eligible veterans.
At the state level, NY similarly showed increases in dual use among both service-connected (1.48 pp; 95% CI, 0.80 to 2.16) and income-eligible veterans (0.73 pp; 95% CI, 0.39 to 1.07) after Medicaid expansion. However, dual use in AZ increased significantly only among service-connected veterans (0.70 pp; 95% CI, 0.03 to 1.38), not income-eligible veterans (0.31 pp; 95% CI, -0.17 to 0.78).

Among outpatient visits, Medicaid expansion was associated with increased dual use only for income-eligible veterans (0.16 pp; 95% CI, 0.03-0.29), and not service-connected veterans (0.09 pp; 95% CI, -0.04 to 0.21). State-level analyses showed that Medicaid expansion in NY was not associated with changes in dual use for either service-connected (0.01 pp; 95% CI, -0.16 to 0.17) or income-eligible veterans (0.03 pp; 95% CI, -0.12 to 0.18), while expansion in AZ was associated with increases in dual use among both service-connected (0.42 pp; 95% CI, 0.23 to 0.61) and income-eligible veterans (0.83 pp; 95% CI, 0.59 to 1.07).

Concerning per capita utilization of depression care after Medicaid expansion, analyses showed no detectable changes for either inpatient or outpatient services, among both service-connected and income-eligible veterans. However, while this pattern held at the state level among hospitalizations, outpatient visit results showed divergent trends between AZ and NY. In NY, Medicaid expansion was associated with decreased per capita utilization of outpatient depression care among both service-connected (-0.25 visits annually; 95% CI, -0.48 to -0.01) and income-eligible veterans (-0.64 visits annually; 95% CI, -0.93 to -0.35). In AZ, Medicaid expansion was associated with increased per capita utilization of outpatient depression care among both service-connected (0.62 visits annually; 95% CI, 0.32-0.91) and income-eligible veterans (2.32 visits annually; 95% CI, 1.99-2.65).

Discussion

Our study quantified changes in depression-related health care utilization after Medicaid expansions in NY and AZ in 2001. Overall, the balance of evidence indicated that Medicaid expansion was associated with decreased reliance on the VA for depression-related services. There was an exception: income-eligible veterans in AZ did not shift their hospital care away from the VA in a statistically discernible way, although the point estimate was lower. More broadly, these findings concerning veterans’ reliance varied not only in inpatient vs outpatient services and income- vs service-connected eligibility, but also in the state-level contexts of veteran dual users and per capita utilization.

Given that the overall per capita utilization of depression care was unchanged from pre- to postexpansion periods, one might interpret the decreases in VA reliance and increases in Medicaid-VA dual users as a substitution effect from VA care to non-VA care. This could be plausible for hospitalizations where state-level analyses showed similarly stable levels of per capita utilization. However, state-level trends in our outpatient utilization analysis, especially with a substantial 2.32 pp increase in annual per capita visits among income-eligible veterans in AZ, leave open the possibility that in some cases veterans may be complementing VA care with Medicaid-reimbursed services.

The causes underlying these differences in reliance shifts between NY and AZ are likely also influenced by the policy contexts of their respective Medicaid expansions. For example, in 1999, NY passed Kendra’s Law, which established a procedure for obtaining court orders for assisted outpatient mental health treatment for individuals deemed unlikely to survive safely in the community.²⁶ A reasonable inference is that there was less unfulfilled outpatient mental health need in NY under the existing accessibility provisioned by Kendra’s Law. In addition, while both states extended coverage to childless adults under 100% of the Federal Poverty level (FPL), the AZ Medicaid expansion was via a voters’ initiative and extended family coverage to 200% FPL vs 150% FPL for families in NY. Given that the AZ Medicaid expansion enjoyed both broader public participation and generosity in terms of eligibility, its uptake and therefore effect size may have been larger than in NY for nonacute outpatient care.

Our findings contribute to the growing body of literature surrounding the changes in health care utilization after Medicaid expansion, specifically for a newly dual-eligible population of veterans seeking mental health services for depression. While prior research concerning Medicare dual-enrolled veterans has shown high reliance on the VA for both mental health diagnoses and services, scholars have established the association of Medicaid enrollment with decreased VA reliance.^27-29 Our analysis is the first to investigate state-level effects of Medicaid expansion on VA reliance for a single mental health condition using a natural experimental framework. We focus on a population that includes a large portion of veterans who are newly Medicaid-eligible due to a sweeping policy change and use demographically matched nonexpansion states to draw comparisons in VA reliance for depression care. Our findings of Medicaid expansion–associated decreases in VA reliance for depression care complement prior literature that describe Medicaid enrollment–associated decreases in VA reliance for overall mental health care.

Implications

From a systems-level perspective, the implications of shifting services away from the VA are complex and incompletely understood. The VA lacks interoperability with the electronic health records (EHRs) used by Medicaid clinicians. Consequently, significant issues of service duplication and incomplete clinical data exist for veterans seeking treatment outside of the VA system, posing health care quality and safety concerns.³⁰ On one hand, Medicaid access is associated with increased health care utilization attributed to filling unmet needs for Medicare dual enrollees, as well as increased prescription filling for psychiatric medications.^31,32 Furthermore, the only randomized control trial of Medicaid expansion to date was associated with a 9-pp decrease in positive screening rates for depression among those who received access at around 2 years postexpansion.³³ On the other hand, the VA has developed a mental health system tailored to the particular needs of veterans, and health care practitioners at the VA have significantly greater rates of military cultural competency compared to those in nonmilitary settings (70% vs 24% in the TRICARE network and 8% among those with no military or TRICARE affiliation).³⁴ Compared to individuals seeking mental health services with private insurance plans, veterans were about twice as likely to receive appropriate treatment for schizophrenia and depression at the VA.³⁵ These documented strengths of VA mental health care may together help explain the small absolute number of visits that were associated with shifts away from VA overall after Medicaid expansion.

Finally, it is worth considering extrinsic factors that influence utilization among newly dual-eligible veterans. For example, hospitalizations are less likely to be planned than outpatient services, translating to a greater importance of proximity to a nearby medical facility than a veteran’s preference of where to seek care. In the same vein, major VA medical centers are fewer and more distant on average than VA outpatient clinics, therefore reducing the advantage of a Medicaid-reimbursed outpatient clinic in terms of distance.³⁶ These realities may partially explain the proportionally larger shifts away from the VA for hospitalizations compared to outpatient care for depression.

These shifts in utilization after Medicaid expansion may have important implications for VA policymakers. First, more study is needed to know which types of veterans are more likely to use Medicaid instead of VA services—or use both Medicaid and VA services. Our research indicates unsurprisingly that veterans without service-connected disability ratings and eligible for VA services due to low income are more likely to use at least some Medicaid services. Further understanding of who switches will be useful for the VA both tailoring its services to those who prefer VA and for reaching out to specific types of patients who might be better served by staying within the VA system. Finally, VA clinicians and administrators can prioritize improving care coordination for those who chose to use both Medicaid and VA services.

Limitations and Future Directions

Our results should be interpreted within methodological and data limitations. With only 2 states in our sample, NY demonstrably skewed overall results, contributing 1.7 to 3 times more observations than AZ across subanalyses—a challenge also cited by Sommers and colleagues.¹⁹ Our veteran groupings were also unable to distinguish those veterans classified as service-connected who may also have qualified by income-eligible criteria (which would tend to understate the size of results) and those veterans who gained and then lost Medicaid coverage in a given year. Our study also faces limitations in generalizability and establishing causality. First, we included only 2 historical state Medicaid expansions, compared with the 38 states and Washington, DC, that have now expanded Medicaid to date under the ACA. Just in the 2 states from our study, we noted significant heterogeneity in the shifts associated with Medicaid expansion, which makes extrapolating specific trends difficult. Differences in underlying health care resources, legislation, and other external factors may limit the applicability of Medicaid expansion in the era of the ACA, as well as the Veterans Choice and MISSION acts. Second, while we leveraged a difference-in-difference analysis using demographically matched, neighboring comparison states, our findings are nevertheless drawn from observational data obviating causality. VA data for other sources of coverage such as private insurance are limited and not included in our study, and MAX datasets vary by quality across states, translating to potential gaps in our study cohort.²⁸Finally, as in any study using diagnoses, visits addressing care for depression may have been missed if other diagnoses were noted as primary (eg, VA clinicians carrying forward old diagnoses, like PTSD, on the problem list) or nondepression care visits may have been captured if a depression diagnosis was used by default.

Moving forward, our study demonstrates the potential for applying a natural experimental approach to studying dual-eligible veterans at the interface of Medicaid expansion. We focused on changes in VA reliance for the specific condition of depression and, in doing so, invite further inquiry into the impact of state mental health policy on outcomes more proximate to veterans’ outcomes. Clinical indicators, such as rates of antidepressant filling, utilization and duration of psychotherapy, and PHQ-9 scores, can similarly be investigated by natural experimental design. While current limits of administrative data and the siloing of EHRs may pose barriers to some of these avenues of research, multidisciplinary methodologies and data querying innovations such as natural language processing algorithms for clinical notes hold exciting opportunities to bridge the gap between policy and clinical efficacy.

Conclusions

This study applied a difference-in-difference analysis and found that Medicaid expansion is associated with decreases in VA reliance for both inpatient and outpatient services for depression. As additional data are generated from the Medicaid expansions of the ACA, similarly robust methods should be applied to further explore the impacts associated with such policy shifts and open the door to a better understanding of implications at the clinical level.

Acknowledgments

We acknowledge the efforts of Janine Wong, who proofread and formatted the manuscript.

The US Department of Veterans Affairs (VA) is the largest integrated health care system in the United States, providing care for more than 9 million veterans.¹ With veterans experiencing mental health conditions like posttraumatic stress disorder (PTSD), substance use disorders, and other serious mental illnesses (SMI) at higher rates compared with the general population, the VA plays an important role in the provision of mental health services.^2-5 Since the implementation of its Mental Health Strategic Plan in 2004, the VA has overseen the development of a wide array of mental health programs geared toward the complex needs of veterans. Research has demonstrated VA care outperforming Medicaid-reimbursed services in terms of the percentage of veterans filling antidepressants for at least 12 weeks after initiation of treatment for major depressive disorder (MDD), as well as posthospitalization follow-up.⁶

Eligible veterans enrolled in the VA often also seek non-VA care. Medicaid covers nearly 10% of all nonelderly veterans, and of these veterans, 39% rely solely on Medicaid for health care access.⁷ Today, Medicaid is the largest payer for mental health services in the US, providing coverage for approximately 27% of Americans who have SMI and helping fulfill unmet mental health needs.^8,9 Understanding which of these systems veterans choose to use, and under which circumstances, is essential in guiding the allocation of limited health care resources.¹⁰

Beyond Medicaid, alternatives to VA care may include TRICARE, Medicare, Indian Health Services, and employer-based or self-purchased private insurance. While these options potentially increase convenience, choice, and access to health care practitioners (HCPs) and services not available at local VA systems, cross-system utilization with poor integration may cause care coordination and continuity problems, such as medication mismanagement and opioid overdose, unnecessary duplicate utilization, and possible increased mortality.^11-15 As recent national legislative changes, such as the Patient Protection and Affordable Care Act (ACA), Veterans Access, Choice and Accountability Act, and the VA MISSION Act, continue to shift the health care landscape for veterans, questions surrounding how veterans are changing their health care use become significant.^16,17

Here, we approach the impacts of Medicaid expansion on veterans’ reliance on the VA for mental health services with a unique lens. We leverage a difference-in-difference design to study 2 historical Medicaid expansions in Arizona (AZ) and New York (NY), which extended eligibility to childless adults in 2001. Prior Medicaid dual-eligible mental health research investigated reliance shifts during the immediate postenrollment year in a subset of veterans newly enrolled in Medicaid.¹⁸ However, this study took place in a period of relative policy stability. In contrast, we investigate the potential effects of a broad policy shift by analyzing state-level changes in veterans’ reliance over 6 years after a statewide Medicaid expansion. We match expansion states with demographically similar nonexpansion states to account for unobserved trends and confounding effects. Prior studies have used this method to evaluate post-Medicaid expansion mortality changes and changes in veteran dual enrollment and hospitalizations.^10,19 While a study of ACA Medicaid expansion states would be ideal, Medicaid data from most states were only available through 2014 at the time of this analysis. Our study offers a quasi-experimental framework leveraging longitudinal data that can be applied as more post-ACA data become available.

Given the rising incidence of suicide among veterans, understanding care-seeking behaviors for depression among veterans is important as it is the most common psychiatric condition found in those who died by suicide.^20,21 Furthermore, depression may be useful as a clinical proxy for mental health policy impacts, given that the Patient Health Questionnaire-9 (PHQ-9) screening tool is well validated and increasingly research accessible, and it is a chronic condition responsive to both well-managed pharmacologic treatment and psychotherapeutic interventions.^22,23

In this study, we quantify the change in care-seeking behavior for depression among veterans after Medicaid expansion, using a quasi-experimental design. We hypothesize that new access to Medicaid would be associated with a shift away from using VA services for depression. Given the income-dependent eligibility requirements of Medicaid, we also hypothesize that veterans who qualified for VA coverage due to low income, determined by a regional means test (Priority group 5, “income-eligible”), would be more likely to shift care compared with those whose serviced-connected conditions related to their military service (Priority groups 1-4, “service-connected”) provide VA access.

Methods

To investigate the relative changes in veterans’ reliance on the VA for depression care after the 2001 NY and AZ Medicaid expansions We used a retrospective, difference-in-difference analysis. Our comparison pairings, based on prior demographic analyses were as follows: NY with Pennsylvania(PA); AZ with New Mexico and Nevada (NM/NV).¹⁹ The time frame of our analysis was 1999 to 2006, with pre- and postexpansion periods defined as 1999 to 2000 and 2001 to 2006, respectively.

Data

We included veterans aged 18 to 64 years, seeking care for depression from 1999 to 2006, who were also VA-enrolled and residing in our states of interest. We counted veterans as enrolled in Medicaid if they were enrolled at least 1 month in a given year.

Using similar methods like those used in prior studies, we selected patients with encounters documenting depression as the primary outpatient or inpatient diagnosis using International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes: 296.2x for a single episode of major depressive disorder, 296.3x for a recurrent episode of MDD, 300.4 for dysthymia, and 311.0 for depression not otherwise specified.^18,24 We used data from the Medicaid Analytic eXtract files (MAX) for Medicaid data and the VA Corporate Data Warehouse (CDW) for VA data. We chose 1999 as the first study year because it was the earliest year MAX data were available.

Our final sample included 1833 person-years pre-expansion and 7157 postexpansion in our inpatient analysis, as well as 31,767 person-years pre-expansion and 130,382 postexpansion in our outpatient analysis.

Outcomes and Variables

Our primary outcomes were comparative shifts in VA reliance between expansion and nonexpansion states after Medicaid expansion for both inpatient and outpatient depression care. For each year of study, we calculated a veteran’s VA reliance by aggregating the number of days with depression-related encounters at the VA and dividing by the total number of days with a VA or Medicaid depression-related encounters for the year. To provide context to these shifts in VA reliance, we further analyzed the changes in the proportion of annual VA-Medicaid dual users and annual per capita utilization of depression care across the VA and Medicaid. Changes in the proportion would indicate a relative shift in usage between the VA and Medicaid. Annual per capita changes demonstrate changes in the volume of usage. Understanding how proportion and volume interact is critical to understanding likely ramifications for resource management and cost. For example, a relative shift in the proportion of care toward Medicaid might be explained by a substitution effect of increased Medicaid usage and lower VA per capita usage, or an additive (or complementary) effect, with more Medicaid services coming on top of the current VA services.

We conducted subanalyses by income-eligible and service-connected veterans and adjusted our models for age, non-White race, sex, distances to the nearest inpatient and outpatient VA facilities, and VA Relative Risk Score, which is a measure of disease burden and clinical complexity validated specifically for veterans.²⁵

Statistical Analysis

We used fractional logistic regression to model the adjusted effect of Medicaid expansion on VA reliance for depression care. In parallel, we leveraged ordered logit regression and negative binomial regression models to examine the proportion of VA-Medicaid dual users and the per capita utilization of Medicaid and VA depression care, respectively. To estimate the difference-in-difference effects, we used the interaction term of 2 categorical variables—expansion vs nonexpansion states and pre- vs postexpansion status—as the independent variable. We then calculated the average marginal effects with 95% CIs to estimate the differences in outcomes between expansion and nonexpansion states from pre- to postexpansion periods, as well as year-by-year shifts as a robustness check. We conducted these analyses using Stata MP, version 15.

This project was approved by the Baylor College of Medicine Institutional Review Board (IRB # H-40441) and the Michael E. Debakey Veterans Affairs Medical Center Research and Development Committee.

Results

Baseline and postexpansion characteristics

VA Reliance

Overall, we observed postexpansion decreases in VA reliance for depression care

At the state level, reliance on the VA for inpatient depression care in NY decreased by 13.53 pp (95% CI, -22.58 to -4.49) for income-eligible veterans and 16.67 pp (95% CI, -24.53 to -8.80) for service-connected veterans. No relative differences were observed in the outpatient comparisons for both income-eligible (-0.58 pp; 95% CI, -2.13 to 0.98) and service-connected (0.05 pp; 95% CI, -1.00 to 1.10) veterans. In AZ, Medicaid expansion was associated with decreased VA reliance for outpatient depression care among income-eligible veterans (-8.60 pp; 95% CI, -10.60 to -6.61), greater than that for service-connected veterans (-2.89 pp; 95% CI, -4.02 to -1.77). This decrease in VA reliance was significant in the inpatient context only for service-connected veterans (-4.55 pp; 95% CI, -8.14 to -0.97), not income-eligible veterans (-8.38 pp; 95% CI, -17.91 to 1.16).

By applying the aggregate pp changes toward the postexpansion number of visits across both expansion and nonexpansion states, we found that expansion of Medicaid across all our study states would have resulted in 996 fewer hospitalizations and 10,109 fewer outpatient visits for depression at VA in the postexpansion period vs if no states had chosen to expand Medicaid.

Dual Use/Per Capita Utilization

Overall, Medicaid expansion was associated with greater dual use for inpatient depression care—a 0.97-pp (95% CI, 0.46 to 1.48) increase among service-connected veterans and a 0.64-pp (95% CI, 0.35 to 0.94) increase among income-eligible veterans.
At the state level, NY similarly showed increases in dual use among both service-connected (1.48 pp; 95% CI, 0.80 to 2.16) and income-eligible veterans (0.73 pp; 95% CI, 0.39 to 1.07) after Medicaid expansion. However, dual use in AZ increased significantly only among service-connected veterans (0.70 pp; 95% CI, 0.03 to 1.38), not income-eligible veterans (0.31 pp; 95% CI, -0.17 to 0.78).

Among outpatient visits, Medicaid expansion was associated with increased dual use only for income-eligible veterans (0.16 pp; 95% CI, 0.03-0.29), and not service-connected veterans (0.09 pp; 95% CI, -0.04 to 0.21). State-level analyses showed that Medicaid expansion in NY was not associated with changes in dual use for either service-connected (0.01 pp; 95% CI, -0.16 to 0.17) or income-eligible veterans (0.03 pp; 95% CI, -0.12 to 0.18), while expansion in AZ was associated with increases in dual use among both service-connected (0.42 pp; 95% CI, 0.23 to 0.61) and income-eligible veterans (0.83 pp; 95% CI, 0.59 to 1.07).

Concerning per capita utilization of depression care after Medicaid expansion, analyses showed no detectable changes for either inpatient or outpatient services, among both service-connected and income-eligible veterans. However, while this pattern held at the state level among hospitalizations, outpatient visit results showed divergent trends between AZ and NY. In NY, Medicaid expansion was associated with decreased per capita utilization of outpatient depression care among both service-connected (-0.25 visits annually; 95% CI, -0.48 to -0.01) and income-eligible veterans (-0.64 visits annually; 95% CI, -0.93 to -0.35). In AZ, Medicaid expansion was associated with increased per capita utilization of outpatient depression care among both service-connected (0.62 visits annually; 95% CI, 0.32-0.91) and income-eligible veterans (2.32 visits annually; 95% CI, 1.99-2.65).

Discussion

Our study quantified changes in depression-related health care utilization after Medicaid expansions in NY and AZ in 2001. Overall, the balance of evidence indicated that Medicaid expansion was associated with decreased reliance on the VA for depression-related services. There was an exception: income-eligible veterans in AZ did not shift their hospital care away from the VA in a statistically discernible way, although the point estimate was lower. More broadly, these findings concerning veterans’ reliance varied not only in inpatient vs outpatient services and income- vs service-connected eligibility, but also in the state-level contexts of veteran dual users and per capita utilization.

Given that the overall per capita utilization of depression care was unchanged from pre- to postexpansion periods, one might interpret the decreases in VA reliance and increases in Medicaid-VA dual users as a substitution effect from VA care to non-VA care. This could be plausible for hospitalizations where state-level analyses showed similarly stable levels of per capita utilization. However, state-level trends in our outpatient utilization analysis, especially with a substantial 2.32 pp increase in annual per capita visits among income-eligible veterans in AZ, leave open the possibility that in some cases veterans may be complementing VA care with Medicaid-reimbursed services.

The causes underlying these differences in reliance shifts between NY and AZ are likely also influenced by the policy contexts of their respective Medicaid expansions. For example, in 1999, NY passed Kendra’s Law, which established a procedure for obtaining court orders for assisted outpatient mental health treatment for individuals deemed unlikely to survive safely in the community.²⁶ A reasonable inference is that there was less unfulfilled outpatient mental health need in NY under the existing accessibility provisioned by Kendra’s Law. In addition, while both states extended coverage to childless adults under 100% of the Federal Poverty level (FPL), the AZ Medicaid expansion was via a voters’ initiative and extended family coverage to 200% FPL vs 150% FPL for families in NY. Given that the AZ Medicaid expansion enjoyed both broader public participation and generosity in terms of eligibility, its uptake and therefore effect size may have been larger than in NY for nonacute outpatient care.

Our findings contribute to the growing body of literature surrounding the changes in health care utilization after Medicaid expansion, specifically for a newly dual-eligible population of veterans seeking mental health services for depression. While prior research concerning Medicare dual-enrolled veterans has shown high reliance on the VA for both mental health diagnoses and services, scholars have established the association of Medicaid enrollment with decreased VA reliance.^27-29 Our analysis is the first to investigate state-level effects of Medicaid expansion on VA reliance for a single mental health condition using a natural experimental framework. We focus on a population that includes a large portion of veterans who are newly Medicaid-eligible due to a sweeping policy change and use demographically matched nonexpansion states to draw comparisons in VA reliance for depression care. Our findings of Medicaid expansion–associated decreases in VA reliance for depression care complement prior literature that describe Medicaid enrollment–associated decreases in VA reliance for overall mental health care.

Implications

From a systems-level perspective, the implications of shifting services away from the VA are complex and incompletely understood. The VA lacks interoperability with the electronic health records (EHRs) used by Medicaid clinicians. Consequently, significant issues of service duplication and incomplete clinical data exist for veterans seeking treatment outside of the VA system, posing health care quality and safety concerns.³⁰ On one hand, Medicaid access is associated with increased health care utilization attributed to filling unmet needs for Medicare dual enrollees, as well as increased prescription filling for psychiatric medications.^31,32 Furthermore, the only randomized control trial of Medicaid expansion to date was associated with a 9-pp decrease in positive screening rates for depression among those who received access at around 2 years postexpansion.³³ On the other hand, the VA has developed a mental health system tailored to the particular needs of veterans, and health care practitioners at the VA have significantly greater rates of military cultural competency compared to those in nonmilitary settings (70% vs 24% in the TRICARE network and 8% among those with no military or TRICARE affiliation).³⁴ Compared to individuals seeking mental health services with private insurance plans, veterans were about twice as likely to receive appropriate treatment for schizophrenia and depression at the VA.³⁵ These documented strengths of VA mental health care may together help explain the small absolute number of visits that were associated with shifts away from VA overall after Medicaid expansion.

Finally, it is worth considering extrinsic factors that influence utilization among newly dual-eligible veterans. For example, hospitalizations are less likely to be planned than outpatient services, translating to a greater importance of proximity to a nearby medical facility than a veteran’s preference of where to seek care. In the same vein, major VA medical centers are fewer and more distant on average than VA outpatient clinics, therefore reducing the advantage of a Medicaid-reimbursed outpatient clinic in terms of distance.³⁶ These realities may partially explain the proportionally larger shifts away from the VA for hospitalizations compared to outpatient care for depression.

These shifts in utilization after Medicaid expansion may have important implications for VA policymakers. First, more study is needed to know which types of veterans are more likely to use Medicaid instead of VA services—or use both Medicaid and VA services. Our research indicates unsurprisingly that veterans without service-connected disability ratings and eligible for VA services due to low income are more likely to use at least some Medicaid services. Further understanding of who switches will be useful for the VA both tailoring its services to those who prefer VA and for reaching out to specific types of patients who might be better served by staying within the VA system. Finally, VA clinicians and administrators can prioritize improving care coordination for those who chose to use both Medicaid and VA services.

Limitations and Future Directions

Our results should be interpreted within methodological and data limitations. With only 2 states in our sample, NY demonstrably skewed overall results, contributing 1.7 to 3 times more observations than AZ across subanalyses—a challenge also cited by Sommers and colleagues.¹⁹ Our veteran groupings were also unable to distinguish those veterans classified as service-connected who may also have qualified by income-eligible criteria (which would tend to understate the size of results) and those veterans who gained and then lost Medicaid coverage in a given year. Our study also faces limitations in generalizability and establishing causality. First, we included only 2 historical state Medicaid expansions, compared with the 38 states and Washington, DC, that have now expanded Medicaid to date under the ACA. Just in the 2 states from our study, we noted significant heterogeneity in the shifts associated with Medicaid expansion, which makes extrapolating specific trends difficult. Differences in underlying health care resources, legislation, and other external factors may limit the applicability of Medicaid expansion in the era of the ACA, as well as the Veterans Choice and MISSION acts. Second, while we leveraged a difference-in-difference analysis using demographically matched, neighboring comparison states, our findings are nevertheless drawn from observational data obviating causality. VA data for other sources of coverage such as private insurance are limited and not included in our study, and MAX datasets vary by quality across states, translating to potential gaps in our study cohort.²⁸Finally, as in any study using diagnoses, visits addressing care for depression may have been missed if other diagnoses were noted as primary (eg, VA clinicians carrying forward old diagnoses, like PTSD, on the problem list) or nondepression care visits may have been captured if a depression diagnosis was used by default.

Moving forward, our study demonstrates the potential for applying a natural experimental approach to studying dual-eligible veterans at the interface of Medicaid expansion. We focused on changes in VA reliance for the specific condition of depression and, in doing so, invite further inquiry into the impact of state mental health policy on outcomes more proximate to veterans’ outcomes. Clinical indicators, such as rates of antidepressant filling, utilization and duration of psychotherapy, and PHQ-9 scores, can similarly be investigated by natural experimental design. While current limits of administrative data and the siloing of EHRs may pose barriers to some of these avenues of research, multidisciplinary methodologies and data querying innovations such as natural language processing algorithms for clinical notes hold exciting opportunities to bridge the gap between policy and clinical efficacy.

Conclusions

This study applied a difference-in-difference analysis and found that Medicaid expansion is associated with decreases in VA reliance for both inpatient and outpatient services for depression. As additional data are generated from the Medicaid expansions of the ACA, similarly robust methods should be applied to further explore the impacts associated with such policy shifts and open the door to a better understanding of implications at the clinical level.

Acknowledgments

We acknowledge the efforts of Janine Wong, who proofread and formatted the manuscript.

The US Department of Veterans Affairs (VA) is the largest integrated health care system in the United States, providing care for more than 9 million veterans.¹ With veterans experiencing mental health conditions like posttraumatic stress disorder (PTSD), substance use disorders, and other serious mental illnesses (SMI) at higher rates compared with the general population, the VA plays an important role in the provision of mental health services.^2-5 Since the implementation of its Mental Health Strategic Plan in 2004, the VA has overseen the development of a wide array of mental health programs geared toward the complex needs of veterans. Research has demonstrated VA care outperforming Medicaid-reimbursed services in terms of the percentage of veterans filling antidepressants for at least 12 weeks after initiation of treatment for major depressive disorder (MDD), as well as posthospitalization follow-up.⁶

Eligible veterans enrolled in the VA often also seek non-VA care. Medicaid covers nearly 10% of all nonelderly veterans, and of these veterans, 39% rely solely on Medicaid for health care access.⁷ Today, Medicaid is the largest payer for mental health services in the US, providing coverage for approximately 27% of Americans who have SMI and helping fulfill unmet mental health needs.^8,9 Understanding which of these systems veterans choose to use, and under which circumstances, is essential in guiding the allocation of limited health care resources.¹⁰

Beyond Medicaid, alternatives to VA care may include TRICARE, Medicare, Indian Health Services, and employer-based or self-purchased private insurance. While these options potentially increase convenience, choice, and access to health care practitioners (HCPs) and services not available at local VA systems, cross-system utilization with poor integration may cause care coordination and continuity problems, such as medication mismanagement and opioid overdose, unnecessary duplicate utilization, and possible increased mortality.^11-15 As recent national legislative changes, such as the Patient Protection and Affordable Care Act (ACA), Veterans Access, Choice and Accountability Act, and the VA MISSION Act, continue to shift the health care landscape for veterans, questions surrounding how veterans are changing their health care use become significant.^16,17

Here, we approach the impacts of Medicaid expansion on veterans’ reliance on the VA for mental health services with a unique lens. We leverage a difference-in-difference design to study 2 historical Medicaid expansions in Arizona (AZ) and New York (NY), which extended eligibility to childless adults in 2001. Prior Medicaid dual-eligible mental health research investigated reliance shifts during the immediate postenrollment year in a subset of veterans newly enrolled in Medicaid.¹⁸ However, this study took place in a period of relative policy stability. In contrast, we investigate the potential effects of a broad policy shift by analyzing state-level changes in veterans’ reliance over 6 years after a statewide Medicaid expansion. We match expansion states with demographically similar nonexpansion states to account for unobserved trends and confounding effects. Prior studies have used this method to evaluate post-Medicaid expansion mortality changes and changes in veteran dual enrollment and hospitalizations.^10,19 While a study of ACA Medicaid expansion states would be ideal, Medicaid data from most states were only available through 2014 at the time of this analysis. Our study offers a quasi-experimental framework leveraging longitudinal data that can be applied as more post-ACA data become available.

Given the rising incidence of suicide among veterans, understanding care-seeking behaviors for depression among veterans is important as it is the most common psychiatric condition found in those who died by suicide.^20,21 Furthermore, depression may be useful as a clinical proxy for mental health policy impacts, given that the Patient Health Questionnaire-9 (PHQ-9) screening tool is well validated and increasingly research accessible, and it is a chronic condition responsive to both well-managed pharmacologic treatment and psychotherapeutic interventions.^22,23

In this study, we quantify the change in care-seeking behavior for depression among veterans after Medicaid expansion, using a quasi-experimental design. We hypothesize that new access to Medicaid would be associated with a shift away from using VA services for depression. Given the income-dependent eligibility requirements of Medicaid, we also hypothesize that veterans who qualified for VA coverage due to low income, determined by a regional means test (Priority group 5, “income-eligible”), would be more likely to shift care compared with those whose serviced-connected conditions related to their military service (Priority groups 1-4, “service-connected”) provide VA access.

Methods

To investigate the relative changes in veterans’ reliance on the VA for depression care after the 2001 NY and AZ Medicaid expansions We used a retrospective, difference-in-difference analysis. Our comparison pairings, based on prior demographic analyses were as follows: NY with Pennsylvania(PA); AZ with New Mexico and Nevada (NM/NV).¹⁹ The time frame of our analysis was 1999 to 2006, with pre- and postexpansion periods defined as 1999 to 2000 and 2001 to 2006, respectively.

Data

We included veterans aged 18 to 64 years, seeking care for depression from 1999 to 2006, who were also VA-enrolled and residing in our states of interest. We counted veterans as enrolled in Medicaid if they were enrolled at least 1 month in a given year.

Using similar methods like those used in prior studies, we selected patients with encounters documenting depression as the primary outpatient or inpatient diagnosis using International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes: 296.2x for a single episode of major depressive disorder, 296.3x for a recurrent episode of MDD, 300.4 for dysthymia, and 311.0 for depression not otherwise specified.^18,24 We used data from the Medicaid Analytic eXtract files (MAX) for Medicaid data and the VA Corporate Data Warehouse (CDW) for VA data. We chose 1999 as the first study year because it was the earliest year MAX data were available.

Our final sample included 1833 person-years pre-expansion and 7157 postexpansion in our inpatient analysis, as well as 31,767 person-years pre-expansion and 130,382 postexpansion in our outpatient analysis.

Outcomes and Variables

Our primary outcomes were comparative shifts in VA reliance between expansion and nonexpansion states after Medicaid expansion for both inpatient and outpatient depression care. For each year of study, we calculated a veteran’s VA reliance by aggregating the number of days with depression-related encounters at the VA and dividing by the total number of days with a VA or Medicaid depression-related encounters for the year. To provide context to these shifts in VA reliance, we further analyzed the changes in the proportion of annual VA-Medicaid dual users and annual per capita utilization of depression care across the VA and Medicaid. Changes in the proportion would indicate a relative shift in usage between the VA and Medicaid. Annual per capita changes demonstrate changes in the volume of usage. Understanding how proportion and volume interact is critical to understanding likely ramifications for resource management and cost. For example, a relative shift in the proportion of care toward Medicaid might be explained by a substitution effect of increased Medicaid usage and lower VA per capita usage, or an additive (or complementary) effect, with more Medicaid services coming on top of the current VA services.

