Federal Practitioner

Valproic acid (VPA) and its derivative, divalproex (DVP) are prescribed for a variety of indications, commonly for seizure control in patients with epilepsy, mood stabilization in patients with bipolar disorder, and migraine prophylaxis. Gastrointestinal distress and sedation are among the most reported adverse effects (AEs) with DVP therapy.¹ Although serious hepatic and hematologic AEs are rare, monitoring is still recommended. DVP can cause various hematologic dyscrasias, the most common being thrombocytopenia.^1,2 Neutropenia and leukopenia have been reported in isolated cases, most occurring in pediatric patients or patients with epilepsy.^3-14

Several case reports of DVP-related neutropenia (absolute neutrophil count [ANC] < 1.50 10³/mcL) and leukopenia (white blood cell count [WBC] < 4.0 10³/mcL) were reviewed during our literature search, some caused by DVP monotherapy; others were thought to be related to concomitant use of DVP and another drug.^15-25 Quetiapine was the antipsychotic most commonly implicated in causing hematologic abnormalities when combined with DVP. We report a case of neutropenia and leukopenia that presented after a cross taper from quetiapine to DVP for the treatment of borderline personality disorder (BPD).

Although no medications have been approved by the US Food and Drug Administration (FDA) for the treatment of BPD, mood stabilizers, including DVP, have literature to support their use for the treatment of affective dysregulation and impulsive behavioral dyscontrol.^26-28 A therapeutic range for DVP in the treatment of BPD has not been defined; therefore, for this case report, the generally accepted range of 50 to 100 µg/mL will be considered therapeutic.¹

Case Presentation

A 34-year-old male patient presented to the mental health clinic pharmacist reporting that his current psychotropic medication regimen was not effective. His medical history included posttraumatic stress disorder (PTSD), opioid use disorder, alcohol use disorder, stimulant use disorder, cannabis use, BPD, hypertension, hyperlipidemia, prediabetes, gastroesophageal reflex disease, and a pulmonary nodule. On initial presentation, the patient was prescribed buprenorphine 24 mg/naloxone 6 mg, quetiapine 400 mg, duloxetine 120 mg, and prazosin 15 mg per day. At the time of pharmacy consultation, last reported alcohol or nonprescribed opioid use was about 6 months prior, and methamphetamine use about 1 month prior, with ongoing cannabis use. The patient had a history of participating in cognitive processing therapy, dialectical behavior therapy (DBT), and residential treatment for both PTSD and substance use. Additionally, he was actively participating in contingency management for stimulant use disorder and self-management and recovery training group.

The patient reported ongoing mood lability, hypervigilance, and oversedation with current psychotropic regimen. The prescriber of his medication for opioid use disorder also reported the patient experienced labile mood, impulsive behavior, and anger outbursts. In the setting of intolerability due to oversedation with quetiapine, cardiometabolic risk, and lack of clear indication for use, the patient and health care practitioner (HCP) agreed to taper quetiapine and initiate a trial of DVP for affective dysregulation and impulsive-behavioral dyscontrol. To prevent cholinergic rebound and insomnia with abrupt discontinuation of quetiapine, DVP and quetiapine were cross tapered. The following cross taper was prescribed: quetiapine 300 mg and DVP 500 mg per day for week 1; quetiapine 200 mg and DVP 500 mg per day for week 2; quetiapine 100 mg and DVP 1000 mg per day for week 3; quetiapine 50 mg and DVP 1000 mg per day for week 4; followed by DVP 1000 mg per day and discontinuation of quetiapine.

During a 4-week follow-up appointment, the patient reported appropriate completion of cross taper but stopped taking the DVP 3 days prior to the appointment due to self-reported lack of efficacy. For this reason, serum VPA level was not obtained. After discussion with his HCP, the patient restarted DVP 1000 mg per day without retitration with plans to get laboratory tests in 1 week. The next week, laboratory tests were notable for VPA level 28.74 (reference range, 50-100) µg/mL, low WBC 3.51 (reference range, 4.00-10.00) 10³/mcL, platelets 169 (reference range, 150-420) 10³/mcL, and low ANC 1.00 (reference range, 1.50-7.40) 10³/mcL (Table). This raised clinical concern as the patient had no history of documented neutropenia or leukopenia, with most recent complete blood count (CBC) prior to DVP initiation 3 months earlier while prescribed quetiapine.

Discussion

Our case report describes isolated neutropenia and leukopenia that developed after a cross taper from quetiapine to DVP. Hematologic abnormalities resolved after discontinuation of DVP, suggesting a likely correlation. DVP has a well-established, dose-related prevalence of thrombocytopenia occurring in up to 27% of patients.¹ Fewer case reports exist on neutropenia and leukopenia. DVP-induced neutropenia is thought to be a result of direct bone marrow suppression, whereas the more commonly occurring blood dyscrasia, thrombocytopenia, is thought to be caused by an antibody-mediated destruction of platelets.⁶

Management of DVP-induced thrombocytopenia is often dependent on the severity of the reaction. In mild-to-moderate cases, intervention may not be necessary as thrombocytopenia has been shown to resolve without adjustment to DVP therapy.¹ In more severe or symptomatic cases, dose reduction or discontinuation of the offending agent is recommended, typically resulting in resolution shortly following pharmacologic intervention.

Guidance on the management of other drug-induced hematologic abnormalities, such as neutropenia and leukopenia are not as well established. A 2019 systematic review of idiosyncratic drug-induced neutropenia suggested that continuing the offending drug with strict monitoring could be considered in cases of mild neutropenia. In cases of moderate neutropenia, the author suggests temporary cessation of the drug and reinstatement once neutrophil count normalizes and definitive cessation of the drug in severe cases.³⁰

In our case, continuing the offending agent with close monitoring was considered, similar to the well-established management of clozapine-induced neutropenia. However, due to the concern that the ANC was bordering moderate neutropenia in the absence of a therapeutic VPA level as well as a significant reduction in platelets, although not meeting criteria for thrombocytopenia, the decision was made to err on the side of caution and discontinue the most likely offending agent.

It is important to highlight that DVP was replacing quetiapine in the form of a cross taper. Quetiapine is structurally similar to clozapine. While clozapine has strict monitoring requirements related to neutropenia, blood dyscrasias with quetiapine therapy are rare. Quetiapine-induced hematologic abnormalities may be due to direct toxicity or to an immune-mediated mechanism, leading to bone marrow suppression.²⁰ Case reports documenting blood dyscrasias with the combination of DVP and quetiapine were identified during literature review.^15-19 Despite these case reports, we believe DVP was the primary offending agent in our case as the patient’s last dose of quetiapine was 2 weeks before obtaining the abnormal CBC. There was no history of blood dyscrasias with quetiapine monotherapy; however, the effect of the combination of DVP and quetiapine is unknown as no CBC was obtained during the cross-taper period.

Although there are no FDA-approved medications for the treatment of BPD, mood stabilizers, including DVP, have some research to support their use for the treatment of affective dysregulation and impulsive-behavioral dyscontrol.^26-28 In our case, DVP was selected due to the evidence for use in BPD and ability to assess adherence with therapeutic monitoring. Although polypharmacy is a concern in patients with BPD, in our case we believed that the patient’s ongoing mood lability and impulsive behaviors warranted pharmacologic intervention. Additionally, DVP provided an advantage in its ability to quickly titrate to therapeutic dose when compared with lamotrigine and a lower risk of cognitive AEs when compared with topiramate.

Conclusions

To our knowledge, this case report demonstrates the first published case of neutropenia and leukopenia related to DVP therapy for the treatment of BPD. Routine CBC monitoring is recommended with DVP therapy, and our case highlights the importance of evaluating for not only thrombocytopenia, but also other blood dyscrasias during the titration phase even in the absence of a therapeutic VPA level. Further studies are warranted to determine incidence of DVP-related neutropenia and leukopenia and to evaluate the safety of continuing DVP in cases of mild-to-moderate neutropenia with close monitoring.

Valproic acid (VPA) and its derivative, divalproex (DVP) are prescribed for a variety of indications, commonly for seizure control in patients with epilepsy, mood stabilization in patients with bipolar disorder, and migraine prophylaxis. Gastrointestinal distress and sedation are among the most reported adverse effects (AEs) with DVP therapy.¹ Although serious hepatic and hematologic AEs are rare, monitoring is still recommended. DVP can cause various hematologic dyscrasias, the most common being thrombocytopenia.^1,2 Neutropenia and leukopenia have been reported in isolated cases, most occurring in pediatric patients or patients with epilepsy.^3-14

Several case reports of DVP-related neutropenia (absolute neutrophil count [ANC] < 1.50 10³/mcL) and leukopenia (white blood cell count [WBC] < 4.0 10³/mcL) were reviewed during our literature search, some caused by DVP monotherapy; others were thought to be related to concomitant use of DVP and another drug.^15-25 Quetiapine was the antipsychotic most commonly implicated in causing hematologic abnormalities when combined with DVP. We report a case of neutropenia and leukopenia that presented after a cross taper from quetiapine to DVP for the treatment of borderline personality disorder (BPD).

Although no medications have been approved by the US Food and Drug Administration (FDA) for the treatment of BPD, mood stabilizers, including DVP, have literature to support their use for the treatment of affective dysregulation and impulsive behavioral dyscontrol.^26-28 A therapeutic range for DVP in the treatment of BPD has not been defined; therefore, for this case report, the generally accepted range of 50 to 100 µg/mL will be considered therapeutic.¹

Case Presentation

A 34-year-old male patient presented to the mental health clinic pharmacist reporting that his current psychotropic medication regimen was not effective. His medical history included posttraumatic stress disorder (PTSD), opioid use disorder, alcohol use disorder, stimulant use disorder, cannabis use, BPD, hypertension, hyperlipidemia, prediabetes, gastroesophageal reflex disease, and a pulmonary nodule. On initial presentation, the patient was prescribed buprenorphine 24 mg/naloxone 6 mg, quetiapine 400 mg, duloxetine 120 mg, and prazosin 15 mg per day. At the time of pharmacy consultation, last reported alcohol or nonprescribed opioid use was about 6 months prior, and methamphetamine use about 1 month prior, with ongoing cannabis use. The patient had a history of participating in cognitive processing therapy, dialectical behavior therapy (DBT), and residential treatment for both PTSD and substance use. Additionally, he was actively participating in contingency management for stimulant use disorder and self-management and recovery training group.

The patient reported ongoing mood lability, hypervigilance, and oversedation with current psychotropic regimen. The prescriber of his medication for opioid use disorder also reported the patient experienced labile mood, impulsive behavior, and anger outbursts. In the setting of intolerability due to oversedation with quetiapine, cardiometabolic risk, and lack of clear indication for use, the patient and health care practitioner (HCP) agreed to taper quetiapine and initiate a trial of DVP for affective dysregulation and impulsive-behavioral dyscontrol. To prevent cholinergic rebound and insomnia with abrupt discontinuation of quetiapine, DVP and quetiapine were cross tapered. The following cross taper was prescribed: quetiapine 300 mg and DVP 500 mg per day for week 1; quetiapine 200 mg and DVP 500 mg per day for week 2; quetiapine 100 mg and DVP 1000 mg per day for week 3; quetiapine 50 mg and DVP 1000 mg per day for week 4; followed by DVP 1000 mg per day and discontinuation of quetiapine.

During a 4-week follow-up appointment, the patient reported appropriate completion of cross taper but stopped taking the DVP 3 days prior to the appointment due to self-reported lack of efficacy. For this reason, serum VPA level was not obtained. After discussion with his HCP, the patient restarted DVP 1000 mg per day without retitration with plans to get laboratory tests in 1 week. The next week, laboratory tests were notable for VPA level 28.74 (reference range, 50-100) µg/mL, low WBC 3.51 (reference range, 4.00-10.00) 10³/mcL, platelets 169 (reference range, 150-420) 10³/mcL, and low ANC 1.00 (reference range, 1.50-7.40) 10³/mcL (Table). This raised clinical concern as the patient had no history of documented neutropenia or leukopenia, with most recent complete blood count (CBC) prior to DVP initiation 3 months earlier while prescribed quetiapine.