We conducted subanalyses by income-eligible and service-connected veterans and adjusted our models for age, non-White race, sex, distances to the nearest inpatient and outpatient VA facilities, and VA Relative Risk Score, which is a measure of disease burden and clinical complexity validated specifically for veterans.²⁵

Statistical Analysis

We used fractional logistic regression to model the adjusted effect of Medicaid expansion on VA reliance for depression care. In parallel, we leveraged ordered logit regression and negative binomial regression models to examine the proportion of VA-Medicaid dual users and the per capita utilization of Medicaid and VA depression care, respectively. To estimate the difference-in-difference effects, we used the interaction term of 2 categorical variables—expansion vs nonexpansion states and pre- vs postexpansion status—as the independent variable. We then calculated the average marginal effects with 95% CIs to estimate the differences in outcomes between expansion and nonexpansion states from pre- to postexpansion periods, as well as year-by-year shifts as a robustness check. We conducted these analyses using Stata MP, version 15.

This project was approved by the Baylor College of Medicine Institutional Review Board (IRB # H-40441) and the Michael E. Debakey Veterans Affairs Medical Center Research and Development Committee.

Results

Baseline and postexpansion characteristics

VA Reliance

Overall, we observed postexpansion decreases in VA reliance for depression care

At the state level, reliance on the VA for inpatient depression care in NY decreased by 13.53 pp (95% CI, -22.58 to -4.49) for income-eligible veterans and 16.67 pp (95% CI, -24.53 to -8.80) for service-connected veterans. No relative differences were observed in the outpatient comparisons for both income-eligible (-0.58 pp; 95% CI, -2.13 to 0.98) and service-connected (0.05 pp; 95% CI, -1.00 to 1.10) veterans. In AZ, Medicaid expansion was associated with decreased VA reliance for outpatient depression care among income-eligible veterans (-8.60 pp; 95% CI, -10.60 to -6.61), greater than that for service-connected veterans (-2.89 pp; 95% CI, -4.02 to -1.77). This decrease in VA reliance was significant in the inpatient context only for service-connected veterans (-4.55 pp; 95% CI, -8.14 to -0.97), not income-eligible veterans (-8.38 pp; 95% CI, -17.91 to 1.16).

By applying the aggregate pp changes toward the postexpansion number of visits across both expansion and nonexpansion states, we found that expansion of Medicaid across all our study states would have resulted in 996 fewer hospitalizations and 10,109 fewer outpatient visits for depression at VA in the postexpansion period vs if no states had chosen to expand Medicaid.

Dual Use/Per Capita Utilization

Overall, Medicaid expansion was associated with greater dual use for inpatient depression care—a 0.97-pp (95% CI, 0.46 to 1.48) increase among service-connected veterans and a 0.64-pp (95% CI, 0.35 to 0.94) increase among income-eligible veterans.
At the state level, NY similarly showed increases in dual use among both service-connected (1.48 pp; 95% CI, 0.80 to 2.16) and income-eligible veterans (0.73 pp; 95% CI, 0.39 to 1.07) after Medicaid expansion. However, dual use in AZ increased significantly only among service-connected veterans (0.70 pp; 95% CI, 0.03 to 1.38), not income-eligible veterans (0.31 pp; 95% CI, -0.17 to 0.78).

Among outpatient visits, Medicaid expansion was associated with increased dual use only for income-eligible veterans (0.16 pp; 95% CI, 0.03-0.29), and not service-connected veterans (0.09 pp; 95% CI, -0.04 to 0.21). State-level analyses showed that Medicaid expansion in NY was not associated with changes in dual use for either service-connected (0.01 pp; 95% CI, -0.16 to 0.17) or income-eligible veterans (0.03 pp; 95% CI, -0.12 to 0.18), while expansion in AZ was associated with increases in dual use among both service-connected (0.42 pp; 95% CI, 0.23 to 0.61) and income-eligible veterans (0.83 pp; 95% CI, 0.59 to 1.07).

Concerning per capita utilization of depression care after Medicaid expansion, analyses showed no detectable changes for either inpatient or outpatient services, among both service-connected and income-eligible veterans. However, while this pattern held at the state level among hospitalizations, outpatient visit results showed divergent trends between AZ and NY. In NY, Medicaid expansion was associated with decreased per capita utilization of outpatient depression care among both service-connected (-0.25 visits annually; 95% CI, -0.48 to -0.01) and income-eligible veterans (-0.64 visits annually; 95% CI, -0.93 to -0.35). In AZ, Medicaid expansion was associated with increased per capita utilization of outpatient depression care among both service-connected (0.62 visits annually; 95% CI, 0.32-0.91) and income-eligible veterans (2.32 visits annually; 95% CI, 1.99-2.65).

Discussion

Our study quantified changes in depression-related health care utilization after Medicaid expansions in NY and AZ in 2001. Overall, the balance of evidence indicated that Medicaid expansion was associated with decreased reliance on the VA for depression-related services. There was an exception: income-eligible veterans in AZ did not shift their hospital care away from the VA in a statistically discernible way, although the point estimate was lower. More broadly, these findings concerning veterans’ reliance varied not only in inpatient vs outpatient services and income- vs service-connected eligibility, but also in the state-level contexts of veteran dual users and per capita utilization.

Given that the overall per capita utilization of depression care was unchanged from pre- to postexpansion periods, one might interpret the decreases in VA reliance and increases in Medicaid-VA dual users as a substitution effect from VA care to non-VA care. This could be plausible for hospitalizations where state-level analyses showed similarly stable levels of per capita utilization. However, state-level trends in our outpatient utilization analysis, especially with a substantial 2.32 pp increase in annual per capita visits among income-eligible veterans in AZ, leave open the possibility that in some cases veterans may be complementing VA care with Medicaid-reimbursed services.

The causes underlying these differences in reliance shifts between NY and AZ are likely also influenced by the policy contexts of their respective Medicaid expansions. For example, in 1999, NY passed Kendra’s Law, which established a procedure for obtaining court orders for assisted outpatient mental health treatment for individuals deemed unlikely to survive safely in the community.²⁶ A reasonable inference is that there was less unfulfilled outpatient mental health need in NY under the existing accessibility provisioned by Kendra’s Law. In addition, while both states extended coverage to childless adults under 100% of the Federal Poverty level (FPL), the AZ Medicaid expansion was via a voters’ initiative and extended family coverage to 200% FPL vs 150% FPL for families in NY. Given that the AZ Medicaid expansion enjoyed both broader public participation and generosity in terms of eligibility, its uptake and therefore effect size may have been larger than in NY for nonacute outpatient care.

Our findings contribute to the growing body of literature surrounding the changes in health care utilization after Medicaid expansion, specifically for a newly dual-eligible population of veterans seeking mental health services for depression. While prior research concerning Medicare dual-enrolled veterans has shown high reliance on the VA for both mental health diagnoses and services, scholars have established the association of Medicaid enrollment with decreased VA reliance.^27-29 Our analysis is the first to investigate state-level effects of Medicaid expansion on VA reliance for a single mental health condition using a natural experimental framework. We focus on a population that includes a large portion of veterans who are newly Medicaid-eligible due to a sweeping policy change and use demographically matched nonexpansion states to draw comparisons in VA reliance for depression care. Our findings of Medicaid expansion–associated decreases in VA reliance for depression care complement prior literature that describe Medicaid enrollment–associated decreases in VA reliance for overall mental health care.

Implications

From a systems-level perspective, the implications of shifting services away from the VA are complex and incompletely understood. The VA lacks interoperability with the electronic health records (EHRs) used by Medicaid clinicians. Consequently, significant issues of service duplication and incomplete clinical data exist for veterans seeking treatment outside of the VA system, posing health care quality and safety concerns.³⁰ On one hand, Medicaid access is associated with increased health care utilization attributed to filling unmet needs for Medicare dual enrollees, as well as increased prescription filling for psychiatric medications.^31,32 Furthermore, the only randomized control trial of Medicaid expansion to date was associated with a 9-pp decrease in positive screening rates for depression among those who received access at around 2 years postexpansion.³³ On the other hand, the VA has developed a mental health system tailored to the particular needs of veterans, and health care practitioners at the VA have significantly greater rates of military cultural competency compared to those in nonmilitary settings (70% vs 24% in the TRICARE network and 8% among those with no military or TRICARE affiliation).³⁴ Compared to individuals seeking mental health services with private insurance plans, veterans were about twice as likely to receive appropriate treatment for schizophrenia and depression at the VA.³⁵ These documented strengths of VA mental health care may together help explain the small absolute number of visits that were associated with shifts away from VA overall after Medicaid expansion.

Finally, it is worth considering extrinsic factors that influence utilization among newly dual-eligible veterans. For example, hospitalizations are less likely to be planned than outpatient services, translating to a greater importance of proximity to a nearby medical facility than a veteran’s preference of where to seek care. In the same vein, major VA medical centers are fewer and more distant on average than VA outpatient clinics, therefore reducing the advantage of a Medicaid-reimbursed outpatient clinic in terms of distance.³⁶ These realities may partially explain the proportionally larger shifts away from the VA for hospitalizations compared to outpatient care for depression.

These shifts in utilization after Medicaid expansion may have important implications for VA policymakers. First, more study is needed to know which types of veterans are more likely to use Medicaid instead of VA services—or use both Medicaid and VA services. Our research indicates unsurprisingly that veterans without service-connected disability ratings and eligible for VA services due to low income are more likely to use at least some Medicaid services. Further understanding of who switches will be useful for the VA both tailoring its services to those who prefer VA and for reaching out to specific types of patients who might be better served by staying within the VA system. Finally, VA clinicians and administrators can prioritize improving care coordination for those who chose to use both Medicaid and VA services.

Limitations and Future Directions

Our results should be interpreted within methodological and data limitations. With only 2 states in our sample, NY demonstrably skewed overall results, contributing 1.7 to 3 times more observations than AZ across subanalyses—a challenge also cited by Sommers and colleagues.¹⁹ Our veteran groupings were also unable to distinguish those veterans classified as service-connected who may also have qualified by income-eligible criteria (which would tend to understate the size of results) and those veterans who gained and then lost Medicaid coverage in a given year. Our study also faces limitations in generalizability and establishing causality. First, we included only 2 historical state Medicaid expansions, compared with the 38 states and Washington, DC, that have now expanded Medicaid to date under the ACA. Just in the 2 states from our study, we noted significant heterogeneity in the shifts associated with Medicaid expansion, which makes extrapolating specific trends difficult. Differences in underlying health care resources, legislation, and other external factors may limit the applicability of Medicaid expansion in the era of the ACA, as well as the Veterans Choice and MISSION acts. Second, while we leveraged a difference-in-difference analysis using demographically matched, neighboring comparison states, our findings are nevertheless drawn from observational data obviating causality. VA data for other sources of coverage such as private insurance are limited and not included in our study, and MAX datasets vary by quality across states, translating to potential gaps in our study cohort.²⁸Finally, as in any study using diagnoses, visits addressing care for depression may have been missed if other diagnoses were noted as primary (eg, VA clinicians carrying forward old diagnoses, like PTSD, on the problem list) or nondepression care visits may have been captured if a depression diagnosis was used by default.

Moving forward, our study demonstrates the potential for applying a natural experimental approach to studying dual-eligible veterans at the interface of Medicaid expansion. We focused on changes in VA reliance for the specific condition of depression and, in doing so, invite further inquiry into the impact of state mental health policy on outcomes more proximate to veterans’ outcomes. Clinical indicators, such as rates of antidepressant filling, utilization and duration of psychotherapy, and PHQ-9 scores, can similarly be investigated by natural experimental design. While current limits of administrative data and the siloing of EHRs may pose barriers to some of these avenues of research, multidisciplinary methodologies and data querying innovations such as natural language processing algorithms for clinical notes hold exciting opportunities to bridge the gap between policy and clinical efficacy.

Conclusions

This study applied a difference-in-difference analysis and found that Medicaid expansion is associated with decreases in VA reliance for both inpatient and outpatient services for depression. As additional data are generated from the Medicaid expansions of the ACA, similarly robust methods should be applied to further explore the impacts associated with such policy shifts and open the door to a better understanding of implications at the clinical level.

Acknowledgments

We acknowledge the efforts of Janine Wong, who proofread and formatted the manuscript.

About 700,000 US military personnel were deployed in Operation Desert Storm (August 1990 to March 1991).¹ Almost 30 years since the war, a large number of these veterans continue to experience a complex of symptoms of unknown etiology called Gulf War illness (GWI), which significantly affects health and quality of life (QOL). The lack of clear etiology of the illness has impaired research to find specific treatments and has further exacerbated the stress among veterans. GWI typically includes a mixture of chronic headache, cognitive difficulties, widespread pain, unexplained fatigue, memory and concentration problems, as well as chronic respiratory and gastrointestinal (GI) symptoms.² Abdominal pain and alteration of bowel habits are also symptoms typical of irritable bowel syndrome (IBS). It has been estimated that IBS occurs in up to 30% of Gulf War veterans.³

The etiology of IBS is unknown. Possible mechanisms include visceral hypersensitivity, altered gut motor function, aberrant brain-gut interaction, and psychological factors, perhaps with a genetic predisposition.⁴ Gastroenteritis has been reported as a triggering mechanism in up to one-third of patients with IBS.⁵ Gastroenteritis can alter the gut microbiota and has been reported to be a significant risk factor for the development of IBS.⁶ In one study of Operation Desert Shield soldiers, > 50% of military personnel developed acute gastroenteritis while on duty.⁷ A high prevalence of extra-intestinal symptoms also has been reported, including fatigue, headache, joint pains, and anxiety, in Gulf War veterans with IBS. These extra-intestinal symptoms of IBS are consistent with the reported GWI symptoms. Change in gut microbiota also has been associated with many of the extra-intestinal symptoms of IBS, especially fatigue.^8,9 Gut microbiota are known to change with travel, stress, and a change in diet, all potential factors that are relevant to Gulf War veterans. This would suggest that an imbalance in the gut microbiota, ie, dysbiosis, may play a role in the pathogenesis of both IBS and GWI. Dysbiosis could be a risk factor for or alternatively a consequence of GWI.

A systematic review highlighted the heterogeneity of the gut microbiota in patients with IBS.¹⁰ Overall, Enterobacteriaceae, Lactobacillaceae, and Bacteroides were increased, whereas Clostridiales, Faecalibacterium, and Bifidobacterium were decreased in patients with IBS compared with controls. Gut microbiota also has been associated with cognitive changes, anxiety, and depression—symptoms associated with IBS and are part of the GWI.

If altered gut microbiota contributes to the etiopathogenesis of IBS, its restoration of with probiotics should help. Probiotics are live organisms that when ingested may improve health by promoting the growth of naturally occurring flora and establishing a healthy gut flora. Probiotics have several mechanisms of actions. Probiotics work in the lumen of the gut by producing antibacterial molecules and enhancing the mucosal barrier.¹¹ Probiotics also may produce metabolic compounds that alter the intestinal microbiota and improve intestinal barrier function.¹² Probiotics also have been shown to activate receptors in the enteric nervous system with the potential to promote pain relief in the setting of visceral hyperalgesia.^13,14 The anti-inflammatory properties of probiotics potentially could modulate the basic pathophysiology of IBS and improve motility, visceral hypersensitivity, and brain-gut interaction.¹⁵ Furthermore, significant gut dysbiosis has been shown with GWI; suggesting that probiotics may have a role in its management.^16,17

Probiotics have not been studied in Gulf War veterans with IBS. We performed a prospective, double-blind placebo-controlled study to determine the efficacy of a commercially available probiotic containing 8 strains of bacteria (De Simone Formulation; formally known as VSL#3 and Visbiome) on symptoms of IBS and GWI. This probiotic was selected as the overall literature suggested benefit of combination probiotics in IBS, and VSL#3 has been shown to be efficacious in ulcerative colitis and microscopic colitis.^18-20

Methods

Veterans who served in Operation Desert Storm (August 1990 to March 1991) and enrolled at the George E. Wahlen Veterans Affairs (VA) Medical Center (GEWVAMC), Salt Lake City, Utah, were eligible for the study. The inclusion criteria were: veterans aged ≥ 35 years; ≥ 2 nonintestinal GWI symptoms (eg, fatigue, joint pains, insomnia, general stiffness, and headache); IBS diagnosis based on the Rome III criteria; IBS symptoms > 6 months; normal gross appearance of the colonic mucosa; negative markers for celiac disease and inflammatory bowel disease (IBD); normal thyroid function; and serum calcium levels.²¹ Those who had a clinically significant cardiac, pulmonary, hepatic or renal dysfunction; history of/or presence of systemic malignancy; current evidence of celiac disease or IBD; unstable/significant psychiatric disease; recent change in GI medications; current pregnancy; or use of antibiotics or probiotics within the past 1 month were excluded. Subjects were enrolled from a list of veterans with GWI from the GEWVAMC Gulf War registry; referrals to gastroenterology clinics for IBS from internal medicine clinics; and posted advertisements.

Protocol

After written informed consent was obtained, each veteran was verified to have IBS and ≥ 2 GWI symptoms. All veterans had the following tests and panels: complete blood count, erythrocyte sedimentation rate, serum comprehensive metabolic panel, thyroid-stimulating hormone, tissue transglutaminase, stool test for ova and parasite, giardia antigen, and clostridia toxins to exclude organic cause of GI symptoms. Colonoscopy was performed in all veterans to exclude IBD, and to rule out microscopic or lymphocytic colitis.

Randomization was computer generated and maintained by the study pharmacist so that study personnel and patients were blinded to the trial groups. All investigators were blinded and allocation was concealed. The medication was supplied in a numbered container by the pharmacist after patient enrollment. After a 2-week run-in period, veterans were randomized (1:1) to receive either 1 sachet of probiotic (De Simone Formulation; formally known as VSL#3 and Visbiome) or placebo once daily for 8 weeks.

Each probiotic packet contains 900 billion probiotic bacteria per sachet.¹¹ This formulation contained 8 viable strains of bacteria: 4 strains of Lactobacillus (L acidophilus, L plantarum, L paracasei, L delbrueckii subsp. bulgaricus); 3 strains of Bifidobacteria (Bifidobacterium breve, B lactis, B infantis); and 1 strain of Streptococcus thermophilus. This formulation had been commercialized and studied as VSL#3 and is currently available in the United States under the Visbiome trade name. While branding changed during the study, the formulation did not. The investigational medicine (VSL#3, Visbiome, and placebo) were shipped from the manufacturer Dupont/Danisco in Madison, Wisconsin. The subjects received placebo or probiotic (VSL#3/Visbiome) and both were identical in appearance. The medication was supplied in a numbered container by the pharmacist after patient enrollment.

Measures

Veterans completed the bowel disease questionnaire to record baseline bowel habits.²² All veterans recorded daily bowel symptoms to confirm the presence of IBS during the 2-week pretreatment period, at baseline, and at the end of the 8-week treatment. The symptoms assessed included severity of abdominal pain (0, none to 100, severe); severity of bloating (0, none to 100, severe); stool frequency; Bristol stool scale (1, very hard to 7, watery); severity of diarrhea (0, none to 100, severe); severity of constipation (0, none to 100, severe); satisfaction with bowel habits (0, none to 100, severe); and IBS affecting or interfering with life (0, none to 100, severe). The bowel symptom score is the sum of the 5 symptom scores.^23,24

IBS-specific QOL (IBS-QOL) was recorded at baseline and at the end of treatment.²⁵ The IBS-QOL consists of a 34-item validated disease-specific questionnaire that measures 8 domains relevant to subjects with IBS: dysphoria, interference with activity, body image, health worry, food avoidance, social reaction, sexual life, and relationships. We used the Somatic Symptom Checklist to detect the following extra-intestinal symptoms that are common among veterans with GWI: headache, backache, wheeziness, insomnia, bad breath, fatigue, general stiffness, dizziness, weakness, sensitivity to hot and cold, palpitation, and tightness in chest. Subjects rated symptoms on a scale of 1 to 5: how often (1, none; 2, monthly; 3, once weekly; 4, several times weekly; 5, daily), and how bothersome (1, not at all to 5, extremely).²⁶

Subjects completed the Posttraumatic Stress Disorder (PTSD) Checklist–Military, which is specific to military experience with 17 items on a 1 to 5 scale (1, not at all to 5, extremely). Scores were summed to produce a total symptom severity score (range, 17-85).²⁷ Subjects also completed the Brief Symptom Inventory 18 (BSI-18) during the baseline evaluation.²⁸ BSI-18 measures subjects’ reported overall psychological distress. It assesses 3 symptoms dimensions (somatization, depression, and anxiety) and a global severity index. The raw scores were transferred to normative T scores based on samples of nonpatient normal men and women.

Statistical Methods

Comparisons of the probiotic vs placebo groups for demographic variable were analyzed using a 2-sample t test for continuous variables, and with a χ² test or Fisher exact test for categorical variables. The primary and secondary outcome variables were recorded daily for 2 weeks as pretreatment baseline and for 2 weeks at the end of treatment. These symptoms were recorded as ordered categorical variables, which were then averaged across the week to produce a continuous measurement for statistical analysis. For the primary outcome of GI symptoms, posttreatment comparisons were made between the study groups using a 2-sample t test of the baseline vs posttreatment values. All P values were calculated for 2-sided comparisons. The planned sample size in our study protocol was to recruit 40 individuals per group in order to achieve 80% power to detect a 30% improvement between baseline and end of treatment in the primary bowel symptom score. This study recruited 53 subjects. With this sample size, the study had 80% power to detect a 0.8 SD in any of the outcomes.

Results

We screened 101 veterans with IBS and GWI; 39 veterans did not fulfill the inclusion/exclusion criteria, 22 declined to participate or did not complete the screening questionnaires and tests, and 9 were lost to follow-up. Sixty-two participants were randomized in a double-blind placebo-controlled study design; 9 dropped out before the end of the study. Data were analyzed from 53 veterans who completed the study, 29 in the placebo group and 24 in the probiotic group (Figure 1). The cohort was primarily male with a mean (SD) age of 55 (8) years (range, 42-73) (Table 1).

Overall, the treatment was well tolerated. All subjects were contacted every 2 weeks during the study to check for adverse effects, but no serious events were reported. There were no differences at baseline in any of the BSI-18 subscale scores in veterans between the groups. There was a greater mean (SEM) improvement of diarrhea severity in the probiotic group compared with the placebo group: 18 (6), a 31% improvement, vs 6 (5), a 13% improvement, respectively; however, the difference was not statistically significance (P = .13) (Table 2). There also was a greater mean (SEM) improvement in satisfaction of bowel habits in the probiotic group compared with the placebo group: 16 (7), a 35% improvement vs 4 (9), an 8% worsening; this also was not statistically significant (P = .09). There was no difference in the change of IBS-QOL before and after treatment in either group (Figure 2). There was no improvement in any of the symptoms of GWI (all P ≥ .06) (Appendix).

Discussion

GWI is a complex multisystem illness of unknown etiology. There was high prevalence of diarrhea during deployment, and veterans were exposed to several physical, environmental, and mental stresses of the war.³ A change in gut microbiota can occur during deployment due to diet changes, environmental and physical stress, and GI infections.²⁹ These changes would suggest that manipulation of gut microbiota might offer a new modality of treatment of IBS and GWI. We evaluated the effect of a high-potency multistrain probiotic in veterans with IBS and GWI. We did not detect any statistically significant differences between the probiotic and placebo groups on bowel symptom score and individual symptoms of IBS and on QOL. Also, there was no improvement for the other symptoms of GWI. To our knowledge, this is the first study evaluating the effect of probiotics in veterans with IBS and GWI. Our results are consistent with the literature on probiotics and IBS.

The probiotic formulation used in our study has been evaluated in patients with IBS previously. Kim and colleagues found that after 8 weeks of treatment of patients with diarrhea-predominant IBS with VSL#3, there was improvement in bloating, but no effect was found on abdominal pain, gas, or urgency.³⁰ A subsequent study by the same investigators on patients with all types of IBS found that VSL#3 showed no effect on abdominal pain, stool frequency and consistency, or on bloating, but there was improvement in flatulence.³¹ Another study that evaluated the effect of VSL#3 on symptoms of diarrhea-predominant IBS and QOL found improvement in IBS symptoms from baseline in both the probiotic and the placebo groups, but the difference between the 2 groups was not statistically significant.³² Similarly, Wong and colleagues performed a double-blind, placebo-controlled mechanistic study to evaluate the effect of VSL#3. They found improvement in bowel symptom score, abdominal pain intensity, and satisfaction with bowel habits with both the VSL#3 and placebo group but similar to our study, the differences were not statistically significant.

Several reviews have evaluated the efficacy of probiotics for IBS. A 2010 review found evidence that probiotics trended toward improved IBS symptoms compared with placebo.³³ The 2014 follow-up by the same authors demonstrated that overall, probiotics improved global symptoms of IBS and multistrain probiotics were more effective.²⁰ A third meta-analysis from the same group found evidence that multistrain probiotics seemed to have a beneficial effect but could not definitively conclude that probiotics are efficacious in improving IBS symptoms.³⁴ Other authors also have seen inconsistent effects of probiotics compared with placebo on global symptoms, abdominal pain, and bloating after performing systematic reviews of the literature.^35-38 Although several reviews support that multistrain probiotics are more effective, they fail to conclude which combinations are more efficacious.

The effect of probiotics on QOL has not been investigated by many studies.³⁷ In our study, we did not find significant improvement in QOL in the probiotic group, which is in line with 2 previous studies that showed no effect on IBS QOL of VSL#3 vs placebo.^32,39 Most of the research reports that multistrain probiotics are more effective than using a single strain.^34,35,40Bifidobacterium and Lactobacillus are the most commonly used bacteria in the multistrain probiotics that have shown their positive effect on IBS.^35,41 The probiotic used in our study contained other species along with these 2 microorganisms.

The dose and duration of treatment of probiotics also has been debated. In one meta-analysis, the investigators found that studies of ≥ 8 weeks were more likely to show a positive effect; 4 of the 7 studies with statistically significant improvement in IBS symptoms were longer than 8 weeks.³⁵ However, another meta-analysis based on 35 randomized controlled trials found that there was not a statistically significant difference between groups treated for > 4 weeks vs < 4 weeks.⁴² In addition, another meta-analysis of VSL#3 on IBS in children and adults also found no difference in results based on the duration of treatment of probiotics.⁴³ Similar to our study, 3 other studies of VSL#3 treated patients for 8 weeks and found no statistically significant effect.^30-32 In the past, VSL#3 has been used at dosages of 450 or 900 billion bacteria per day.

An individual’s response to probiotics may depend on the subtype of IBS. However, most of the studies, like ours, included groups of all subtypes. It may be that probiotics are more effective in patients with moderate-to-severe symptoms. Most of our patients had milder symptoms, and we cannot discount how subjects with more severe disease may have responded to the drug. Interestingly, one study demonstrated that Lactobacillus was more effective in patients with moderately severe abdominal pain compared with mild symptoms.⁴⁴

In our study, the probiotic did not improve PTSD symptoms or other extra-intestinal symptoms common in IBS and GWI. Similar to our study, Wong and colleagues did not find significant improvement of psychological and sleep scores after treatment with VSL#3.⁶ Similarly, there is evidence that alteration in gut microbiota is associated with health and diseases, but what specific alterations occur and whether they can be improved with probiotics remains unknown.⁴⁵

Limitations

The inconsistent response to probiotics in various studies may be due to IBS heterogeneity. Furthermore, there are demographic differences between Gulf War veterans and patients enrolled in other studies: Gulf War veterans are predominantly male, many were deployed abroad and had a history of gastroenteritis during deployment, and were exposed to stressful situations.⁴⁶ These factors may be involved in triggering or maintaining IBS in Gulf War veterans. A further limitation of our randomized trial is the relatively small sample size.

Conclusions

This study did not demonstrate statistically significant improvement in symptoms of IBS or improvement in QOL after treatment with a multistrain probiotic. We also did not find any improvement in symptoms of GWI or PTSD. There was no difference in psychological scores between the placebo and treatment groups, and it is unlikely that psychological factors confounded the response to treatment in this study.

The effectiveness of a probiotic may depend on the baseline gut microbiome of the individual and depend on the strain, amount, and frequency of bacteria used. A lack of response of the probiotics does not exclude gut viruses and fungi having a role in exacerbating GWI symptoms. It is also possible that the bacteria present or the dose of the probiotic used was not sufficient to improve symptoms. So far, the definitive benefit of probiotics has been demonstrated for only a few preparations, and none are approved by the US Food and Drug Administration for any disease. More research is needed to determine whether probiotics have any role in the treatment of IBS and GWI.

Acknowledgments

AKT received grant support from the US Department of Veterans Affairs and the US Department of Defense (W81XWH-10-1-0593, W81XWH-15-1-0636). We thank Keith G. Tolman, MD, for assistance in editing the initial proposal and for periodic consultation. We thank the manufacturer of the probiotic for supplying the active drug and the placebo. The manufacture of the probiotic had no role in the design and conduct of the study, analysis and interpretation of the data, and in the preparation of the manuscript.

About 700,000 US military personnel were deployed in Operation Desert Storm (August 1990 to March 1991).¹ Almost 30 years since the war, a large number of these veterans continue to experience a complex of symptoms of unknown etiology called Gulf War illness (GWI), which significantly affects health and quality of life (QOL). The lack of clear etiology of the illness has impaired research to find specific treatments and has further exacerbated the stress among veterans. GWI typically includes a mixture of chronic headache, cognitive difficulties, widespread pain, unexplained fatigue, memory and concentration problems, as well as chronic respiratory and gastrointestinal (GI) symptoms.² Abdominal pain and alteration of bowel habits are also symptoms typical of irritable bowel syndrome (IBS). It has been estimated that IBS occurs in up to 30% of Gulf War veterans.³

The etiology of IBS is unknown. Possible mechanisms include visceral hypersensitivity, altered gut motor function, aberrant brain-gut interaction, and psychological factors, perhaps with a genetic predisposition.⁴ Gastroenteritis has been reported as a triggering mechanism in up to one-third of patients with IBS.⁵ Gastroenteritis can alter the gut microbiota and has been reported to be a significant risk factor for the development of IBS.⁶ In one study of Operation Desert Shield soldiers, > 50% of military personnel developed acute gastroenteritis while on duty.⁷ A high prevalence of extra-intestinal symptoms also has been reported, including fatigue, headache, joint pains, and anxiety, in Gulf War veterans with IBS. These extra-intestinal symptoms of IBS are consistent with the reported GWI symptoms. Change in gut microbiota also has been associated with many of the extra-intestinal symptoms of IBS, especially fatigue.^8,9 Gut microbiota are known to change with travel, stress, and a change in diet, all potential factors that are relevant to Gulf War veterans. This would suggest that an imbalance in the gut microbiota, ie, dysbiosis, may play a role in the pathogenesis of both IBS and GWI. Dysbiosis could be a risk factor for or alternatively a consequence of GWI.

A systematic review highlighted the heterogeneity of the gut microbiota in patients with IBS.¹⁰ Overall, Enterobacteriaceae, Lactobacillaceae, and Bacteroides were increased, whereas Clostridiales, Faecalibacterium, and Bifidobacterium were decreased in patients with IBS compared with controls. Gut microbiota also has been associated with cognitive changes, anxiety, and depression—symptoms associated with IBS and are part of the GWI.

If altered gut microbiota contributes to the etiopathogenesis of IBS, its restoration of with probiotics should help. Probiotics are live organisms that when ingested may improve health by promoting the growth of naturally occurring flora and establishing a healthy gut flora. Probiotics have several mechanisms of actions. Probiotics work in the lumen of the gut by producing antibacterial molecules and enhancing the mucosal barrier.¹¹ Probiotics also may produce metabolic compounds that alter the intestinal microbiota and improve intestinal barrier function.¹² Probiotics also have been shown to activate receptors in the enteric nervous system with the potential to promote pain relief in the setting of visceral hyperalgesia.^13,14 The anti-inflammatory properties of probiotics potentially could modulate the basic pathophysiology of IBS and improve motility, visceral hypersensitivity, and brain-gut interaction.¹⁵ Furthermore, significant gut dysbiosis has been shown with GWI; suggesting that probiotics may have a role in its management.^16,17

Probiotics have not been studied in Gulf War veterans with IBS. We performed a prospective, double-blind placebo-controlled study to determine the efficacy of a commercially available probiotic containing 8 strains of bacteria (De Simone Formulation; formally known as VSL#3 and Visbiome) on symptoms of IBS and GWI. This probiotic was selected as the overall literature suggested benefit of combination probiotics in IBS, and VSL#3 has been shown to be efficacious in ulcerative colitis and microscopic colitis.^18-20

Methods

Veterans who served in Operation Desert Storm (August 1990 to March 1991) and enrolled at the George E. Wahlen Veterans Affairs (VA) Medical Center (GEWVAMC), Salt Lake City, Utah, were eligible for the study. The inclusion criteria were: veterans aged ≥ 35 years; ≥ 2 nonintestinal GWI symptoms (eg, fatigue, joint pains, insomnia, general stiffness, and headache); IBS diagnosis based on the Rome III criteria; IBS symptoms > 6 months; normal gross appearance of the colonic mucosa; negative markers for celiac disease and inflammatory bowel disease (IBD); normal thyroid function; and serum calcium levels.²¹ Those who had a clinically significant cardiac, pulmonary, hepatic or renal dysfunction; history of/or presence of systemic malignancy; current evidence of celiac disease or IBD; unstable/significant psychiatric disease; recent change in GI medications; current pregnancy; or use of antibiotics or probiotics within the past 1 month were excluded. Subjects were enrolled from a list of veterans with GWI from the GEWVAMC Gulf War registry; referrals to gastroenterology clinics for IBS from internal medicine clinics; and posted advertisements.

Protocol

After written informed consent was obtained, each veteran was verified to have IBS and ≥ 2 GWI symptoms. All veterans had the following tests and panels: complete blood count, erythrocyte sedimentation rate, serum comprehensive metabolic panel, thyroid-stimulating hormone, tissue transglutaminase, stool test for ova and parasite, giardia antigen, and clostridia toxins to exclude organic cause of GI symptoms. Colonoscopy was performed in all veterans to exclude IBD, and to rule out microscopic or lymphocytic colitis.

Randomization was computer generated and maintained by the study pharmacist so that study personnel and patients were blinded to the trial groups. All investigators were blinded and allocation was concealed. The medication was supplied in a numbered container by the pharmacist after patient enrollment. After a 2-week run-in period, veterans were randomized (1:1) to receive either 1 sachet of probiotic (De Simone Formulation; formally known as VSL#3 and Visbiome) or placebo once daily for 8 weeks.