Discussion

Our case report describes isolated neutropenia and leukopenia that developed after a cross taper from quetiapine to DVP. Hematologic abnormalities resolved after discontinuation of DVP, suggesting a likely correlation. DVP has a well-established, dose-related prevalence of thrombocytopenia occurring in up to 27% of patients.¹ Fewer case reports exist on neutropenia and leukopenia. DVP-induced neutropenia is thought to be a result of direct bone marrow suppression, whereas the more commonly occurring blood dyscrasia, thrombocytopenia, is thought to be caused by an antibody-mediated destruction of platelets.⁶

Management of DVP-induced thrombocytopenia is often dependent on the severity of the reaction. In mild-to-moderate cases, intervention may not be necessary as thrombocytopenia has been shown to resolve without adjustment to DVP therapy.¹ In more severe or symptomatic cases, dose reduction or discontinuation of the offending agent is recommended, typically resulting in resolution shortly following pharmacologic intervention.

Guidance on the management of other drug-induced hematologic abnormalities, such as neutropenia and leukopenia are not as well established. A 2019 systematic review of idiosyncratic drug-induced neutropenia suggested that continuing the offending drug with strict monitoring could be considered in cases of mild neutropenia. In cases of moderate neutropenia, the author suggests temporary cessation of the drug and reinstatement once neutrophil count normalizes and definitive cessation of the drug in severe cases.³⁰

In our case, continuing the offending agent with close monitoring was considered, similar to the well-established management of clozapine-induced neutropenia. However, due to the concern that the ANC was bordering moderate neutropenia in the absence of a therapeutic VPA level as well as a significant reduction in platelets, although not meeting criteria for thrombocytopenia, the decision was made to err on the side of caution and discontinue the most likely offending agent.

It is important to highlight that DVP was replacing quetiapine in the form of a cross taper. Quetiapine is structurally similar to clozapine. While clozapine has strict monitoring requirements related to neutropenia, blood dyscrasias with quetiapine therapy are rare. Quetiapine-induced hematologic abnormalities may be due to direct toxicity or to an immune-mediated mechanism, leading to bone marrow suppression.²⁰ Case reports documenting blood dyscrasias with the combination of DVP and quetiapine were identified during literature review.^15-19 Despite these case reports, we believe DVP was the primary offending agent in our case as the patient’s last dose of quetiapine was 2 weeks before obtaining the abnormal CBC. There was no history of blood dyscrasias with quetiapine monotherapy; however, the effect of the combination of DVP and quetiapine is unknown as no CBC was obtained during the cross-taper period.

Although there are no FDA-approved medications for the treatment of BPD, mood stabilizers, including DVP, have some research to support their use for the treatment of affective dysregulation and impulsive-behavioral dyscontrol.^26-28 In our case, DVP was selected due to the evidence for use in BPD and ability to assess adherence with therapeutic monitoring. Although polypharmacy is a concern in patients with BPD, in our case we believed that the patient’s ongoing mood lability and impulsive behaviors warranted pharmacologic intervention. Additionally, DVP provided an advantage in its ability to quickly titrate to therapeutic dose when compared with lamotrigine and a lower risk of cognitive AEs when compared with topiramate.

Conclusions

To our knowledge, this case report demonstrates the first published case of neutropenia and leukopenia related to DVP therapy for the treatment of BPD. Routine CBC monitoring is recommended with DVP therapy, and our case highlights the importance of evaluating for not only thrombocytopenia, but also other blood dyscrasias during the titration phase even in the absence of a therapeutic VPA level. Further studies are warranted to determine incidence of DVP-related neutropenia and leukopenia and to evaluate the safety of continuing DVP in cases of mild-to-moderate neutropenia with close monitoring.

Valproic acid (VPA) and its derivative, divalproex (DVP) are prescribed for a variety of indications, commonly for seizure control in patients with epilepsy, mood stabilization in patients with bipolar disorder, and migraine prophylaxis. Gastrointestinal distress and sedation are among the most reported adverse effects (AEs) with DVP therapy.¹ Although serious hepatic and hematologic AEs are rare, monitoring is still recommended. DVP can cause various hematologic dyscrasias, the most common being thrombocytopenia.^1,2 Neutropenia and leukopenia have been reported in isolated cases, most occurring in pediatric patients or patients with epilepsy.^3-14

Several case reports of DVP-related neutropenia (absolute neutrophil count [ANC] < 1.50 10³/mcL) and leukopenia (white blood cell count [WBC] < 4.0 10³/mcL) were reviewed during our literature search, some caused by DVP monotherapy; others were thought to be related to concomitant use of DVP and another drug.^15-25 Quetiapine was the antipsychotic most commonly implicated in causing hematologic abnormalities when combined with DVP. We report a case of neutropenia and leukopenia that presented after a cross taper from quetiapine to DVP for the treatment of borderline personality disorder (BPD).

Although no medications have been approved by the US Food and Drug Administration (FDA) for the treatment of BPD, mood stabilizers, including DVP, have literature to support their use for the treatment of affective dysregulation and impulsive behavioral dyscontrol.^26-28 A therapeutic range for DVP in the treatment of BPD has not been defined; therefore, for this case report, the generally accepted range of 50 to 100 µg/mL will be considered therapeutic.¹

Case Presentation

A 34-year-old male patient presented to the mental health clinic pharmacist reporting that his current psychotropic medication regimen was not effective. His medical history included posttraumatic stress disorder (PTSD), opioid use disorder, alcohol use disorder, stimulant use disorder, cannabis use, BPD, hypertension, hyperlipidemia, prediabetes, gastroesophageal reflex disease, and a pulmonary nodule. On initial presentation, the patient was prescribed buprenorphine 24 mg/naloxone 6 mg, quetiapine 400 mg, duloxetine 120 mg, and prazosin 15 mg per day. At the time of pharmacy consultation, last reported alcohol or nonprescribed opioid use was about 6 months prior, and methamphetamine use about 1 month prior, with ongoing cannabis use. The patient had a history of participating in cognitive processing therapy, dialectical behavior therapy (DBT), and residential treatment for both PTSD and substance use. Additionally, he was actively participating in contingency management for stimulant use disorder and self-management and recovery training group.

The patient reported ongoing mood lability, hypervigilance, and oversedation with current psychotropic regimen. The prescriber of his medication for opioid use disorder also reported the patient experienced labile mood, impulsive behavior, and anger outbursts. In the setting of intolerability due to oversedation with quetiapine, cardiometabolic risk, and lack of clear indication for use, the patient and health care practitioner (HCP) agreed to taper quetiapine and initiate a trial of DVP for affective dysregulation and impulsive-behavioral dyscontrol. To prevent cholinergic rebound and insomnia with abrupt discontinuation of quetiapine, DVP and quetiapine were cross tapered. The following cross taper was prescribed: quetiapine 300 mg and DVP 500 mg per day for week 1; quetiapine 200 mg and DVP 500 mg per day for week 2; quetiapine 100 mg and DVP 1000 mg per day for week 3; quetiapine 50 mg and DVP 1000 mg per day for week 4; followed by DVP 1000 mg per day and discontinuation of quetiapine.

During a 4-week follow-up appointment, the patient reported appropriate completion of cross taper but stopped taking the DVP 3 days prior to the appointment due to self-reported lack of efficacy. For this reason, serum VPA level was not obtained. After discussion with his HCP, the patient restarted DVP 1000 mg per day without retitration with plans to get laboratory tests in 1 week. The next week, laboratory tests were notable for VPA level 28.74 (reference range, 50-100) µg/mL, low WBC 3.51 (reference range, 4.00-10.00) 10³/mcL, platelets 169 (reference range, 150-420) 10³/mcL, and low ANC 1.00 (reference range, 1.50-7.40) 10³/mcL (Table). This raised clinical concern as the patient had no history of documented neutropenia or leukopenia, with most recent complete blood count (CBC) prior to DVP initiation 3 months earlier while prescribed quetiapine.

Discussion

Our case report describes isolated neutropenia and leukopenia that developed after a cross taper from quetiapine to DVP. Hematologic abnormalities resolved after discontinuation of DVP, suggesting a likely correlation. DVP has a well-established, dose-related prevalence of thrombocytopenia occurring in up to 27% of patients.¹ Fewer case reports exist on neutropenia and leukopenia. DVP-induced neutropenia is thought to be a result of direct bone marrow suppression, whereas the more commonly occurring blood dyscrasia, thrombocytopenia, is thought to be caused by an antibody-mediated destruction of platelets.⁶

Management of DVP-induced thrombocytopenia is often dependent on the severity of the reaction. In mild-to-moderate cases, intervention may not be necessary as thrombocytopenia has been shown to resolve without adjustment to DVP therapy.¹ In more severe or symptomatic cases, dose reduction or discontinuation of the offending agent is recommended, typically resulting in resolution shortly following pharmacologic intervention.

Guidance on the management of other drug-induced hematologic abnormalities, such as neutropenia and leukopenia are not as well established. A 2019 systematic review of idiosyncratic drug-induced neutropenia suggested that continuing the offending drug with strict monitoring could be considered in cases of mild neutropenia. In cases of moderate neutropenia, the author suggests temporary cessation of the drug and reinstatement once neutrophil count normalizes and definitive cessation of the drug in severe cases.³⁰

In our case, continuing the offending agent with close monitoring was considered, similar to the well-established management of clozapine-induced neutropenia. However, due to the concern that the ANC was bordering moderate neutropenia in the absence of a therapeutic VPA level as well as a significant reduction in platelets, although not meeting criteria for thrombocytopenia, the decision was made to err on the side of caution and discontinue the most likely offending agent.

It is important to highlight that DVP was replacing quetiapine in the form of a cross taper. Quetiapine is structurally similar to clozapine. While clozapine has strict monitoring requirements related to neutropenia, blood dyscrasias with quetiapine therapy are rare. Quetiapine-induced hematologic abnormalities may be due to direct toxicity or to an immune-mediated mechanism, leading to bone marrow suppression.²⁰ Case reports documenting blood dyscrasias with the combination of DVP and quetiapine were identified during literature review.^15-19 Despite these case reports, we believe DVP was the primary offending agent in our case as the patient’s last dose of quetiapine was 2 weeks before obtaining the abnormal CBC. There was no history of blood dyscrasias with quetiapine monotherapy; however, the effect of the combination of DVP and quetiapine is unknown as no CBC was obtained during the cross-taper period.

Although there are no FDA-approved medications for the treatment of BPD, mood stabilizers, including DVP, have some research to support their use for the treatment of affective dysregulation and impulsive-behavioral dyscontrol.^26-28 In our case, DVP was selected due to the evidence for use in BPD and ability to assess adherence with therapeutic monitoring. Although polypharmacy is a concern in patients with BPD, in our case we believed that the patient’s ongoing mood lability and impulsive behaviors warranted pharmacologic intervention. Additionally, DVP provided an advantage in its ability to quickly titrate to therapeutic dose when compared with lamotrigine and a lower risk of cognitive AEs when compared with topiramate.

Conclusions

To our knowledge, this case report demonstrates the first published case of neutropenia and leukopenia related to DVP therapy for the treatment of BPD. Routine CBC monitoring is recommended with DVP therapy, and our case highlights the importance of evaluating for not only thrombocytopenia, but also other blood dyscrasias during the titration phase even in the absence of a therapeutic VPA level. Further studies are warranted to determine incidence of DVP-related neutropenia and leukopenia and to evaluate the safety of continuing DVP in cases of mild-to-moderate neutropenia with close monitoring.

At the James A. Haley Veterans’ Hospital (JAHVH) in Tampa, Florida, the prevalence of amputations among patients at the spinal cord injury (SCI) center seems high. Despite limited data demonstrating altered hemodynamics in the lower extremities (LEs) among the SCI population and increased frequency of peripheral arterial disease (PAD), amputations among patients with SCI have received little attention in research.^1-3

In the United States, most amputations are caused by vascular disease related to peripheral arterial disease (PAD) and diabetes mellitus (DM).⁴ PAD primarily affects the LEs and is caused by atherosclerotic obstruction leading to insufficient blood flow. PAD can present clinically as LE pain, nonhealing ulcers, nonpalpable distal pulses, shiny or cold skin, absence of hair on the LE, or distal extremity pallor when the affected extremity is elevated. However, PAD is often asymptomatic. The diagnosis of PAD is typically made with an ankle-brachial index (ABI) ≤ 0.9.⁵ The prevalence of PAD is about 4.3% in Americans aged ≥ 40 years, increases with age, and is almost twice as common among Black Americans compared with that of White Americans.⁶ Many studies in SCI populations have documented an increased prevalence of DM, dyslipidemia, obesity, hypertension (HTN), and cigarette smoking.^7-9 PAD shares these risk factors with coronary artery disease (CAD), but relative to CAD, tobacco smoking was a more substantial causative factor for PAD.¹⁰ Given the preponderance of associated risk factors in this population, PAD is likely more prevalent among patients with SCI than in the population without disabilities. Beyond these known risk factors, researchers hypothesized that SCI contributes to vascular disease by altering arterial function. However, this is still a topic of debate.^11-13 Trauma also is a common cause of amputation, accounting for 45% of amputations in 2005.⁴ Patients with SCI may experience traumatic amputations simultaneously as their SCI, but they may also be predisposed to traumatic amputations related to osteopenia and impaired sensation.

Since amputation is an invasive surgery, knowing the severity of this issue is important in the SCI population. This study quantifies the prevalence of amputations of the LEs among the patients at our SCI center. It then characterizes these amputations’ etiology, their relationship with medical comorbidities, and certain SCI classifications.

Methods

This retrospective cohort study used the US Department of Veterans Affairs (VA) Computerized Patient Record System. The cohort was defined as all patients who received an annual examination at our SCI center over 4 years from October 1, 2009 to September 30, 2013. Annual examination includes a physical examination, relevant surveillance laboratory tests, and imaging, such as renal ultrasound for those with indwelling urinary catheters. One characteristic of the patient population in the VA system is that diagnoses, such as multiple sclerosis (MS) and amyotrophic lateral sclerosis (ALS), that involve spinal cord lesions causing symptoms are included in the registry, besides those with other traumatic or nontraumatic SCI. October 1 to September 30 was chosen based on the VA fiscal year (FY).

During this period, 1678 patients had an annual examination. Of those, 299 patients had an SCI etiology of ALS or MS, and 41 had nonfocal SCI etiology that could not be assessed using the American Spinal Injury Association Impairment Scale (AIS) and were excluded. Also excluded were 283 patients who did not have an annual examination during the specified time span. Some patients do not have an annual examination every year; for those with multiple annual examinations during that time frame, the most recent was used.

One thousand fifty-five patients were included in the statistical analysis. Date of birth, sex, race, ethnicity, date of death, smoking status, DM diagnosis, HTN diagnosis, use of an antiplatelet, antihypertensive, or lipid-lowering agent, blood pressure, hemoglobin A_1c, and lipid panel were collected. The amputation level and etiology were noted. The levels of amputation were classified as toe/partial foot, transtibial amputation (TTA), or transfemoral amputation (TFA). Hip and knee disarticulations were classified at a TFA level. The etiology was classified as dysvascular, traumatic, other, or unknown. Dysvascular included a range of clinical etiologies, including DM, PAD, infection, and poor wound healing. These etiologies were grouped because patients tended to have an overlap of etiologies in the medical record. This collective dysvascular category is consistently used in amputation research, even though the difficulty of identifying this group of etiology can be challenging.^4,14 In the setting of peripheral vascular disease, there may be decreased oxygen delivery, nutrients, or antibiotics that could impair wound healing, leading to infection. Additionally, infection causes microthrombi formation that could lead to worsening ischemia, necrosis, and gangrene.¹⁵ The traumatic classification was applied if the amputation was related to a traumatic event or fracture, including those who failed conservative management of a fracture. The other classification included amputations for cancer.

Statistical Analysis

Descriptive data were summarized as the median and IQR for continuous variables or the number and percentage for categorical variables. The χ² test was used to analyze the association between categorical variables and amputation status. A nonparametric Wilcoxon test was used to investigate the distribution of continuous variables across patients with amputation and patients without amputation. Binary logistic regression analysis was used to investigate amputation risk factors. We report goodness of fit using the Hosmer and Lemeshow test and the area under the curve (AUC) for the multivariate model. Statistical significance was prespecified at a 2-sided P < .05. SAS version 9.4 was used for all statistical analyses.

Results

Mean age was approximately 61 years for the 91 patients at the time of the most recent amputation (Table 1). Among those with amputation, 63% were paraplegic and 37% were tetraplegic.

Of 1055 patients with SCI, 91 (8.6%) patients had an amputation. Of those, 70 (76.1%) were from nontraumatic causes (dysvascular), 17 (18.5%) were traumatic, 4 (4.3%) were from other causes (ie, cancer), and only 1 (1.1%) was of unknown cause.

Injury by Impairment Scale Level

Forty-nine (11.5%) of 426 patients with AIS level A SCI had undergone amputation. In order of prevalence, 23 (46.9%) were unilateral TFA, 17 (34.6%) were bilateral TFA, 10.2% were partial foot/toe, 4.1% were unilateral TTA, and 4.1% were a TTA/TFA combination. Both hip and knee disarticulations were classified in the TFA category.

Sixteen (13.0%) of 123 patients with AIS level B SCI had undergone amputation; 5 (31.3%) of those amputations were unilateral TFA, 6 (37.5%) were bilateral TFA, 3 (18.8%) were partial toe or foot, and 1 (6.3%) was for unilateral and bilateral TTA each.

Twelve (8.4%) of 143 patients with AIS level C SCI had undergone amputation: 6 (50.0%) were bilateral TFA; 3 (25.0%) were unilateral TFA; and 3 (25.0%) were unilateral TTA.

Fourteen (3.9%) of 356 patients with AIS level D SCI had undergone amputation. Of those 6 (42.9%) underwent a partial foot/toe amputation; 5 (35.7%) had undergone a unilateral TTA, and 1 (7.1%) underwent amputation in each of the following categories: bilateral TTA, unilateral TFA, and bilateral TFA each.

None of the 7 individuals with AIS E level SCI had undergone amputation.

Health Risk Factors

Of the 91 patients with amputation, the majority (81.3%) were either former or current smokers. Thirty-six percent of those who had undergone amputation had a diagnosis of DM, while only 21% of those who had not undergone amputation had a diagnosis of DM.

At the time of their annual examination 532 patients had a diagnosis of HTN while 523 patients did not. Among patients with amputations, 59 (64.8%) had HTN, while 32 (35.2%) did not. Of the 964 patients without amputation, the prevalence of HTN was 50.9%

.Of 1055 patients with SCI, only 103 (9.8%) had a PAD diagnosis, including 38 (41.9%) patients with amputation. Just 65 (6.7%) patients with SCI without amputation had PAD (P < .001). PAD is highly correlated with dysvascular causes of amputation. Among those with amputations due to dysvascular etiology, 50.0% (35/70) had PAD, but for the 21 amputations due to nondysvascular etiology, only 3 (14.3%) had PAD (P = .004).

Amputation Predictive Model

A multivariate logistic regression analysis was used to build a predictive model for amputation among patients with SCI while controlling for covariates. In our multivariate analysis, high-density lipoprotein cholesterol (HDL-C), tetraplegia, and PAD were predictive factors for amputation. Patients with SCI who had PAD were 8.6 times more likely to undergo amputation compared to those without PAD (odds ratio [OR], 9.8; P < .001; 95% CI, 5.9-16.3). Every unit of HDL-C decreased the odds of amputation by 5% (OR, 0.95; P < .001; 95% CI, 0.93-0.98).

Having tetraplegia decreased the odds of amputation by 43%, compared with those with paraplegia (OR, 0.57; P = .02; 95% CI, 0.36 - 0.92). AUC was 0.76, and the Hosmer and Lemeshow goodness of model fit test P value was .66, indicating the good predictive power of the model (Table 3).

Discussion

In the US, 54 to 82% of amputations occur secondary to chronic vascular disease. Our study showed similar results: 76.1% of amputations were dysvascular.^4,16 Even in a 2019 systematic review, the most recent prevalence of amputation data was in 2005.¹⁷ The study concluded that among the general population in the US, prevalence of amputation was estimated to be 1 in 190 people, or about 0.5% of the population.⁴ We found that the prevalence of amputation among the SCI population in this study was 8.7%. This result is consistent with our initial hypothesis that the prevalence of amputation would be higher among the people with SCI. Using a different case acquisition method, Svircev and colleagues reported that about a 4% prevalence of LE amputation among veterans with chronic SCI (over 1 year from the initial SCI), with an emphasis that it was not a study of amputation incidence.¹⁸ In comparison, we calculated a 7.5% prevalence of amputation during the chronic SCI stage, which showed institutional variation and a consistent observation that LE amputations occurred more frequently in the SCI population.

Our results showed a positive correlation between the completeness of injury and the prevalence of amputation. Those individuals with a motor complete injury, AIS A (40.3%) or AIS B (11.7%) account for approximately half of all amputations in our population with SCI. Another finding was that proximal amputations were more frequent with more neurologically complete SCIs. Of those with an injury classified as AIS A and an amputation, 42 of 49 subjects underwent at least 1 TFA (23 were unilateral TFA, 17 were bilateral TFA, 2 were a TFA/TTA combination). Of those with an AIS B injury and an amputation, 11 of 16 subjects (68.8%) had at least 1 TFA (5 unilateral TFA and 6 bilateral TFA). Among patients with AIS C injury and amputation, 75% had a TFA. At the same time, only 13.3% of all amputations were at the transfemoral level in those with an AIS D injury. None of the participants with an injury classified as AIS E had undergone an amputation.

Given a paucity of literature available regarding amputation levels in patients with SCI, a discussion with a JAHVH vascular surgeon helped explain the rationale behind different levels of amputation among the SCI population—TFA was performed in 64 of 91 cases (70%). Institutionally, TFAs were performed more often because this level had the greatest chance of healing, avoiding infection, and eliminating knee contracture issues, which may affect quality of life. This was believed to be the best option in those individuals who were already nonambulatory. Although this study did not collect data on ambulatory status, this helps explain why those with an SCI classification of AIS D were more likely to have had a more distal amputation to preserve current or a future chance of ambulation, provided that whether the limb is salvageable is the priority of surgical decision.

The prevalence of PAD among veterans is generally higher than it is in the nonveteran population. Studies show that the prevalence of PAD risk factors in the veteran population exceeds national estimates. Nearly two-thirds of veterans have HTN, 1 in 4 has DM, and 1 in 4 is a current smoker, placing veterans at a significantly increased risk of PADand, therefore, amputation.^19,20 These rates were about the same or greater in our SCI population: 50.4% had HTN, 22.3% had a diagnosis of DM, and 71.8% smoked previously or currently smoked. In 3 large studies, HTN was second only to current smoking as the most attributable risk factor for PAD.²¹

Ongoing research by JAHVH vascular surgeons suggests that patients with SCI were younger and less likely to have HTN, PAD, and/or CAD compared with patients undergoing TFA without SCI. Additionally, patients with SCI had better postoperative outcomes in terms of 30-day mortality, 3-year mortality, and had no increased rate of surgical revisions, strokes, or wound-healing complications. This supports the previous thought that the AIS classification plays a large role in determining amputation levels.