Each probiotic packet contains 900 billion probiotic bacteria per sachet.¹¹ This formulation contained 8 viable strains of bacteria: 4 strains of Lactobacillus (L acidophilus, L plantarum, L paracasei, L delbrueckii subsp. bulgaricus); 3 strains of Bifidobacteria (Bifidobacterium breve, B lactis, B infantis); and 1 strain of Streptococcus thermophilus. This formulation had been commercialized and studied as VSL#3 and is currently available in the United States under the Visbiome trade name. While branding changed during the study, the formulation did not. The investigational medicine (VSL#3, Visbiome, and placebo) were shipped from the manufacturer Dupont/Danisco in Madison, Wisconsin. The subjects received placebo or probiotic (VSL#3/Visbiome) and both were identical in appearance. The medication was supplied in a numbered container by the pharmacist after patient enrollment.

Measures

Veterans completed the bowel disease questionnaire to record baseline bowel habits.²² All veterans recorded daily bowel symptoms to confirm the presence of IBS during the 2-week pretreatment period, at baseline, and at the end of the 8-week treatment. The symptoms assessed included severity of abdominal pain (0, none to 100, severe); severity of bloating (0, none to 100, severe); stool frequency; Bristol stool scale (1, very hard to 7, watery); severity of diarrhea (0, none to 100, severe); severity of constipation (0, none to 100, severe); satisfaction with bowel habits (0, none to 100, severe); and IBS affecting or interfering with life (0, none to 100, severe). The bowel symptom score is the sum of the 5 symptom scores.^23,24

IBS-specific QOL (IBS-QOL) was recorded at baseline and at the end of treatment.²⁵ The IBS-QOL consists of a 34-item validated disease-specific questionnaire that measures 8 domains relevant to subjects with IBS: dysphoria, interference with activity, body image, health worry, food avoidance, social reaction, sexual life, and relationships. We used the Somatic Symptom Checklist to detect the following extra-intestinal symptoms that are common among veterans with GWI: headache, backache, wheeziness, insomnia, bad breath, fatigue, general stiffness, dizziness, weakness, sensitivity to hot and cold, palpitation, and tightness in chest. Subjects rated symptoms on a scale of 1 to 5: how often (1, none; 2, monthly; 3, once weekly; 4, several times weekly; 5, daily), and how bothersome (1, not at all to 5, extremely).²⁶

Subjects completed the Posttraumatic Stress Disorder (PTSD) Checklist–Military, which is specific to military experience with 17 items on a 1 to 5 scale (1, not at all to 5, extremely). Scores were summed to produce a total symptom severity score (range, 17-85).²⁷ Subjects also completed the Brief Symptom Inventory 18 (BSI-18) during the baseline evaluation.²⁸ BSI-18 measures subjects’ reported overall psychological distress. It assesses 3 symptoms dimensions (somatization, depression, and anxiety) and a global severity index. The raw scores were transferred to normative T scores based on samples of nonpatient normal men and women.

Statistical Methods

Comparisons of the probiotic vs placebo groups for demographic variable were analyzed using a 2-sample t test for continuous variables, and with a χ² test or Fisher exact test for categorical variables. The primary and secondary outcome variables were recorded daily for 2 weeks as pretreatment baseline and for 2 weeks at the end of treatment. These symptoms were recorded as ordered categorical variables, which were then averaged across the week to produce a continuous measurement for statistical analysis. For the primary outcome of GI symptoms, posttreatment comparisons were made between the study groups using a 2-sample t test of the baseline vs posttreatment values. All P values were calculated for 2-sided comparisons. The planned sample size in our study protocol was to recruit 40 individuals per group in order to achieve 80% power to detect a 30% improvement between baseline and end of treatment in the primary bowel symptom score. This study recruited 53 subjects. With this sample size, the study had 80% power to detect a 0.8 SD in any of the outcomes.

Results

We screened 101 veterans with IBS and GWI; 39 veterans did not fulfill the inclusion/exclusion criteria, 22 declined to participate or did not complete the screening questionnaires and tests, and 9 were lost to follow-up. Sixty-two participants were randomized in a double-blind placebo-controlled study design; 9 dropped out before the end of the study. Data were analyzed from 53 veterans who completed the study, 29 in the placebo group and 24 in the probiotic group (Figure 1). The cohort was primarily male with a mean (SD) age of 55 (8) years (range, 42-73) (Table 1).

Overall, the treatment was well tolerated. All subjects were contacted every 2 weeks during the study to check for adverse effects, but no serious events were reported. There were no differences at baseline in any of the BSI-18 subscale scores in veterans between the groups. There was a greater mean (SEM) improvement of diarrhea severity in the probiotic group compared with the placebo group: 18 (6), a 31% improvement, vs 6 (5), a 13% improvement, respectively; however, the difference was not statistically significance (P = .13) (Table 2). There also was a greater mean (SEM) improvement in satisfaction of bowel habits in the probiotic group compared with the placebo group: 16 (7), a 35% improvement vs 4 (9), an 8% worsening; this also was not statistically significant (P = .09). There was no difference in the change of IBS-QOL before and after treatment in either group (Figure 2). There was no improvement in any of the symptoms of GWI (all P ≥ .06) (Appendix).

Discussion

GWI is a complex multisystem illness of unknown etiology. There was high prevalence of diarrhea during deployment, and veterans were exposed to several physical, environmental, and mental stresses of the war.³ A change in gut microbiota can occur during deployment due to diet changes, environmental and physical stress, and GI infections.²⁹ These changes would suggest that manipulation of gut microbiota might offer a new modality of treatment of IBS and GWI. We evaluated the effect of a high-potency multistrain probiotic in veterans with IBS and GWI. We did not detect any statistically significant differences between the probiotic and placebo groups on bowel symptom score and individual symptoms of IBS and on QOL. Also, there was no improvement for the other symptoms of GWI. To our knowledge, this is the first study evaluating the effect of probiotics in veterans with IBS and GWI. Our results are consistent with the literature on probiotics and IBS.

The probiotic formulation used in our study has been evaluated in patients with IBS previously. Kim and colleagues found that after 8 weeks of treatment of patients with diarrhea-predominant IBS with VSL#3, there was improvement in bloating, but no effect was found on abdominal pain, gas, or urgency.³⁰ A subsequent study by the same investigators on patients with all types of IBS found that VSL#3 showed no effect on abdominal pain, stool frequency and consistency, or on bloating, but there was improvement in flatulence.³¹ Another study that evaluated the effect of VSL#3 on symptoms of diarrhea-predominant IBS and QOL found improvement in IBS symptoms from baseline in both the probiotic and the placebo groups, but the difference between the 2 groups was not statistically significant.³² Similarly, Wong and colleagues performed a double-blind, placebo-controlled mechanistic study to evaluate the effect of VSL#3. They found improvement in bowel symptom score, abdominal pain intensity, and satisfaction with bowel habits with both the VSL#3 and placebo group but similar to our study, the differences were not statistically significant.

Several reviews have evaluated the efficacy of probiotics for IBS. A 2010 review found evidence that probiotics trended toward improved IBS symptoms compared with placebo.³³ The 2014 follow-up by the same authors demonstrated that overall, probiotics improved global symptoms of IBS and multistrain probiotics were more effective.²⁰ A third meta-analysis from the same group found evidence that multistrain probiotics seemed to have a beneficial effect but could not definitively conclude that probiotics are efficacious in improving IBS symptoms.³⁴ Other authors also have seen inconsistent effects of probiotics compared with placebo on global symptoms, abdominal pain, and bloating after performing systematic reviews of the literature.^35-38 Although several reviews support that multistrain probiotics are more effective, they fail to conclude which combinations are more efficacious.

The effect of probiotics on QOL has not been investigated by many studies.³⁷ In our study, we did not find significant improvement in QOL in the probiotic group, which is in line with 2 previous studies that showed no effect on IBS QOL of VSL#3 vs placebo.^32,39 Most of the research reports that multistrain probiotics are more effective than using a single strain.^34,35,40Bifidobacterium and Lactobacillus are the most commonly used bacteria in the multistrain probiotics that have shown their positive effect on IBS.^35,41 The probiotic used in our study contained other species along with these 2 microorganisms.

The dose and duration of treatment of probiotics also has been debated. In one meta-analysis, the investigators found that studies of ≥ 8 weeks were more likely to show a positive effect; 4 of the 7 studies with statistically significant improvement in IBS symptoms were longer than 8 weeks.³⁵ However, another meta-analysis based on 35 randomized controlled trials found that there was not a statistically significant difference between groups treated for > 4 weeks vs < 4 weeks.⁴² In addition, another meta-analysis of VSL#3 on IBS in children and adults also found no difference in results based on the duration of treatment of probiotics.⁴³ Similar to our study, 3 other studies of VSL#3 treated patients for 8 weeks and found no statistically significant effect.^30-32 In the past, VSL#3 has been used at dosages of 450 or 900 billion bacteria per day.

An individual’s response to probiotics may depend on the subtype of IBS. However, most of the studies, like ours, included groups of all subtypes. It may be that probiotics are more effective in patients with moderate-to-severe symptoms. Most of our patients had milder symptoms, and we cannot discount how subjects with more severe disease may have responded to the drug. Interestingly, one study demonstrated that Lactobacillus was more effective in patients with moderately severe abdominal pain compared with mild symptoms.⁴⁴

In our study, the probiotic did not improve PTSD symptoms or other extra-intestinal symptoms common in IBS and GWI. Similar to our study, Wong and colleagues did not find significant improvement of psychological and sleep scores after treatment with VSL#3.⁶ Similarly, there is evidence that alteration in gut microbiota is associated with health and diseases, but what specific alterations occur and whether they can be improved with probiotics remains unknown.⁴⁵

Limitations

The inconsistent response to probiotics in various studies may be due to IBS heterogeneity. Furthermore, there are demographic differences between Gulf War veterans and patients enrolled in other studies: Gulf War veterans are predominantly male, many were deployed abroad and had a history of gastroenteritis during deployment, and were exposed to stressful situations.⁴⁶ These factors may be involved in triggering or maintaining IBS in Gulf War veterans. A further limitation of our randomized trial is the relatively small sample size.

Conclusions

This study did not demonstrate statistically significant improvement in symptoms of IBS or improvement in QOL after treatment with a multistrain probiotic. We also did not find any improvement in symptoms of GWI or PTSD. There was no difference in psychological scores between the placebo and treatment groups, and it is unlikely that psychological factors confounded the response to treatment in this study.

The effectiveness of a probiotic may depend on the baseline gut microbiome of the individual and depend on the strain, amount, and frequency of bacteria used. A lack of response of the probiotics does not exclude gut viruses and fungi having a role in exacerbating GWI symptoms. It is also possible that the bacteria present or the dose of the probiotic used was not sufficient to improve symptoms. So far, the definitive benefit of probiotics has been demonstrated for only a few preparations, and none are approved by the US Food and Drug Administration for any disease. More research is needed to determine whether probiotics have any role in the treatment of IBS and GWI.

Acknowledgments

AKT received grant support from the US Department of Veterans Affairs and the US Department of Defense (W81XWH-10-1-0593, W81XWH-15-1-0636). We thank Keith G. Tolman, MD, for assistance in editing the initial proposal and for periodic consultation. We thank the manufacturer of the probiotic for supplying the active drug and the placebo. The manufacture of the probiotic had no role in the design and conduct of the study, analysis and interpretation of the data, and in the preparation of the manuscript.

About 700,000 US military personnel were deployed in Operation Desert Storm (August 1990 to March 1991).¹ Almost 30 years since the war, a large number of these veterans continue to experience a complex of symptoms of unknown etiology called Gulf War illness (GWI), which significantly affects health and quality of life (QOL). The lack of clear etiology of the illness has impaired research to find specific treatments and has further exacerbated the stress among veterans. GWI typically includes a mixture of chronic headache, cognitive difficulties, widespread pain, unexplained fatigue, memory and concentration problems, as well as chronic respiratory and gastrointestinal (GI) symptoms.² Abdominal pain and alteration of bowel habits are also symptoms typical of irritable bowel syndrome (IBS). It has been estimated that IBS occurs in up to 30% of Gulf War veterans.³

The etiology of IBS is unknown. Possible mechanisms include visceral hypersensitivity, altered gut motor function, aberrant brain-gut interaction, and psychological factors, perhaps with a genetic predisposition.⁴ Gastroenteritis has been reported as a triggering mechanism in up to one-third of patients with IBS.⁵ Gastroenteritis can alter the gut microbiota and has been reported to be a significant risk factor for the development of IBS.⁶ In one study of Operation Desert Shield soldiers, > 50% of military personnel developed acute gastroenteritis while on duty.⁷ A high prevalence of extra-intestinal symptoms also has been reported, including fatigue, headache, joint pains, and anxiety, in Gulf War veterans with IBS. These extra-intestinal symptoms of IBS are consistent with the reported GWI symptoms. Change in gut microbiota also has been associated with many of the extra-intestinal symptoms of IBS, especially fatigue.^8,9 Gut microbiota are known to change with travel, stress, and a change in diet, all potential factors that are relevant to Gulf War veterans. This would suggest that an imbalance in the gut microbiota, ie, dysbiosis, may play a role in the pathogenesis of both IBS and GWI. Dysbiosis could be a risk factor for or alternatively a consequence of GWI.

A systematic review highlighted the heterogeneity of the gut microbiota in patients with IBS.¹⁰ Overall, Enterobacteriaceae, Lactobacillaceae, and Bacteroides were increased, whereas Clostridiales, Faecalibacterium, and Bifidobacterium were decreased in patients with IBS compared with controls. Gut microbiota also has been associated with cognitive changes, anxiety, and depression—symptoms associated with IBS and are part of the GWI.

If altered gut microbiota contributes to the etiopathogenesis of IBS, its restoration of with probiotics should help. Probiotics are live organisms that when ingested may improve health by promoting the growth of naturally occurring flora and establishing a healthy gut flora. Probiotics have several mechanisms of actions. Probiotics work in the lumen of the gut by producing antibacterial molecules and enhancing the mucosal barrier.¹¹ Probiotics also may produce metabolic compounds that alter the intestinal microbiota and improve intestinal barrier function.¹² Probiotics also have been shown to activate receptors in the enteric nervous system with the potential to promote pain relief in the setting of visceral hyperalgesia.^13,14 The anti-inflammatory properties of probiotics potentially could modulate the basic pathophysiology of IBS and improve motility, visceral hypersensitivity, and brain-gut interaction.¹⁵ Furthermore, significant gut dysbiosis has been shown with GWI; suggesting that probiotics may have a role in its management.^16,17

Probiotics have not been studied in Gulf War veterans with IBS. We performed a prospective, double-blind placebo-controlled study to determine the efficacy of a commercially available probiotic containing 8 strains of bacteria (De Simone Formulation; formally known as VSL#3 and Visbiome) on symptoms of IBS and GWI. This probiotic was selected as the overall literature suggested benefit of combination probiotics in IBS, and VSL#3 has been shown to be efficacious in ulcerative colitis and microscopic colitis.^18-20

Methods

Veterans who served in Operation Desert Storm (August 1990 to March 1991) and enrolled at the George E. Wahlen Veterans Affairs (VA) Medical Center (GEWVAMC), Salt Lake City, Utah, were eligible for the study. The inclusion criteria were: veterans aged ≥ 35 years; ≥ 2 nonintestinal GWI symptoms (eg, fatigue, joint pains, insomnia, general stiffness, and headache); IBS diagnosis based on the Rome III criteria; IBS symptoms > 6 months; normal gross appearance of the colonic mucosa; negative markers for celiac disease and inflammatory bowel disease (IBD); normal thyroid function; and serum calcium levels.²¹ Those who had a clinically significant cardiac, pulmonary, hepatic or renal dysfunction; history of/or presence of systemic malignancy; current evidence of celiac disease or IBD; unstable/significant psychiatric disease; recent change in GI medications; current pregnancy; or use of antibiotics or probiotics within the past 1 month were excluded. Subjects were enrolled from a list of veterans with GWI from the GEWVAMC Gulf War registry; referrals to gastroenterology clinics for IBS from internal medicine clinics; and posted advertisements.

Protocol

After written informed consent was obtained, each veteran was verified to have IBS and ≥ 2 GWI symptoms. All veterans had the following tests and panels: complete blood count, erythrocyte sedimentation rate, serum comprehensive metabolic panel, thyroid-stimulating hormone, tissue transglutaminase, stool test for ova and parasite, giardia antigen, and clostridia toxins to exclude organic cause of GI symptoms. Colonoscopy was performed in all veterans to exclude IBD, and to rule out microscopic or lymphocytic colitis.

Randomization was computer generated and maintained by the study pharmacist so that study personnel and patients were blinded to the trial groups. All investigators were blinded and allocation was concealed. The medication was supplied in a numbered container by the pharmacist after patient enrollment. After a 2-week run-in period, veterans were randomized (1:1) to receive either 1 sachet of probiotic (De Simone Formulation; formally known as VSL#3 and Visbiome) or placebo once daily for 8 weeks.

Each probiotic packet contains 900 billion probiotic bacteria per sachet.¹¹ This formulation contained 8 viable strains of bacteria: 4 strains of Lactobacillus (L acidophilus, L plantarum, L paracasei, L delbrueckii subsp. bulgaricus); 3 strains of Bifidobacteria (Bifidobacterium breve, B lactis, B infantis); and 1 strain of Streptococcus thermophilus. This formulation had been commercialized and studied as VSL#3 and is currently available in the United States under the Visbiome trade name. While branding changed during the study, the formulation did not. The investigational medicine (VSL#3, Visbiome, and placebo) were shipped from the manufacturer Dupont/Danisco in Madison, Wisconsin. The subjects received placebo or probiotic (VSL#3/Visbiome) and both were identical in appearance. The medication was supplied in a numbered container by the pharmacist after patient enrollment.

Measures

Veterans completed the bowel disease questionnaire to record baseline bowel habits.²² All veterans recorded daily bowel symptoms to confirm the presence of IBS during the 2-week pretreatment period, at baseline, and at the end of the 8-week treatment. The symptoms assessed included severity of abdominal pain (0, none to 100, severe); severity of bloating (0, none to 100, severe); stool frequency; Bristol stool scale (1, very hard to 7, watery); severity of diarrhea (0, none to 100, severe); severity of constipation (0, none to 100, severe); satisfaction with bowel habits (0, none to 100, severe); and IBS affecting or interfering with life (0, none to 100, severe). The bowel symptom score is the sum of the 5 symptom scores.^23,24

IBS-specific QOL (IBS-QOL) was recorded at baseline and at the end of treatment.²⁵ The IBS-QOL consists of a 34-item validated disease-specific questionnaire that measures 8 domains relevant to subjects with IBS: dysphoria, interference with activity, body image, health worry, food avoidance, social reaction, sexual life, and relationships. We used the Somatic Symptom Checklist to detect the following extra-intestinal symptoms that are common among veterans with GWI: headache, backache, wheeziness, insomnia, bad breath, fatigue, general stiffness, dizziness, weakness, sensitivity to hot and cold, palpitation, and tightness in chest. Subjects rated symptoms on a scale of 1 to 5: how often (1, none; 2, monthly; 3, once weekly; 4, several times weekly; 5, daily), and how bothersome (1, not at all to 5, extremely).²⁶

Subjects completed the Posttraumatic Stress Disorder (PTSD) Checklist–Military, which is specific to military experience with 17 items on a 1 to 5 scale (1, not at all to 5, extremely). Scores were summed to produce a total symptom severity score (range, 17-85).²⁷ Subjects also completed the Brief Symptom Inventory 18 (BSI-18) during the baseline evaluation.²⁸ BSI-18 measures subjects’ reported overall psychological distress. It assesses 3 symptoms dimensions (somatization, depression, and anxiety) and a global severity index. The raw scores were transferred to normative T scores based on samples of nonpatient normal men and women.

Statistical Methods

Comparisons of the probiotic vs placebo groups for demographic variable were analyzed using a 2-sample t test for continuous variables, and with a χ² test or Fisher exact test for categorical variables. The primary and secondary outcome variables were recorded daily for 2 weeks as pretreatment baseline and for 2 weeks at the end of treatment. These symptoms were recorded as ordered categorical variables, which were then averaged across the week to produce a continuous measurement for statistical analysis. For the primary outcome of GI symptoms, posttreatment comparisons were made between the study groups using a 2-sample t test of the baseline vs posttreatment values. All P values were calculated for 2-sided comparisons. The planned sample size in our study protocol was to recruit 40 individuals per group in order to achieve 80% power to detect a 30% improvement between baseline and end of treatment in the primary bowel symptom score. This study recruited 53 subjects. With this sample size, the study had 80% power to detect a 0.8 SD in any of the outcomes.

Results

We screened 101 veterans with IBS and GWI; 39 veterans did not fulfill the inclusion/exclusion criteria, 22 declined to participate or did not complete the screening questionnaires and tests, and 9 were lost to follow-up. Sixty-two participants were randomized in a double-blind placebo-controlled study design; 9 dropped out before the end of the study. Data were analyzed from 53 veterans who completed the study, 29 in the placebo group and 24 in the probiotic group (Figure 1). The cohort was primarily male with a mean (SD) age of 55 (8) years (range, 42-73) (Table 1).

Overall, the treatment was well tolerated. All subjects were contacted every 2 weeks during the study to check for adverse effects, but no serious events were reported. There were no differences at baseline in any of the BSI-18 subscale scores in veterans between the groups. There was a greater mean (SEM) improvement of diarrhea severity in the probiotic group compared with the placebo group: 18 (6), a 31% improvement, vs 6 (5), a 13% improvement, respectively; however, the difference was not statistically significance (P = .13) (Table 2). There also was a greater mean (SEM) improvement in satisfaction of bowel habits in the probiotic group compared with the placebo group: 16 (7), a 35% improvement vs 4 (9), an 8% worsening; this also was not statistically significant (P = .09). There was no difference in the change of IBS-QOL before and after treatment in either group (Figure 2). There was no improvement in any of the symptoms of GWI (all P ≥ .06) (Appendix).

Discussion

GWI is a complex multisystem illness of unknown etiology. There was high prevalence of diarrhea during deployment, and veterans were exposed to several physical, environmental, and mental stresses of the war.³ A change in gut microbiota can occur during deployment due to diet changes, environmental and physical stress, and GI infections.²⁹ These changes would suggest that manipulation of gut microbiota might offer a new modality of treatment of IBS and GWI. We evaluated the effect of a high-potency multistrain probiotic in veterans with IBS and GWI. We did not detect any statistically significant differences between the probiotic and placebo groups on bowel symptom score and individual symptoms of IBS and on QOL. Also, there was no improvement for the other symptoms of GWI. To our knowledge, this is the first study evaluating the effect of probiotics in veterans with IBS and GWI. Our results are consistent with the literature on probiotics and IBS.

The probiotic formulation used in our study has been evaluated in patients with IBS previously. Kim and colleagues found that after 8 weeks of treatment of patients with diarrhea-predominant IBS with VSL#3, there was improvement in bloating, but no effect was found on abdominal pain, gas, or urgency.³⁰ A subsequent study by the same investigators on patients with all types of IBS found that VSL#3 showed no effect on abdominal pain, stool frequency and consistency, or on bloating, but there was improvement in flatulence.³¹ Another study that evaluated the effect of VSL#3 on symptoms of diarrhea-predominant IBS and QOL found improvement in IBS symptoms from baseline in both the probiotic and the placebo groups, but the difference between the 2 groups was not statistically significant.³² Similarly, Wong and colleagues performed a double-blind, placebo-controlled mechanistic study to evaluate the effect of VSL#3. They found improvement in bowel symptom score, abdominal pain intensity, and satisfaction with bowel habits with both the VSL#3 and placebo group but similar to our study, the differences were not statistically significant.

Several reviews have evaluated the efficacy of probiotics for IBS. A 2010 review found evidence that probiotics trended toward improved IBS symptoms compared with placebo.³³ The 2014 follow-up by the same authors demonstrated that overall, probiotics improved global symptoms of IBS and multistrain probiotics were more effective.²⁰ A third meta-analysis from the same group found evidence that multistrain probiotics seemed to have a beneficial effect but could not definitively conclude that probiotics are efficacious in improving IBS symptoms.³⁴ Other authors also have seen inconsistent effects of probiotics compared with placebo on global symptoms, abdominal pain, and bloating after performing systematic reviews of the literature.^35-38 Although several reviews support that multistrain probiotics are more effective, they fail to conclude which combinations are more efficacious.

The effect of probiotics on QOL has not been investigated by many studies.³⁷ In our study, we did not find significant improvement in QOL in the probiotic group, which is in line with 2 previous studies that showed no effect on IBS QOL of VSL#3 vs placebo.^32,39 Most of the research reports that multistrain probiotics are more effective than using a single strain.^34,35,40Bifidobacterium and Lactobacillus are the most commonly used bacteria in the multistrain probiotics that have shown their positive effect on IBS.^35,41 The probiotic used in our study contained other species along with these 2 microorganisms.

The dose and duration of treatment of probiotics also has been debated. In one meta-analysis, the investigators found that studies of ≥ 8 weeks were more likely to show a positive effect; 4 of the 7 studies with statistically significant improvement in IBS symptoms were longer than 8 weeks.³⁵ However, another meta-analysis based on 35 randomized controlled trials found that there was not a statistically significant difference between groups treated for > 4 weeks vs < 4 weeks.⁴² In addition, another meta-analysis of VSL#3 on IBS in children and adults also found no difference in results based on the duration of treatment of probiotics.⁴³ Similar to our study, 3 other studies of VSL#3 treated patients for 8 weeks and found no statistically significant effect.^30-32 In the past, VSL#3 has been used at dosages of 450 or 900 billion bacteria per day.

An individual’s response to probiotics may depend on the subtype of IBS. However, most of the studies, like ours, included groups of all subtypes. It may be that probiotics are more effective in patients with moderate-to-severe symptoms. Most of our patients had milder symptoms, and we cannot discount how subjects with more severe disease may have responded to the drug. Interestingly, one study demonstrated that Lactobacillus was more effective in patients with moderately severe abdominal pain compared with mild symptoms.⁴⁴

In our study, the probiotic did not improve PTSD symptoms or other extra-intestinal symptoms common in IBS and GWI. Similar to our study, Wong and colleagues did not find significant improvement of psychological and sleep scores after treatment with VSL#3.⁶ Similarly, there is evidence that alteration in gut microbiota is associated with health and diseases, but what specific alterations occur and whether they can be improved with probiotics remains unknown.⁴⁵

Limitations

The inconsistent response to probiotics in various studies may be due to IBS heterogeneity. Furthermore, there are demographic differences between Gulf War veterans and patients enrolled in other studies: Gulf War veterans are predominantly male, many were deployed abroad and had a history of gastroenteritis during deployment, and were exposed to stressful situations.⁴⁶ These factors may be involved in triggering or maintaining IBS in Gulf War veterans. A further limitation of our randomized trial is the relatively small sample size.

Conclusions

This study did not demonstrate statistically significant improvement in symptoms of IBS or improvement in QOL after treatment with a multistrain probiotic. We also did not find any improvement in symptoms of GWI or PTSD. There was no difference in psychological scores between the placebo and treatment groups, and it is unlikely that psychological factors confounded the response to treatment in this study.

The effectiveness of a probiotic may depend on the baseline gut microbiome of the individual and depend on the strain, amount, and frequency of bacteria used. A lack of response of the probiotics does not exclude gut viruses and fungi having a role in exacerbating GWI symptoms. It is also possible that the bacteria present or the dose of the probiotic used was not sufficient to improve symptoms. So far, the definitive benefit of probiotics has been demonstrated for only a few preparations, and none are approved by the US Food and Drug Administration for any disease. More research is needed to determine whether probiotics have any role in the treatment of IBS and GWI.

Acknowledgments

AKT received grant support from the US Department of Veterans Affairs and the US Department of Defense (W81XWH-10-1-0593, W81XWH-15-1-0636). We thank Keith G. Tolman, MD, for assistance in editing the initial proposal and for periodic consultation. We thank the manufacturer of the probiotic for supplying the active drug and the placebo. The manufacture of the probiotic had no role in the design and conduct of the study, analysis and interpretation of the data, and in the preparation of the manuscript.

Mohs micrographic surgery (MMS) is most commonly used for the surgical management of squamous cell carcinomas (SCCs) and basal cell carcinomas (BCCs) in high-risk locations. The ability for 100% margin evaluation with MMS also has shown lower recurrence rates compared with wide local excision for less common and/or more aggressive tumors. However, there is a lack of standardization on initial and subsequent margin size when treating these less common skin tumors, such as dermatofibrosarcoma protuberans (DFSP), atypical fibroxanthoma (AFX), and sebaceous carcinoma.

Because Mohs surgeons must balance normal tissue preservation with the importance of tumor clearance in the context of comprehensive margin control, we aimed to assess the practice patterns of Mohs surgeons regarding margin size for these unique tumors. The average margin size for each Mohs layer has been reported to be 1 to 3 mm for BCC compared with 3 to 6 mm or larger for other skin cancers, such as melanoma in situ (MIS).^1-3 We hypothesized that the initial margin size would vary among surgeons and likely be greater for more aggressive and rarer malignancies as well as for lesions on the trunk and extremities.

Methods

A descriptive survey was created using SurveyMonkey and distributed to members of the American College of Mohs Surgery (ACMS). Survey participants and their responses were anonymous. Demographic information on survey participants was collected in addition to initial and subsequent MMS margin size for DFSP, AFX, MIS, invasive melanoma, sebaceous carcinoma, microcystic adnexal carcinoma (MAC), poorly differentiated SCC, Merkel cell carcinoma, extramammary Paget disease, leiomyosarcoma, and endocrine mucin-producing sweat gland carcinoma. Survey participants were asked to choose from a range of margin sizes: 1 to 3 mm, 4 to 6 mm, 7 to 9 mm, and greater than 9 mm. This study was approved by the University of Texas Southwest Medical Center (Dallas, Texas) institutional review board.

Results

Eighty-seven respondents from the ACMS listserve completed the survey (response rate <10%). Of these, 58 respondents (66.7%) reported practicing for more than 5 years, and 58 (66.7%) were male. Practice setting was primarily private/community (71.3% [62/87]), and survey respondents were located across the United States. More than 50% of survey respondents treated the following tumors on the head and neck in their respective practices: DFSP (80.9% [55/68]), AFX (95.6% [65/68]), MIS (67.7% [46/68]), sebaceous carcinoma (92.7% [63/68]), MAC (83.8% [57/68]), poorly differentiated SCC (97.1% [66/68]), and endocrine mucin-producing sweat gland carcinoma (51.5% [35/68]). More than 50% of survey respondents treated the following tumors on the trunk and extremities: DFSP (90.3% [47/52]), AFX (86.4% [45/52]), MIS (55.8% [29/52]), sebaceous carcinoma (80.8% [42/52]), MAC (73.1% [38/52]), poorly differentiated SCC (94.2% [49/52]), and extramammary Paget disease (53.9% [28/52]). Invasive melanoma, Merkel cell carcinoma, and leiomyosarcoma were overall less commonly treated.

In general, respondent Mohs surgeons were more likely to take larger initial and subsequent margins for tumors treated on the trunk and extremities compared with the head and neck (Table). In addition, initial margin size often was larger than the 1- to 3-mm margin commonly used in Mohs surgery for BCCs and less aggressive SCCs (Table). A larger initial margin size (>9 mm) and subsequent margin size (4–6 mm) was more commonly reported for certain tumors known to be more aggressive and/or have extensive subclinical extension, such as DFSP and invasive melanoma. Of note, most respondents performed 4- to 6-mm margins (37/67 [55.2%]) for poorly differentiated SCC. Overall, there was a high range of margin size variability among Mohs surgeons for these unique and/or more aggressive skin tumors.

Comment

Given that no guidelines exist on margins with MMS for less commonly treated skin tumors, this study helps give Mohs surgeons perspective on current practice patterns for both initial and subsequent Mohs margin sizes. High margin-size variability among Mohs surgeons is expected, as surgeons also need to account for high-risk features of the tumor or specific locations where tissue sparing is critical. Overall, Mohs surgeons are more likely to take larger initial margins for these less common skin tumors compared with BCCs or SCCs. Initial margin size was consistently larger on the trunk and extremities where tissue sparing often is less critical.

Our survey was limited by a small sample size and incomplete response of the ACMS membership. In addition, most respondents practiced in a private/community setting, which may have led to bias, as academic centers may manage rare malignancies more commonly and/or have increased access to immunostains and multispecialty care. Future registries for rare skin malignancies will hopefully be developed that will allow for further consensus on standardized margins. Additional studies on the average number of stages required to clear these less common tumors also are warranted.

Mohs micrographic surgery (MMS) is most commonly used for the surgical management of squamous cell carcinomas (SCCs) and basal cell carcinomas (BCCs) in high-risk locations. The ability for 100% margin evaluation with MMS also has shown lower recurrence rates compared with wide local excision for less common and/or more aggressive tumors. However, there is a lack of standardization on initial and subsequent margin size when treating these less common skin tumors, such as dermatofibrosarcoma protuberans (DFSP), atypical fibroxanthoma (AFX), and sebaceous carcinoma.

Because Mohs surgeons must balance normal tissue preservation with the importance of tumor clearance in the context of comprehensive margin control, we aimed to assess the practice patterns of Mohs surgeons regarding margin size for these unique tumors. The average margin size for each Mohs layer has been reported to be 1 to 3 mm for BCC compared with 3 to 6 mm or larger for other skin cancers, such as melanoma in situ (MIS).^1-3 We hypothesized that the initial margin size would vary among surgeons and likely be greater for more aggressive and rarer malignancies as well as for lesions on the trunk and extremities.

Methods

A descriptive survey was created using SurveyMonkey and distributed to members of the American College of Mohs Surgery (ACMS). Survey participants and their responses were anonymous. Demographic information on survey participants was collected in addition to initial and subsequent MMS margin size for DFSP, AFX, MIS, invasive melanoma, sebaceous carcinoma, microcystic adnexal carcinoma (MAC), poorly differentiated SCC, Merkel cell carcinoma, extramammary Paget disease, leiomyosarcoma, and endocrine mucin-producing sweat gland carcinoma. Survey participants were asked to choose from a range of margin sizes: 1 to 3 mm, 4 to 6 mm, 7 to 9 mm, and greater than 9 mm. This study was approved by the University of Texas Southwest Medical Center (Dallas, Texas) institutional review board.

Results

Eighty-seven respondents from the ACMS listserve completed the survey (response rate <10%). Of these, 58 respondents (66.7%) reported practicing for more than 5 years, and 58 (66.7%) were male. Practice setting was primarily private/community (71.3% [62/87]), and survey respondents were located across the United States. More than 50% of survey respondents treated the following tumors on the head and neck in their respective practices: DFSP (80.9% [55/68]), AFX (95.6% [65/68]), MIS (67.7% [46/68]), sebaceous carcinoma (92.7% [63/68]), MAC (83.8% [57/68]), poorly differentiated SCC (97.1% [66/68]), and endocrine mucin-producing sweat gland carcinoma (51.5% [35/68]). More than 50% of survey respondents treated the following tumors on the trunk and extremities: DFSP (90.3% [47/52]), AFX (86.4% [45/52]), MIS (55.8% [29/52]), sebaceous carcinoma (80.8% [42/52]), MAC (73.1% [38/52]), poorly differentiated SCC (94.2% [49/52]), and extramammary Paget disease (53.9% [28/52]). Invasive melanoma, Merkel cell carcinoma, and leiomyosarcoma were overall less commonly treated.

In general, respondent Mohs surgeons were more likely to take larger initial and subsequent margins for tumors treated on the trunk and extremities compared with the head and neck (Table). In addition, initial margin size often was larger than the 1- to 3-mm margin commonly used in Mohs surgery for BCCs and less aggressive SCCs (Table). A larger initial margin size (>9 mm) and subsequent margin size (4–6 mm) was more commonly reported for certain tumors known to be more aggressive and/or have extensive subclinical extension, such as DFSP and invasive melanoma. Of note, most respondents performed 4- to 6-mm margins (37/67 [55.2%]) for poorly differentiated SCC. Overall, there was a high range of margin size variability among Mohs surgeons for these unique and/or more aggressive skin tumors.

Comment

Given that no guidelines exist on margins with MMS for less commonly treated skin tumors, this study helps give Mohs surgeons perspective on current practice patterns for both initial and subsequent Mohs margin sizes. High margin-size variability among Mohs surgeons is expected, as surgeons also need to account for high-risk features of the tumor or specific locations where tissue sparing is critical. Overall, Mohs surgeons are more likely to take larger initial margins for these less common skin tumors compared with BCCs or SCCs. Initial margin size was consistently larger on the trunk and extremities where tissue sparing often is less critical.

Our survey was limited by a small sample size and incomplete response of the ACMS membership. In addition, most respondents practiced in a private/community setting, which may have led to bias, as academic centers may manage rare malignancies more commonly and/or have increased access to immunostains and multispecialty care. Future registries for rare skin malignancies will hopefully be developed that will allow for further consensus on standardized margins. Additional studies on the average number of stages required to clear these less common tumors also are warranted.

Mohs micrographic surgery (MMS) is most commonly used for the surgical management of squamous cell carcinomas (SCCs) and basal cell carcinomas (BCCs) in high-risk locations. The ability for 100% margin evaluation with MMS also has shown lower recurrence rates compared with wide local excision for less common and/or more aggressive tumors. However, there is a lack of standardization on initial and subsequent margin size when treating these less common skin tumors, such as dermatofibrosarcoma protuberans (DFSP), atypical fibroxanthoma (AFX), and sebaceous carcinoma.

Because Mohs surgeons must balance normal tissue preservation with the importance of tumor clearance in the context of comprehensive margin control, we aimed to assess the practice patterns of Mohs surgeons regarding margin size for these unique tumors. The average margin size for each Mohs layer has been reported to be 1 to 3 mm for BCC compared with 3 to 6 mm or larger for other skin cancers, such as melanoma in situ (MIS).^1-3 We hypothesized that the initial margin size would vary among surgeons and likely be greater for more aggressive and rarer malignancies as well as for lesions on the trunk and extremities.

Methods

A descriptive survey was created using SurveyMonkey and distributed to members of the American College of Mohs Surgery (ACMS). Survey participants and their responses were anonymous. Demographic information on survey participants was collected in addition to initial and subsequent MMS margin size for DFSP, AFX, MIS, invasive melanoma, sebaceous carcinoma, microcystic adnexal carcinoma (MAC), poorly differentiated SCC, Merkel cell carcinoma, extramammary Paget disease, leiomyosarcoma, and endocrine mucin-producing sweat gland carcinoma. Survey participants were asked to choose from a range of margin sizes: 1 to 3 mm, 4 to 6 mm, 7 to 9 mm, and greater than 9 mm. This study was approved by the University of Texas Southwest Medical Center (Dallas, Texas) institutional review board.

Results

Eighty-seven respondents from the ACMS listserve completed the survey (response rate <10%). Of these, 58 respondents (66.7%) reported practicing for more than 5 years, and 58 (66.7%) were male. Practice setting was primarily private/community (71.3% [62/87]), and survey respondents were located across the United States. More than 50% of survey respondents treated the following tumors on the head and neck in their respective practices: DFSP (80.9% [55/68]), AFX (95.6% [65/68]), MIS (67.7% [46/68]), sebaceous carcinoma (92.7% [63/68]), MAC (83.8% [57/68]), poorly differentiated SCC (97.1% [66/68]), and endocrine mucin-producing sweat gland carcinoma (51.5% [35/68]). More than 50% of survey respondents treated the following tumors on the trunk and extremities: DFSP (90.3% [47/52]), AFX (86.4% [45/52]), MIS (55.8% [29/52]), sebaceous carcinoma (80.8% [42/52]), MAC (73.1% [38/52]), poorly differentiated SCC (94.2% [49/52]), and extramammary Paget disease (53.9% [28/52]). Invasive melanoma, Merkel cell carcinoma, and leiomyosarcoma were overall less commonly treated.

In general, respondent Mohs surgeons were more likely to take larger initial and subsequent margins for tumors treated on the trunk and extremities compared with the head and neck (Table). In addition, initial margin size often was larger than the 1- to 3-mm margin commonly used in Mohs surgery for BCCs and less aggressive SCCs (Table). A larger initial margin size (>9 mm) and subsequent margin size (4–6 mm) was more commonly reported for certain tumors known to be more aggressive and/or have extensive subclinical extension, such as DFSP and invasive melanoma. Of note, most respondents performed 4- to 6-mm margins (37/67 [55.2%]) for poorly differentiated SCC. Overall, there was a high range of margin size variability among Mohs surgeons for these unique and/or more aggressive skin tumors.

Comment

Given that no guidelines exist on margins with MMS for less commonly treated skin tumors, this study helps give Mohs surgeons perspective on current practice patterns for both initial and subsequent Mohs margin sizes. High margin-size variability among Mohs surgeons is expected, as surgeons also need to account for high-risk features of the tumor or specific locations where tissue sparing is critical. Overall, Mohs surgeons are more likely to take larger initial margins for these less common skin tumors compared with BCCs or SCCs. Initial margin size was consistently larger on the trunk and extremities where tissue sparing often is less critical.

Our survey was limited by a small sample size and incomplete response of the ACMS membership. In addition, most respondents practiced in a private/community setting, which may have led to bias, as academic centers may manage rare malignancies more commonly and/or have increased access to immunostains and multispecialty care. Future registries for rare skin malignancies will hopefully be developed that will allow for further consensus on standardized margins. Additional studies on the average number of stages required to clear these less common tumors also are warranted.

To the Editor:

Brodalumab, a monoclonal antibody targeting IL-17RA, was approved by the US Food and Drug Administration (FDA) in 2017 for the treatment of moderate to severe chronic plaque psoriasis. The drug is the only biologic agent available for the treatment of psoriasis for which a psoriasis area severity index score of 100 is a primary end point.^1,2 Brodalumab is associated with an FDA boxed warning due to an increased risk for suicidal ideation and behavior (SIB), including completed suicides, during clinical trials.

We sought to characterize national utilization of this effective yet underutilized drug among Medicare beneficiaries by surveying the Medicare Part D Prescriber dataset.³ We tabulated brodalumab utilization statistics and characteristics of high-volume prescribers who had 11 or more annual claims for brodalumab.

Despite its associated boxed warning, the number of Medicare D claims for brodalumab increased by 1756 from 2017 to 2019, surpassing $7 million in costs by 2019. The number of beneficiaries also increased from 11 to 292—a 415.2% annual increase in beneficiaries for whom brodalumab was prescribed (Table 1).

In addition, states in the West and South had the highest utilization rates of brodalumab in 2019. There also was an increasing trend toward high-volume prescribers of brodalumab, with private practice clinicians constituting the majority (Table 2).

There was a substantial increase in advanced practice providers including nurse practitioners and physician assistants who were brodalumab prescribers. Although this trend might promote greater access to brodalumab, it is vital to ensure that advanced practice providers receive targeted training to properly understand the complexities of treatment with brodalumab.

Although the utilization of brodalumab has increased since 2017 (P<.001), it is still underutilized compared to the other IL-17 inhibitors secukinumab and ixekizumab. Secukinumab was FDA approved for the treatment of moderate to severe plaque psoriasis in 2015, followed by ixekizumab in 2016.⁴

According to the Medicare Part D database, both secukinumab and ixekizumab had a higher number of total claims and prescribers compared to brodalumab in the years of their debut.³ In 2015, there were 3593 claims for and 862 prescribers of secukinumab; in 2016, there were 1731 claims for and 681 prescribers of ixekizumab. In contrast, there were only 29 claims for and 11 prescribers of brodalumab in 2017, the year that the drug was approved by the FDA. During the same 3-year period, secukinumab and ixekizumab had a substantially greater number of claims—totals of 176,823 and 55,289, respectively—than brodalumab. The higher number of claims for secukinumab and ixekizumab compared to brodalumab may reflect clinicians’ increasing confidence in prescribing those drugs, given their long-term safety and efficacy. In addition, secukinumab and ixekizumab do not require completion of a Risk Evaluation and Mitigation Strategy (REMS) program, which makes them more readily prescribable.³

Overall, most experts agree that there is no increase in the risk for suicide associated with brodalumab compared to the general population. A 2-year pharmacovigilance report on brodalumab supports the safety of this drug.⁵ All participants who completed suicide during the clinical trials harbored an underlying psychiatric disorder or stressor(s).⁶

Although causation between brodalumab and SIB has not been demonstrated, it remains imperative that prescribers diligently assess patients’ risk of SIB and subsequently their access to appropriate psychiatric services as a precaution, if necessary. This is particularly important for private practice prescribers, who constitute the majority of Medicare D brodalumab claims, because they must ensure collaboration with a multidisciplinary team involving mental health providers. Lastly, considering that the highest number of brodalumab Medicare D claims were in western and southern states, it is critical to note that those 2 regions also harbor comparatively fewer mental health facilities that accept Medicare than other regions of the country.⁷ Prescribers in western and southern states must be mindful of mental health coverage limitations when treating psoriasis patients with brodalumab.

The increase in the number of claims, beneficiaries, and prescribers of brodalumab during its first 3 years of availability might be attributed to its efficacy and safety. On the other hand, the boxed warning and REMS associated with brodalumab might have led to underutilization of this drug compared to other IL-17 inhibitors.

Our analysis is limited by its representative restriction to Medicare patients. There also are limited data on brodalumab given its novelty. Individual attributes of prescribers with fewer than 11 annual claims for brodalumab could not be obtained because of dataset regulations; however, aggregated utilization statistics provide an indication of brodalumab prescribing patterns among all providers. Furthermore, during this analysis, data on the Medicare D database were limited to 2013 through 2020. Studies are needed to determine prescribing patterns of brodalumab since this study period.

To the Editor:

Brodalumab, a monoclonal antibody targeting IL-17RA, was approved by the US Food and Drug Administration (FDA) in 2017 for the treatment of moderate to severe chronic plaque psoriasis. The drug is the only biologic agent available for the treatment of psoriasis for which a psoriasis area severity index score of 100 is a primary end point.^1,2 Brodalumab is associated with an FDA boxed warning due to an increased risk for suicidal ideation and behavior (SIB), including completed suicides, during clinical trials.

We sought to characterize national utilization of this effective yet underutilized drug among Medicare beneficiaries by surveying the Medicare Part D Prescriber dataset.³ We tabulated brodalumab utilization statistics and characteristics of high-volume prescribers who had 11 or more annual claims for brodalumab.

Despite its associated boxed warning, the number of Medicare D claims for brodalumab increased by 1756 from 2017 to 2019, surpassing $7 million in costs by 2019. The number of beneficiaries also increased from 11 to 292—a 415.2% annual increase in beneficiaries for whom brodalumab was prescribed (Table 1).

In addition, states in the West and South had the highest utilization rates of brodalumab in 2019. There also was an increasing trend toward high-volume prescribers of brodalumab, with private practice clinicians constituting the majority (Table 2).

There was a substantial increase in advanced practice providers including nurse practitioners and physician assistants who were brodalumab prescribers. Although this trend might promote greater access to brodalumab, it is vital to ensure that advanced practice providers receive targeted training to properly understand the complexities of treatment with brodalumab.

Although the utilization of brodalumab has increased since 2017 (P<.001), it is still underutilized compared to the other IL-17 inhibitors secukinumab and ixekizumab. Secukinumab was FDA approved for the treatment of moderate to severe plaque psoriasis in 2015, followed by ixekizumab in 2016.⁴

According to the Medicare Part D database, both secukinumab and ixekizumab had a higher number of total claims and prescribers compared to brodalumab in the years of their debut.³ In 2015, there were 3593 claims for and 862 prescribers of secukinumab; in 2016, there were 1731 claims for and 681 prescribers of ixekizumab. In contrast, there were only 29 claims for and 11 prescribers of brodalumab in 2017, the year that the drug was approved by the FDA. During the same 3-year period, secukinumab and ixekizumab had a substantially greater number of claims—totals of 176,823 and 55,289, respectively—than brodalumab. The higher number of claims for secukinumab and ixekizumab compared to brodalumab may reflect clinicians’ increasing confidence in prescribing those drugs, given their long-term safety and efficacy. In addition, secukinumab and ixekizumab do not require completion of a Risk Evaluation and Mitigation Strategy (REMS) program, which makes them more readily prescribable.³

Overall, most experts agree that there is no increase in the risk for suicide associated with brodalumab compared to the general population. A 2-year pharmacovigilance report on brodalumab supports the safety of this drug.⁵ All participants who completed suicide during the clinical trials harbored an underlying psychiatric disorder or stressor(s).⁶

Although causation between brodalumab and SIB has not been demonstrated, it remains imperative that prescribers diligently assess patients’ risk of SIB and subsequently their access to appropriate psychiatric services as a precaution, if necessary. This is particularly important for private practice prescribers, who constitute the majority of Medicare D brodalumab claims, because they must ensure collaboration with a multidisciplinary team involving mental health providers. Lastly, considering that the highest number of brodalumab Medicare D claims were in western and southern states, it is critical to note that those 2 regions also harbor comparatively fewer mental health facilities that accept Medicare than other regions of the country.⁷ Prescribers in western and southern states must be mindful of mental health coverage limitations when treating psoriasis patients with brodalumab.

The increase in the number of claims, beneficiaries, and prescribers of brodalumab during its first 3 years of availability might be attributed to its efficacy and safety. On the other hand, the boxed warning and REMS associated with brodalumab might have led to underutilization of this drug compared to other IL-17 inhibitors.

Our analysis is limited by its representative restriction to Medicare patients. There also are limited data on brodalumab given its novelty. Individual attributes of prescribers with fewer than 11 annual claims for brodalumab could not be obtained because of dataset regulations; however, aggregated utilization statistics provide an indication of brodalumab prescribing patterns among all providers. Furthermore, during this analysis, data on the Medicare D database were limited to 2013 through 2020. Studies are needed to determine prescribing patterns of brodalumab since this study period.

To the Editor:

Brodalumab, a monoclonal antibody targeting IL-17RA, was approved by the US Food and Drug Administration (FDA) in 2017 for the treatment of moderate to severe chronic plaque psoriasis. The drug is the only biologic agent available for the treatment of psoriasis for which a psoriasis area severity index score of 100 is a primary end point.^1,2 Brodalumab is associated with an FDA boxed warning due to an increased risk for suicidal ideation and behavior (SIB), including completed suicides, during clinical trials.

We sought to characterize national utilization of this effective yet underutilized drug among Medicare beneficiaries by surveying the Medicare Part D Prescriber dataset.³ We tabulated brodalumab utilization statistics and characteristics of high-volume prescribers who had 11 or more annual claims for brodalumab.

Despite its associated boxed warning, the number of Medicare D claims for brodalumab increased by 1756 from 2017 to 2019, surpassing $7 million in costs by 2019. The number of beneficiaries also increased from 11 to 292—a 415.2% annual increase in beneficiaries for whom brodalumab was prescribed (Table 1).

In addition, states in the West and South had the highest utilization rates of brodalumab in 2019. There also was an increasing trend toward high-volume prescribers of brodalumab, with private practice clinicians constituting the majority (Table 2).

There was a substantial increase in advanced practice providers including nurse practitioners and physician assistants who were brodalumab prescribers. Although this trend might promote greater access to brodalumab, it is vital to ensure that advanced practice providers receive targeted training to properly understand the complexities of treatment with brodalumab.

Although the utilization of brodalumab has increased since 2017 (P<.001), it is still underutilized compared to the other IL-17 inhibitors secukinumab and ixekizumab. Secukinumab was FDA approved for the treatment of moderate to severe plaque psoriasis in 2015, followed by ixekizumab in 2016.⁴

According to the Medicare Part D database, both secukinumab and ixekizumab had a higher number of total claims and prescribers compared to brodalumab in the years of their debut.³ In 2015, there were 3593 claims for and 862 prescribers of secukinumab; in 2016, there were 1731 claims for and 681 prescribers of ixekizumab. In contrast, there were only 29 claims for and 11 prescribers of brodalumab in 2017, the year that the drug was approved by the FDA. During the same 3-year period, secukinumab and ixekizumab had a substantially greater number of claims—totals of 176,823 and 55,289, respectively—than brodalumab. The higher number of claims for secukinumab and ixekizumab compared to brodalumab may reflect clinicians’ increasing confidence in prescribing those drugs, given their long-term safety and efficacy. In addition, secukinumab and ixekizumab do not require completion of a Risk Evaluation and Mitigation Strategy (REMS) program, which makes them more readily prescribable.³

Overall, most experts agree that there is no increase in the risk for suicide associated with brodalumab compared to the general population. A 2-year pharmacovigilance report on brodalumab supports the safety of this drug.⁵ All participants who completed suicide during the clinical trials harbored an underlying psychiatric disorder or stressor(s).⁶

Although causation between brodalumab and SIB has not been demonstrated, it remains imperative that prescribers diligently assess patients’ risk of SIB and subsequently their access to appropriate psychiatric services as a precaution, if necessary. This is particularly important for private practice prescribers, who constitute the majority of Medicare D brodalumab claims, because they must ensure collaboration with a multidisciplinary team involving mental health providers. Lastly, considering that the highest number of brodalumab Medicare D claims were in western and southern states, it is critical to note that those 2 regions also harbor comparatively fewer mental health facilities that accept Medicare than other regions of the country.⁷ Prescribers in western and southern states must be mindful of mental health coverage limitations when treating psoriasis patients with brodalumab.

The increase in the number of claims, beneficiaries, and prescribers of brodalumab during its first 3 years of availability might be attributed to its efficacy and safety. On the other hand, the boxed warning and REMS associated with brodalumab might have led to underutilization of this drug compared to other IL-17 inhibitors.

Our analysis is limited by its representative restriction to Medicare patients. There also are limited data on brodalumab given its novelty. Individual attributes of prescribers with fewer than 11 annual claims for brodalumab could not be obtained because of dataset regulations; however, aggregated utilization statistics provide an indication of brodalumab prescribing patterns among all providers. Furthermore, during this analysis, data on the Medicare D database were limited to 2013 through 2020. Studies are needed to determine prescribing patterns of brodalumab since this study period.

Approximately 4.5% of adults within the United States identify as members of the lesbian, gay, bisexual, transgender (LGBT) community.¹ This is an umbrella term inclusive of all individuals identifying as nonheterosexual or noncisgender. Although the LGBT community has increasingly become more recognized and accepted by society over time, health care disparities persist and have been well documented in the literature.^2-4 Dermatologists have the potential to greatly impact LGBT health, as many health concerns in this population are cutaneous, such as sun-protection behaviors, side effects of gender-affirming hormone therapy and gender-affirming procedures, and cutaneous manifestations of sexually transmitted infections.^5-7

An education gap has been demonstrated in both medical students and resident physicians regarding LGBT health and cultural competency. In a large-scale, multi-institutional survey study published in 2015, approximately two-thirds of medical students rated their schools’ LGBT curriculum as fair, poor, or very poor.⁸ Additional studies have echoed these results and have demonstrated not only the need but the desire for additional training on LGBT issues in medical school.^9-11 The Association of American Medical Colleges has begun implementing curricular and institutional changes to fulfill this need.^12,13

The LGBT education gap has been shown to extend into residency training. Multiple studies performed within a variety of medical specialties have demonstrated that resident physicians receive insufficient training in LGBT health issues, lack comfort in caring for LGBT patients, and would benefit from dedicated curricula on these topics.^14-18 Currently, the 2022 Accreditation Council for Graduate Medical Education (ACGME) guidelines related to LGBT health are minimal and nonspecific.¹⁹

Ensuring that dermatology trainees are well equipped to manage these issues while providing culturally competent care to LGBT patients is paramount. However, research suggests that dedicated training on these topics likely is insufficient. A survey study of dermatology residency program directors (N=90) revealed that although 81% (72/89) viewed training in LGBT health as either very important or somewhat important, 46% (41/90) of programs did not dedicate any time to this content and 37% (33/90) only dedicated 1 to 2 hours per year.²⁰

To further explore this potential education gap, we surveyed dermatology residents directly to better understand LGBT education within residency training, resident preparedness to care for LGBT patients, and outness/discrimination of LGBT-identifying residents. We believe this study should drive future research on the development and implementation of LGBT-specific curricula in dermatology training programs.

Methods

A cross-sectional survey study of dermatology residents in the United States was conducted. The study was deemed exempt from review by The Ohio State University (Columbus, Ohio) institutional review board. Survey responses were collected from October 7, 2020, to November 13, 2020. Qualtrics software was used to create the 20-question survey, which included a combination of categorical, dichotomous, and optional free-text questions related to patient demographics, LGBT training experiences, perceived areas of curriculum improvement, comfort level managing LGBT health issues, and personal experiences. Some questions were adapted from prior surveys.^15,21 Validated survey tools used included the 2020 US Census to collect information regarding race and ethnicity, the Mohr and Fassinger Outness Inventory to measure outness regarding sexual orientation, and select questions from the 2020 Association of American Medical Colleges Medical School Graduation Questionnaire regarding discrimination.^22-24

The survey was distributed to current allopathic and osteopathic dermatology residents by a variety of methods, including emails to program director and program coordinator listserves. The survey also was posted in the American Academy of Dermatology Expert Resource Group on LGBTQ Health October 2020 newsletter, as well as dermatology social media groups, including a messaging forum limited to dermatology residents, a Facebook group open to dermatologists and dermatology residents, and the Facebook group of the Gay and Lesbian Dermatology Association. Current dermatology residents, including those in combined dermatology and internal medicine programs, were included. Individuals who had been accepted to dermatology training programs but had not yet started were excluded. A follow-up email was sent to the program director listserve approximately 3 weeks after the initial distribution.

Statistical Analysis—The data were analyzed in Qualtrics and Microsoft Excel using descriptive statistics. Stata software (Stata 15.1, StataCorp) was used to perform a Kruskal-Wallis equality-of-populations rank test to compare the means of education level and feelings of preparedness.

Results

Demographics of Respondents—A total of 126 responses were recorded, 12 of which were blank and were removed from the database. A total of 114 dermatology residents’ responses were collected in Qualtrics and analyzed; 91 completed the entire survey (an 80% completion rate). Based on the 2020-2021 ACGME data listing, there were 1612 dermatology residents in the United States, which is an estimated response rate of 7% (114/1612).²⁵ The eTable outlines the demographics of the survey respondents. Most were cisgender females (60%), followed by cisgender males (35%); the remainder preferred not to answer. Regarding sexual orientation, 77% identified as straight or heterosexual; 17% as gay, lesbian, or homosexual; 1% as queer; and 1% as bisexual. The training programs were in 26 states, the majority of which were in the Midwest (34%) and in urban settings (69%). A wide range of postgraduate levels and residency sizes were represented in the survey.

LGBT Education—Fifty-one percent of respondents reported that their programs offer 1 hour or less of LGBT-related curricula per year; 34% reported no time dedicated to this topic. A small portion of residents (5%) reported 10 or more hours of LGBT education per year. Residents also were asked the average number of hours of LGBT education they thought they should receive. The discrepancy between these measures can be visualized in Figure 1. The median hours of education received was 1 hour (IQR, 0–4 hours), whereas the median hours of education desired was 4 hours (IQR, 2–5 hours). The most common and most helpful methods of education reported were clinical experiences with faculty or patients and live lectures.

Overall, 45% of survey respondents felt that LGBT topics were covered poorly or not at all in dermatology residency, whereas 26% thought the coverage was good or excellent. The topics that residents were most likely to report receiving good or excellent coverage were dermatologic manifestations of HIV/AIDS (70%) and sexually transmitted diseases in LGBT patients (48%). The topics that were most likely to be reported as not taught or poorly taught included dermatologic concerns associated with puberty blockers (71%), body image (58%), dermatologic concerns associated with gender-affirming surgery (55%), skin cancer risk (53%), taking an LGBT-oriented history and physical examination (52%), and effects of gender-affirming hormone therapy on the skin (50%). A detailed breakdown of coverage level by topic can be found in Figure 2.

Preparedness to Care for LGBT Patients—Only 68% of survey respondents agreed or strongly agreed that they feel comfortable treating LGBT patients. Furthermore, 49% of dermatology residents reported that they feel not at all prepared or insufficiently prepared to provide care to LGBT individuals (Figure 2), and 60% believed that LGBT training needed to be improved at their residency programs.

There was a significant association between reported level of education and feelings of preparedness. A high ranking of provided education was associated with higher levels of feeling prepared to care for LGBT patients (Kruskal-Wallis rank test, P<.001).

Discrimination/Outness—Approximately one-fourth (24%; 4/17) of nonheterosexual dermatology residents reported that they had been subjected to offensive remarks about their sexual orientation in the workplace. One respondent commented that they were less “out” at their residency program due to fear of discrimination. Nearly one-third of the overall group of dermatology residents surveyed (29%; 27/92) reported that they had witnessed inappropriate or discriminatory comments about LGBT persons made by employees or staff at their programs. Most residents surveyed (96%; 88/92) agreed or strongly agreed that they feel comfortable working alongside LGBT physicians.

There were 18 nonheterosexual dermatologyresidents who completed the Mohr and Fassinger Outness Inventory.²³ In general, respondents reported that they were more “out” with friends and family than work peers and were least “out” with work supervisors and strangers.

Comment

Dermatology Residents Desire More Time on LGBT Health—This cross-sectional survey study explored dermatology residents’ educational experiences with LGBT health during residency training. Similar studies have been performed in other specialties, including a study from 2019 surveying emergency medicine residents that demonstrated residents find caring for LGBT patients more challenging.¹⁵ Another 2019 study surveying psychiatry residents found that 42.4% (N=99) reported no coverage of LGBT topics.¹⁸ Our study is unique in that it surveyed dermatology residents directly regarding this topic. Although most dermatology program directors view LGBT dermatologic health as an important topic, a prior study revealed that many programs are lacking dedicated LGBT educational experiences. The most common barriers reported were insufficient time in the didactic schedule and lack of experienced faculty.²⁰

Our study revealed that dermatology residents overall tend to agree with residents from other specialties and dermatology program directors. Most of the dermatology residents surveyed reported desiring more time per year spent on LGBT health education than they receive, and 60% expressed that LGBT educational experiences need to be improved at their residency programs. Education on and subsequent comfort level with LGBT health issues varied by subtopic, with most residents feeling comfortable dealing with dermatologic manifestations of HIV/AIDS and other sexually transmitted diseases and less comfortable with topics such as puberty blockers, gender-affirming surgery and hormone therapy, body image, and skin cancer risk.

Overall, LGBT health training is viewed as important and in need of improvement by both program directors and residents, yet implementation lags at many programs. A small proportion of the represented programs are excelling in this area—just over 5% of respondents reported receiving 10 or more hours of LGBT-relevant education per year, and approximately 26% of residents felt that LGBT coverage was good or excellent at their programs. Our study showed a clear relationship between feelings of preparedness and education level. The lack of LGBT education at some dermatology residency programs translated into a large portion of dermatology residents feeling ill equipped to care for LGBT patients after graduation—nearly 50% of those surveyed reported feeling insufficiently prepared to care for the LGBT community.

Discrimination in Residency Programs—Dermatology residency programs also are not free from sexual orientation–related and gender identity–related workplace discrimination. Although 96% of dermatology residents reported that they feel comfortable working alongside LGBT physicians, 24% of nonheterosexual respondents stated they had been subjected to offensive remarks about their sexual orientation, and 29% of the overall group of dermatology residents had witnessed discriminatory comments to LGBT individuals at their programs. In addition, some nonheterosexual dermatology residents reported being less “out” with their workplace supervisors and strangers, such as patients, than with their family and friends, and 50% of this group reported that their sexual identity was not openly discussed with their workplace supervisors. It has been demonstrated that individuals are more likely to “come out” in perceived LGBT-friendly workplace environments and that being “out” positively impacts psychological health because of the effects of perceived social support and self-coherence.^26,27

Study Strengths and Limitations—Strengths of this study include the modest sample size of dermatology residents that participated, high completion rate, and the anonymity of the survey. Limitations include the risk of sampling bias by posting the survey on LGBT-specific groups. The survey also took place in the fall, so the results may not accurately reflect programs that cover this material later in the academic year. Lastly, not all survey questions were validated.

Implementing Change in Residency Programs—Although the results of this study exposed the need for increasing LGBT education in dermatology residency, they do not provide guidelines for the best strategy to begin implementing change. A study from 2020 provides some guidance for incorporating LGBT health training into dermatology residency programs through a combination of curricular modifications and climate optimization.²⁸ Additional future research should focus on the best methods for preparing dermatology residents to care for this population. In this study, residents reported that the most effective teaching methods were real encounters with LGBT patients or faculty educated on LGBT health as well as live lectures from experts. There also appeared to be a correlation between hours spent on LGBT health, including various subtopics, and residents’ perceived preparedness in these areas. Potential actionable items include clarifying the ACGME guidelines on LGBT health topics; increasing the sexual and gender diversity of the faculty, staff, residents, and patients; and dedicating additional didactic and clinical time to LGBT topics and experiences.

Conclusion

This survey study of dermatology residents regarding LGBT learning experiences in residency training provided evidence that dermatology residents as a whole are not adequately taught LGBT health topics and therefore feel unprepared to take care of this patient population. Additionally, most residents desire improvement of LGBT health education and training. Further studies focusing on the best methods for implementing LGBT-specific curricula are needed.

Approximately 4.5% of adults within the United States identify as members of the lesbian, gay, bisexual, transgender (LGBT) community.¹ This is an umbrella term inclusive of all individuals identifying as nonheterosexual or noncisgender. Although the LGBT community has increasingly become more recognized and accepted by society over time, health care disparities persist and have been well documented in the literature.^2-4 Dermatologists have the potential to greatly impact LGBT health, as many health concerns in this population are cutaneous, such as sun-protection behaviors, side effects of gender-affirming hormone therapy and gender-affirming procedures, and cutaneous manifestations of sexually transmitted infections.^5-7

An education gap has been demonstrated in both medical students and resident physicians regarding LGBT health and cultural competency. In a large-scale, multi-institutional survey study published in 2015, approximately two-thirds of medical students rated their schools’ LGBT curriculum as fair, poor, or very poor.⁸ Additional studies have echoed these results and have demonstrated not only the need but the desire for additional training on LGBT issues in medical school.^9-11 The Association of American Medical Colleges has begun implementing curricular and institutional changes to fulfill this need.^12,13

The LGBT education gap has been shown to extend into residency training. Multiple studies performed within a variety of medical specialties have demonstrated that resident physicians receive insufficient training in LGBT health issues, lack comfort in caring for LGBT patients, and would benefit from dedicated curricula on these topics.^14-18 Currently, the 2022 Accreditation Council for Graduate Medical Education (ACGME) guidelines related to LGBT health are minimal and nonspecific.¹⁹

Ensuring that dermatology trainees are well equipped to manage these issues while providing culturally competent care to LGBT patients is paramount. However, research suggests that dedicated training on these topics likely is insufficient. A survey study of dermatology residency program directors (N=90) revealed that although 81% (72/89) viewed training in LGBT health as either very important or somewhat important, 46% (41/90) of programs did not dedicate any time to this content and 37% (33/90) only dedicated 1 to 2 hours per year.²⁰

To further explore this potential education gap, we surveyed dermatology residents directly to better understand LGBT education within residency training, resident preparedness to care for LGBT patients, and outness/discrimination of LGBT-identifying residents. We believe this study should drive future research on the development and implementation of LGBT-specific curricula in dermatology training programs.

Methods

A cross-sectional survey study of dermatology residents in the United States was conducted. The study was deemed exempt from review by The Ohio State University (Columbus, Ohio) institutional review board. Survey responses were collected from October 7, 2020, to November 13, 2020. Qualtrics software was used to create the 20-question survey, which included a combination of categorical, dichotomous, and optional free-text questions related to patient demographics, LGBT training experiences, perceived areas of curriculum improvement, comfort level managing LGBT health issues, and personal experiences. Some questions were adapted from prior surveys.^15,21 Validated survey tools used included the 2020 US Census to collect information regarding race and ethnicity, the Mohr and Fassinger Outness Inventory to measure outness regarding sexual orientation, and select questions from the 2020 Association of American Medical Colleges Medical School Graduation Questionnaire regarding discrimination.^22-24

The survey was distributed to current allopathic and osteopathic dermatology residents by a variety of methods, including emails to program director and program coordinator listserves. The survey also was posted in the American Academy of Dermatology Expert Resource Group on LGBTQ Health October 2020 newsletter, as well as dermatology social media groups, including a messaging forum limited to dermatology residents, a Facebook group open to dermatologists and dermatology residents, and the Facebook group of the Gay and Lesbian Dermatology Association. Current dermatology residents, including those in combined dermatology and internal medicine programs, were included. Individuals who had been accepted to dermatology training programs but had not yet started were excluded. A follow-up email was sent to the program director listserve approximately 3 weeks after the initial distribution.