Limitations

As a single-center study carried out at an SCI specialized center of a VA hospital, this study's finding may not be generalizable. Incomplete documentation in the health record may have led to underreporting of amputations and other information. The practice of the vascular surgeons at JAHVH may not represent the approach of vascular surgeons nationwide. Another limitation of this study is that the duration of SCI was not considered when looking at health risk factors associated with amputation in the SCI population (ie, total cholesterol, hemoglobin A_1c, etc). Finally, the medication regimens were not reviewed to determine whether they meet the standard of care in relation to eventual diagnosis of PAD.

A prospective study comparing the prevalence of amputation between veterans with SCI vs veterans without SCI could better investigate the difference in amputation risks. This study only compared our veterans with SCI in reference to the general population. Veterans are more likely to be smokers than the general population, contributing to PAD.¹⁷ In addition, data regarding patients’ functional status in regard to transferring and ambulation before and after amputation were not collected, which would have contributed to an understanding of how amputation affects functional status in this population.

Conclusions

There is an increased prevalence of amputation among veterans with SCI compared with that of the nationwide population and a plurality were TFAs. This data suggest that those with a motor complete SCI are more likely to undergo a more proximal amputation. This is likely secondary to a lower likelihood of ambulation with more neurologically complete injuries along with a greater chance of healing with a more proximal amputation. It is challenging to correlate any variables specific to SCI (ie, immobility, time since injury, level of injury, etc) with an increased risk of amputation as the known comorbidities associated with PAD are highly prevalent in this population. Having PAD, low HDL-C (< 40 mg/dL), and paraplegia instead of tetraplegia were independent predictors of amputation.

Health care professionals need to be aware of the high prevalence of amputation in the SCI population. Comorbidities should be aggressively treated as PAD, in addition to being associated with amputation, has been linked with increased mortality.²⁵ Studies using a larger population and multiple centers are needed to confirm such a concerning finding.

Acknowledgments

This material is based on work supported (or supported in part) with resources and the use of facilities at the James A. Haley Veterans’ Hospital (JAHVH). Authors gratefully acknowledge the inputs and support of Dr. James Brooks, MD, RPVI, assistant professor of surgery, University of South Florida (USF), and attending surgeon, vascular surgery service, medical director of the peripheral vascular laboratory, JAHVH; and Dr. Kevin White, MD, assistant professor, USF, and Chief of Spinal Cord Injury Center, JAHVH.

At the James A. Haley Veterans’ Hospital (JAHVH) in Tampa, Florida, the prevalence of amputations among patients at the spinal cord injury (SCI) center seems high. Despite limited data demonstrating altered hemodynamics in the lower extremities (LEs) among the SCI population and increased frequency of peripheral arterial disease (PAD), amputations among patients with SCI have received little attention in research.^1-3

In the United States, most amputations are caused by vascular disease related to peripheral arterial disease (PAD) and diabetes mellitus (DM).⁴ PAD primarily affects the LEs and is caused by atherosclerotic obstruction leading to insufficient blood flow. PAD can present clinically as LE pain, nonhealing ulcers, nonpalpable distal pulses, shiny or cold skin, absence of hair on the LE, or distal extremity pallor when the affected extremity is elevated. However, PAD is often asymptomatic. The diagnosis of PAD is typically made with an ankle-brachial index (ABI) ≤ 0.9.⁵ The prevalence of PAD is about 4.3% in Americans aged ≥ 40 years, increases with age, and is almost twice as common among Black Americans compared with that of White Americans.⁶ Many studies in SCI populations have documented an increased prevalence of DM, dyslipidemia, obesity, hypertension (HTN), and cigarette smoking.^7-9 PAD shares these risk factors with coronary artery disease (CAD), but relative to CAD, tobacco smoking was a more substantial causative factor for PAD.¹⁰ Given the preponderance of associated risk factors in this population, PAD is likely more prevalent among patients with SCI than in the population without disabilities. Beyond these known risk factors, researchers hypothesized that SCI contributes to vascular disease by altering arterial function. However, this is still a topic of debate.^11-13 Trauma also is a common cause of amputation, accounting for 45% of amputations in 2005.⁴ Patients with SCI may experience traumatic amputations simultaneously as their SCI, but they may also be predisposed to traumatic amputations related to osteopenia and impaired sensation.

Since amputation is an invasive surgery, knowing the severity of this issue is important in the SCI population. This study quantifies the prevalence of amputations of the LEs among the patients at our SCI center. It then characterizes these amputations’ etiology, their relationship with medical comorbidities, and certain SCI classifications.

Methods

This retrospective cohort study used the US Department of Veterans Affairs (VA) Computerized Patient Record System. The cohort was defined as all patients who received an annual examination at our SCI center over 4 years from October 1, 2009 to September 30, 2013. Annual examination includes a physical examination, relevant surveillance laboratory tests, and imaging, such as renal ultrasound for those with indwelling urinary catheters. One characteristic of the patient population in the VA system is that diagnoses, such as multiple sclerosis (MS) and amyotrophic lateral sclerosis (ALS), that involve spinal cord lesions causing symptoms are included in the registry, besides those with other traumatic or nontraumatic SCI. October 1 to September 30 was chosen based on the VA fiscal year (FY).

During this period, 1678 patients had an annual examination. Of those, 299 patients had an SCI etiology of ALS or MS, and 41 had nonfocal SCI etiology that could not be assessed using the American Spinal Injury Association Impairment Scale (AIS) and were excluded. Also excluded were 283 patients who did not have an annual examination during the specified time span. Some patients do not have an annual examination every year; for those with multiple annual examinations during that time frame, the most recent was used.

One thousand fifty-five patients were included in the statistical analysis. Date of birth, sex, race, ethnicity, date of death, smoking status, DM diagnosis, HTN diagnosis, use of an antiplatelet, antihypertensive, or lipid-lowering agent, blood pressure, hemoglobin A_1c, and lipid panel were collected. The amputation level and etiology were noted. The levels of amputation were classified as toe/partial foot, transtibial amputation (TTA), or transfemoral amputation (TFA). Hip and knee disarticulations were classified at a TFA level. The etiology was classified as dysvascular, traumatic, other, or unknown. Dysvascular included a range of clinical etiologies, including DM, PAD, infection, and poor wound healing. These etiologies were grouped because patients tended to have an overlap of etiologies in the medical record. This collective dysvascular category is consistently used in amputation research, even though the difficulty of identifying this group of etiology can be challenging.^4,14 In the setting of peripheral vascular disease, there may be decreased oxygen delivery, nutrients, or antibiotics that could impair wound healing, leading to infection. Additionally, infection causes microthrombi formation that could lead to worsening ischemia, necrosis, and gangrene.¹⁵ The traumatic classification was applied if the amputation was related to a traumatic event or fracture, including those who failed conservative management of a fracture. The other classification included amputations for cancer.

Statistical Analysis

Descriptive data were summarized as the median and IQR for continuous variables or the number and percentage for categorical variables. The χ² test was used to analyze the association between categorical variables and amputation status. A nonparametric Wilcoxon test was used to investigate the distribution of continuous variables across patients with amputation and patients without amputation. Binary logistic regression analysis was used to investigate amputation risk factors. We report goodness of fit using the Hosmer and Lemeshow test and the area under the curve (AUC) for the multivariate model. Statistical significance was prespecified at a 2-sided P < .05. SAS version 9.4 was used for all statistical analyses.

Results

Mean age was approximately 61 years for the 91 patients at the time of the most recent amputation (Table 1). Among those with amputation, 63% were paraplegic and 37% were tetraplegic.

Of 1055 patients with SCI, 91 (8.6%) patients had an amputation. Of those, 70 (76.1%) were from nontraumatic causes (dysvascular), 17 (18.5%) were traumatic, 4 (4.3%) were from other causes (ie, cancer), and only 1 (1.1%) was of unknown cause.

Injury by Impairment Scale Level

Forty-nine (11.5%) of 426 patients with AIS level A SCI had undergone amputation. In order of prevalence, 23 (46.9%) were unilateral TFA, 17 (34.6%) were bilateral TFA, 10.2% were partial foot/toe, 4.1% were unilateral TTA, and 4.1% were a TTA/TFA combination. Both hip and knee disarticulations were classified in the TFA category.

Sixteen (13.0%) of 123 patients with AIS level B SCI had undergone amputation; 5 (31.3%) of those amputations were unilateral TFA, 6 (37.5%) were bilateral TFA, 3 (18.8%) were partial toe or foot, and 1 (6.3%) was for unilateral and bilateral TTA each.

Twelve (8.4%) of 143 patients with AIS level C SCI had undergone amputation: 6 (50.0%) were bilateral TFA; 3 (25.0%) were unilateral TFA; and 3 (25.0%) were unilateral TTA.

Fourteen (3.9%) of 356 patients with AIS level D SCI had undergone amputation. Of those 6 (42.9%) underwent a partial foot/toe amputation; 5 (35.7%) had undergone a unilateral TTA, and 1 (7.1%) underwent amputation in each of the following categories: bilateral TTA, unilateral TFA, and bilateral TFA each.

None of the 7 individuals with AIS E level SCI had undergone amputation.

Health Risk Factors

Of the 91 patients with amputation, the majority (81.3%) were either former or current smokers. Thirty-six percent of those who had undergone amputation had a diagnosis of DM, while only 21% of those who had not undergone amputation had a diagnosis of DM.

At the time of their annual examination 532 patients had a diagnosis of HTN while 523 patients did not. Among patients with amputations, 59 (64.8%) had HTN, while 32 (35.2%) did not. Of the 964 patients without amputation, the prevalence of HTN was 50.9%

.Of 1055 patients with SCI, only 103 (9.8%) had a PAD diagnosis, including 38 (41.9%) patients with amputation. Just 65 (6.7%) patients with SCI without amputation had PAD (P < .001). PAD is highly correlated with dysvascular causes of amputation. Among those with amputations due to dysvascular etiology, 50.0% (35/70) had PAD, but for the 21 amputations due to nondysvascular etiology, only 3 (14.3%) had PAD (P = .004).

Amputation Predictive Model

A multivariate logistic regression analysis was used to build a predictive model for amputation among patients with SCI while controlling for covariates. In our multivariate analysis, high-density lipoprotein cholesterol (HDL-C), tetraplegia, and PAD were predictive factors for amputation. Patients with SCI who had PAD were 8.6 times more likely to undergo amputation compared to those without PAD (odds ratio [OR], 9.8; P < .001; 95% CI, 5.9-16.3). Every unit of HDL-C decreased the odds of amputation by 5% (OR, 0.95; P < .001; 95% CI, 0.93-0.98).

Having tetraplegia decreased the odds of amputation by 43%, compared with those with paraplegia (OR, 0.57; P = .02; 95% CI, 0.36 - 0.92). AUC was 0.76, and the Hosmer and Lemeshow goodness of model fit test P value was .66, indicating the good predictive power of the model (Table 3).

Discussion

In the US, 54 to 82% of amputations occur secondary to chronic vascular disease. Our study showed similar results: 76.1% of amputations were dysvascular.^4,16 Even in a 2019 systematic review, the most recent prevalence of amputation data was in 2005.¹⁷ The study concluded that among the general population in the US, prevalence of amputation was estimated to be 1 in 190 people, or about 0.5% of the population.⁴ We found that the prevalence of amputation among the SCI population in this study was 8.7%. This result is consistent with our initial hypothesis that the prevalence of amputation would be higher among the people with SCI. Using a different case acquisition method, Svircev and colleagues reported that about a 4% prevalence of LE amputation among veterans with chronic SCI (over 1 year from the initial SCI), with an emphasis that it was not a study of amputation incidence.¹⁸ In comparison, we calculated a 7.5% prevalence of amputation during the chronic SCI stage, which showed institutional variation and a consistent observation that LE amputations occurred more frequently in the SCI population.