Statistical Analysis—The data were analyzed in Qualtrics and Microsoft Excel using descriptive statistics. Stata software (Stata 15.1, StataCorp) was used to perform a Kruskal-Wallis equality-of-populations rank test to compare the means of education level and feelings of preparedness.

Results

Demographics of Respondents—A total of 126 responses were recorded, 12 of which were blank and were removed from the database. A total of 114 dermatology residents’ responses were collected in Qualtrics and analyzed; 91 completed the entire survey (an 80% completion rate). Based on the 2020-2021 ACGME data listing, there were 1612 dermatology residents in the United States, which is an estimated response rate of 7% (114/1612).²⁵ The eTable outlines the demographics of the survey respondents. Most were cisgender females (60%), followed by cisgender males (35%); the remainder preferred not to answer. Regarding sexual orientation, 77% identified as straight or heterosexual; 17% as gay, lesbian, or homosexual; 1% as queer; and 1% as bisexual. The training programs were in 26 states, the majority of which were in the Midwest (34%) and in urban settings (69%). A wide range of postgraduate levels and residency sizes were represented in the survey.

LGBT Education—Fifty-one percent of respondents reported that their programs offer 1 hour or less of LGBT-related curricula per year; 34% reported no time dedicated to this topic. A small portion of residents (5%) reported 10 or more hours of LGBT education per year. Residents also were asked the average number of hours of LGBT education they thought they should receive. The discrepancy between these measures can be visualized in Figure 1. The median hours of education received was 1 hour (IQR, 0–4 hours), whereas the median hours of education desired was 4 hours (IQR, 2–5 hours). The most common and most helpful methods of education reported were clinical experiences with faculty or patients and live lectures.

Overall, 45% of survey respondents felt that LGBT topics were covered poorly or not at all in dermatology residency, whereas 26% thought the coverage was good or excellent. The topics that residents were most likely to report receiving good or excellent coverage were dermatologic manifestations of HIV/AIDS (70%) and sexually transmitted diseases in LGBT patients (48%). The topics that were most likely to be reported as not taught or poorly taught included dermatologic concerns associated with puberty blockers (71%), body image (58%), dermatologic concerns associated with gender-affirming surgery (55%), skin cancer risk (53%), taking an LGBT-oriented history and physical examination (52%), and effects of gender-affirming hormone therapy on the skin (50%). A detailed breakdown of coverage level by topic can be found in Figure 2.

Preparedness to Care for LGBT Patients—Only 68% of survey respondents agreed or strongly agreed that they feel comfortable treating LGBT patients. Furthermore, 49% of dermatology residents reported that they feel not at all prepared or insufficiently prepared to provide care to LGBT individuals (Figure 2), and 60% believed that LGBT training needed to be improved at their residency programs.

There was a significant association between reported level of education and feelings of preparedness. A high ranking of provided education was associated with higher levels of feeling prepared to care for LGBT patients (Kruskal-Wallis rank test, P<.001).

Discrimination/Outness—Approximately one-fourth (24%; 4/17) of nonheterosexual dermatology residents reported that they had been subjected to offensive remarks about their sexual orientation in the workplace. One respondent commented that they were less “out” at their residency program due to fear of discrimination. Nearly one-third of the overall group of dermatology residents surveyed (29%; 27/92) reported that they had witnessed inappropriate or discriminatory comments about LGBT persons made by employees or staff at their programs. Most residents surveyed (96%; 88/92) agreed or strongly agreed that they feel comfortable working alongside LGBT physicians.

There were 18 nonheterosexual dermatologyresidents who completed the Mohr and Fassinger Outness Inventory.²³ In general, respondents reported that they were more “out” with friends and family than work peers and were least “out” with work supervisors and strangers.

Comment

Dermatology Residents Desire More Time on LGBT Health—This cross-sectional survey study explored dermatology residents’ educational experiences with LGBT health during residency training. Similar studies have been performed in other specialties, including a study from 2019 surveying emergency medicine residents that demonstrated residents find caring for LGBT patients more challenging.¹⁵ Another 2019 study surveying psychiatry residents found that 42.4% (N=99) reported no coverage of LGBT topics.¹⁸ Our study is unique in that it surveyed dermatology residents directly regarding this topic. Although most dermatology program directors view LGBT dermatologic health as an important topic, a prior study revealed that many programs are lacking dedicated LGBT educational experiences. The most common barriers reported were insufficient time in the didactic schedule and lack of experienced faculty.²⁰

Our study revealed that dermatology residents overall tend to agree with residents from other specialties and dermatology program directors. Most of the dermatology residents surveyed reported desiring more time per year spent on LGBT health education than they receive, and 60% expressed that LGBT educational experiences need to be improved at their residency programs. Education on and subsequent comfort level with LGBT health issues varied by subtopic, with most residents feeling comfortable dealing with dermatologic manifestations of HIV/AIDS and other sexually transmitted diseases and less comfortable with topics such as puberty blockers, gender-affirming surgery and hormone therapy, body image, and skin cancer risk.

Overall, LGBT health training is viewed as important and in need of improvement by both program directors and residents, yet implementation lags at many programs. A small proportion of the represented programs are excelling in this area—just over 5% of respondents reported receiving 10 or more hours of LGBT-relevant education per year, and approximately 26% of residents felt that LGBT coverage was good or excellent at their programs. Our study showed a clear relationship between feelings of preparedness and education level. The lack of LGBT education at some dermatology residency programs translated into a large portion of dermatology residents feeling ill equipped to care for LGBT patients after graduation—nearly 50% of those surveyed reported feeling insufficiently prepared to care for the LGBT community.

Discrimination in Residency Programs—Dermatology residency programs also are not free from sexual orientation–related and gender identity–related workplace discrimination. Although 96% of dermatology residents reported that they feel comfortable working alongside LGBT physicians, 24% of nonheterosexual respondents stated they had been subjected to offensive remarks about their sexual orientation, and 29% of the overall group of dermatology residents had witnessed discriminatory comments to LGBT individuals at their programs. In addition, some nonheterosexual dermatology residents reported being less “out” with their workplace supervisors and strangers, such as patients, than with their family and friends, and 50% of this group reported that their sexual identity was not openly discussed with their workplace supervisors. It has been demonstrated that individuals are more likely to “come out” in perceived LGBT-friendly workplace environments and that being “out” positively impacts psychological health because of the effects of perceived social support and self-coherence.^26,27

Study Strengths and Limitations—Strengths of this study include the modest sample size of dermatology residents that participated, high completion rate, and the anonymity of the survey. Limitations include the risk of sampling bias by posting the survey on LGBT-specific groups. The survey also took place in the fall, so the results may not accurately reflect programs that cover this material later in the academic year. Lastly, not all survey questions were validated.

Implementing Change in Residency Programs—Although the results of this study exposed the need for increasing LGBT education in dermatology residency, they do not provide guidelines for the best strategy to begin implementing change. A study from 2020 provides some guidance for incorporating LGBT health training into dermatology residency programs through a combination of curricular modifications and climate optimization.²⁸ Additional future research should focus on the best methods for preparing dermatology residents to care for this population. In this study, residents reported that the most effective teaching methods were real encounters with LGBT patients or faculty educated on LGBT health as well as live lectures from experts. There also appeared to be a correlation between hours spent on LGBT health, including various subtopics, and residents’ perceived preparedness in these areas. Potential actionable items include clarifying the ACGME guidelines on LGBT health topics; increasing the sexual and gender diversity of the faculty, staff, residents, and patients; and dedicating additional didactic and clinical time to LGBT topics and experiences.

Conclusion

This survey study of dermatology residents regarding LGBT learning experiences in residency training provided evidence that dermatology residents as a whole are not adequately taught LGBT health topics and therefore feel unprepared to take care of this patient population. Additionally, most residents desire improvement of LGBT health education and training. Further studies focusing on the best methods for implementing LGBT-specific curricula are needed.

Approximately 4.5% of adults within the United States identify as members of the lesbian, gay, bisexual, transgender (LGBT) community.¹ This is an umbrella term inclusive of all individuals identifying as nonheterosexual or noncisgender. Although the LGBT community has increasingly become more recognized and accepted by society over time, health care disparities persist and have been well documented in the literature.^2-4 Dermatologists have the potential to greatly impact LGBT health, as many health concerns in this population are cutaneous, such as sun-protection behaviors, side effects of gender-affirming hormone therapy and gender-affirming procedures, and cutaneous manifestations of sexually transmitted infections.^5-7

An education gap has been demonstrated in both medical students and resident physicians regarding LGBT health and cultural competency. In a large-scale, multi-institutional survey study published in 2015, approximately two-thirds of medical students rated their schools’ LGBT curriculum as fair, poor, or very poor.⁸ Additional studies have echoed these results and have demonstrated not only the need but the desire for additional training on LGBT issues in medical school.^9-11 The Association of American Medical Colleges has begun implementing curricular and institutional changes to fulfill this need.^12,13

The LGBT education gap has been shown to extend into residency training. Multiple studies performed within a variety of medical specialties have demonstrated that resident physicians receive insufficient training in LGBT health issues, lack comfort in caring for LGBT patients, and would benefit from dedicated curricula on these topics.^14-18 Currently, the 2022 Accreditation Council for Graduate Medical Education (ACGME) guidelines related to LGBT health are minimal and nonspecific.¹⁹

Ensuring that dermatology trainees are well equipped to manage these issues while providing culturally competent care to LGBT patients is paramount. However, research suggests that dedicated training on these topics likely is insufficient. A survey study of dermatology residency program directors (N=90) revealed that although 81% (72/89) viewed training in LGBT health as either very important or somewhat important, 46% (41/90) of programs did not dedicate any time to this content and 37% (33/90) only dedicated 1 to 2 hours per year.²⁰

To further explore this potential education gap, we surveyed dermatology residents directly to better understand LGBT education within residency training, resident preparedness to care for LGBT patients, and outness/discrimination of LGBT-identifying residents. We believe this study should drive future research on the development and implementation of LGBT-specific curricula in dermatology training programs.

Methods

A cross-sectional survey study of dermatology residents in the United States was conducted. The study was deemed exempt from review by The Ohio State University (Columbus, Ohio) institutional review board. Survey responses were collected from October 7, 2020, to November 13, 2020. Qualtrics software was used to create the 20-question survey, which included a combination of categorical, dichotomous, and optional free-text questions related to patient demographics, LGBT training experiences, perceived areas of curriculum improvement, comfort level managing LGBT health issues, and personal experiences. Some questions were adapted from prior surveys.^15,21 Validated survey tools used included the 2020 US Census to collect information regarding race and ethnicity, the Mohr and Fassinger Outness Inventory to measure outness regarding sexual orientation, and select questions from the 2020 Association of American Medical Colleges Medical School Graduation Questionnaire regarding discrimination.^22-24

The survey was distributed to current allopathic and osteopathic dermatology residents by a variety of methods, including emails to program director and program coordinator listserves. The survey also was posted in the American Academy of Dermatology Expert Resource Group on LGBTQ Health October 2020 newsletter, as well as dermatology social media groups, including a messaging forum limited to dermatology residents, a Facebook group open to dermatologists and dermatology residents, and the Facebook group of the Gay and Lesbian Dermatology Association. Current dermatology residents, including those in combined dermatology and internal medicine programs, were included. Individuals who had been accepted to dermatology training programs but had not yet started were excluded. A follow-up email was sent to the program director listserve approximately 3 weeks after the initial distribution.

Statistical Analysis—The data were analyzed in Qualtrics and Microsoft Excel using descriptive statistics. Stata software (Stata 15.1, StataCorp) was used to perform a Kruskal-Wallis equality-of-populations rank test to compare the means of education level and feelings of preparedness.

Results

Demographics of Respondents—A total of 126 responses were recorded, 12 of which were blank and were removed from the database. A total of 114 dermatology residents’ responses were collected in Qualtrics and analyzed; 91 completed the entire survey (an 80% completion rate). Based on the 2020-2021 ACGME data listing, there were 1612 dermatology residents in the United States, which is an estimated response rate of 7% (114/1612).²⁵ The eTable outlines the demographics of the survey respondents. Most were cisgender females (60%), followed by cisgender males (35%); the remainder preferred not to answer. Regarding sexual orientation, 77% identified as straight or heterosexual; 17% as gay, lesbian, or homosexual; 1% as queer; and 1% as bisexual. The training programs were in 26 states, the majority of which were in the Midwest (34%) and in urban settings (69%). A wide range of postgraduate levels and residency sizes were represented in the survey.

LGBT Education—Fifty-one percent of respondents reported that their programs offer 1 hour or less of LGBT-related curricula per year; 34% reported no time dedicated to this topic. A small portion of residents (5%) reported 10 or more hours of LGBT education per year. Residents also were asked the average number of hours of LGBT education they thought they should receive. The discrepancy between these measures can be visualized in Figure 1. The median hours of education received was 1 hour (IQR, 0–4 hours), whereas the median hours of education desired was 4 hours (IQR, 2–5 hours). The most common and most helpful methods of education reported were clinical experiences with faculty or patients and live lectures.

Overall, 45% of survey respondents felt that LGBT topics were covered poorly or not at all in dermatology residency, whereas 26% thought the coverage was good or excellent. The topics that residents were most likely to report receiving good or excellent coverage were dermatologic manifestations of HIV/AIDS (70%) and sexually transmitted diseases in LGBT patients (48%). The topics that were most likely to be reported as not taught or poorly taught included dermatologic concerns associated with puberty blockers (71%), body image (58%), dermatologic concerns associated with gender-affirming surgery (55%), skin cancer risk (53%), taking an LGBT-oriented history and physical examination (52%), and effects of gender-affirming hormone therapy on the skin (50%). A detailed breakdown of coverage level by topic can be found in Figure 2.

Preparedness to Care for LGBT Patients—Only 68% of survey respondents agreed or strongly agreed that they feel comfortable treating LGBT patients. Furthermore, 49% of dermatology residents reported that they feel not at all prepared or insufficiently prepared to provide care to LGBT individuals (Figure 2), and 60% believed that LGBT training needed to be improved at their residency programs.

There was a significant association between reported level of education and feelings of preparedness. A high ranking of provided education was associated with higher levels of feeling prepared to care for LGBT patients (Kruskal-Wallis rank test, P<.001).

Discrimination/Outness—Approximately one-fourth (24%; 4/17) of nonheterosexual dermatology residents reported that they had been subjected to offensive remarks about their sexual orientation in the workplace. One respondent commented that they were less “out” at their residency program due to fear of discrimination. Nearly one-third of the overall group of dermatology residents surveyed (29%; 27/92) reported that they had witnessed inappropriate or discriminatory comments about LGBT persons made by employees or staff at their programs. Most residents surveyed (96%; 88/92) agreed or strongly agreed that they feel comfortable working alongside LGBT physicians.

There were 18 nonheterosexual dermatologyresidents who completed the Mohr and Fassinger Outness Inventory.²³ In general, respondents reported that they were more “out” with friends and family than work peers and were least “out” with work supervisors and strangers.

Comment

Dermatology Residents Desire More Time on LGBT Health—This cross-sectional survey study explored dermatology residents’ educational experiences with LGBT health during residency training. Similar studies have been performed in other specialties, including a study from 2019 surveying emergency medicine residents that demonstrated residents find caring for LGBT patients more challenging.¹⁵ Another 2019 study surveying psychiatry residents found that 42.4% (N=99) reported no coverage of LGBT topics.¹⁸ Our study is unique in that it surveyed dermatology residents directly regarding this topic. Although most dermatology program directors view LGBT dermatologic health as an important topic, a prior study revealed that many programs are lacking dedicated LGBT educational experiences. The most common barriers reported were insufficient time in the didactic schedule and lack of experienced faculty.²⁰

Our study revealed that dermatology residents overall tend to agree with residents from other specialties and dermatology program directors. Most of the dermatology residents surveyed reported desiring more time per year spent on LGBT health education than they receive, and 60% expressed that LGBT educational experiences need to be improved at their residency programs. Education on and subsequent comfort level with LGBT health issues varied by subtopic, with most residents feeling comfortable dealing with dermatologic manifestations of HIV/AIDS and other sexually transmitted diseases and less comfortable with topics such as puberty blockers, gender-affirming surgery and hormone therapy, body image, and skin cancer risk.

Overall, LGBT health training is viewed as important and in need of improvement by both program directors and residents, yet implementation lags at many programs. A small proportion of the represented programs are excelling in this area—just over 5% of respondents reported receiving 10 or more hours of LGBT-relevant education per year, and approximately 26% of residents felt that LGBT coverage was good or excellent at their programs. Our study showed a clear relationship between feelings of preparedness and education level. The lack of LGBT education at some dermatology residency programs translated into a large portion of dermatology residents feeling ill equipped to care for LGBT patients after graduation—nearly 50% of those surveyed reported feeling insufficiently prepared to care for the LGBT community.

Discrimination in Residency Programs—Dermatology residency programs also are not free from sexual orientation–related and gender identity–related workplace discrimination. Although 96% of dermatology residents reported that they feel comfortable working alongside LGBT physicians, 24% of nonheterosexual respondents stated they had been subjected to offensive remarks about their sexual orientation, and 29% of the overall group of dermatology residents had witnessed discriminatory comments to LGBT individuals at their programs. In addition, some nonheterosexual dermatology residents reported being less “out” with their workplace supervisors and strangers, such as patients, than with their family and friends, and 50% of this group reported that their sexual identity was not openly discussed with their workplace supervisors. It has been demonstrated that individuals are more likely to “come out” in perceived LGBT-friendly workplace environments and that being “out” positively impacts psychological health because of the effects of perceived social support and self-coherence.^26,27

Study Strengths and Limitations—Strengths of this study include the modest sample size of dermatology residents that participated, high completion rate, and the anonymity of the survey. Limitations include the risk of sampling bias by posting the survey on LGBT-specific groups. The survey also took place in the fall, so the results may not accurately reflect programs that cover this material later in the academic year. Lastly, not all survey questions were validated.

Implementing Change in Residency Programs—Although the results of this study exposed the need for increasing LGBT education in dermatology residency, they do not provide guidelines for the best strategy to begin implementing change. A study from 2020 provides some guidance for incorporating LGBT health training into dermatology residency programs through a combination of curricular modifications and climate optimization.²⁸ Additional future research should focus on the best methods for preparing dermatology residents to care for this population. In this study, residents reported that the most effective teaching methods were real encounters with LGBT patients or faculty educated on LGBT health as well as live lectures from experts. There also appeared to be a correlation between hours spent on LGBT health, including various subtopics, and residents’ perceived preparedness in these areas. Potential actionable items include clarifying the ACGME guidelines on LGBT health topics; increasing the sexual and gender diversity of the faculty, staff, residents, and patients; and dedicating additional didactic and clinical time to LGBT topics and experiences.

Conclusion

This survey study of dermatology residents regarding LGBT learning experiences in residency training provided evidence that dermatology residents as a whole are not adequately taught LGBT health topics and therefore feel unprepared to take care of this patient population. Additionally, most residents desire improvement of LGBT health education and training. Further studies focusing on the best methods for implementing LGBT-specific curricula are needed.

To the Editor:

In a 2010 landmark paper, researchers reported that the Preexposure Prophylaxis Initiative (iPrEx) trial demonstrated that once-daily pre-exposure prophylaxis (PrEP) with emtricitabine plus tenofovir disoproxil fumarate, which was approved by the US Food and Drug Administration (FDA) and packaged together as Truvada (Gilead Sciences, Inc), achieved a 44% reduction in the incidence of HIV infection compared to the placebo arm of the study (64/1248 HIV infections in the placebo group vs 36/1251 in the intervention group).¹ Subsequently, the US Department of Health and Human Services proposed an initiative to reduce new HIV infections by 90% by 2030.² The Centers for Disease Control and Prevention estimates that 1.1 million Americans have an indication for PrEP, yet only approximately 400,000 individuals currently take PrEP.^3,4

Increasing awareness of PrEP and its indications is essential because PrEP exerts its greatest benefit when used broadly. Awareness among primary care and infectious disease physicians was reported at 76%⁵; awareness among other medical specialists remains unknown. Awareness of PrEP among dermatologists is important because dermatologists play an important role in the diagnosis and treatment of many sexually transmitted infections (STIs), which are a risk factor for transmission of HIV. As providers who treat STIs, dermatologists are in a prime position to educate patients about PrEP, refer them for treatment, and prescribe the regimen. We conducted a survey to assess dermatologists’ knowledge about and attitudes toward PrEP. We also provide a brief summary of prescribing information about common PrEP regimens to fill in the knowledge gap among dermatologists as a way to promote its utilization.

An electronic survey was distributed to 486 members of the Association of Professors of Dermatology based in the United States using the web-based survey application REDCap. The study was approved by the New York University Grossman School of Medicine (New York, New York) institutional review board. Eighty-one anonymous survey responses were completed and returned (response rate, 16.6%). Data were analyzed using descriptive statistics.

The mean age (SD) of respondents was 39.1 (9.7) years; 49.4% (40/81) were male; and 74.1% (60/81) were attending physicians, with a mean (SD) of 9.4 (8.6) years of practice. Clinical practices were predominantly from the northeast (46.9% [38/81]) and mostly in an academic setting (74.1% [60/81]). As shown in Table 1, most surveyed dermatologists reported being aware of PrEP (93.8% [76/81]), but a minority (42.0% [34/81]) were familiar with indications for its use; even fewer (4.9% [4/81]) were current prescribers. Referral to other physicians for PrEP was reported by 58.0% (47/81) of respondents.

Despite respondents’ awareness of PrEP as a preventive measure (93.8% [76/81]) and their willingness to prescribe it (67.9% [55/81]), many reported being largely unfamiliar with its indications (58.0% [47/81]) and uncomfortable discussing its adverse effects (72.8% [59/81]), conducting appropriate laboratory monitoring (84.0% [68/81]), and refilling existing prescriptions (77.8% [63/81]). Respondents’ lack of education about PrEP was a barrier to prescribing (51.9% [42/81] to 59.3% [48/81]) and explains why a small minority (4.9% [4/81]) currently prescribe the regimen.

Our study sought to characterize current clinical knowledge about and practice patterns of PrEP among dermatologists. Dermatologists often encounter patients who present with an STI, which is a risk factor for HIV infection, but our survey respondents reported several barriers to utilizing PrEP. The difference in the degree of respondents’ willingness to prescribe PrEP (67.9%) and those who self-identified as prescribers (4.9%) suggests a role for dermatologists in prescribing or discussing PrEP with their patients—albeit a currently undefined role.

The results of our study suggested that half (41/81) of dermatologists believe that PrEP prescription is out of their scope of practice, likely due to a combination of scheduling, laboratory monitoring, and medicolegal concerns. For dermatologists who are interested in being PrEP prescribers, our results suggested that closing the knowledge gap around PrEP among dermatologists through training and education could improve comfort with this medication and lead to changes in practice to prevent the spread of HIV infection.

PrEP is indicated for HIV-negative patients who have HIV-positive sexual partners, utilize barrier protection methods inconsistently, or had a diagnosis of an STI in the last 6 months.⁶ In 2012, the FDA approved once-daily use of emtricitabine plus tenofovir for primary prevention of HIV infection. Post hoc analysis of iPrEx trial data revealed that once-daily PrEP taken regularly had a 92% to 100% protective effect against HIV.⁷

Regrettably, real-world uptake of PrEP has been slower than desired. The most recent data (2021) show that nearly 1 million individuals worldwide take PrEP; however, this represents only approximately one-third of those eligible.⁸ Utilization is notably lower among Black and Latino populations who stand to gain the most from PrEP given their higher risk of contracting HIV compared to their White counterparts.⁹ As such, improving access to PrEP through expanded provider awareness is essential to decrease the risk for HIV infection and transmission.

Emtricitabine plus tenofovir is safe and well tolerated; more common adverse effects are headache, nausea, vomiting, rash, and loss of appetite. Tenofovir likely decreases bone mineral density, even in HIV-negative patients¹⁰; mineralization seems to recover after the medication is discontinued.¹¹ Rarely, tenofovir can increase the level of creatinine and hepatic transaminases; a recent report on its long-term side effects has shown small nonprogressive decreases in glomerular filtration rate.¹² Monitoring kidney function is a component of prescribing PrEP (Table 2).

In 2019, emtricitabine plus tenofovir was reformulated with tenofovir alafenamide; the new combination regimen received FDA approval for once-daily PrEP under the brand name Descovy (Gilead Sciences, Inc). The new formulation results in a lower blood concentration of tenofovir and has been reported to present less of a risk for bone and kidney toxicity.^13,14

Notably, emtricitabine plus tenofovir alafenamide might accumulate faster in peripheral lymphatic tissue than emtricitabine plus tenofovir disoproxil fumarate. This property has led to a new regimen known as “on-demand PrEP,” which follows a 2-1-1 dosing regimen: Patients take a double dose 2 to 24 hours before sexual activity, 1 dose on the day of sexual activity, and 1 dose the day after sexual activity.¹⁵ Because some patients at risk for HIV infection might not be consistently sexually active, on-demand PrEP allows them to cycle on and off the medication. Barriers to implementing on-demand PrEP include requiring that sexual activity be planned and an adverse effect profile similar to daily-use PrEP.¹⁶

The FDA recently approved a long-acting, once-monthly combination injectable PrEP of cabotegravir and rilpivirine.¹⁷ The long duration of action of this PrEP will benefit patients who report problems with medication adherence.

Our study demonstrates low frequency in prescribing patterns of PrEP among dermatologists and suggests that an addressable barrier to such prescribing is the lack of knowledge on how to prescribe it safely, which warrants further clinical investigation. We summarize an approach to prescribing PrEP in Table 2. Our study was limited by a small sample of mostly academic dermatologists and selection bias, which may diminish the generalizability of findings. A study of a larger, more representative group of dermatologists likely would show different prescribing patterns and degrees of knowledge about PrEP. Research is needed to study the impact of educational interventions that aim to increase both knowledge and prescribing of PrEP among dermatologists.

To the Editor:

In a 2010 landmark paper, researchers reported that the Preexposure Prophylaxis Initiative (iPrEx) trial demonstrated that once-daily pre-exposure prophylaxis (PrEP) with emtricitabine plus tenofovir disoproxil fumarate, which was approved by the US Food and Drug Administration (FDA) and packaged together as Truvada (Gilead Sciences, Inc), achieved a 44% reduction in the incidence of HIV infection compared to the placebo arm of the study (64/1248 HIV infections in the placebo group vs 36/1251 in the intervention group).¹ Subsequently, the US Department of Health and Human Services proposed an initiative to reduce new HIV infections by 90% by 2030.² The Centers for Disease Control and Prevention estimates that 1.1 million Americans have an indication for PrEP, yet only approximately 400,000 individuals currently take PrEP.^3,4

Increasing awareness of PrEP and its indications is essential because PrEP exerts its greatest benefit when used broadly. Awareness among primary care and infectious disease physicians was reported at 76%⁵; awareness among other medical specialists remains unknown. Awareness of PrEP among dermatologists is important because dermatologists play an important role in the diagnosis and treatment of many sexually transmitted infections (STIs), which are a risk factor for transmission of HIV. As providers who treat STIs, dermatologists are in a prime position to educate patients about PrEP, refer them for treatment, and prescribe the regimen. We conducted a survey to assess dermatologists’ knowledge about and attitudes toward PrEP. We also provide a brief summary of prescribing information about common PrEP regimens to fill in the knowledge gap among dermatologists as a way to promote its utilization.

An electronic survey was distributed to 486 members of the Association of Professors of Dermatology based in the United States using the web-based survey application REDCap. The study was approved by the New York University Grossman School of Medicine (New York, New York) institutional review board. Eighty-one anonymous survey responses were completed and returned (response rate, 16.6%). Data were analyzed using descriptive statistics.

The mean age (SD) of respondents was 39.1 (9.7) years; 49.4% (40/81) were male; and 74.1% (60/81) were attending physicians, with a mean (SD) of 9.4 (8.6) years of practice. Clinical practices were predominantly from the northeast (46.9% [38/81]) and mostly in an academic setting (74.1% [60/81]). As shown in Table 1, most surveyed dermatologists reported being aware of PrEP (93.8% [76/81]), but a minority (42.0% [34/81]) were familiar with indications for its use; even fewer (4.9% [4/81]) were current prescribers. Referral to other physicians for PrEP was reported by 58.0% (47/81) of respondents.

Despite respondents’ awareness of PrEP as a preventive measure (93.8% [76/81]) and their willingness to prescribe it (67.9% [55/81]), many reported being largely unfamiliar with its indications (58.0% [47/81]) and uncomfortable discussing its adverse effects (72.8% [59/81]), conducting appropriate laboratory monitoring (84.0% [68/81]), and refilling existing prescriptions (77.8% [63/81]). Respondents’ lack of education about PrEP was a barrier to prescribing (51.9% [42/81] to 59.3% [48/81]) and explains why a small minority (4.9% [4/81]) currently prescribe the regimen.

Our study sought to characterize current clinical knowledge about and practice patterns of PrEP among dermatologists. Dermatologists often encounter patients who present with an STI, which is a risk factor for HIV infection, but our survey respondents reported several barriers to utilizing PrEP. The difference in the degree of respondents’ willingness to prescribe PrEP (67.9%) and those who self-identified as prescribers (4.9%) suggests a role for dermatologists in prescribing or discussing PrEP with their patients—albeit a currently undefined role.

The results of our study suggested that half (41/81) of dermatologists believe that PrEP prescription is out of their scope of practice, likely due to a combination of scheduling, laboratory monitoring, and medicolegal concerns. For dermatologists who are interested in being PrEP prescribers, our results suggested that closing the knowledge gap around PrEP among dermatologists through training and education could improve comfort with this medication and lead to changes in practice to prevent the spread of HIV infection.

PrEP is indicated for HIV-negative patients who have HIV-positive sexual partners, utilize barrier protection methods inconsistently, or had a diagnosis of an STI in the last 6 months.⁶ In 2012, the FDA approved once-daily use of emtricitabine plus tenofovir for primary prevention of HIV infection. Post hoc analysis of iPrEx trial data revealed that once-daily PrEP taken regularly had a 92% to 100% protective effect against HIV.⁷

Regrettably, real-world uptake of PrEP has been slower than desired. The most recent data (2021) show that nearly 1 million individuals worldwide take PrEP; however, this represents only approximately one-third of those eligible.⁸ Utilization is notably lower among Black and Latino populations who stand to gain the most from PrEP given their higher risk of contracting HIV compared to their White counterparts.⁹ As such, improving access to PrEP through expanded provider awareness is essential to decrease the risk for HIV infection and transmission.

Emtricitabine plus tenofovir is safe and well tolerated; more common adverse effects are headache, nausea, vomiting, rash, and loss of appetite. Tenofovir likely decreases bone mineral density, even in HIV-negative patients¹⁰; mineralization seems to recover after the medication is discontinued.¹¹ Rarely, tenofovir can increase the level of creatinine and hepatic transaminases; a recent report on its long-term side effects has shown small nonprogressive decreases in glomerular filtration rate.¹² Monitoring kidney function is a component of prescribing PrEP (Table 2).

In 2019, emtricitabine plus tenofovir was reformulated with tenofovir alafenamide; the new combination regimen received FDA approval for once-daily PrEP under the brand name Descovy (Gilead Sciences, Inc). The new formulation results in a lower blood concentration of tenofovir and has been reported to present less of a risk for bone and kidney toxicity.^13,14

Notably, emtricitabine plus tenofovir alafenamide might accumulate faster in peripheral lymphatic tissue than emtricitabine plus tenofovir disoproxil fumarate. This property has led to a new regimen known as “on-demand PrEP,” which follows a 2-1-1 dosing regimen: Patients take a double dose 2 to 24 hours before sexual activity, 1 dose on the day of sexual activity, and 1 dose the day after sexual activity.¹⁵ Because some patients at risk for HIV infection might not be consistently sexually active, on-demand PrEP allows them to cycle on and off the medication. Barriers to implementing on-demand PrEP include requiring that sexual activity be planned and an adverse effect profile similar to daily-use PrEP.¹⁶

The FDA recently approved a long-acting, once-monthly combination injectable PrEP of cabotegravir and rilpivirine.¹⁷ The long duration of action of this PrEP will benefit patients who report problems with medication adherence.

Our study demonstrates low frequency in prescribing patterns of PrEP among dermatologists and suggests that an addressable barrier to such prescribing is the lack of knowledge on how to prescribe it safely, which warrants further clinical investigation. We summarize an approach to prescribing PrEP in Table 2. Our study was limited by a small sample of mostly academic dermatologists and selection bias, which may diminish the generalizability of findings. A study of a larger, more representative group of dermatologists likely would show different prescribing patterns and degrees of knowledge about PrEP. Research is needed to study the impact of educational interventions that aim to increase both knowledge and prescribing of PrEP among dermatologists.

To the Editor:

In a 2010 landmark paper, researchers reported that the Preexposure Prophylaxis Initiative (iPrEx) trial demonstrated that once-daily pre-exposure prophylaxis (PrEP) with emtricitabine plus tenofovir disoproxil fumarate, which was approved by the US Food and Drug Administration (FDA) and packaged together as Truvada (Gilead Sciences, Inc), achieved a 44% reduction in the incidence of HIV infection compared to the placebo arm of the study (64/1248 HIV infections in the placebo group vs 36/1251 in the intervention group).¹ Subsequently, the US Department of Health and Human Services proposed an initiative to reduce new HIV infections by 90% by 2030.² The Centers for Disease Control and Prevention estimates that 1.1 million Americans have an indication for PrEP, yet only approximately 400,000 individuals currently take PrEP.^3,4

Increasing awareness of PrEP and its indications is essential because PrEP exerts its greatest benefit when used broadly. Awareness among primary care and infectious disease physicians was reported at 76%⁵; awareness among other medical specialists remains unknown. Awareness of PrEP among dermatologists is important because dermatologists play an important role in the diagnosis and treatment of many sexually transmitted infections (STIs), which are a risk factor for transmission of HIV. As providers who treat STIs, dermatologists are in a prime position to educate patients about PrEP, refer them for treatment, and prescribe the regimen. We conducted a survey to assess dermatologists’ knowledge about and attitudes toward PrEP. We also provide a brief summary of prescribing information about common PrEP regimens to fill in the knowledge gap among dermatologists as a way to promote its utilization.

An electronic survey was distributed to 486 members of the Association of Professors of Dermatology based in the United States using the web-based survey application REDCap. The study was approved by the New York University Grossman School of Medicine (New York, New York) institutional review board. Eighty-one anonymous survey responses were completed and returned (response rate, 16.6%). Data were analyzed using descriptive statistics.

The mean age (SD) of respondents was 39.1 (9.7) years; 49.4% (40/81) were male; and 74.1% (60/81) were attending physicians, with a mean (SD) of 9.4 (8.6) years of practice. Clinical practices were predominantly from the northeast (46.9% [38/81]) and mostly in an academic setting (74.1% [60/81]). As shown in Table 1, most surveyed dermatologists reported being aware of PrEP (93.8% [76/81]), but a minority (42.0% [34/81]) were familiar with indications for its use; even fewer (4.9% [4/81]) were current prescribers. Referral to other physicians for PrEP was reported by 58.0% (47/81) of respondents.

Despite respondents’ awareness of PrEP as a preventive measure (93.8% [76/81]) and their willingness to prescribe it (67.9% [55/81]), many reported being largely unfamiliar with its indications (58.0% [47/81]) and uncomfortable discussing its adverse effects (72.8% [59/81]), conducting appropriate laboratory monitoring (84.0% [68/81]), and refilling existing prescriptions (77.8% [63/81]). Respondents’ lack of education about PrEP was a barrier to prescribing (51.9% [42/81] to 59.3% [48/81]) and explains why a small minority (4.9% [4/81]) currently prescribe the regimen.

Our study sought to characterize current clinical knowledge about and practice patterns of PrEP among dermatologists. Dermatologists often encounter patients who present with an STI, which is a risk factor for HIV infection, but our survey respondents reported several barriers to utilizing PrEP. The difference in the degree of respondents’ willingness to prescribe PrEP (67.9%) and those who self-identified as prescribers (4.9%) suggests a role for dermatologists in prescribing or discussing PrEP with their patients—albeit a currently undefined role.

The results of our study suggested that half (41/81) of dermatologists believe that PrEP prescription is out of their scope of practice, likely due to a combination of scheduling, laboratory monitoring, and medicolegal concerns. For dermatologists who are interested in being PrEP prescribers, our results suggested that closing the knowledge gap around PrEP among dermatologists through training and education could improve comfort with this medication and lead to changes in practice to prevent the spread of HIV infection.

PrEP is indicated for HIV-negative patients who have HIV-positive sexual partners, utilize barrier protection methods inconsistently, or had a diagnosis of an STI in the last 6 months.⁶ In 2012, the FDA approved once-daily use of emtricitabine plus tenofovir for primary prevention of HIV infection. Post hoc analysis of iPrEx trial data revealed that once-daily PrEP taken regularly had a 92% to 100% protective effect against HIV.⁷

Regrettably, real-world uptake of PrEP has been slower than desired. The most recent data (2021) show that nearly 1 million individuals worldwide take PrEP; however, this represents only approximately one-third of those eligible.⁸ Utilization is notably lower among Black and Latino populations who stand to gain the most from PrEP given their higher risk of contracting HIV compared to their White counterparts.⁹ As such, improving access to PrEP through expanded provider awareness is essential to decrease the risk for HIV infection and transmission.

Emtricitabine plus tenofovir is safe and well tolerated; more common adverse effects are headache, nausea, vomiting, rash, and loss of appetite. Tenofovir likely decreases bone mineral density, even in HIV-negative patients¹⁰; mineralization seems to recover after the medication is discontinued.¹¹ Rarely, tenofovir can increase the level of creatinine and hepatic transaminases; a recent report on its long-term side effects has shown small nonprogressive decreases in glomerular filtration rate.¹² Monitoring kidney function is a component of prescribing PrEP (Table 2).

In 2019, emtricitabine plus tenofovir was reformulated with tenofovir alafenamide; the new combination regimen received FDA approval for once-daily PrEP under the brand name Descovy (Gilead Sciences, Inc). The new formulation results in a lower blood concentration of tenofovir and has been reported to present less of a risk for bone and kidney toxicity.^13,14

Notably, emtricitabine plus tenofovir alafenamide might accumulate faster in peripheral lymphatic tissue than emtricitabine plus tenofovir disoproxil fumarate. This property has led to a new regimen known as “on-demand PrEP,” which follows a 2-1-1 dosing regimen: Patients take a double dose 2 to 24 hours before sexual activity, 1 dose on the day of sexual activity, and 1 dose the day after sexual activity.¹⁵ Because some patients at risk for HIV infection might not be consistently sexually active, on-demand PrEP allows them to cycle on and off the medication. Barriers to implementing on-demand PrEP include requiring that sexual activity be planned and an adverse effect profile similar to daily-use PrEP.¹⁶

The FDA recently approved a long-acting, once-monthly combination injectable PrEP of cabotegravir and rilpivirine.¹⁷ The long duration of action of this PrEP will benefit patients who report problems with medication adherence.

Our study demonstrates low frequency in prescribing patterns of PrEP among dermatologists and suggests that an addressable barrier to such prescribing is the lack of knowledge on how to prescribe it safely, which warrants further clinical investigation. We summarize an approach to prescribing PrEP in Table 2. Our study was limited by a small sample of mostly academic dermatologists and selection bias, which may diminish the generalizability of findings. A study of a larger, more representative group of dermatologists likely would show different prescribing patterns and degrees of knowledge about PrEP. Research is needed to study the impact of educational interventions that aim to increase both knowledge and prescribing of PrEP among dermatologists.

In 2019, diabetes mellitus (DM) was the seventh leading cause of death in the United States, and currently, about 11% of the American population has a DM diagnosis.¹ Most have a diagnosis of type 2 diabetes (T2DM), which has a strong genetic predisposition, and the risk of developing T2DM increases with age, obesity, and lack of physical activity.^1,2 Nearly one-quarter of veterans have a diagnosis of DM, and DM is the leading cause of comorbidities, such as blindness, end-stage renal disease, and amputation for patients receiving care from the Veterans Health Administration (VHA).² The elevated incidence of DM in the veteran population is attributed to a variety of factors, including exposure to herbicides, such as Agent Orange, advanced age, increased risk of obesity, and limited access to high-quality food.³

After diagnosis, both the American Diabetes Association (ADA) and the American Association of Clinical Endocrinologists and American College of Endocrinology (AACE/ACE) emphasize the appropriate use of lifestyle management and pharmacologic therapy for DM care. The use of pharmacologic agents (oral medications, insulin, or noninsulin injectables) is often determined by efficacy, cost, potential adverse effects (AEs), and patient factors and comorbidities.^4,5

The initial recommendation for pharmacologic treatment for T2DM differs slightly between expert guidelines. The ADA and AACE/ACE recommend any of the following as initial monotherapy, listed in order to represent a hierarchy of usage: metformin, glucagon-like peptide-1 receptor agonists (GLP-1 RAs), sodium-glucose cotransporter 2 (SGLT-2) inhibitors, or dipeptidyl peptidase-4 (DPP-4) inhibitors, with the first 3 agents carrying the strongest recommendations.^4,5 For patients with established atherosclerotic cardiovascular disease (CVD), chronic kidney disease, or heart failure, it is recommended to start a long-acting GLP-1 RA or SGLT-2 inhibitor. For patients with T2DM and hemoglobin A_1c (HbA_1c) between 7.5% and 9.0% at diagnosis, the AACE/ACE recommend initiation of dual therapy using metformin alongside another first-line agent and recommend the addition of another antidiabetic agent if glycemic goals are not met after regular follow-up. AACE/ACE recommend the consideration of insulin therapy in symptomatic patients with HbA_1c > 9.0%.⁵ In contrast, the ADA recommends metformin as first-line therapy for all patients with T2DM and recommends dual therapy using metformin and another preferred agent (selection based on comorbidities) when HbA_1c is 1.5% to 2% above target. The ADA recommends the consideration of insulin with HbA_1c > 10% or with evidence of ongoing catabolism or symptoms of hyperglycemia.⁴ There are several reasons why insulin may be initiated prior to GLP-1 RAs, including profound hyperglycemia at time of diagnosis or implementation of insulin agents prior to commercial availability of GLP-1 RA.

GLP-1 RAs are analogs of the hormone incretin, which increases glucose-dependent insulin secretion, decreases postprandial glucagon secretion, increases satiety, and slows gastric emptying.^6,7 When used in combination with noninsulin agents, GLP-1 RAs have demonstrated HbA_1c reductions of 0.5% to 1.5%.⁸ The use of GLP-1 RAs with basal insulin also has been studied extensively.^6,8-10 When the combination of GLP-1 RAs and basal insulin was compared with basal/bolus insulin regimens, the use of the GLP-1 RAs resulted in lower HbA_1c levels and lower incidence of hypoglycemia.^6,9 Data have demonstrated the complementary mechanisms of using basal insulin and GLP 1 RAs in decreasing HbA_1c levels, insulin requirements, and weight compared with using basal insulin monotherapy and basal/bolus combinations.^6,9-13 Moreover, 3 GLP-1 RA medications currently on the market (liraglutide, dulaglutide, and semaglutide) have displayed cardiovascular and renal benefits, further supporting the use of these medications.^2,5

Despite these benefits, GLP-1 RAs may have bothersome AEs and are associated with a high cost.⁶ In addition, some studies have found that as the length of therapy increases, the positive effects of these agents may diminish.^9,11 In one study, which looked at the impact of the addition of exenatide to patients taking basal or basal/bolus insulin regimens, mean changes in weight were −2.4 kg at 0 to 6 months, −4.3 kg at 6 to 12 months, −6.2 kg at 12 to 18 months, and −5.5 kg at 18 to 27 months. After 18 months, an increase in weight was observed, but the increase remained lower than baseline.¹¹ Another study, conducted over 12 months, found no significant decrease in weight or total daily dose (TDD) of insulin when exenatide or liraglutide were added to various insulin regimens (basal or basal/bolus).¹³ To date, minimal published data exist regarding the addition of newer GLP-1 RAs and the long-term use of these agents beyond 12 months in patients taking basal/bolus insulin regimens. The primary goal of this study was to evaluate the effect of adding GLP-1 RAs to basal/bolus insulin regimens over a 24-month period.

Methods

This study was a retrospective, electronic health record review of all patients on basal and bolus insulin regimens who received additional therapy with a GLP-1 RA at Veteran Health Indiana in Indianapolis from September 1, 2015, to June 30, 2019. Patients meeting inclusion criteria served as their own control. The primary outcome was change in HbA_1c at 3, 6, 12, 18, and 24 months after initiation of the GLP-1 RA. Secondary outcomes included change in weight and TDD of insulin at 3, 6, 12, 18, and 24 months after the initiation of the GLP-1 RAs and incidence of patient-reported or laboratory-confirmed hypoglycemia and other AEs.

Patients were included if they were aged ≥ 18 years with a diagnosis of T2DM, had concomitant prescriptions for both a basal insulin (glargine, detemir, or NPH) and a bolus insulin (aspart, lispro, or regular) before receiving add-on therapy with a GLP-1 RA (exenatide, liraglutide, albiglutide, lixisenatide, dulaglutide, or semaglutide) from September 1, 2015, to June 30, 2019, and had baseline and subsequent Hb A_1c measurements available in the electronic health record. Patients were excluded if they had a diagnosis of T1DM, were followed by an outside clinician for DM care, or if the GLP-1 RA was discontinued before subsequent HbA_1c measurement. The study protocol was approved by the Research and Development Office of Veteran Health Indiana, and the project was deemed exempt from review by the Indiana University Institutional Review Board due to the retrospective nature of the study.

Data analysis was performed using Excel. Change from baseline for each interval was computed, and 1 sample t tests (2-tailed) compared change from baseline to no change. Due to the disparity in the number of patients with data available at each of the time intervals, a mean plot was presented for each group of patients within each interval, allowing mean changes in individual groups to be observed over time.

Results

One hundred twenty-three subjects met inclusion criteria; 16 patients were excluded due to GLP-1 RA discontinuation before follow-up measurement of HbA_1c; 14 were excluded due to patients being managed by a clinician outside of the facility; 1 patient was excluded for lack of documentation regarding baseline and subsequent insulin doses. Ninety-two patient charts were reviewed. Participants had a mean age of 64 years, 95% were male, and 89% were White. Mean baseline Hb A_1c was 9.2%, mean body mass index was 38.9, and the mean TDD of insulin was 184 units. Mean duration of DM was 10 years, and mean use of basal/bolus insulin regimen was 6.1 years. Most participants (91%) used an insulin regimen containing insulin glargine and insulin aspart; the remaining participants used insulin detemir and insulin aspart. Semaglutide and liraglutide were the most commonly used GLP-1 RAs (44% and 39%, respectively) (Table 1).

Since some patients switched between GLP-1 RAs throughout the study and there was variation in timing of laboratory and clinic follow-up, a different number of patient charts were available for review at each period (Table 2). Glycemic control was significantly improved at all time points when compared with baseline, but over time the benefit declined. The mean change in HbA_1c was −1.1% (95% CI, −1.3 to −0.8; P < .001) at 3 months; −1.0% (95% CI, −1.3 to −0.7; P < .001) at 6 months; −0.9% (95% CI, −1.3 to −0.6; P < .001) at 12 months; −0.9% (95% CI −1.4 to −0.3; P = .002) at 18 months; and −0.7% (95% CI, −1.4 to 0.1; P = .07) at 24 months (Figure 1). Mean weight decreased from baseline −2.7 kg (95% CI, −3.7 to −1.6; P < .001); −4.4 kg (95% CI −5.7 to −3.2; P < .001) at 6 months; −3.9 kg (95% CI −6.0 to −1.9; P < .001) at 12 months; −4.7 kg (95% CI −6.7 to −2.6; P < .001) at 18 months; and −2.8 kg (95% CI, −5.9 to 0.3; P = .07) at 24 months (Figure 2). Mean TDD decreased at 3 months −12 units (95% CI, −19 to −5; P < .001); −18 units (95% CI, −27 to −9; P < .001) at 6 months; −14 units (95% CI, −24 to −5; P = .004) at 12 months; −9 units (95% CI, −21 to 3; P = .15) at 18 months; and −18 units (95% CI, −43 to 5 units; P = .12) at 24 months (Figure 3). The most common AEs were hypoglycemia (30%), diarrhea (11%), nausea (4%), and abdominal pain (3%).

Discussion

Adding a GLP-1 RA to basal/bolus insulin regimens was associated with a statistically significant decrease in HbA_1c at each time point through 18 months. The greatest improvement in glycemic control from baseline was seen at 3 months, with improvements in HbA_1c diminishing at each subsequent period. The study also demonstrated a significant decrease in weight at each time point through 18 months. The greatest decrease in weight was observed at both 6 and 12 months. Statistically significant decreases in TDD were observed at 3, 6, and 12 months. Insulin changes after 12 months were not found to be statistically significant.

Few studies have previously evaluated the use of GLP-1 RAs in patients with T2DM who are already taking basal/bolus insulin regimens. Gyorffy and colleagues reported significant improvements in glycemic control at 3 and 6 months in a sample of 54 patients taking basal/bolus insulin when liraglutide or exenatide was added, although statistical significance was not found at the final 12-month time point.¹³ That study also found a significant decrease in weight at 6 months; however there was not a significant reduction in weight at both 3 and 12 months of GLP-1 RA therapy. There was not a significant decrease in TDD at any of the collected time points. Nonetheless, Gyorffy and colleagues concluded that reduction in TDD leveled off after 12 months, which is consistent with this study’s findings. The small size of the study may have limited the ability to detect statistical significance; however, this study was conducted in a population that was racially diverse and included a higher proportion of women, though average age was similar.¹³

Yoon and colleagues reported weight loss through 18 months, then saw weight increase, though weights did remain lover than baseline. The study also showed no significant change in TDD of insulin after 12 months of concomitant exenatide and insulin therapy.¹¹ Although these results mirror the outcomes observed in this study, Yoon and colleagues did not differentiate results between basal and basal/bolus insulin groups.¹¹ Seino and colleagues observed no significant change in weight after 36 weeks of GLP-1 RA therapy in Japanese patients when used with basal and basal/bolus insulin regimens. Despite the consideration that the population in the study was not overweight (mean body mass index was 25.6), the results of these studies support the idea that effects of GLP-1 RAs on weight and TDD may diminish over time.¹⁴

Within the VHA, GLP-1 RAs are nonformulary medications. Patients must meet certain criteria in order to be approved for these agents, which may include diagnosis of CVD, renal disease, or failure to reach glycemic control with the use of oral agents or insulin. Therefore, participants of this study represent a particular subset of VHA patients, many of whom may have been selected for consideration due to long-standing or uncontrolled T2DM and failure of previous therapies. The baseline demographics support this idea, given poor glycemic control at baseline and high insulin requirements. Once approved for GLP-1 RA therapy, semaglutide is currently the preferred agent within the VHA, with other agents being available for select considerations. It should be noted that albiglutide, which was the primary agent selected for some of the patients included in this study, was removed from the market in 2017 for economic considerations.¹⁵ In the case for these patients, a conversion to a formulary-preferred GLP-1 RA was made.

Most of the patients included in this study (70%) were maintained on metformin from baseline throughout the study period. Fifty-seven percent of patients were taking TDD of insulin > 150 units. Considering the significant cost of concentrated insulins, the addition of GLP-1 RAs to standard insulin may prove to be beneficial from a cost standpoint. Additional research in this area may be warranted to establish more data regarding this potential benefit of GLP-1 RAs as add-on therapy.

Limitations and Strengths

Limitations of this study include a small patient population and a gradual reduction in available data as time periods progressed, making even smaller sample sizes for subsequent time periods. A majority of participants were older males of White race. This could have limited the determination of statistical significance and applicability of the results to other patient populations. Another potential limitation was the retrospective nature of the study design, which may have limited reporting of hypoglycemia and other AEs based on the documentation of the clinician.

Strengths included the length of study duration and the diversity of GLP-1 RAs used by participants, as the impact of many of these agents has not yet been assessed in the literature. In addition, the retrospective nature of the study allows for a more realistic representation of patient adherence, education, and motivation, which are likely different from those of patients included in prospective clinical trials.

There are no clear guidelines dictating the optimal duration of concomitant GLP-1 RA and insulin therapy; however, our study suggests that there may be continued benefits past short-term use. Also our study suggests that patients with T2DM treated with basal/bolus insulin regimens may glean additional benefit from adding GLP-1 RAs; however, further randomized, controlled studies are warranted, particularly in poorly controlled patients requiring even more aggressive treatment regimens, such as concentrated insulins.

Conclusions

In our study, adding GLP-1 RA to basal/bolus insulin was associated with a significant decrease in HbA_1c from baseline through 18 months. An overall decrease in weight and TDD of insulin was observed through 24 months, but the change in weight was not significant past 18 months, and the change in insulin requirement was not significant past 12 months. Hypoglycemia was observed in almost one-third of patients, and gastrointestinal symptoms were the most common AE observed as a result adding GLP-1 RAs. More studies are needed to better evaluate the durability and cost benefit of GLP-1 RAs, especially in patients with high insulin requirements.

Acknowledgments

This material is the result of work supported with resources and facilities at Veteran Health Indiana in Indianapolis. Study data were collected and managed using REDCap electronic data capture tools hosted at Veteran Health Indiana. The authors also acknowledge George Eckert for his assistance with data analysis.

In 2019, diabetes mellitus (DM) was the seventh leading cause of death in the United States, and currently, about 11% of the American population has a DM diagnosis.¹ Most have a diagnosis of type 2 diabetes (T2DM), which has a strong genetic predisposition, and the risk of developing T2DM increases with age, obesity, and lack of physical activity.^1,2 Nearly one-quarter of veterans have a diagnosis of DM, and DM is the leading cause of comorbidities, such as blindness, end-stage renal disease, and amputation for patients receiving care from the Veterans Health Administration (VHA).² The elevated incidence of DM in the veteran population is attributed to a variety of factors, including exposure to herbicides, such as Agent Orange, advanced age, increased risk of obesity, and limited access to high-quality food.³

After diagnosis, both the American Diabetes Association (ADA) and the American Association of Clinical Endocrinologists and American College of Endocrinology (AACE/ACE) emphasize the appropriate use of lifestyle management and pharmacologic therapy for DM care. The use of pharmacologic agents (oral medications, insulin, or noninsulin injectables) is often determined by efficacy, cost, potential adverse effects (AEs), and patient factors and comorbidities.^4,5

The initial recommendation for pharmacologic treatment for T2DM differs slightly between expert guidelines. The ADA and AACE/ACE recommend any of the following as initial monotherapy, listed in order to represent a hierarchy of usage: metformin, glucagon-like peptide-1 receptor agonists (GLP-1 RAs), sodium-glucose cotransporter 2 (SGLT-2) inhibitors, or dipeptidyl peptidase-4 (DPP-4) inhibitors, with the first 3 agents carrying the strongest recommendations.^4,5 For patients with established atherosclerotic cardiovascular disease (CVD), chronic kidney disease, or heart failure, it is recommended to start a long-acting GLP-1 RA or SGLT-2 inhibitor. For patients with T2DM and hemoglobin A_1c (HbA_1c) between 7.5% and 9.0% at diagnosis, the AACE/ACE recommend initiation of dual therapy using metformin alongside another first-line agent and recommend the addition of another antidiabetic agent if glycemic goals are not met after regular follow-up. AACE/ACE recommend the consideration of insulin therapy in symptomatic patients with HbA_1c > 9.0%.⁵ In contrast, the ADA recommends metformin as first-line therapy for all patients with T2DM and recommends dual therapy using metformin and another preferred agent (selection based on comorbidities) when HbA_1c is 1.5% to 2% above target. The ADA recommends the consideration of insulin with HbA_1c > 10% or with evidence of ongoing catabolism or symptoms of hyperglycemia.⁴ There are several reasons why insulin may be initiated prior to GLP-1 RAs, including profound hyperglycemia at time of diagnosis or implementation of insulin agents prior to commercial availability of GLP-1 RA.

GLP-1 RAs are analogs of the hormone incretin, which increases glucose-dependent insulin secretion, decreases postprandial glucagon secretion, increases satiety, and slows gastric emptying.^6,7 When used in combination with noninsulin agents, GLP-1 RAs have demonstrated HbA_1c reductions of 0.5% to 1.5%.⁸ The use of GLP-1 RAs with basal insulin also has been studied extensively.^6,8-10 When the combination of GLP-1 RAs and basal insulin was compared with basal/bolus insulin regimens, the use of the GLP-1 RAs resulted in lower HbA_1c levels and lower incidence of hypoglycemia.^6,9 Data have demonstrated the complementary mechanisms of using basal insulin and GLP 1 RAs in decreasing HbA_1c levels, insulin requirements, and weight compared with using basal insulin monotherapy and basal/bolus combinations.^6,9-13 Moreover, 3 GLP-1 RA medications currently on the market (liraglutide, dulaglutide, and semaglutide) have displayed cardiovascular and renal benefits, further supporting the use of these medications.^2,5

Despite these benefits, GLP-1 RAs may have bothersome AEs and are associated with a high cost.⁶ In addition, some studies have found that as the length of therapy increases, the positive effects of these agents may diminish.^9,11 In one study, which looked at the impact of the addition of exenatide to patients taking basal or basal/bolus insulin regimens, mean changes in weight were −2.4 kg at 0 to 6 months, −4.3 kg at 6 to 12 months, −6.2 kg at 12 to 18 months, and −5.5 kg at 18 to 27 months. After 18 months, an increase in weight was observed, but the increase remained lower than baseline.¹¹ Another study, conducted over 12 months, found no significant decrease in weight or total daily dose (TDD) of insulin when exenatide or liraglutide were added to various insulin regimens (basal or basal/bolus).¹³ To date, minimal published data exist regarding the addition of newer GLP-1 RAs and the long-term use of these agents beyond 12 months in patients taking basal/bolus insulin regimens. The primary goal of this study was to evaluate the effect of adding GLP-1 RAs to basal/bolus insulin regimens over a 24-month period.

Methods

This study was a retrospective, electronic health record review of all patients on basal and bolus insulin regimens who received additional therapy with a GLP-1 RA at Veteran Health Indiana in Indianapolis from September 1, 2015, to June 30, 2019. Patients meeting inclusion criteria served as their own control. The primary outcome was change in HbA_1c at 3, 6, 12, 18, and 24 months after initiation of the GLP-1 RA. Secondary outcomes included change in weight and TDD of insulin at 3, 6, 12, 18, and 24 months after the initiation of the GLP-1 RAs and incidence of patient-reported or laboratory-confirmed hypoglycemia and other AEs.

Patients were included if they were aged ≥ 18 years with a diagnosis of T2DM, had concomitant prescriptions for both a basal insulin (glargine, detemir, or NPH) and a bolus insulin (aspart, lispro, or regular) before receiving add-on therapy with a GLP-1 RA (exenatide, liraglutide, albiglutide, lixisenatide, dulaglutide, or semaglutide) from September 1, 2015, to June 30, 2019, and had baseline and subsequent Hb A_1c measurements available in the electronic health record. Patients were excluded if they had a diagnosis of T1DM, were followed by an outside clinician for DM care, or if the GLP-1 RA was discontinued before subsequent HbA_1c measurement. The study protocol was approved by the Research and Development Office of Veteran Health Indiana, and the project was deemed exempt from review by the Indiana University Institutional Review Board due to the retrospective nature of the study.

Data analysis was performed using Excel. Change from baseline for each interval was computed, and 1 sample t tests (2-tailed) compared change from baseline to no change. Due to the disparity in the number of patients with data available at each of the time intervals, a mean plot was presented for each group of patients within each interval, allowing mean changes in individual groups to be observed over time.

Results

One hundred twenty-three subjects met inclusion criteria; 16 patients were excluded due to GLP-1 RA discontinuation before follow-up measurement of HbA_1c; 14 were excluded due to patients being managed by a clinician outside of the facility; 1 patient was excluded for lack of documentation regarding baseline and subsequent insulin doses. Ninety-two patient charts were reviewed. Participants had a mean age of 64 years, 95% were male, and 89% were White. Mean baseline Hb A_1c was 9.2%, mean body mass index was 38.9, and the mean TDD of insulin was 184 units. Mean duration of DM was 10 years, and mean use of basal/bolus insulin regimen was 6.1 years. Most participants (91%) used an insulin regimen containing insulin glargine and insulin aspart; the remaining participants used insulin detemir and insulin aspart. Semaglutide and liraglutide were the most commonly used GLP-1 RAs (44% and 39%, respectively) (Table 1).

Since some patients switched between GLP-1 RAs throughout the study and there was variation in timing of laboratory and clinic follow-up, a different number of patient charts were available for review at each period (Table 2). Glycemic control was significantly improved at all time points when compared with baseline, but over time the benefit declined. The mean change in HbA_1c was −1.1% (95% CI, −1.3 to −0.8; P < .001) at 3 months; −1.0% (95% CI, −1.3 to −0.7; P < .001) at 6 months; −0.9% (95% CI, −1.3 to −0.6; P < .001) at 12 months; −0.9% (95% CI −1.4 to −0.3; P = .002) at 18 months; and −0.7% (95% CI, −1.4 to 0.1; P = .07) at 24 months (Figure 1). Mean weight decreased from baseline −2.7 kg (95% CI, −3.7 to −1.6; P < .001); −4.4 kg (95% CI −5.7 to −3.2; P < .001) at 6 months; −3.9 kg (95% CI −6.0 to −1.9; P < .001) at 12 months; −4.7 kg (95% CI −6.7 to −2.6; P < .001) at 18 months; and −2.8 kg (95% CI, −5.9 to 0.3; P = .07) at 24 months (Figure 2). Mean TDD decreased at 3 months −12 units (95% CI, −19 to −5; P < .001); −18 units (95% CI, −27 to −9; P < .001) at 6 months; −14 units (95% CI, −24 to −5; P = .004) at 12 months; −9 units (95% CI, −21 to 3; P = .15) at 18 months; and −18 units (95% CI, −43 to 5 units; P = .12) at 24 months (Figure 3). The most common AEs were hypoglycemia (30%), diarrhea (11%), nausea (4%), and abdominal pain (3%).

Discussion

Adding a GLP-1 RA to basal/bolus insulin regimens was associated with a statistically significant decrease in HbA_1c at each time point through 18 months. The greatest improvement in glycemic control from baseline was seen at 3 months, with improvements in HbA_1c diminishing at each subsequent period. The study also demonstrated a significant decrease in weight at each time point through 18 months. The greatest decrease in weight was observed at both 6 and 12 months. Statistically significant decreases in TDD were observed at 3, 6, and 12 months. Insulin changes after 12 months were not found to be statistically significant.

Few studies have previously evaluated the use of GLP-1 RAs in patients with T2DM who are already taking basal/bolus insulin regimens. Gyorffy and colleagues reported significant improvements in glycemic control at 3 and 6 months in a sample of 54 patients taking basal/bolus insulin when liraglutide or exenatide was added, although statistical significance was not found at the final 12-month time point.¹³ That study also found a significant decrease in weight at 6 months; however there was not a significant reduction in weight at both 3 and 12 months of GLP-1 RA therapy. There was not a significant decrease in TDD at any of the collected time points. Nonetheless, Gyorffy and colleagues concluded that reduction in TDD leveled off after 12 months, which is consistent with this study’s findings. The small size of the study may have limited the ability to detect statistical significance; however, this study was conducted in a population that was racially diverse and included a higher proportion of women, though average age was similar.¹³

Yoon and colleagues reported weight loss through 18 months, then saw weight increase, though weights did remain lover than baseline. The study also showed no significant change in TDD of insulin after 12 months of concomitant exenatide and insulin therapy.¹¹ Although these results mirror the outcomes observed in this study, Yoon and colleagues did not differentiate results between basal and basal/bolus insulin groups.¹¹ Seino and colleagues observed no significant change in weight after 36 weeks of GLP-1 RA therapy in Japanese patients when used with basal and basal/bolus insulin regimens. Despite the consideration that the population in the study was not overweight (mean body mass index was 25.6), the results of these studies support the idea that effects of GLP-1 RAs on weight and TDD may diminish over time.¹⁴

Within the VHA, GLP-1 RAs are nonformulary medications. Patients must meet certain criteria in order to be approved for these agents, which may include diagnosis of CVD, renal disease, or failure to reach glycemic control with the use of oral agents or insulin. Therefore, participants of this study represent a particular subset of VHA patients, many of whom may have been selected for consideration due to long-standing or uncontrolled T2DM and failure of previous therapies. The baseline demographics support this idea, given poor glycemic control at baseline and high insulin requirements. Once approved for GLP-1 RA therapy, semaglutide is currently the preferred agent within the VHA, with other agents being available for select considerations. It should be noted that albiglutide, which was the primary agent selected for some of the patients included in this study, was removed from the market in 2017 for economic considerations.¹⁵ In the case for these patients, a conversion to a formulary-preferred GLP-1 RA was made.

Most of the patients included in this study (70%) were maintained on metformin from baseline throughout the study period. Fifty-seven percent of patients were taking TDD of insulin > 150 units. Considering the significant cost of concentrated insulins, the addition of GLP-1 RAs to standard insulin may prove to be beneficial from a cost standpoint. Additional research in this area may be warranted to establish more data regarding this potential benefit of GLP-1 RAs as add-on therapy.

Limitations and Strengths

Limitations of this study include a small patient population and a gradual reduction in available data as time periods progressed, making even smaller sample sizes for subsequent time periods. A majority of participants were older males of White race. This could have limited the determination of statistical significance and applicability of the results to other patient populations. Another potential limitation was the retrospective nature of the study design, which may have limited reporting of hypoglycemia and other AEs based on the documentation of the clinician.

Strengths included the length of study duration and the diversity of GLP-1 RAs used by participants, as the impact of many of these agents has not yet been assessed in the literature. In addition, the retrospective nature of the study allows for a more realistic representation of patient adherence, education, and motivation, which are likely different from those of patients included in prospective clinical trials.

There are no clear guidelines dictating the optimal duration of concomitant GLP-1 RA and insulin therapy; however, our study suggests that there may be continued benefits past short-term use. Also our study suggests that patients with T2DM treated with basal/bolus insulin regimens may glean additional benefit from adding GLP-1 RAs; however, further randomized, controlled studies are warranted, particularly in poorly controlled patients requiring even more aggressive treatment regimens, such as concentrated insulins.

Conclusions

In our study, adding GLP-1 RA to basal/bolus insulin was associated with a significant decrease in HbA_1c from baseline through 18 months. An overall decrease in weight and TDD of insulin was observed through 24 months, but the change in weight was not significant past 18 months, and the change in insulin requirement was not significant past 12 months. Hypoglycemia was observed in almost one-third of patients, and gastrointestinal symptoms were the most common AE observed as a result adding GLP-1 RAs. More studies are needed to better evaluate the durability and cost benefit of GLP-1 RAs, especially in patients with high insulin requirements.

Acknowledgments

This material is the result of work supported with resources and facilities at Veteran Health Indiana in Indianapolis. Study data were collected and managed using REDCap electronic data capture tools hosted at Veteran Health Indiana. The authors also acknowledge George Eckert for his assistance with data analysis.