Our results showed a positive correlation between the completeness of injury and the prevalence of amputation. Those individuals with a motor complete injury, AIS A (40.3%) or AIS B (11.7%) account for approximately half of all amputations in our population with SCI. Another finding was that proximal amputations were more frequent with more neurologically complete SCIs. Of those with an injury classified as AIS A and an amputation, 42 of 49 subjects underwent at least 1 TFA (23 were unilateral TFA, 17 were bilateral TFA, 2 were a TFA/TTA combination). Of those with an AIS B injury and an amputation, 11 of 16 subjects (68.8%) had at least 1 TFA (5 unilateral TFA and 6 bilateral TFA). Among patients with AIS C injury and amputation, 75% had a TFA. At the same time, only 13.3% of all amputations were at the transfemoral level in those with an AIS D injury. None of the participants with an injury classified as AIS E had undergone an amputation.

Given a paucity of literature available regarding amputation levels in patients with SCI, a discussion with a JAHVH vascular surgeon helped explain the rationale behind different levels of amputation among the SCI population—TFA was performed in 64 of 91 cases (70%). Institutionally, TFAs were performed more often because this level had the greatest chance of healing, avoiding infection, and eliminating knee contracture issues, which may affect quality of life. This was believed to be the best option in those individuals who were already nonambulatory. Although this study did not collect data on ambulatory status, this helps explain why those with an SCI classification of AIS D were more likely to have had a more distal amputation to preserve current or a future chance of ambulation, provided that whether the limb is salvageable is the priority of surgical decision.

The prevalence of PAD among veterans is generally higher than it is in the nonveteran population. Studies show that the prevalence of PAD risk factors in the veteran population exceeds national estimates. Nearly two-thirds of veterans have HTN, 1 in 4 has DM, and 1 in 4 is a current smoker, placing veterans at a significantly increased risk of PADand, therefore, amputation.^19,20 These rates were about the same or greater in our SCI population: 50.4% had HTN, 22.3% had a diagnosis of DM, and 71.8% smoked previously or currently smoked. In 3 large studies, HTN was second only to current smoking as the most attributable risk factor for PAD.²¹

Ongoing research by JAHVH vascular surgeons suggests that patients with SCI were younger and less likely to have HTN, PAD, and/or CAD compared with patients undergoing TFA without SCI. Additionally, patients with SCI had better postoperative outcomes in terms of 30-day mortality, 3-year mortality, and had no increased rate of surgical revisions, strokes, or wound-healing complications. This supports the previous thought that the AIS classification plays a large role in determining amputation levels.

Limitations

As a single-center study carried out at an SCI specialized center of a VA hospital, this study's finding may not be generalizable. Incomplete documentation in the health record may have led to underreporting of amputations and other information. The practice of the vascular surgeons at JAHVH may not represent the approach of vascular surgeons nationwide. Another limitation of this study is that the duration of SCI was not considered when looking at health risk factors associated with amputation in the SCI population (ie, total cholesterol, hemoglobin A_1c, etc). Finally, the medication regimens were not reviewed to determine whether they meet the standard of care in relation to eventual diagnosis of PAD.

A prospective study comparing the prevalence of amputation between veterans with SCI vs veterans without SCI could better investigate the difference in amputation risks. This study only compared our veterans with SCI in reference to the general population. Veterans are more likely to be smokers than the general population, contributing to PAD.¹⁷ In addition, data regarding patients’ functional status in regard to transferring and ambulation before and after amputation were not collected, which would have contributed to an understanding of how amputation affects functional status in this population.

Conclusions

There is an increased prevalence of amputation among veterans with SCI compared with that of the nationwide population and a plurality were TFAs. This data suggest that those with a motor complete SCI are more likely to undergo a more proximal amputation. This is likely secondary to a lower likelihood of ambulation with more neurologically complete injuries along with a greater chance of healing with a more proximal amputation. It is challenging to correlate any variables specific to SCI (ie, immobility, time since injury, level of injury, etc) with an increased risk of amputation as the known comorbidities associated with PAD are highly prevalent in this population. Having PAD, low HDL-C (< 40 mg/dL), and paraplegia instead of tetraplegia were independent predictors of amputation.

Health care professionals need to be aware of the high prevalence of amputation in the SCI population. Comorbidities should be aggressively treated as PAD, in addition to being associated with amputation, has been linked with increased mortality.²⁵ Studies using a larger population and multiple centers are needed to confirm such a concerning finding.

Acknowledgments

This material is based on work supported (or supported in part) with resources and the use of facilities at the James A. Haley Veterans’ Hospital (JAHVH). Authors gratefully acknowledge the inputs and support of Dr. James Brooks, MD, RPVI, assistant professor of surgery, University of South Florida (USF), and attending surgeon, vascular surgery service, medical director of the peripheral vascular laboratory, JAHVH; and Dr. Kevin White, MD, assistant professor, USF, and Chief of Spinal Cord Injury Center, JAHVH.

At the James A. Haley Veterans’ Hospital (JAHVH) in Tampa, Florida, the prevalence of amputations among patients at the spinal cord injury (SCI) center seems high. Despite limited data demonstrating altered hemodynamics in the lower extremities (LEs) among the SCI population and increased frequency of peripheral arterial disease (PAD), amputations among patients with SCI have received little attention in research.^1-3

In the United States, most amputations are caused by vascular disease related to peripheral arterial disease (PAD) and diabetes mellitus (DM).⁴ PAD primarily affects the LEs and is caused by atherosclerotic obstruction leading to insufficient blood flow. PAD can present clinically as LE pain, nonhealing ulcers, nonpalpable distal pulses, shiny or cold skin, absence of hair on the LE, or distal extremity pallor when the affected extremity is elevated. However, PAD is often asymptomatic. The diagnosis of PAD is typically made with an ankle-brachial index (ABI) ≤ 0.9.⁵ The prevalence of PAD is about 4.3% in Americans aged ≥ 40 years, increases with age, and is almost twice as common among Black Americans compared with that of White Americans.⁶ Many studies in SCI populations have documented an increased prevalence of DM, dyslipidemia, obesity, hypertension (HTN), and cigarette smoking.^7-9 PAD shares these risk factors with coronary artery disease (CAD), but relative to CAD, tobacco smoking was a more substantial causative factor for PAD.¹⁰ Given the preponderance of associated risk factors in this population, PAD is likely more prevalent among patients with SCI than in the population without disabilities. Beyond these known risk factors, researchers hypothesized that SCI contributes to vascular disease by altering arterial function. However, this is still a topic of debate.^11-13 Trauma also is a common cause of amputation, accounting for 45% of amputations in 2005.⁴ Patients with SCI may experience traumatic amputations simultaneously as their SCI, but they may also be predisposed to traumatic amputations related to osteopenia and impaired sensation.

Since amputation is an invasive surgery, knowing the severity of this issue is important in the SCI population. This study quantifies the prevalence of amputations of the LEs among the patients at our SCI center. It then characterizes these amputations’ etiology, their relationship with medical comorbidities, and certain SCI classifications.

Methods

This retrospective cohort study used the US Department of Veterans Affairs (VA) Computerized Patient Record System. The cohort was defined as all patients who received an annual examination at our SCI center over 4 years from October 1, 2009 to September 30, 2013. Annual examination includes a physical examination, relevant surveillance laboratory tests, and imaging, such as renal ultrasound for those with indwelling urinary catheters. One characteristic of the patient population in the VA system is that diagnoses, such as multiple sclerosis (MS) and amyotrophic lateral sclerosis (ALS), that involve spinal cord lesions causing symptoms are included in the registry, besides those with other traumatic or nontraumatic SCI. October 1 to September 30 was chosen based on the VA fiscal year (FY).

During this period, 1678 patients had an annual examination. Of those, 299 patients had an SCI etiology of ALS or MS, and 41 had nonfocal SCI etiology that could not be assessed using the American Spinal Injury Association Impairment Scale (AIS) and were excluded. Also excluded were 283 patients who did not have an annual examination during the specified time span. Some patients do not have an annual examination every year; for those with multiple annual examinations during that time frame, the most recent was used.

One thousand fifty-five patients were included in the statistical analysis. Date of birth, sex, race, ethnicity, date of death, smoking status, DM diagnosis, HTN diagnosis, use of an antiplatelet, antihypertensive, or lipid-lowering agent, blood pressure, hemoglobin A_1c, and lipid panel were collected. The amputation level and etiology were noted. The levels of amputation were classified as toe/partial foot, transtibial amputation (TTA), or transfemoral amputation (TFA). Hip and knee disarticulations were classified at a TFA level. The etiology was classified as dysvascular, traumatic, other, or unknown. Dysvascular included a range of clinical etiologies, including DM, PAD, infection, and poor wound healing. These etiologies were grouped because patients tended to have an overlap of etiologies in the medical record. This collective dysvascular category is consistently used in amputation research, even though the difficulty of identifying this group of etiology can be challenging.^4,14 In the setting of peripheral vascular disease, there may be decreased oxygen delivery, nutrients, or antibiotics that could impair wound healing, leading to infection. Additionally, infection causes microthrombi formation that could lead to worsening ischemia, necrosis, and gangrene.¹⁵ The traumatic classification was applied if the amputation was related to a traumatic event or fracture, including those who failed conservative management of a fracture. The other classification included amputations for cancer.

Statistical Analysis

Descriptive data were summarized as the median and IQR for continuous variables or the number and percentage for categorical variables. The χ² test was used to analyze the association between categorical variables and amputation status. A nonparametric Wilcoxon test was used to investigate the distribution of continuous variables across patients with amputation and patients without amputation. Binary logistic regression analysis was used to investigate amputation risk factors. We report goodness of fit using the Hosmer and Lemeshow test and the area under the curve (AUC) for the multivariate model. Statistical significance was prespecified at a 2-sided P < .05. SAS version 9.4 was used for all statistical analyses.

Results

Mean age was approximately 61 years for the 91 patients at the time of the most recent amputation (Table 1). Among those with amputation, 63% were paraplegic and 37% were tetraplegic.

Of 1055 patients with SCI, 91 (8.6%) patients had an amputation. Of those, 70 (76.1%) were from nontraumatic causes (dysvascular), 17 (18.5%) were traumatic, 4 (4.3%) were from other causes (ie, cancer), and only 1 (1.1%) was of unknown cause.

Injury by Impairment Scale Level

Forty-nine (11.5%) of 426 patients with AIS level A SCI had undergone amputation. In order of prevalence, 23 (46.9%) were unilateral TFA, 17 (34.6%) were bilateral TFA, 10.2% were partial foot/toe, 4.1% were unilateral TTA, and 4.1% were a TTA/TFA combination. Both hip and knee disarticulations were classified in the TFA category.

Sixteen (13.0%) of 123 patients with AIS level B SCI had undergone amputation; 5 (31.3%) of those amputations were unilateral TFA, 6 (37.5%) were bilateral TFA, 3 (18.8%) were partial toe or foot, and 1 (6.3%) was for unilateral and bilateral TTA each.

Twelve (8.4%) of 143 patients with AIS level C SCI had undergone amputation: 6 (50.0%) were bilateral TFA; 3 (25.0%) were unilateral TFA; and 3 (25.0%) were unilateral TTA.

Fourteen (3.9%) of 356 patients with AIS level D SCI had undergone amputation. Of those 6 (42.9%) underwent a partial foot/toe amputation; 5 (35.7%) had undergone a unilateral TTA, and 1 (7.1%) underwent amputation in each of the following categories: bilateral TTA, unilateral TFA, and bilateral TFA each.