In 2019, diabetes mellitus (DM) was the seventh leading cause of death in the United States, and currently, about 11% of the American population has a DM diagnosis.¹ Most have a diagnosis of type 2 diabetes (T2DM), which has a strong genetic predisposition, and the risk of developing T2DM increases with age, obesity, and lack of physical activity.^1,2 Nearly one-quarter of veterans have a diagnosis of DM, and DM is the leading cause of comorbidities, such as blindness, end-stage renal disease, and amputation for patients receiving care from the Veterans Health Administration (VHA).² The elevated incidence of DM in the veteran population is attributed to a variety of factors, including exposure to herbicides, such as Agent Orange, advanced age, increased risk of obesity, and limited access to high-quality food.³

After diagnosis, both the American Diabetes Association (ADA) and the American Association of Clinical Endocrinologists and American College of Endocrinology (AACE/ACE) emphasize the appropriate use of lifestyle management and pharmacologic therapy for DM care. The use of pharmacologic agents (oral medications, insulin, or noninsulin injectables) is often determined by efficacy, cost, potential adverse effects (AEs), and patient factors and comorbidities.^4,5

The initial recommendation for pharmacologic treatment for T2DM differs slightly between expert guidelines. The ADA and AACE/ACE recommend any of the following as initial monotherapy, listed in order to represent a hierarchy of usage: metformin, glucagon-like peptide-1 receptor agonists (GLP-1 RAs), sodium-glucose cotransporter 2 (SGLT-2) inhibitors, or dipeptidyl peptidase-4 (DPP-4) inhibitors, with the first 3 agents carrying the strongest recommendations.^4,5 For patients with established atherosclerotic cardiovascular disease (CVD), chronic kidney disease, or heart failure, it is recommended to start a long-acting GLP-1 RA or SGLT-2 inhibitor. For patients with T2DM and hemoglobin A_1c (HbA_1c) between 7.5% and 9.0% at diagnosis, the AACE/ACE recommend initiation of dual therapy using metformin alongside another first-line agent and recommend the addition of another antidiabetic agent if glycemic goals are not met after regular follow-up. AACE/ACE recommend the consideration of insulin therapy in symptomatic patients with HbA_1c > 9.0%.⁵ In contrast, the ADA recommends metformin as first-line therapy for all patients with T2DM and recommends dual therapy using metformin and another preferred agent (selection based on comorbidities) when HbA_1c is 1.5% to 2% above target. The ADA recommends the consideration of insulin with HbA_1c > 10% or with evidence of ongoing catabolism or symptoms of hyperglycemia.⁴ There are several reasons why insulin may be initiated prior to GLP-1 RAs, including profound hyperglycemia at time of diagnosis or implementation of insulin agents prior to commercial availability of GLP-1 RA.

GLP-1 RAs are analogs of the hormone incretin, which increases glucose-dependent insulin secretion, decreases postprandial glucagon secretion, increases satiety, and slows gastric emptying.^6,7 When used in combination with noninsulin agents, GLP-1 RAs have demonstrated HbA_1c reductions of 0.5% to 1.5%.⁸ The use of GLP-1 RAs with basal insulin also has been studied extensively.^6,8-10 When the combination of GLP-1 RAs and basal insulin was compared with basal/bolus insulin regimens, the use of the GLP-1 RAs resulted in lower HbA_1c levels and lower incidence of hypoglycemia.^6,9 Data have demonstrated the complementary mechanisms of using basal insulin and GLP 1 RAs in decreasing HbA_1c levels, insulin requirements, and weight compared with using basal insulin monotherapy and basal/bolus combinations.^6,9-13 Moreover, 3 GLP-1 RA medications currently on the market (liraglutide, dulaglutide, and semaglutide) have displayed cardiovascular and renal benefits, further supporting the use of these medications.^2,5

Despite these benefits, GLP-1 RAs may have bothersome AEs and are associated with a high cost.⁶ In addition, some studies have found that as the length of therapy increases, the positive effects of these agents may diminish.^9,11 In one study, which looked at the impact of the addition of exenatide to patients taking basal or basal/bolus insulin regimens, mean changes in weight were −2.4 kg at 0 to 6 months, −4.3 kg at 6 to 12 months, −6.2 kg at 12 to 18 months, and −5.5 kg at 18 to 27 months. After 18 months, an increase in weight was observed, but the increase remained lower than baseline.¹¹ Another study, conducted over 12 months, found no significant decrease in weight or total daily dose (TDD) of insulin when exenatide or liraglutide were added to various insulin regimens (basal or basal/bolus).¹³ To date, minimal published data exist regarding the addition of newer GLP-1 RAs and the long-term use of these agents beyond 12 months in patients taking basal/bolus insulin regimens. The primary goal of this study was to evaluate the effect of adding GLP-1 RAs to basal/bolus insulin regimens over a 24-month period.

Methods

This study was a retrospective, electronic health record review of all patients on basal and bolus insulin regimens who received additional therapy with a GLP-1 RA at Veteran Health Indiana in Indianapolis from September 1, 2015, to June 30, 2019. Patients meeting inclusion criteria served as their own control. The primary outcome was change in HbA_1c at 3, 6, 12, 18, and 24 months after initiation of the GLP-1 RA. Secondary outcomes included change in weight and TDD of insulin at 3, 6, 12, 18, and 24 months after the initiation of the GLP-1 RAs and incidence of patient-reported or laboratory-confirmed hypoglycemia and other AEs.

Patients were included if they were aged ≥ 18 years with a diagnosis of T2DM, had concomitant prescriptions for both a basal insulin (glargine, detemir, or NPH) and a bolus insulin (aspart, lispro, or regular) before receiving add-on therapy with a GLP-1 RA (exenatide, liraglutide, albiglutide, lixisenatide, dulaglutide, or semaglutide) from September 1, 2015, to June 30, 2019, and had baseline and subsequent Hb A_1c measurements available in the electronic health record. Patients were excluded if they had a diagnosis of T1DM, were followed by an outside clinician for DM care, or if the GLP-1 RA was discontinued before subsequent HbA_1c measurement. The study protocol was approved by the Research and Development Office of Veteran Health Indiana, and the project was deemed exempt from review by the Indiana University Institutional Review Board due to the retrospective nature of the study.

Data analysis was performed using Excel. Change from baseline for each interval was computed, and 1 sample t tests (2-tailed) compared change from baseline to no change. Due to the disparity in the number of patients with data available at each of the time intervals, a mean plot was presented for each group of patients within each interval, allowing mean changes in individual groups to be observed over time.

Results

One hundred twenty-three subjects met inclusion criteria; 16 patients were excluded due to GLP-1 RA discontinuation before follow-up measurement of HbA_1c; 14 were excluded due to patients being managed by a clinician outside of the facility; 1 patient was excluded for lack of documentation regarding baseline and subsequent insulin doses. Ninety-two patient charts were reviewed. Participants had a mean age of 64 years, 95% were male, and 89% were White. Mean baseline Hb A_1c was 9.2%, mean body mass index was 38.9, and the mean TDD of insulin was 184 units. Mean duration of DM was 10 years, and mean use of basal/bolus insulin regimen was 6.1 years. Most participants (91%) used an insulin regimen containing insulin glargine and insulin aspart; the remaining participants used insulin detemir and insulin aspart. Semaglutide and liraglutide were the most commonly used GLP-1 RAs (44% and 39%, respectively) (Table 1).

Since some patients switched between GLP-1 RAs throughout the study and there was variation in timing of laboratory and clinic follow-up, a different number of patient charts were available for review at each period (Table 2). Glycemic control was significantly improved at all time points when compared with baseline, but over time the benefit declined. The mean change in HbA_1c was −1.1% (95% CI, −1.3 to −0.8; P < .001) at 3 months; −1.0% (95% CI, −1.3 to −0.7; P < .001) at 6 months; −0.9% (95% CI, −1.3 to −0.6; P < .001) at 12 months; −0.9% (95% CI −1.4 to −0.3; P = .002) at 18 months; and −0.7% (95% CI, −1.4 to 0.1; P = .07) at 24 months (Figure 1). Mean weight decreased from baseline −2.7 kg (95% CI, −3.7 to −1.6; P < .001); −4.4 kg (95% CI −5.7 to −3.2; P < .001) at 6 months; −3.9 kg (95% CI −6.0 to −1.9; P < .001) at 12 months; −4.7 kg (95% CI −6.7 to −2.6; P < .001) at 18 months; and −2.8 kg (95% CI, −5.9 to 0.3; P = .07) at 24 months (Figure 2). Mean TDD decreased at 3 months −12 units (95% CI, −19 to −5; P < .001); −18 units (95% CI, −27 to −9; P < .001) at 6 months; −14 units (95% CI, −24 to −5; P = .004) at 12 months; −9 units (95% CI, −21 to 3; P = .15) at 18 months; and −18 units (95% CI, −43 to 5 units; P = .12) at 24 months (Figure 3). The most common AEs were hypoglycemia (30%), diarrhea (11%), nausea (4%), and abdominal pain (3%).

Discussion

Adding a GLP-1 RA to basal/bolus insulin regimens was associated with a statistically significant decrease in HbA_1c at each time point through 18 months. The greatest improvement in glycemic control from baseline was seen at 3 months, with improvements in HbA_1c diminishing at each subsequent period. The study also demonstrated a significant decrease in weight at each time point through 18 months. The greatest decrease in weight was observed at both 6 and 12 months. Statistically significant decreases in TDD were observed at 3, 6, and 12 months. Insulin changes after 12 months were not found to be statistically significant.

Few studies have previously evaluated the use of GLP-1 RAs in patients with T2DM who are already taking basal/bolus insulin regimens. Gyorffy and colleagues reported significant improvements in glycemic control at 3 and 6 months in a sample of 54 patients taking basal/bolus insulin when liraglutide or exenatide was added, although statistical significance was not found at the final 12-month time point.¹³ That study also found a significant decrease in weight at 6 months; however there was not a significant reduction in weight at both 3 and 12 months of GLP-1 RA therapy. There was not a significant decrease in TDD at any of the collected time points. Nonetheless, Gyorffy and colleagues concluded that reduction in TDD leveled off after 12 months, which is consistent with this study’s findings. The small size of the study may have limited the ability to detect statistical significance; however, this study was conducted in a population that was racially diverse and included a higher proportion of women, though average age was similar.¹³

Yoon and colleagues reported weight loss through 18 months, then saw weight increase, though weights did remain lover than baseline. The study also showed no significant change in TDD of insulin after 12 months of concomitant exenatide and insulin therapy.¹¹ Although these results mirror the outcomes observed in this study, Yoon and colleagues did not differentiate results between basal and basal/bolus insulin groups.¹¹ Seino and colleagues observed no significant change in weight after 36 weeks of GLP-1 RA therapy in Japanese patients when used with basal and basal/bolus insulin regimens. Despite the consideration that the population in the study was not overweight (mean body mass index was 25.6), the results of these studies support the idea that effects of GLP-1 RAs on weight and TDD may diminish over time.¹⁴

Within the VHA, GLP-1 RAs are nonformulary medications. Patients must meet certain criteria in order to be approved for these agents, which may include diagnosis of CVD, renal disease, or failure to reach glycemic control with the use of oral agents or insulin. Therefore, participants of this study represent a particular subset of VHA patients, many of whom may have been selected for consideration due to long-standing or uncontrolled T2DM and failure of previous therapies. The baseline demographics support this idea, given poor glycemic control at baseline and high insulin requirements. Once approved for GLP-1 RA therapy, semaglutide is currently the preferred agent within the VHA, with other agents being available for select considerations. It should be noted that albiglutide, which was the primary agent selected for some of the patients included in this study, was removed from the market in 2017 for economic considerations.¹⁵ In the case for these patients, a conversion to a formulary-preferred GLP-1 RA was made.

Most of the patients included in this study (70%) were maintained on metformin from baseline throughout the study period. Fifty-seven percent of patients were taking TDD of insulin > 150 units. Considering the significant cost of concentrated insulins, the addition of GLP-1 RAs to standard insulin may prove to be beneficial from a cost standpoint. Additional research in this area may be warranted to establish more data regarding this potential benefit of GLP-1 RAs as add-on therapy.

Limitations and Strengths

Limitations of this study include a small patient population and a gradual reduction in available data as time periods progressed, making even smaller sample sizes for subsequent time periods. A majority of participants were older males of White race. This could have limited the determination of statistical significance and applicability of the results to other patient populations. Another potential limitation was the retrospective nature of the study design, which may have limited reporting of hypoglycemia and other AEs based on the documentation of the clinician.

Strengths included the length of study duration and the diversity of GLP-1 RAs used by participants, as the impact of many of these agents has not yet been assessed in the literature. In addition, the retrospective nature of the study allows for a more realistic representation of patient adherence, education, and motivation, which are likely different from those of patients included in prospective clinical trials.

There are no clear guidelines dictating the optimal duration of concomitant GLP-1 RA and insulin therapy; however, our study suggests that there may be continued benefits past short-term use. Also our study suggests that patients with T2DM treated with basal/bolus insulin regimens may glean additional benefit from adding GLP-1 RAs; however, further randomized, controlled studies are warranted, particularly in poorly controlled patients requiring even more aggressive treatment regimens, such as concentrated insulins.

Conclusions

In our study, adding GLP-1 RA to basal/bolus insulin was associated with a significant decrease in HbA_1c from baseline through 18 months. An overall decrease in weight and TDD of insulin was observed through 24 months, but the change in weight was not significant past 18 months, and the change in insulin requirement was not significant past 12 months. Hypoglycemia was observed in almost one-third of patients, and gastrointestinal symptoms were the most common AE observed as a result adding GLP-1 RAs. More studies are needed to better evaluate the durability and cost benefit of GLP-1 RAs, especially in patients with high insulin requirements.

Acknowledgments

This material is the result of work supported with resources and facilities at Veteran Health Indiana in Indianapolis. Study data were collected and managed using REDCap electronic data capture tools hosted at Veteran Health Indiana. The authors also acknowledge George Eckert for his assistance with data analysis.

Even with recent pharmacologic treatment advances, narrowband UVB (NB-UVB) phototherapy remains a versatile, safe, and efficacious adjunctive or exclusive treatment for multiple dermatologic conditions, including psoriasis and atopic dermatitis. ^1-9 Some providers choose NB-UVB phototherapy as a first-line treatment for older adult patients who frequently use multiple treatment modalities for more than 1 health condition. Older adults with atopic dermatitis and psoriasis are at higher risk for comorbidities such as autoimmune disorders, diabetes mellitus, dyslipidemia, sleep disorders, neuropsychiatric disorders, and cardiovascular disease that can complicate treatment compared with their peers without these dermatologic diagnoses. ^10-12 Polypharmacy (ie, the use of 5 or more daily medications), frequently associated with these conditions, contributes to prescribers pursuing NB-UVB phototherapy as a nonpharmacologic treatment, but some providers wonder if it is as effective and safe for their older patients compared with younger patients.

In a prior study, Matthews et al¹³ reported that 96% (50/52) of patients older than 65 years achieved medium to high levels of clearance with NB-UVB phototherapy. Nonetheless, 2 other findings in this study related to the number of treatments required to achieve clearance (ie, clearance rates) and erythema rates prompted further investigation. The first finding was higher-than-expected clearance rates. Older adults had a clearance rate with a mean of 33 treatments compared to prior studies featuring mean clearance rates of 20 to 28 treatments.^7,8,14-16 This finding resembled a study in the United Kingdom¹⁷ with a median clearance rate in older adults of 30 treatments. In contrast, the median clearance rate from a study in Turkey¹⁸ was 42 treatments in older adults. We hypothesized that more photosensitizing medications used in older vs younger adults prompted more dose adjustments with NB-UVB phototherapy to avoid burning (ie, erythema) at baseline and throughout the treatment course. These dose adjustments may have increased the overall clearance rates. If true, we predicted that younger adults treated with the same protocol would have cleared more quickly, either because of age-related differences or because they likely had fewer comorbidities and therefore fewer medications.

The second finding from Matthews et al¹³ that warranted further investigation was a higher erythema rate compared to the older adult study from the United Kingdom.¹⁷ We hypothesized that potentially greater use of photosensitizing medications in the United States could explain the higher erythema rates. Although medication-induced photosensitivity is less likely with NB-UVB phototherapy than with UVA, certain medications can cause UVB photosensitivity, including thiazides, quinidine, calcium channel antagonists, phenothiazines, and nonsteroidal anti-inflammatory drugs.^8,19,20 Therefore, photosensitizing medication use either at baseline or during a course of NB-UVB phototherapy could increase the risk for erythema. Age-related skin changes also have been considered as a potential cause for erythema. One study found that the skin of older patients was more sensitive than younger patients, resulting in a lower minimal erythema dose (MED)¹⁴—the lowest UV dose that results in erythema.²¹ Others, however, found similar MEDs across age groups, but older adults experienced more intense erythema in the late phase of NB-UVB treatment.^22,23 Such conflicting findings indicate that questions remain regarding the risk for erythema in older patients and if photosensitizing medications are responsible for an increased risk.

This retrospective study aimed to determine if NB-UVB phototherapy is equally effective in both older and younger adults treated with the same protocol; to examine the association between the use of photosensitizing medications and clearance rates in both older and younger adults; and to examine the association between the use of photosensitizing medications and erythema rates in older vs younger adults.

Methods

Study Design and Patients—This retrospective cohort study used billing records to identify patients who received NB-UVB phototherapy at 3 different clinical sites within a large US health care system in Washington (Group Health Cooperative, now Kaiser Permanente Washington), serving more than 600,000 patients between January 1, 2012, and December 31, 2016. The institutional review board of Kaiser Permanente Washington Health Research Institute approved this study (IRB 1498087-4). Younger adults were classified as those 64 years or younger and older adults as those 65 years and older at the start of their phototherapy regimen. A power analysis determined that the optimal sample size for this study was 250 patients.

Individuals were excluded if they had fewer than 6 phototherapy treatments; a diagnosis of vitiligo, photosensitivity dermatitis, morphea, or pityriasis rubra pilaris; and/or treatment of the hands or feet only.

Phototherapy Protocol—Using a 48-lamp NB-UVB unit, trained phototherapy nurses provided all treatments following standardized treatment protocols¹³ based on previously published phototherapy guidelines.²⁴ Nurses determined each patient’s disease clearance level using a 3-point clearance scale (high, medium, low).¹³ Each patient’s starting dose was determined based on the estimated MED for their skin phototype. If the patient was using photosensitizing medications, the protocol indicated a need for a decreased starting dose—down 25% to 50%—depending on the presumed level of photosensitivity. All clinical sites used the same protocol, but decisions about adjustments within this range were made by individual registered nurses and dermatologists, which could lead to variability across sites. Protocols also directed nurses to query patients about specific treatment responses, including erythema, tenderness, or itching; how their condition was responding; use of photosensitizing medications; missed treatments; and placement of shielding. Doses were adjusted accordingly.

Statistical Analysis—Data were analyzed using Stata statistical software (StataCorp LLC). Univariate analyses were used to examine the data and identify outliers, bad values, and missing data, as well as to calculate descriptive statistics. Pearson χ² and Fisher exact statistics were used to calculate differences in categorical variables. Linear multivariate regression models and logistic multivariate models were used to examine statistical relationships between variables. Statistical significance was defined as P≤.05.

Results

Patient Characteristics—Medical records were reviewed for 172 patients who received phototherapy between 2012 and 2016. Patients ranged in age from 23 to 91 years, with 102 patients 64 years and younger and 70 patients 65 years and older. Tables 1 and 2 outline the patient characteristics and conditions treated.

Phototherapy Effectiveness—Narrowband UVB phototherapy was found to be equally effective in older vs younger adults, with 82.9% of older adults (n=58) achieving a high level of clearance vs 80.4% (n=82) of younger adults, and 5.7% (n=4) of older adults achieved a medium level of clearance vs 10% (n=10) of younger adults (Table 3). Although older adults had slightly faster clearance rates on average (34.6 vs 37.2 treatments), these differences were not significant.

Photosensitizing Medications, Clearance Levels, and Clearance Rates—There was no significant association between clearance levels and number of photosensitizing medications in either younger (Figure 1) or older (Figure 2) adults. There was a wide range of clearance rates in both groups (Table 3), but no relationship was identified between clearance rates and photosensitizing medications or age (Figure 3). Clinic C had higher overall clearance rates for both age groups compared to the other clinics (Figure 4), but the clearance levels were still equivalent. No consistent pattern emerged indicating that age was a factor for the slower clearance at this site, and no relationship was identified between taking photosensitizing medications and clearance levels (Fisher exact test, P=.467) or clearance rates (t[149]=0.75; P=.45).

Frequency of Treatments and Clearance Rates—Older adults more consistently completed the recommended frequency of treatments—3 times weekly—compared to younger adults (74.3% vs 58.5%). However, all patients who completed 3 treatments per week required a similar number of treatments to clear (older adults, mean [SD]: 35.7 [21.6]; younger adults, mean [SD]: 34.7 [19.0]; P=.85). Among patients completing 2 or fewer treatments per week, older adults required a mean (SD) of only 31 (9.0) treatments to clear vs 41.5 (21.3) treatments to clear for younger adults, but the difference was not statistically significant (P=.08). However, even those with suboptimal frequency ultimately achieved similar clearance levels.

Photosensitizing Medications and Erythema Rates—Many patients in both age groups took medications that listed photosensitivity as a potential side effect (77.1% of older adults and 60.8% of younger adults). Of them, most patients took only 1 or 2 photosensitizing medications. However, significantly more older patients took 3 or more photosensitizing medications (28.6% vs 12.7%; P=.01)(Table 3). Asymptomatic (grade 1) erythema was unrelated to medication use and quite common in all adults (48.6% of older adults and 60.8% of younger adults). Most patients had only a few episodes of grade 1 erythema (mean [SD], 1.2 [2.9] in older adults and 1.6 [2.2] in younger adults). More older adults had grade 2 erythema (28.6%) compared to younger adults (17.6%). Patients using 3 or more photosensitizing medications were twice as likely to experience grade 2 erythema. Grades 3 and 4 erythema were extremely rare; none of the patients stopped phototherapy because they experienced erythema.

Overall, phototherapy nurses adjusted the starting dose according to the phototype-based protocol an average of 69% of the time for patients on medications with photosensitivity listed as a potential side effect. However, the frequency depended significantly on the clinic (clinic A, 24%; clinic B, 92%; clinic C, 87%)(P≤.001). Nurses across all clinics consistently decreased the treatment dose when patients reported starting new photosensitizing medications. Patients with adjusted starting doses had slightly but not significantly higher clearance rates compared to those without (mean, 37.8 vs 35.5; t(104)=0.58; P=.56).

Comment

Comparisons to Prior Studies—This study confirmed that phototherapy is equally effective for older and younger adults, with approximately 90% reaching medium to high clearance levels with approximately 35 treatments in both groups. Prior studies of all age groups found that patients typically cleared with an average of 20 to 28 treatments.^7,8,14-16 In contrast, the findings in older adults from this study were similar to the older adult study from the United Kingdom that reported a 91% clear/near clear rate with an average of 30 treatments.¹⁷ The clearance level also was similar to the older adult study in Turkey¹⁸ that reported 73.7% (70/95) of patients with psoriasis achieved a minimum psoriasis area severity index of 75, indicating 75% improvement from baseline.

Impact of Photosensitizing Medications on Clearance—Photosensitizing medications and treatment frequency were 2 factors that might explain the slower clearance rates in younger adults. In this study, both groups of patients used similar numbers of photosensitizing medications, but more older adults were taking 3 or more medications (Table 3). We found no statistically significant relationship between taking photosensitizing medications and either the clearance rates or the level of clearance achieved in either age group.

Impact of Treatment Frequency—Weekly treatment frequency also was examined. One prior study demonstrated that treatments 3 times weekly led to a faster clearance time and higher clearance levels compared with twice-weekly treatment.⁷ When patients completed treatments twice weekly, it took an average of 1.5 times more days to clear, which impacted cost and clinical resource availability. The patients ranged in age from 17 to 80 years, but outcomes in older patients were not described separately.⁷ Interestingly, our study seemed to find a difference between age groups when the impact of treatment frequency was examined. Older adults completed nearly 4 fewer mean treatments to clear when treating less often, with more than 80% achieving high levels of clearance, whereas the younger adults required almost 7 more treatments to clear when they came in less frequently, with approximately 80% achieving a high level of clearance. As a result, our study found that in both age groups, slowing the treatment frequency extended the treatment time to clearance—more for the younger adults than the older adults—but did not significantly change the percentage of individuals reaching full clearance in either group.

Erythema Rates—There was no association between photosensitizing medications and erythema rates except when patients were taking at least 3 medications. Most medications that listed photosensitivity as a possible side effect did not specify their relevant range of UV radiation; therefore, all such medications were examined during this analysis. Prior research has shown UVB range photosensitizing medications include thiazides, quinidine, calcium channel antagonists, phenothiazines, and nonsteroidal anti-inflammatory drugs.¹⁹ A sensitivity analysis that focused only on these medications found no association between them and any particular grade of erythema. However, patients taking 3 or more of any medications listing photosensitivity as a side effect had an increased risk for grade 2 erythema.

Erythema rates in this study were consistent with a 2013 systematic review that reported 57% of patients with asymptomatic grade 1 erythema.²⁵ In the 2 other comparative older adult studies, erythema rates varied widely: 35% in a study from Turkey¹⁸compared to only1.89% in a study from the United Kingdom.¹⁷

The starting dose for NB-UVB may drive erythema rates. The current study’s protocols were based on an estimated MED that is subjectively determined by the dermatology provider’s assessment of the patient’s skin sensitivity via examination and questions to the patient about their response to environmental sun exposure (ie, burning and tanning)²⁶ and is frequently used to determine the starting dose and subsequent dose escalation. Certain medications have been found to increase photosensitivity and erythema,²⁰ which can change an individual’s MED. If photosensitizing medications are started prior to or during a course of NB-UVB without a pretreatment MED, they might increase the risk for erythema. This study did not identify specific erythema-inducing medications but did find that taking 3 or more photosensitizing medications was associated with increased episodes of grade 2 erythema. Similarly, Harrop et al⁸ found that patients who were taking photosensitizing medications were more likely to have grade 2 or higher erythema, despite baseline MED testing, which is an established safety mechanism to reduce the risk and severity of erythema.^14,20,27 The authors of a recent study of older adults in Taiwan specifically recommended MED testing due to the unpredictable influence of polypharmacy on MED calculations in this population.²⁸ Therefore, this study’s use of an estimated MED in older adults may have influenced the starting dose as well as the incidence and severity of erythemic events. Age-related skin changes likely are ruled out as a consideration for mild erythema by the similarity of grade 1 erythema rates in both older and younger adults. Other studies have identified differences between the age groups, where older patients experienced more intense erythema in the late phase of UVB treatments.^22,23 This phenomenon could increase the risk for a grade 2 erythema, which may correspond with this study’s findings.

Other potential causes of erythema were ruled out during our study, including erythema related to missed treatments and shielding mishaps. Other factors, however, may impact the level of sensitivity each patient has to phototherapy, including genetics, epigenetics, and cumulative sun damage. With NB-UVB, near-erythemogenic doses are optimal to achieve effective treatments but require a delicate balance to achieve, which may be more problematic for older adults, especially those taking several medications.

Study Limitations—Our study design made it difficult to draw conclusions about rarer dermatologic conditions. Some patients received treatments over years that were not included in the study period. Finally, power calculations suggested that our actual sample size was too small, with approximately one-third of the required sample missing.

Practical Implications—The goals of phototherapy are to achieve a high level of disease clearance with the fewest number of treatments possible and minimal side effects. Skin phototype–driven standardized doses based on estimated MED may be conservatively low to minimize the risk of side effects (eg, erythema), which could slow the treatment progression. Thus, basing the starting dose on individual MED assessments may improve clearance rates. This study also confirmed that phototherapy is safe with minimal erythema in adults of all ages. The erythema episodes that patients experienced were few and mild, but because of greater rates of grade 2 erythema in patients on 3 or more photosensitizing medications, consideration of MED testing in both age groups might optimize doses at baseline and prompt caution for subsequent dose titration in this subset of patients.

The extra staff training and patient monitoring required for MED testing likely is to add value and preserve resources if faster clearance rates could be achieved and may warrant further investigation. Phototherapy centers require standardized treatment protocols, diligent well-trained staff, and program monitoring to ensure consistent care to all patients. This study highlighted the ongoing opportunity for health care organizations to conduct evidence-based practice inquiries to continually optimize care for their patients.

Even with recent pharmacologic treatment advances, narrowband UVB (NB-UVB) phototherapy remains a versatile, safe, and efficacious adjunctive or exclusive treatment for multiple dermatologic conditions, including psoriasis and atopic dermatitis. ^1-9 Some providers choose NB-UVB phototherapy as a first-line treatment for older adult patients who frequently use multiple treatment modalities for more than 1 health condition. Older adults with atopic dermatitis and psoriasis are at higher risk for comorbidities such as autoimmune disorders, diabetes mellitus, dyslipidemia, sleep disorders, neuropsychiatric disorders, and cardiovascular disease that can complicate treatment compared with their peers without these dermatologic diagnoses. ^10-12 Polypharmacy (ie, the use of 5 or more daily medications), frequently associated with these conditions, contributes to prescribers pursuing NB-UVB phototherapy as a nonpharmacologic treatment, but some providers wonder if it is as effective and safe for their older patients compared with younger patients.

In a prior study, Matthews et al¹³ reported that 96% (50/52) of patients older than 65 years achieved medium to high levels of clearance with NB-UVB phototherapy. Nonetheless, 2 other findings in this study related to the number of treatments required to achieve clearance (ie, clearance rates) and erythema rates prompted further investigation. The first finding was higher-than-expected clearance rates. Older adults had a clearance rate with a mean of 33 treatments compared to prior studies featuring mean clearance rates of 20 to 28 treatments.^7,8,14-16 This finding resembled a study in the United Kingdom¹⁷ with a median clearance rate in older adults of 30 treatments. In contrast, the median clearance rate from a study in Turkey¹⁸ was 42 treatments in older adults. We hypothesized that more photosensitizing medications used in older vs younger adults prompted more dose adjustments with NB-UVB phototherapy to avoid burning (ie, erythema) at baseline and throughout the treatment course. These dose adjustments may have increased the overall clearance rates. If true, we predicted that younger adults treated with the same protocol would have cleared more quickly, either because of age-related differences or because they likely had fewer comorbidities and therefore fewer medications.

The second finding from Matthews et al¹³ that warranted further investigation was a higher erythema rate compared to the older adult study from the United Kingdom.¹⁷ We hypothesized that potentially greater use of photosensitizing medications in the United States could explain the higher erythema rates. Although medication-induced photosensitivity is less likely with NB-UVB phototherapy than with UVA, certain medications can cause UVB photosensitivity, including thiazides, quinidine, calcium channel antagonists, phenothiazines, and nonsteroidal anti-inflammatory drugs.^8,19,20 Therefore, photosensitizing medication use either at baseline or during a course of NB-UVB phototherapy could increase the risk for erythema. Age-related skin changes also have been considered as a potential cause for erythema. One study found that the skin of older patients was more sensitive than younger patients, resulting in a lower minimal erythema dose (MED)¹⁴—the lowest UV dose that results in erythema.²¹ Others, however, found similar MEDs across age groups, but older adults experienced more intense erythema in the late phase of NB-UVB treatment.^22,23 Such conflicting findings indicate that questions remain regarding the risk for erythema in older patients and if photosensitizing medications are responsible for an increased risk.

This retrospective study aimed to determine if NB-UVB phototherapy is equally effective in both older and younger adults treated with the same protocol; to examine the association between the use of photosensitizing medications and clearance rates in both older and younger adults; and to examine the association between the use of photosensitizing medications and erythema rates in older vs younger adults.

Methods

Study Design and Patients—This retrospective cohort study used billing records to identify patients who received NB-UVB phototherapy at 3 different clinical sites within a large US health care system in Washington (Group Health Cooperative, now Kaiser Permanente Washington), serving more than 600,000 patients between January 1, 2012, and December 31, 2016. The institutional review board of Kaiser Permanente Washington Health Research Institute approved this study (IRB 1498087-4). Younger adults were classified as those 64 years or younger and older adults as those 65 years and older at the start of their phototherapy regimen. A power analysis determined that the optimal sample size for this study was 250 patients.

Individuals were excluded if they had fewer than 6 phototherapy treatments; a diagnosis of vitiligo, photosensitivity dermatitis, morphea, or pityriasis rubra pilaris; and/or treatment of the hands or feet only.

Phototherapy Protocol—Using a 48-lamp NB-UVB unit, trained phototherapy nurses provided all treatments following standardized treatment protocols¹³ based on previously published phototherapy guidelines.²⁴ Nurses determined each patient’s disease clearance level using a 3-point clearance scale (high, medium, low).¹³ Each patient’s starting dose was determined based on the estimated MED for their skin phototype. If the patient was using photosensitizing medications, the protocol indicated a need for a decreased starting dose—down 25% to 50%—depending on the presumed level of photosensitivity. All clinical sites used the same protocol, but decisions about adjustments within this range were made by individual registered nurses and dermatologists, which could lead to variability across sites. Protocols also directed nurses to query patients about specific treatment responses, including erythema, tenderness, or itching; how their condition was responding; use of photosensitizing medications; missed treatments; and placement of shielding. Doses were adjusted accordingly.

Statistical Analysis—Data were analyzed using Stata statistical software (StataCorp LLC). Univariate analyses were used to examine the data and identify outliers, bad values, and missing data, as well as to calculate descriptive statistics. Pearson χ² and Fisher exact statistics were used to calculate differences in categorical variables. Linear multivariate regression models and logistic multivariate models were used to examine statistical relationships between variables. Statistical significance was defined as P≤.05.

Results

Patient Characteristics—Medical records were reviewed for 172 patients who received phototherapy between 2012 and 2016. Patients ranged in age from 23 to 91 years, with 102 patients 64 years and younger and 70 patients 65 years and older. Tables 1 and 2 outline the patient characteristics and conditions treated.

Phototherapy Effectiveness—Narrowband UVB phototherapy was found to be equally effective in older vs younger adults, with 82.9% of older adults (n=58) achieving a high level of clearance vs 80.4% (n=82) of younger adults, and 5.7% (n=4) of older adults achieved a medium level of clearance vs 10% (n=10) of younger adults (Table 3). Although older adults had slightly faster clearance rates on average (34.6 vs 37.2 treatments), these differences were not significant.