None of the 7 individuals with AIS E level SCI had undergone amputation.

Health Risk Factors

Of the 91 patients with amputation, the majority (81.3%) were either former or current smokers. Thirty-six percent of those who had undergone amputation had a diagnosis of DM, while only 21% of those who had not undergone amputation had a diagnosis of DM.

At the time of their annual examination 532 patients had a diagnosis of HTN while 523 patients did not. Among patients with amputations, 59 (64.8%) had HTN, while 32 (35.2%) did not. Of the 964 patients without amputation, the prevalence of HTN was 50.9%

.Of 1055 patients with SCI, only 103 (9.8%) had a PAD diagnosis, including 38 (41.9%) patients with amputation. Just 65 (6.7%) patients with SCI without amputation had PAD (P < .001). PAD is highly correlated with dysvascular causes of amputation. Among those with amputations due to dysvascular etiology, 50.0% (35/70) had PAD, but for the 21 amputations due to nondysvascular etiology, only 3 (14.3%) had PAD (P = .004).

Amputation Predictive Model

A multivariate logistic regression analysis was used to build a predictive model for amputation among patients with SCI while controlling for covariates. In our multivariate analysis, high-density lipoprotein cholesterol (HDL-C), tetraplegia, and PAD were predictive factors for amputation. Patients with SCI who had PAD were 8.6 times more likely to undergo amputation compared to those without PAD (odds ratio [OR], 9.8; P < .001; 95% CI, 5.9-16.3). Every unit of HDL-C decreased the odds of amputation by 5% (OR, 0.95; P < .001; 95% CI, 0.93-0.98).

Having tetraplegia decreased the odds of amputation by 43%, compared with those with paraplegia (OR, 0.57; P = .02; 95% CI, 0.36 - 0.92). AUC was 0.76, and the Hosmer and Lemeshow goodness of model fit test P value was .66, indicating the good predictive power of the model (Table 3).

Discussion

In the US, 54 to 82% of amputations occur secondary to chronic vascular disease. Our study showed similar results: 76.1% of amputations were dysvascular.^4,16 Even in a 2019 systematic review, the most recent prevalence of amputation data was in 2005.¹⁷ The study concluded that among the general population in the US, prevalence of amputation was estimated to be 1 in 190 people, or about 0.5% of the population.⁴ We found that the prevalence of amputation among the SCI population in this study was 8.7%. This result is consistent with our initial hypothesis that the prevalence of amputation would be higher among the people with SCI. Using a different case acquisition method, Svircev and colleagues reported that about a 4% prevalence of LE amputation among veterans with chronic SCI (over 1 year from the initial SCI), with an emphasis that it was not a study of amputation incidence.¹⁸ In comparison, we calculated a 7.5% prevalence of amputation during the chronic SCI stage, which showed institutional variation and a consistent observation that LE amputations occurred more frequently in the SCI population.

Our results showed a positive correlation between the completeness of injury and the prevalence of amputation. Those individuals with a motor complete injury, AIS A (40.3%) or AIS B (11.7%) account for approximately half of all amputations in our population with SCI. Another finding was that proximal amputations were more frequent with more neurologically complete SCIs. Of those with an injury classified as AIS A and an amputation, 42 of 49 subjects underwent at least 1 TFA (23 were unilateral TFA, 17 were bilateral TFA, 2 were a TFA/TTA combination). Of those with an AIS B injury and an amputation, 11 of 16 subjects (68.8%) had at least 1 TFA (5 unilateral TFA and 6 bilateral TFA). Among patients with AIS C injury and amputation, 75% had a TFA. At the same time, only 13.3% of all amputations were at the transfemoral level in those with an AIS D injury. None of the participants with an injury classified as AIS E had undergone an amputation.

Given a paucity of literature available regarding amputation levels in patients with SCI, a discussion with a JAHVH vascular surgeon helped explain the rationale behind different levels of amputation among the SCI population—TFA was performed in 64 of 91 cases (70%). Institutionally, TFAs were performed more often because this level had the greatest chance of healing, avoiding infection, and eliminating knee contracture issues, which may affect quality of life. This was believed to be the best option in those individuals who were already nonambulatory. Although this study did not collect data on ambulatory status, this helps explain why those with an SCI classification of AIS D were more likely to have had a more distal amputation to preserve current or a future chance of ambulation, provided that whether the limb is salvageable is the priority of surgical decision.

The prevalence of PAD among veterans is generally higher than it is in the nonveteran population. Studies show that the prevalence of PAD risk factors in the veteran population exceeds national estimates. Nearly two-thirds of veterans have HTN, 1 in 4 has DM, and 1 in 4 is a current smoker, placing veterans at a significantly increased risk of PADand, therefore, amputation.^19,20 These rates were about the same or greater in our SCI population: 50.4% had HTN, 22.3% had a diagnosis of DM, and 71.8% smoked previously or currently smoked. In 3 large studies, HTN was second only to current smoking as the most attributable risk factor for PAD.²¹

Ongoing research by JAHVH vascular surgeons suggests that patients with SCI were younger and less likely to have HTN, PAD, and/or CAD compared with patients undergoing TFA without SCI. Additionally, patients with SCI had better postoperative outcomes in terms of 30-day mortality, 3-year mortality, and had no increased rate of surgical revisions, strokes, or wound-healing complications. This supports the previous thought that the AIS classification plays a large role in determining amputation levels.

Limitations

As a single-center study carried out at an SCI specialized center of a VA hospital, this study's finding may not be generalizable. Incomplete documentation in the health record may have led to underreporting of amputations and other information. The practice of the vascular surgeons at JAHVH may not represent the approach of vascular surgeons nationwide. Another limitation of this study is that the duration of SCI was not considered when looking at health risk factors associated with amputation in the SCI population (ie, total cholesterol, hemoglobin A_1c, etc). Finally, the medication regimens were not reviewed to determine whether they meet the standard of care in relation to eventual diagnosis of PAD.

A prospective study comparing the prevalence of amputation between veterans with SCI vs veterans without SCI could better investigate the difference in amputation risks. This study only compared our veterans with SCI in reference to the general population. Veterans are more likely to be smokers than the general population, contributing to PAD.¹⁷ In addition, data regarding patients’ functional status in regard to transferring and ambulation before and after amputation were not collected, which would have contributed to an understanding of how amputation affects functional status in this population.

Conclusions

There is an increased prevalence of amputation among veterans with SCI compared with that of the nationwide population and a plurality were TFAs. This data suggest that those with a motor complete SCI are more likely to undergo a more proximal amputation. This is likely secondary to a lower likelihood of ambulation with more neurologically complete injuries along with a greater chance of healing with a more proximal amputation. It is challenging to correlate any variables specific to SCI (ie, immobility, time since injury, level of injury, etc) with an increased risk of amputation as the known comorbidities associated with PAD are highly prevalent in this population. Having PAD, low HDL-C (< 40 mg/dL), and paraplegia instead of tetraplegia were independent predictors of amputation.

Health care professionals need to be aware of the high prevalence of amputation in the SCI population. Comorbidities should be aggressively treated as PAD, in addition to being associated with amputation, has been linked with increased mortality.²⁵ Studies using a larger population and multiple centers are needed to confirm such a concerning finding.

Acknowledgments

This material is based on work supported (or supported in part) with resources and the use of facilities at the James A. Haley Veterans’ Hospital (JAHVH). Authors gratefully acknowledge the inputs and support of Dr. James Brooks, MD, RPVI, assistant professor of surgery, University of South Florida (USF), and attending surgeon, vascular surgery service, medical director of the peripheral vascular laboratory, JAHVH; and Dr. Kevin White, MD, assistant professor, USF, and Chief of Spinal Cord Injury Center, JAHVH.

In the Veterans Health Administration (VHA), the current prevalence of veterans with dementia is estimated to be about 10%.¹ A 2013 report from the VHA Office of Policy and Planning projected a 22% increase in patients with dementia between 2020 and 2033. That increase amounts to between 276,000 and 335,000 additional veterans enrolled in the US Department of Veterans Affairs (VA) health care.² Of course, these alarming statistics can in no way begin to convey the devastating biopsychosocial impact of Alzheimer disease and other dementias on veterans and their families. In many cases, veterans’ service to their country resulted in injuries and illnesses that increased the risk that they would develop dementia, such as traumatic brain injuries and posttraumatic stress disorder.³

Confronted with these concerning statistics, why didn’t VA Pharmacy Benefits Management (PBM) follow the US Food and Drug Administration (FDA) approval of aduncanumab-avwa for patients with dementia? Instead, PBM issued a monograph in July 2021 that recommended against providing aduncanumab-avwa to patients with Alzheimer dementia (mild or otherwise) or mild cognitive impairment, “given the lack of evidence of a robust and meaningful clinical benefit and the known safety signal.”⁴

In this editorial, I examine the reasons for the PBM recommendation, explain how the VA denial of approval for this new drug for dementia contravened that of the FDA and the ethical implications of this decision for veterans with dementia and the health care professionals (HCPs) who treat them.

The VA PBM national drug monographs are scientific reviews of clinical data supporting the potential inclusion of new medications in the VHA formulary.Aducanumab-avwa is a human monoclonal antibody. Its mechanism of action is to stimulate clearance of β-amyloid plaques from the brains of patients with Alzheimer disease. β-amyloid is a protein byproduct of amyloid precursor protein. Abnormal levels of β-amyloid build up in the brain of a patient with Alzheimer disease, forming clumps that disrupt neuronal connections that enable information transmission and other functions contributing to the death of brain cells.⁵

The FDA approved aducanumab on June 7, 2021, through the accelerated approval pathway.⁶ Drugs approved through the regular pathway must show a clinical benefit. Because detecting and demonstrating clinical benefit through research can be a lengthy process, in 1992 the FDA initiated the accelerated approval pathway. This alternative regulatory option permits the agency to approve a drug that “filled an unmet medical need” for a serious or life-threatening condition based on a surrogate endpoint.⁷ Examples of such endpoints are laboratory values, imaging evidence, physical signs, or other objective findings that are believed to predict a clinical benefit. In 2012, the FDA Safety Innovations Act expanded the basis for approval to an intermediate clinical endpoint: a measure of a therapeutic effect that demonstrates a “reasonable likelihood” of predicting clinical benefit.⁷

The FDA, unlike the PBM, found that aducanumab “provided a meaningful therapeutic advantage over existing treatments.” The FDA underscored that unlike other medications currently available to treat Alzheimer dementias that target symptoms, aducanumab acts on the underlying neurophysiology and neuropathology of the disease based on the decrease in β-amyloid plaques in participants in 2 large clinical trials. The FDA approved the drug for the treatment of patients with either mild cognitive impairment or in the mild state of Alzheimer dementia.

From the time of its announcement, the FDA decision to approve the drug was controversial and criticized in both professional articles and the news media. A particular poignant charge by Largent and Lynch was that the FDA had exploited the desperation of vulnerable patients with dementia and their families willing to try medications with unclear value and uncertain risk precisely because they believed they had no other viable options.⁸ Critics charged that the FDA took the unusual step of overruling the recommendations of a council of its senior advisors, claiming that there was insufficient evidence for approval; that there was a potential conflict of interest between the agency and the pharmaceutical industry; and that the FDA inappropriately used the accelerated approval pathway.⁹ In August 2021, in response to these critiques, the Office of the Inspector General announced that it would review the process the agency used in approving the drug.¹⁰ Nor was the VA alone in its refusal: The Centers for Medicare and Medicaid (CMS) has proposed to cover the drug for its beneficiaries enrolled in CMS-endorsed clinical trials with the caveat that the drug’s manufacturer, Biogen, must continue to conduct studies on the safety and effectiveness of the drug.¹¹

Why did VHA come to a different scientific conclusion than that of the FDA? In reviewing the data from the 2 major studies, PBM did not find that this surrogate endpoint of reduction in β-amyloid plaques was a valid measure of a meaningful clinical benefit. Further, this lack of valid therapeutic change could not outweigh the risks of the amyloid-related imaging abnormalities (ARIA) in research participants. ARIA include cerebral vasogenic edema, effusions in sulci, microhemorrhages in the brain, and/or localized superficial siderosis. These findings are thought to be the result of the antibody binding to β-amyloid deposits that in turn increase cerebrovascular permeability.⁵

In the Veterans Health Administration (VHA), the current prevalence of veterans with dementia is estimated to be about 10%.¹ A 2013 report from the VHA Office of Policy and Planning projected a 22% increase in patients with dementia between 2020 and 2033. That increase amounts to between 276,000 and 335,000 additional veterans enrolled in the US Department of Veterans Affairs (VA) health care.² Of course, these alarming statistics can in no way begin to convey the devastating biopsychosocial impact of Alzheimer disease and other dementias on veterans and their families. In many cases, veterans’ service to their country resulted in injuries and illnesses that increased the risk that they would develop dementia, such as traumatic brain injuries and posttraumatic stress disorder.³

Confronted with these concerning statistics, why didn’t VA Pharmacy Benefits Management (PBM) follow the US Food and Drug Administration (FDA) approval of aduncanumab-avwa for patients with dementia? Instead, PBM issued a monograph in July 2021 that recommended against providing aduncanumab-avwa to patients with Alzheimer dementia (mild or otherwise) or mild cognitive impairment, “given the lack of evidence of a robust and meaningful clinical benefit and the known safety signal.”⁴

In this editorial, I examine the reasons for the PBM recommendation, explain how the VA denial of approval for this new drug for dementia contravened that of the FDA and the ethical implications of this decision for veterans with dementia and the health care professionals (HCPs) who treat them.

The VA PBM national drug monographs are scientific reviews of clinical data supporting the potential inclusion of new medications in the VHA formulary.Aducanumab-avwa is a human monoclonal antibody. Its mechanism of action is to stimulate clearance of β-amyloid plaques from the brains of patients with Alzheimer disease. β-amyloid is a protein byproduct of amyloid precursor protein. Abnormal levels of β-amyloid build up in the brain of a patient with Alzheimer disease, forming clumps that disrupt neuronal connections that enable information transmission and other functions contributing to the death of brain cells.⁵

The FDA approved aducanumab on June 7, 2021, through the accelerated approval pathway.⁶ Drugs approved through the regular pathway must show a clinical benefit. Because detecting and demonstrating clinical benefit through research can be a lengthy process, in 1992 the FDA initiated the accelerated approval pathway. This alternative regulatory option permits the agency to approve a drug that “filled an unmet medical need” for a serious or life-threatening condition based on a surrogate endpoint.⁷ Examples of such endpoints are laboratory values, imaging evidence, physical signs, or other objective findings that are believed to predict a clinical benefit. In 2012, the FDA Safety Innovations Act expanded the basis for approval to an intermediate clinical endpoint: a measure of a therapeutic effect that demonstrates a “reasonable likelihood” of predicting clinical benefit.⁷

The FDA, unlike the PBM, found that aducanumab “provided a meaningful therapeutic advantage over existing treatments.” The FDA underscored that unlike other medications currently available to treat Alzheimer dementias that target symptoms, aducanumab acts on the underlying neurophysiology and neuropathology of the disease based on the decrease in β-amyloid plaques in participants in 2 large clinical trials. The FDA approved the drug for the treatment of patients with either mild cognitive impairment or in the mild state of Alzheimer dementia.

From the time of its announcement, the FDA decision to approve the drug was controversial and criticized in both professional articles and the news media. A particular poignant charge by Largent and Lynch was that the FDA had exploited the desperation of vulnerable patients with dementia and their families willing to try medications with unclear value and uncertain risk precisely because they believed they had no other viable options.⁸ Critics charged that the FDA took the unusual step of overruling the recommendations of a council of its senior advisors, claiming that there was insufficient evidence for approval; that there was a potential conflict of interest between the agency and the pharmaceutical industry; and that the FDA inappropriately used the accelerated approval pathway.⁹ In August 2021, in response to these critiques, the Office of the Inspector General announced that it would review the process the agency used in approving the drug.¹⁰ Nor was the VA alone in its refusal: The Centers for Medicare and Medicaid (CMS) has proposed to cover the drug for its beneficiaries enrolled in CMS-endorsed clinical trials with the caveat that the drug’s manufacturer, Biogen, must continue to conduct studies on the safety and effectiveness of the drug.¹¹

Why did VHA come to a different scientific conclusion than that of the FDA? In reviewing the data from the 2 major studies, PBM did not find that this surrogate endpoint of reduction in β-amyloid plaques was a valid measure of a meaningful clinical benefit. Further, this lack of valid therapeutic change could not outweigh the risks of the amyloid-related imaging abnormalities (ARIA) in research participants. ARIA include cerebral vasogenic edema, effusions in sulci, microhemorrhages in the brain, and/or localized superficial siderosis. These findings are thought to be the result of the antibody binding to β-amyloid deposits that in turn increase cerebrovascular permeability.⁵

In the Veterans Health Administration (VHA), the current prevalence of veterans with dementia is estimated to be about 10%.¹ A 2013 report from the VHA Office of Policy and Planning projected a 22% increase in patients with dementia between 2020 and 2033. That increase amounts to between 276,000 and 335,000 additional veterans enrolled in the US Department of Veterans Affairs (VA) health care.² Of course, these alarming statistics can in no way begin to convey the devastating biopsychosocial impact of Alzheimer disease and other dementias on veterans and their families. In many cases, veterans’ service to their country resulted in injuries and illnesses that increased the risk that they would develop dementia, such as traumatic brain injuries and posttraumatic stress disorder.³

Confronted with these concerning statistics, why didn’t VA Pharmacy Benefits Management (PBM) follow the US Food and Drug Administration (FDA) approval of aduncanumab-avwa for patients with dementia? Instead, PBM issued a monograph in July 2021 that recommended against providing aduncanumab-avwa to patients with Alzheimer dementia (mild or otherwise) or mild cognitive impairment, “given the lack of evidence of a robust and meaningful clinical benefit and the known safety signal.”⁴

In this editorial, I examine the reasons for the PBM recommendation, explain how the VA denial of approval for this new drug for dementia contravened that of the FDA and the ethical implications of this decision for veterans with dementia and the health care professionals (HCPs) who treat them.

The VA PBM national drug monographs are scientific reviews of clinical data supporting the potential inclusion of new medications in the VHA formulary.Aducanumab-avwa is a human monoclonal antibody. Its mechanism of action is to stimulate clearance of β-amyloid plaques from the brains of patients with Alzheimer disease. β-amyloid is a protein byproduct of amyloid precursor protein. Abnormal levels of β-amyloid build up in the brain of a patient with Alzheimer disease, forming clumps that disrupt neuronal connections that enable information transmission and other functions contributing to the death of brain cells.⁵

The FDA approved aducanumab on June 7, 2021, through the accelerated approval pathway.⁶ Drugs approved through the regular pathway must show a clinical benefit. Because detecting and demonstrating clinical benefit through research can be a lengthy process, in 1992 the FDA initiated the accelerated approval pathway. This alternative regulatory option permits the agency to approve a drug that “filled an unmet medical need” for a serious or life-threatening condition based on a surrogate endpoint.⁷ Examples of such endpoints are laboratory values, imaging evidence, physical signs, or other objective findings that are believed to predict a clinical benefit. In 2012, the FDA Safety Innovations Act expanded the basis for approval to an intermediate clinical endpoint: a measure of a therapeutic effect that demonstrates a “reasonable likelihood” of predicting clinical benefit.⁷

The FDA, unlike the PBM, found that aducanumab “provided a meaningful therapeutic advantage over existing treatments.” The FDA underscored that unlike other medications currently available to treat Alzheimer dementias that target symptoms, aducanumab acts on the underlying neurophysiology and neuropathology of the disease based on the decrease in β-amyloid plaques in participants in 2 large clinical trials. The FDA approved the drug for the treatment of patients with either mild cognitive impairment or in the mild state of Alzheimer dementia.

From the time of its announcement, the FDA decision to approve the drug was controversial and criticized in both professional articles and the news media. A particular poignant charge by Largent and Lynch was that the FDA had exploited the desperation of vulnerable patients with dementia and their families willing to try medications with unclear value and uncertain risk precisely because they believed they had no other viable options.⁸ Critics charged that the FDA took the unusual step of overruling the recommendations of a council of its senior advisors, claiming that there was insufficient evidence for approval; that there was a potential conflict of interest between the agency and the pharmaceutical industry; and that the FDA inappropriately used the accelerated approval pathway.⁹ In August 2021, in response to these critiques, the Office of the Inspector General announced that it would review the process the agency used in approving the drug.¹⁰ Nor was the VA alone in its refusal: The Centers for Medicare and Medicaid (CMS) has proposed to cover the drug for its beneficiaries enrolled in CMS-endorsed clinical trials with the caveat that the drug’s manufacturer, Biogen, must continue to conduct studies on the safety and effectiveness of the drug.¹¹

Why did VHA come to a different scientific conclusion than that of the FDA? In reviewing the data from the 2 major studies, PBM did not find that this surrogate endpoint of reduction in β-amyloid plaques was a valid measure of a meaningful clinical benefit. Further, this lack of valid therapeutic change could not outweigh the risks of the amyloid-related imaging abnormalities (ARIA) in research participants. ARIA include cerebral vasogenic edema, effusions in sulci, microhemorrhages in the brain, and/or localized superficial siderosis. These findings are thought to be the result of the antibody binding to β-amyloid deposits that in turn increase cerebrovascular permeability.⁵

Physical exercise offers preventative and therapeutic benefits for a range of chronic health conditions, including cardiovascular disease, type 2 diabetes mellitus, Alzheimer disease, and depression.^1,2 Exercise has been well studied for its antidepressant effects, its ability to reduce risk of aging-related dementia, and favorable effects on a range of cognitive functions.² Lesser evidence exists regarding the impact of exercise on other mental health concerns. Therefore, an accurate understanding of whether physical exercise may ameliorate other conditions is important.

A small meta-analysis by Rosenbaum and colleagues found that exercise interventions were superior to control conditions for symptom reduction in study participants with posttraumatic stress disorder (PTSD).³ This meta-analysis included 4 randomized clinical trials representing 200 cases. The trial included a variety of physical activities (eg, yoga, aerobic, and strength-building exercises) and control conditions, and participants recruited from online, community, inpatient, and outpatient settings. The standardized mean difference (SMD) produced by the analysis indicated a small-to-medium effect (Hedges g, -0.35), with the authors reporting no evidence of publication bias, although an assessment of potential bias associated with individual trial design characteristics was not conducted. Of note, a meta-analysis by Watts and colleagues found that effect sizes for PTSD treatments tend to be smaller in veteran populations.⁴ Therefore, how much the mean effect size estimate in the study is applicable to veterans with PTSD is unknown.³

Veterans represent a unique subpopulation in which PTSD is common, although no meta-analysis yet published has synthesized the effects of exercise interventions from trials of veterans with PTSD.⁵ A recent systematic review by Whitworth and Ciccolo concluded that exercise may be associated with reduced risk of PTSD, a briefer course of PTSD symptoms, and/or reduced sleep- and depression-related difficulties.⁶ However, that review primarily included observational, cross-sectional, and qualitative works. No trials included in our meta-analysis were included in that review.⁶

Evidence-based psychotherapies like cognitive processing therapy and prolonged exposure have been shown to be effective for treating PTSD in veterans; however, these modalities are accompanied by high rates of dropout (eg, 40-60%), thereby limiting their clinical utility.⁷ The use of complementary and alternative approaches for treatment in the United States has increased in recent years, and exercise represents an important complementary treatment option.⁸ In a study by Baldwin and colleagues, nearly 50% of veterans reported using complementary or alternative approaches, and veterans with PTSD were among those likely to use such approaches.⁹ However, current studies of the effects of exercise interventions on PTSD symptom reduction are mostly small and varied, making determinations difficult regarding the potential utility of exercise for treating this condition in veterans.