Photosensitizing Medications, Clearance Levels, and Clearance Rates—There was no significant association between clearance levels and number of photosensitizing medications in either younger (Figure 1) or older (Figure 2) adults. There was a wide range of clearance rates in both groups (Table 3), but no relationship was identified between clearance rates and photosensitizing medications or age (Figure 3). Clinic C had higher overall clearance rates for both age groups compared to the other clinics (Figure 4), but the clearance levels were still equivalent. No consistent pattern emerged indicating that age was a factor for the slower clearance at this site, and no relationship was identified between taking photosensitizing medications and clearance levels (Fisher exact test, P=.467) or clearance rates (t[149]=0.75; P=.45).

Frequency of Treatments and Clearance Rates—Older adults more consistently completed the recommended frequency of treatments—3 times weekly—compared to younger adults (74.3% vs 58.5%). However, all patients who completed 3 treatments per week required a similar number of treatments to clear (older adults, mean [SD]: 35.7 [21.6]; younger adults, mean [SD]: 34.7 [19.0]; P=.85). Among patients completing 2 or fewer treatments per week, older adults required a mean (SD) of only 31 (9.0) treatments to clear vs 41.5 (21.3) treatments to clear for younger adults, but the difference was not statistically significant (P=.08). However, even those with suboptimal frequency ultimately achieved similar clearance levels.

Photosensitizing Medications and Erythema Rates—Many patients in both age groups took medications that listed photosensitivity as a potential side effect (77.1% of older adults and 60.8% of younger adults). Of them, most patients took only 1 or 2 photosensitizing medications. However, significantly more older patients took 3 or more photosensitizing medications (28.6% vs 12.7%; P=.01)(Table 3). Asymptomatic (grade 1) erythema was unrelated to medication use and quite common in all adults (48.6% of older adults and 60.8% of younger adults). Most patients had only a few episodes of grade 1 erythema (mean [SD], 1.2 [2.9] in older adults and 1.6 [2.2] in younger adults). More older adults had grade 2 erythema (28.6%) compared to younger adults (17.6%). Patients using 3 or more photosensitizing medications were twice as likely to experience grade 2 erythema. Grades 3 and 4 erythema were extremely rare; none of the patients stopped phototherapy because they experienced erythema.

Overall, phototherapy nurses adjusted the starting dose according to the phototype-based protocol an average of 69% of the time for patients on medications with photosensitivity listed as a potential side effect. However, the frequency depended significantly on the clinic (clinic A, 24%; clinic B, 92%; clinic C, 87%)(P≤.001). Nurses across all clinics consistently decreased the treatment dose when patients reported starting new photosensitizing medications. Patients with adjusted starting doses had slightly but not significantly higher clearance rates compared to those without (mean, 37.8 vs 35.5; t(104)=0.58; P=.56).

Comment

Comparisons to Prior Studies—This study confirmed that phototherapy is equally effective for older and younger adults, with approximately 90% reaching medium to high clearance levels with approximately 35 treatments in both groups. Prior studies of all age groups found that patients typically cleared with an average of 20 to 28 treatments.^7,8,14-16 In contrast, the findings in older adults from this study were similar to the older adult study from the United Kingdom that reported a 91% clear/near clear rate with an average of 30 treatments.¹⁷ The clearance level also was similar to the older adult study in Turkey¹⁸ that reported 73.7% (70/95) of patients with psoriasis achieved a minimum psoriasis area severity index of 75, indicating 75% improvement from baseline.

Impact of Photosensitizing Medications on Clearance—Photosensitizing medications and treatment frequency were 2 factors that might explain the slower clearance rates in younger adults. In this study, both groups of patients used similar numbers of photosensitizing medications, but more older adults were taking 3 or more medications (Table 3). We found no statistically significant relationship between taking photosensitizing medications and either the clearance rates or the level of clearance achieved in either age group.

Impact of Treatment Frequency—Weekly treatment frequency also was examined. One prior study demonstrated that treatments 3 times weekly led to a faster clearance time and higher clearance levels compared with twice-weekly treatment.⁷ When patients completed treatments twice weekly, it took an average of 1.5 times more days to clear, which impacted cost and clinical resource availability. The patients ranged in age from 17 to 80 years, but outcomes in older patients were not described separately.⁷ Interestingly, our study seemed to find a difference between age groups when the impact of treatment frequency was examined. Older adults completed nearly 4 fewer mean treatments to clear when treating less often, with more than 80% achieving high levels of clearance, whereas the younger adults required almost 7 more treatments to clear when they came in less frequently, with approximately 80% achieving a high level of clearance. As a result, our study found that in both age groups, slowing the treatment frequency extended the treatment time to clearance—more for the younger adults than the older adults—but did not significantly change the percentage of individuals reaching full clearance in either group.

Erythema Rates—There was no association between photosensitizing medications and erythema rates except when patients were taking at least 3 medications. Most medications that listed photosensitivity as a possible side effect did not specify their relevant range of UV radiation; therefore, all such medications were examined during this analysis. Prior research has shown UVB range photosensitizing medications include thiazides, quinidine, calcium channel antagonists, phenothiazines, and nonsteroidal anti-inflammatory drugs.¹⁹ A sensitivity analysis that focused only on these medications found no association between them and any particular grade of erythema. However, patients taking 3 or more of any medications listing photosensitivity as a side effect had an increased risk for grade 2 erythema.

Erythema rates in this study were consistent with a 2013 systematic review that reported 57% of patients with asymptomatic grade 1 erythema.²⁵ In the 2 other comparative older adult studies, erythema rates varied widely: 35% in a study from Turkey¹⁸compared to only1.89% in a study from the United Kingdom.¹⁷

The starting dose for NB-UVB may drive erythema rates. The current study’s protocols were based on an estimated MED that is subjectively determined by the dermatology provider’s assessment of the patient’s skin sensitivity via examination and questions to the patient about their response to environmental sun exposure (ie, burning and tanning)²⁶ and is frequently used to determine the starting dose and subsequent dose escalation. Certain medications have been found to increase photosensitivity and erythema,²⁰ which can change an individual’s MED. If photosensitizing medications are started prior to or during a course of NB-UVB without a pretreatment MED, they might increase the risk for erythema. This study did not identify specific erythema-inducing medications but did find that taking 3 or more photosensitizing medications was associated with increased episodes of grade 2 erythema. Similarly, Harrop et al⁸ found that patients who were taking photosensitizing medications were more likely to have grade 2 or higher erythema, despite baseline MED testing, which is an established safety mechanism to reduce the risk and severity of erythema.^14,20,27 The authors of a recent study of older adults in Taiwan specifically recommended MED testing due to the unpredictable influence of polypharmacy on MED calculations in this population.²⁸ Therefore, this study’s use of an estimated MED in older adults may have influenced the starting dose as well as the incidence and severity of erythemic events. Age-related skin changes likely are ruled out as a consideration for mild erythema by the similarity of grade 1 erythema rates in both older and younger adults. Other studies have identified differences between the age groups, where older patients experienced more intense erythema in the late phase of UVB treatments.^22,23 This phenomenon could increase the risk for a grade 2 erythema, which may correspond with this study’s findings.

Other potential causes of erythema were ruled out during our study, including erythema related to missed treatments and shielding mishaps. Other factors, however, may impact the level of sensitivity each patient has to phototherapy, including genetics, epigenetics, and cumulative sun damage. With NB-UVB, near-erythemogenic doses are optimal to achieve effective treatments but require a delicate balance to achieve, which may be more problematic for older adults, especially those taking several medications.

Study Limitations—Our study design made it difficult to draw conclusions about rarer dermatologic conditions. Some patients received treatments over years that were not included in the study period. Finally, power calculations suggested that our actual sample size was too small, with approximately one-third of the required sample missing.

Practical Implications—The goals of phototherapy are to achieve a high level of disease clearance with the fewest number of treatments possible and minimal side effects. Skin phototype–driven standardized doses based on estimated MED may be conservatively low to minimize the risk of side effects (eg, erythema), which could slow the treatment progression. Thus, basing the starting dose on individual MED assessments may improve clearance rates. This study also confirmed that phototherapy is safe with minimal erythema in adults of all ages. The erythema episodes that patients experienced were few and mild, but because of greater rates of grade 2 erythema in patients on 3 or more photosensitizing medications, consideration of MED testing in both age groups might optimize doses at baseline and prompt caution for subsequent dose titration in this subset of patients.

The extra staff training and patient monitoring required for MED testing likely is to add value and preserve resources if faster clearance rates could be achieved and may warrant further investigation. Phototherapy centers require standardized treatment protocols, diligent well-trained staff, and program monitoring to ensure consistent care to all patients. This study highlighted the ongoing opportunity for health care organizations to conduct evidence-based practice inquiries to continually optimize care for their patients.

Even with recent pharmacologic treatment advances, narrowband UVB (NB-UVB) phototherapy remains a versatile, safe, and efficacious adjunctive or exclusive treatment for multiple dermatologic conditions, including psoriasis and atopic dermatitis. ^1-9 Some providers choose NB-UVB phototherapy as a first-line treatment for older adult patients who frequently use multiple treatment modalities for more than 1 health condition. Older adults with atopic dermatitis and psoriasis are at higher risk for comorbidities such as autoimmune disorders, diabetes mellitus, dyslipidemia, sleep disorders, neuropsychiatric disorders, and cardiovascular disease that can complicate treatment compared with their peers without these dermatologic diagnoses. ^10-12 Polypharmacy (ie, the use of 5 or more daily medications), frequently associated with these conditions, contributes to prescribers pursuing NB-UVB phototherapy as a nonpharmacologic treatment, but some providers wonder if it is as effective and safe for their older patients compared with younger patients.

In a prior study, Matthews et al¹³ reported that 96% (50/52) of patients older than 65 years achieved medium to high levels of clearance with NB-UVB phototherapy. Nonetheless, 2 other findings in this study related to the number of treatments required to achieve clearance (ie, clearance rates) and erythema rates prompted further investigation. The first finding was higher-than-expected clearance rates. Older adults had a clearance rate with a mean of 33 treatments compared to prior studies featuring mean clearance rates of 20 to 28 treatments.^7,8,14-16 This finding resembled a study in the United Kingdom¹⁷ with a median clearance rate in older adults of 30 treatments. In contrast, the median clearance rate from a study in Turkey¹⁸ was 42 treatments in older adults. We hypothesized that more photosensitizing medications used in older vs younger adults prompted more dose adjustments with NB-UVB phototherapy to avoid burning (ie, erythema) at baseline and throughout the treatment course. These dose adjustments may have increased the overall clearance rates. If true, we predicted that younger adults treated with the same protocol would have cleared more quickly, either because of age-related differences or because they likely had fewer comorbidities and therefore fewer medications.

The second finding from Matthews et al¹³ that warranted further investigation was a higher erythema rate compared to the older adult study from the United Kingdom.¹⁷ We hypothesized that potentially greater use of photosensitizing medications in the United States could explain the higher erythema rates. Although medication-induced photosensitivity is less likely with NB-UVB phototherapy than with UVA, certain medications can cause UVB photosensitivity, including thiazides, quinidine, calcium channel antagonists, phenothiazines, and nonsteroidal anti-inflammatory drugs.^8,19,20 Therefore, photosensitizing medication use either at baseline or during a course of NB-UVB phototherapy could increase the risk for erythema. Age-related skin changes also have been considered as a potential cause for erythema. One study found that the skin of older patients was more sensitive than younger patients, resulting in a lower minimal erythema dose (MED)¹⁴—the lowest UV dose that results in erythema.²¹ Others, however, found similar MEDs across age groups, but older adults experienced more intense erythema in the late phase of NB-UVB treatment.^22,23 Such conflicting findings indicate that questions remain regarding the risk for erythema in older patients and if photosensitizing medications are responsible for an increased risk.

This retrospective study aimed to determine if NB-UVB phototherapy is equally effective in both older and younger adults treated with the same protocol; to examine the association between the use of photosensitizing medications and clearance rates in both older and younger adults; and to examine the association between the use of photosensitizing medications and erythema rates in older vs younger adults.

Methods

Study Design and Patients—This retrospective cohort study used billing records to identify patients who received NB-UVB phototherapy at 3 different clinical sites within a large US health care system in Washington (Group Health Cooperative, now Kaiser Permanente Washington), serving more than 600,000 patients between January 1, 2012, and December 31, 2016. The institutional review board of Kaiser Permanente Washington Health Research Institute approved this study (IRB 1498087-4). Younger adults were classified as those 64 years or younger and older adults as those 65 years and older at the start of their phototherapy regimen. A power analysis determined that the optimal sample size for this study was 250 patients.

Individuals were excluded if they had fewer than 6 phototherapy treatments; a diagnosis of vitiligo, photosensitivity dermatitis, morphea, or pityriasis rubra pilaris; and/or treatment of the hands or feet only.

Phototherapy Protocol—Using a 48-lamp NB-UVB unit, trained phototherapy nurses provided all treatments following standardized treatment protocols¹³ based on previously published phototherapy guidelines.²⁴ Nurses determined each patient’s disease clearance level using a 3-point clearance scale (high, medium, low).¹³ Each patient’s starting dose was determined based on the estimated MED for their skin phototype. If the patient was using photosensitizing medications, the protocol indicated a need for a decreased starting dose—down 25% to 50%—depending on the presumed level of photosensitivity. All clinical sites used the same protocol, but decisions about adjustments within this range were made by individual registered nurses and dermatologists, which could lead to variability across sites. Protocols also directed nurses to query patients about specific treatment responses, including erythema, tenderness, or itching; how their condition was responding; use of photosensitizing medications; missed treatments; and placement of shielding. Doses were adjusted accordingly.

Statistical Analysis—Data were analyzed using Stata statistical software (StataCorp LLC). Univariate analyses were used to examine the data and identify outliers, bad values, and missing data, as well as to calculate descriptive statistics. Pearson χ² and Fisher exact statistics were used to calculate differences in categorical variables. Linear multivariate regression models and logistic multivariate models were used to examine statistical relationships between variables. Statistical significance was defined as P≤.05.

Results

Patient Characteristics—Medical records were reviewed for 172 patients who received phototherapy between 2012 and 2016. Patients ranged in age from 23 to 91 years, with 102 patients 64 years and younger and 70 patients 65 years and older. Tables 1 and 2 outline the patient characteristics and conditions treated.

Phototherapy Effectiveness—Narrowband UVB phototherapy was found to be equally effective in older vs younger adults, with 82.9% of older adults (n=58) achieving a high level of clearance vs 80.4% (n=82) of younger adults, and 5.7% (n=4) of older adults achieved a medium level of clearance vs 10% (n=10) of younger adults (Table 3). Although older adults had slightly faster clearance rates on average (34.6 vs 37.2 treatments), these differences were not significant.

Photosensitizing Medications, Clearance Levels, and Clearance Rates—There was no significant association between clearance levels and number of photosensitizing medications in either younger (Figure 1) or older (Figure 2) adults. There was a wide range of clearance rates in both groups (Table 3), but no relationship was identified between clearance rates and photosensitizing medications or age (Figure 3). Clinic C had higher overall clearance rates for both age groups compared to the other clinics (Figure 4), but the clearance levels were still equivalent. No consistent pattern emerged indicating that age was a factor for the slower clearance at this site, and no relationship was identified between taking photosensitizing medications and clearance levels (Fisher exact test, P=.467) or clearance rates (t[149]=0.75; P=.45).

Frequency of Treatments and Clearance Rates—Older adults more consistently completed the recommended frequency of treatments—3 times weekly—compared to younger adults (74.3% vs 58.5%). However, all patients who completed 3 treatments per week required a similar number of treatments to clear (older adults, mean [SD]: 35.7 [21.6]; younger adults, mean [SD]: 34.7 [19.0]; P=.85). Among patients completing 2 or fewer treatments per week, older adults required a mean (SD) of only 31 (9.0) treatments to clear vs 41.5 (21.3) treatments to clear for younger adults, but the difference was not statistically significant (P=.08). However, even those with suboptimal frequency ultimately achieved similar clearance levels.

Photosensitizing Medications and Erythema Rates—Many patients in both age groups took medications that listed photosensitivity as a potential side effect (77.1% of older adults and 60.8% of younger adults). Of them, most patients took only 1 or 2 photosensitizing medications. However, significantly more older patients took 3 or more photosensitizing medications (28.6% vs 12.7%; P=.01)(Table 3). Asymptomatic (grade 1) erythema was unrelated to medication use and quite common in all adults (48.6% of older adults and 60.8% of younger adults). Most patients had only a few episodes of grade 1 erythema (mean [SD], 1.2 [2.9] in older adults and 1.6 [2.2] in younger adults). More older adults had grade 2 erythema (28.6%) compared to younger adults (17.6%). Patients using 3 or more photosensitizing medications were twice as likely to experience grade 2 erythema. Grades 3 and 4 erythema were extremely rare; none of the patients stopped phototherapy because they experienced erythema.

Overall, phototherapy nurses adjusted the starting dose according to the phototype-based protocol an average of 69% of the time for patients on medications with photosensitivity listed as a potential side effect. However, the frequency depended significantly on the clinic (clinic A, 24%; clinic B, 92%; clinic C, 87%)(P≤.001). Nurses across all clinics consistently decreased the treatment dose when patients reported starting new photosensitizing medications. Patients with adjusted starting doses had slightly but not significantly higher clearance rates compared to those without (mean, 37.8 vs 35.5; t(104)=0.58; P=.56).

Comment

Comparisons to Prior Studies—This study confirmed that phototherapy is equally effective for older and younger adults, with approximately 90% reaching medium to high clearance levels with approximately 35 treatments in both groups. Prior studies of all age groups found that patients typically cleared with an average of 20 to 28 treatments.^7,8,14-16 In contrast, the findings in older adults from this study were similar to the older adult study from the United Kingdom that reported a 91% clear/near clear rate with an average of 30 treatments.¹⁷ The clearance level also was similar to the older adult study in Turkey¹⁸ that reported 73.7% (70/95) of patients with psoriasis achieved a minimum psoriasis area severity index of 75, indicating 75% improvement from baseline.

Impact of Photosensitizing Medications on Clearance—Photosensitizing medications and treatment frequency were 2 factors that might explain the slower clearance rates in younger adults. In this study, both groups of patients used similar numbers of photosensitizing medications, but more older adults were taking 3 or more medications (Table 3). We found no statistically significant relationship between taking photosensitizing medications and either the clearance rates or the level of clearance achieved in either age group.

Impact of Treatment Frequency—Weekly treatment frequency also was examined. One prior study demonstrated that treatments 3 times weekly led to a faster clearance time and higher clearance levels compared with twice-weekly treatment.⁷ When patients completed treatments twice weekly, it took an average of 1.5 times more days to clear, which impacted cost and clinical resource availability. The patients ranged in age from 17 to 80 years, but outcomes in older patients were not described separately.⁷ Interestingly, our study seemed to find a difference between age groups when the impact of treatment frequency was examined. Older adults completed nearly 4 fewer mean treatments to clear when treating less often, with more than 80% achieving high levels of clearance, whereas the younger adults required almost 7 more treatments to clear when they came in less frequently, with approximately 80% achieving a high level of clearance. As a result, our study found that in both age groups, slowing the treatment frequency extended the treatment time to clearance—more for the younger adults than the older adults—but did not significantly change the percentage of individuals reaching full clearance in either group.

Erythema Rates—There was no association between photosensitizing medications and erythema rates except when patients were taking at least 3 medications. Most medications that listed photosensitivity as a possible side effect did not specify their relevant range of UV radiation; therefore, all such medications were examined during this analysis. Prior research has shown UVB range photosensitizing medications include thiazides, quinidine, calcium channel antagonists, phenothiazines, and nonsteroidal anti-inflammatory drugs.¹⁹ A sensitivity analysis that focused only on these medications found no association between them and any particular grade of erythema. However, patients taking 3 or more of any medications listing photosensitivity as a side effect had an increased risk for grade 2 erythema.

Erythema rates in this study were consistent with a 2013 systematic review that reported 57% of patients with asymptomatic grade 1 erythema.²⁵ In the 2 other comparative older adult studies, erythema rates varied widely: 35% in a study from Turkey¹⁸compared to only1.89% in a study from the United Kingdom.¹⁷

The starting dose for NB-UVB may drive erythema rates. The current study’s protocols were based on an estimated MED that is subjectively determined by the dermatology provider’s assessment of the patient’s skin sensitivity via examination and questions to the patient about their response to environmental sun exposure (ie, burning and tanning)²⁶ and is frequently used to determine the starting dose and subsequent dose escalation. Certain medications have been found to increase photosensitivity and erythema,²⁰ which can change an individual’s MED. If photosensitizing medications are started prior to or during a course of NB-UVB without a pretreatment MED, they might increase the risk for erythema. This study did not identify specific erythema-inducing medications but did find that taking 3 or more photosensitizing medications was associated with increased episodes of grade 2 erythema. Similarly, Harrop et al⁸ found that patients who were taking photosensitizing medications were more likely to have grade 2 or higher erythema, despite baseline MED testing, which is an established safety mechanism to reduce the risk and severity of erythema.^14,20,27 The authors of a recent study of older adults in Taiwan specifically recommended MED testing due to the unpredictable influence of polypharmacy on MED calculations in this population.²⁸ Therefore, this study’s use of an estimated MED in older adults may have influenced the starting dose as well as the incidence and severity of erythemic events. Age-related skin changes likely are ruled out as a consideration for mild erythema by the similarity of grade 1 erythema rates in both older and younger adults. Other studies have identified differences between the age groups, where older patients experienced more intense erythema in the late phase of UVB treatments.^22,23 This phenomenon could increase the risk for a grade 2 erythema, which may correspond with this study’s findings.

Other potential causes of erythema were ruled out during our study, including erythema related to missed treatments and shielding mishaps. Other factors, however, may impact the level of sensitivity each patient has to phototherapy, including genetics, epigenetics, and cumulative sun damage. With NB-UVB, near-erythemogenic doses are optimal to achieve effective treatments but require a delicate balance to achieve, which may be more problematic for older adults, especially those taking several medications.

Study Limitations—Our study design made it difficult to draw conclusions about rarer dermatologic conditions. Some patients received treatments over years that were not included in the study period. Finally, power calculations suggested that our actual sample size was too small, with approximately one-third of the required sample missing.

Practical Implications—The goals of phototherapy are to achieve a high level of disease clearance with the fewest number of treatments possible and minimal side effects. Skin phototype–driven standardized doses based on estimated MED may be conservatively low to minimize the risk of side effects (eg, erythema), which could slow the treatment progression. Thus, basing the starting dose on individual MED assessments may improve clearance rates. This study also confirmed that phototherapy is safe with minimal erythema in adults of all ages. The erythema episodes that patients experienced were few and mild, but because of greater rates of grade 2 erythema in patients on 3 or more photosensitizing medications, consideration of MED testing in both age groups might optimize doses at baseline and prompt caution for subsequent dose titration in this subset of patients.

The extra staff training and patient monitoring required for MED testing likely is to add value and preserve resources if faster clearance rates could be achieved and may warrant further investigation. Phototherapy centers require standardized treatment protocols, diligent well-trained staff, and program monitoring to ensure consistent care to all patients. This study highlighted the ongoing opportunity for health care organizations to conduct evidence-based practice inquiries to continually optimize care for their patients.

Cigarette smoking is an independent risk factor for lung cancer and atherosclerotic cardiovascular disease (ASCVD).^1-3 The National Lung Screening Trial (NLST) demonstrated both lung cancer mortality reduction with the use of surveillance low-dose computed tomography (LDCT) and ASCVD as the most common cause of death among smokers.^4,5 ASCVD remains the leading cause of death in the lung cancer screening (LCS) population.^2,3 After publication of the NLST results, the US Preventive Services Task Force (USPSTF) established LCS eligibility among smokers and the Center for Medicare and Medicaid Services approved payment for annual LDCT in this group.^1,6,7

Recently LDCT has been proposed as an adjunct diagnostic tool for detecting coronary artery calcium (CAC), which is independently associated with ASCVD and mortality.^8-13 CAC scores have been recommended by the 2019 American College of Cardiology/American Heart Association cholesterol treatment guidelines and shown to be cost-effective in guiding statin therapy for patients with borderline to intermediate ASCVD risk.^14-16 While CAC is conventionally quantified using electrocardiogram (ECG)-gated CT, these scans are not routinely performed in clinical practice because preventive CAC screening is neither recommended by the USPSTF nor covered by most insurance providers.^17,18 LDCT, conversely, is reimbursable and a well-validated ASCVD risk predictor.^18,19

In this study, we aimed to determine the validity of LDCT in identifying CAC among the military LCS population and whether it would impact statin recommendations based on 10-year ASCVD risk.

Methods

Participants were recruited from a retrospective cohort of 563 Military Health System (MHS) beneficiaries who received LCS with LDCT at Naval Medical Center Portsmouth (NMCP) in Virginia between January 1, 2019, and December 31, 2020. The 2013 USPSTF LCS guidelines were followed as the 2021 guidelines had not been published before the start of the study; thus, eligible participants included adults aged 55 to 80 years with at least a 30-pack-year smoking history and currently smoked or had quit within 15 years from the date of study consent.^6,7

Between November 2020 and May 2021, study investigators screened 287 patient records and recruited 190 participants by telephone, starting with individuals who had the most recent LDCT and working backward until reaching the predetermined 170 subjects who had undergone in-office consents before ECG-gated CT scans. Since LDCT was not obtained simultaneously with the ECG-gated CT, participants were required to complete their gated CT within 24 months of their last LDCT. Of the 190 subjects initially recruited, those who were ineligible for LCS (n = 4), had a history of angioplasty, stent, or bypass revascularization procedure (n = 4), did not complete their ECG-gated CT within the specified time frame (n = 8), or withdrew from the study (n = 4) were excluded. While gated CT scans were scored for CAC in the present time, LDCT (previously only read for general lung pathology) was not scored until after participant consent. Patients were peripherally followed, via health record reviews, for 3 months after their gated CT to document any additional imaging ordered by their primary care practitioners. The study was approved by the NMCP Institutional Review Board.

Coronary Artery Calcification Scoring

We performed CT scans using Siemens SOMATOM Flash, a second-generation dual-source scanner; and GE LightSpeed VCT, a single-source, 64-slice scanner. A step-and-shoot prospective trigger technique was used, and contiguous axial images were reconstructed at 2.5-mm or 3-mm intervals for CAC quantification using the Agatston method.²⁰ ECG-gated CT scans were electrocardiographically triggered at mid-diastole (70% of the R-R interval). Radiation dose reduction techniques involved adjustments of the mA according to body mass index and iterative reconstruction. LDCT scans were performed without ECG gating. We reconstructed contiguous axial images at 1-mm intervals for evaluation of the lung parenchyma. Similar dose-reduction techniques were used, to limit radiation exposure for each LDCT scan to < 1.5 mSv, per established guidelines.²¹ CAC on LDCT was also scored using the Agatston method. CAC was scored on the 2 scan types by different blinded reviewers.

Covariates

We reviewed outpatient health records to obtain participants’ age, sex, medical history, statin use, smoking status (current or former), and pack-years. International Classification of Diseases, Tenth Revision codes within medical encounters were used to document prevalent hypertension, hyperlipidemia, and diabetes mellitus. Participants’ most recent low-density lipoprotein value (within 24 months of ECG-gated CT) was recorded and 10-year ASCVD risk scores were calculated using the pooled cohorts equation.

Statistical Analysis

A power analysis performed before study initiation determined that a prospective sample size of 170 would be sufficient to provide strength of correlation between CAC scores calculated from ECG-gated CT and LDCT and achieve a statistical power of at least 80%. The Wilcoxon rank sum and Fisher exact tests were used to evaluate differences in continuous and categorical CAC scores, respectively. Given skewed distributions, Spearman rank correlations and Kendall W coefficient of concordance were respectively used to evaluate correlation and concordance of CAC scores between the 2 scan types. κ statistics were used to rate agreement between categorical CAC scores. Bland-Altman analysis was performed to determine the bias and limits of agreement between ECG-gated CT and LDCT.²² For categorical CAC score analysis, participants were categorized into 5 groups according to standard Agatston score cut-off points. We defined the 5 categories of CAC for both scan types based on previous analysis from Rumberger and colleagues: CAC = 0 (absent), CAC = 1-10 (minimal), CAC = 11-100 (mild), CAC = 101-400 (moderate), CAC > 400 (severe).²³ Of note, LDCT reports at NMCP include a visual CAC score using these qualitative descriptors that were available to LDCT reviewers. Analyses were conducted using SAS version 9.4 and Microsoft Excel; P values < .05 were considered statistically significant.

Results

The 170 participants had a mean (SD) age of 62.1 (4.6) years and were 70.6% male (Table 1). Hyperlipidemia was the most prevalent cardiac risk factor with almost 70% of participants on a statin. There was no incidence of ischemic ASCVD during follow-up, although 1 participant was later diagnosed with lung cancer after evaluation of suspicious pulmonary findings on ECG-gated CT. CAC was identified on both scan types in 126 participants; however, LDCT was discordant with gated CT in identifying CAC in 24 subjects (P < .001).

The correlation between CAC scores on ECG-gated CT and LDCT was 0.945 (P < .001) and the concordance was 0.643, indicating moderate agreement between CAC scores on the 2 different scans (Figure 1). Median CAC scores were significantly higher on ECG-gated CT when compared with LDCT (107.5 vs 48.1 Agatston units, respectively; P < .05). Table 2 shows the CAC score characteristics for both scan types. The κ statistic for agreement between categorical CAC scores on ECG-gated CT compared with LDCT was 0.49 (SEκ= 0.05; 95% CI, -0.73-1.71), and the weighted κ statistic was 0.71, indicating moderate to substantial agreement between the 2 scans using the specified cutoff points. The Bland-Altman analysis presented a mean bias of 111.45 Agatston units, with limits of agreement between -268.64 and 491.54, as shown in Figure 2, suggesting that CAC scores on ECG-gated CT were, on average, about 111 units higher than those on LDCT. Finally, there were 24 participants with CAC seen on ECG-gated CT but none identified on LDCT (P < .001); of this cohort 20 were already on a statin, and of the remaining 4 individuals, 1 met statin criteria based on a > 20% ASCVD risk score alone (regardless of CAC score), 1 with an intermediate risk score met statin criteria based on CAC score reporting, 1 did not meet criteria due to a low-risk score, and the last had no reportable ASCVD risk score.

Discussion

In this study population of recruited MHS beneficiaries, there was a strong correlation and moderate to substantial agreement between CAC scores calculated from LDCT and conventional ECG-gated CT. The number of nonstatin participants who met statin criteria and would have benefited from additional CAC score reporting was statistically significant as compared to their statin counterparts who no longer met the criteria.

CAC screening using nongated CT has become an increasingly available and consistently reproducible means for stratifying ASCVD risk and guiding statin therapy in individuals with equivocal ASCVD risk scores.^24-26 As has been demonstrated in previous studies, our study additionally highlights the effective use of LDCT in not only identifying CAC, but also in beneficially impacting statin decisions in the high-risk smoking population.^24-26 Our results also showed LDCT missed CAC in participants, the majority of which were already on a statin, and only 1 nonstatin individual benefited from additional CAC reporting. CAC scoring on LDCT should be an adjunct, not a substitute, for ASCVD risk stratification to help guide statin management.^25,27

Our results may provide cost considerate implications for preventive CAC screening. While TRICARE covers the cost of ECG-gated CT for MHS beneficiaries, the same is not true of most nonmilitary insurance providers. Concerns about cancer risk from radiation exposure may also lead to hesitation about receiving additional CTs in the smoking population. Since the LCS population already receives annual LDCT, these scans can also be used for CAC scoring to help primary care professionals risk stratify their patients, as has been previously shown.^28-31 Clinicians should consider implementing CAC scoring with annual LDCT scans, which would curtail further risks and expenses from CAC-specified scans.

Although CAC is scored visually and routinely reported in the body of LDCT reports at our facility, this is not a universal practice and was performed in only 44% of subjects with known CAC by a previous study.³² In 2007, there were 600,000 CAC scoring scans and > 9 million routine chest CTs performed in the United States.³³ Based on our results and the growing consensus in the existing literature, CAC scoring on nongated CT is not only valid and reliable, but also can estimate ASCVD risk and subsequent mortality.^34-36 Routine chest CTs remain an available resource for providing additional ASCVD risk stratification.

As we demonstrated, median CAC scores on LDCT were on average significantly lower than those from gated CT. This could be due to slice thickness variability between the GE and Siemens scanners or CAC progression between the time of the retrospective LDCT and prospective ECG-gated CT. Aside from this potential limitation, LDCT has been shown to have a high level of agreement with gated CT in predicting CAC, both visually and by the Agatston technique.^37-39 Our results further support previous recommendations of utilizing CAC score categories when determining ASCVD risk from LDCT and that establishing scoring cutoff points warrants further development for potential standardization.^37-39 Readers should be mindful that LDCT may still be less sensitive and underestimate low CAC levels and that ECG-gated CT may occasionally be more optimal in determining ASCVD risk when considering the negative predictive value of CAC.⁴⁰

Limitations

Our study cohort was composed of MHS beneficiaries. Compared with the general population, these individuals may have greater access to care and be more likely to receive statins after preventive screenings. Additional studies may be required to assess CAC-associated statin eligibility among the general population. As discussed previously LDCT was not performed concomitantly with the ECG-gated CT. Although there was moderate to substantial CAC agreement between the 2 scan types, the timing difference could have led to absolute differences in CAC scores across both scan types and impacted the ability to detect low-level CAC on LDCT. CAC values should be interpreted based on the respective scan type.

Conclusions

LDCT is a reliable diagnostic alternative to ECG-gated CT in predicting CAC. CAC scores from LDCT are highly correlated and concordant with those from gated CT and can help guide statin management in individuals with intermediate ASCVD risk. The proposed duality of LDCT to assess ASCVD risk in addition to lung cancer can reduce the need for unnecessary scans while optimizing preventive clinical care. While coronary calcium and elevated CAC scores can facilitate clinical decision making to initiate statin therapy for intermediate-risk patients, physicians must still determine whether additional cardiac testing is warranted to avoid unnecessary procedures and health care costs. Smokers undergoing annual LDCT may benefit from standardized CAC scoring to help further stratify ASCVD risk while limiting the expense and radiation of additional scans.

Acknowledgments

The authors thank Ms. Lorie Gower for her contributions to the study.