Literature Search

No previous research has synthesized the literature on the effects of exercise on PTSD in the veteran population. The current meta-analysis aims to provide a synthesis of systematically selected studies on this topic to determine whether exercise-based interventions are effective at reducing veterans’ symptoms of PTSD. Our hypothesis was that, when used as a primary or adjuvant intervention for PTSD, physical exercise would be associated with a reduction of PTSD symptom scale scores. We planned a priori to produce separate estimates for single-arm and multi-arm trials. We also wanted to conduct a careful risk of bias assessment—or evaluation of study features that may have systematically influenced results—for included trials, not only to provide context for interpretation of results, but also to inform suggestions for research to advance this field of inquiry.¹⁰

Methods

This study was preregistered on PROSPERO and followed PRISMA guidelines for meta-analyses and systematic reviews.¹¹ Supplementary materials, such as the PRISMA checklist, study data, and funnel plots, are available online (doi.org/10.6084/m9.figshare.c.5618437.v1). Conference abstracts were omitted due to a lack of necessary information. We decided early in the planning process to include both randomized and single-arm trials, expecting the number of completed studies in the area of exercise for PTSD symptom reduction in veterans, and particularly randomized trials of such, would be relatively small.

Studies were included if they met the following criteria: (1) the study was a single- or multi-arm interventional trial; (2) participants were veterans; (3) participants had a current diagnosis of PTSD or exhibited subthreshold PTSD symptoms, as established by authors of the individual studies and supported by a structured clinical interview, semistructured interview, or elevated scores on PTSD symptom self-report measures; (4) the study included an intervention in which exercise (physical activity that is planned, structured, repetitive, and purposive in the sense that improvement or maintenance of physical fitness or health is an objective) was the primary component; (5) PTSD symptom severity was by a clinician-rated or self-report measure; and (6) the study was published in a peer-reviewed journal.¹² Studies were excluded if means, standard deviations, and sample sizes were not available or the full text of the study was not available in English.

Data Collection

Data were extracted from included studies using custom forms and included the following information based on PRISMA guidelines: (1) study design characteristics; (2) intervention details; and (3) PTSD outcome information.¹¹ PTSD symptom severity was the primary outcome of interest. Outcome data were included if they were derived from a measure of PTSD symptoms—equivalency across measures was assumed for meta-analyses. Potential study bias for each outcome was evaluated using the ROBINS-I and Cochrane Collaboration’s RoB 2 tools for single-arm and multi-arm trials, respectively.^13,14 These tools evaluate domains related to the design, conduct, and analysis of studies that are associated with bias (ie, systematic error in findings, such as under- or overestimation of results).¹⁰ Examples include how well authors performed and concealed randomization procedures, addressed missing data, and measured study outcomes.^13,14 The risk of bias (eg, low, moderate, serious) associated with each domain is rated and, based on the domain ratings, each study is then given an overall rating regarding how much risk influences bias.^13,14 Broadly, lower risk of bias corresponds to higher confidence in the validity of results.

Finally, 4 authors (associated with 2 single- and 2 multi-arm studies) were contacted and asked to provide further information. Data for 1 additional multi-arm study were obtained from these communications and included in the final study selection.¹⁵ These authors were also asked for information about any unpublished works of which they were aware, although no additional works were identified.

Statistical Analyses

Analyses were performed with R Studio R 3.6.0 software.¹⁶ An SMD (also known as Hedges g) was calculated for each study outcome: for single-arm trials, this was the SMD between pre- and postintervention scores, whereas for multi-arm trials, this was the SMD between postintervention outcome scores across groups. CIs for each SMD were calculated using a standard normal distribution. Combined SMDs were estimated separately for single- and multi-arm studies, using random-effects meta-analyses. In order to include multiple relevant outcomes from a single trial (ie, for studies using multiple PTSD symptom measures), robust variance estimation was used.¹⁷ Precision was used to weight SMDs.

Correlations between pre- and postintervention scores were not available for 1 single-arm study.¹⁸ A correlation coefficient of 0.8 was imputed to calculate the standard error of the of the SMDs for the Clinician-Administered PTSD Scale (CAPS) and the PTSD Checklist (PCL), as this value is consistent with past findings regarding the test-retest reliability of these measures.^19-22 A sensitivity analysis, using several alternative correlational values, revealed that the choice of correlation coefficient did not impact the overall results of the meta-analysis.

I² was used to evaluate between-study heterogeneity. Values of I² > 25%, 50%, and 75% were selected to reflect low, moderate, and high heterogeneity, respectively, in accordance with guidelines described by Higgins and colleagues.²³ Potential publication bias was assessed via funnel plot and Egger test.²⁴ Finally, although collection of depressive symptom scores was proposed as a secondary outcome in the study protocol, such data were available only for 1 multi-arm study. As a result, this outcome was not evaluated.

Results

Six studies with 101 total participants were included in the single-arm analyses (Table 1).^18,25-29 Participants consisted of veterans with chronic pain, post-9/11 veterans, female veterans of childbearing age, veterans with a history of trauma therapy, and other veterans. Types of exercise included moderate aerobic exercise and yoga. PTSD symptom measures included the CAPS and the PCL (PCL-5 or PCL-M versions). Reported financial sources for included studies included federal grant funding, nonprofit material support, outside organization support, use of US Department of Veterans Affairs (VA) resources, and no reported financial support.

With respect to individual studies, Shivakumar and colleagues found that completion of an aerobic exercise program was associated with reduced scores on 2 different PTSD symptom scales (PCL and CAPS) in 16 women veterans.¹⁸ A trauma-informed yoga intervention study with 18 participants by Cushing and colleagues demonstrated veteran participation to be associated with large reductions in PTSD, anxiety, and depression scale scores.²⁵ In a study with 34 veterans, Chopin and colleagues found that a trauma-informed yoga intervention was associated with a statistically significant reduction in PTSD symptoms, as did a study by Zaccari and colleagues with 17 veterans.^26,29 Justice and Brems also found some evidence that trauma-informed yoga interventions helped PTSD symptoms in a small sample of 4 veterans, although these results were not quantitatively analyzed.²⁷ In contrast, a small pilot study (n = 12) by Staples and colleagues testing a biweekly, 6-week yoga program did not show a significant effect on PTSD symptoms.²⁸

Three studies with 217 total veteran participants were included in the multi-arm analyses (Table 2).^15,30,31 As all multi-arm trials incorporated randomization, they will be referred to as randomized controlled trials (RCTs). On contact, Davis and colleagues provided veteran-specific results for their trial; as such, our data differ from those within the published article.¹⁵ Participants from all included studies were veterans currently experiencing symptoms of PTSD. Types of exercise included yoga and combined methods (eg, aerobic and strength training).^15,30,31 PTSD symptom measures included the CAPS or the PCL-5.^15,30,31 Reported financial sources for included studies included federal grant funding, as well as nonprofit support, private donations, and VA and Department of Defense resources.

Davis and colleagues conducted a recently concluded RCT with > 130 veteran participants and found that a novel manualized yoga program was superior to an attention control in reducing PTSD symptom scale scores for veterans.¹⁵ Goldstein and colleagues found that a program consisting of both aerobic and resistance exercises reduced PTSD symptoms to a greater extent than a waitlist control condition, with 47 veterans randomized in this trial.³⁰ Likewise, Hall and colleagues conducted a pilot RCT in which an intervention that integrated exercise and cognitive behavioral techniques was compared to a waitlist control condition.³¹ For the 48 veterans included in the analyses, the authors reported greater PTSD symptom reduction associated with integrated exercise than that of the control condition; however, the study was not powered to detect statistically significant differences between groups.

Bias Assessment

Results for the risk of bias assessments can be viewed in Tables 3 and 4. For single-arm studies, overall risk of bias was serious for all included trials. Serious risk of bias was found in 2 domains: confounding, due to a lack of accounting for potential preexisting baseline trends (eg, regression to the mean) that could have impacted study results; and measurement, due to the use of a self-report symptom measure (PCL) or CAPS with unblinded assessors. Multiple studies also showed moderate risk in the missing data domain due to participant dropout without appropriate analytic methods to address potential bias.

For RCTs, overall risk of bias ranged from some concerns to high risk. High risk of bias was found in 1 domain, measurement of outcome, due to use of a self-report symptom measure (PCL) with unblinded groups.³¹ The other 2 studies all had some concern of bias in at least 1 of the following domains: randomization, missing data, and measurement of outcome.

Pooled Standardized Mean Differences

Meta-analytic results can be viewed in Figure 2. The pooled SMD for the 6 single-arm studies was -0.60 (df = 4.41, 95% CI, -1.08 to -0.12, P = .03), indicating a statistically significant reduction in PTSD symptoms over the course of an exercise intervention. Combining SMDs for the 3 included RCTs revealed a pooled SMD of -0.40 (df = 1.57, 95% CI, -0.86 to 0.06, P = .06), indicating that exercise did not result in a statistically significant reduction in PTSD symptoms compared with control conditions.

Publication Bias and Heterogeneity

Visual inspection funnel plots and Egger test did not suggest the presence of publication bias for RCTs (t = 1.21, df = 2, P = .35) or single-arm studies (t = -0.36, df = 5, P = .73).

Single-arm studies displayed a high degree of heterogeneity (I² = 81.5%). Including sample size or exercise duration as variables in meta-regressions did not reduce heterogeneity (I² = 85.2% and I² = 83.8%, respectively). Performing a subgroup analysis only on studies using yoga as an intervention also did not reduce heterogeneity (I² = 79.2%). Due to the small number of studies, no further exploration of heterogeneity was conducted on single-arm studies. RCTs did not display any heterogeneity (I² = 0%).

Discussion

Our report represents an early synthesis of the first prospective studies of physical exercise interventions for PTSD in veterans. Results from meta-analyses of 6 single-arm studies (101 participants) and 3 RCTs (217 participants) provide early evidence that exercise may reduce PTSD symptoms in veterans. Yoga was the most common form of exercise used in single-arm studies, whereas RCTs used a wider range of interventions. The pooled SMD of -0.60 for single-arm longitudinal studies suggest a medium decrease in PTSD symptoms for veterans who engage in exercise interventions. Analysis of the RCTs supported this finding, with a pooled SMD of -0.40 reflecting a small-to-medium effect of exercise on PTSD symptoms over control conditions, although this result did not achieve statistical significance. Of note, while the nonsignificant finding for RCTs may have been due to insufficient power caused by the limited number of included studies, possibly exercise was not more efficacious than were the control conditions.

Although RCTs represented a variety of exercise types, PTSD symptom measures, and veteran subgroups, statistical results were not indicative of heterogeneity. However, only the largest and most comprehensive study of exercise for PTSD in veterans to date by Davis and colleagues had a statistically significant SMD.¹⁵ Of note, one of the other 2 RCTs displayed an SMD of a similar magnitude, but this study had a much smaller sample size and was underpowered to detect significance.³⁰ Additionally, risk of bias assessments for single-arm studies and RCTs revealed study characteristics that suggest possible inflation of absolute effect sizes for individual studies. Therefore, the pooled SMDs we report are interpretable but may exceed the true effect of exercise for PTSD symptom reduction in veterans.

Limitations

The present study has several important limitations. First, few studies were found that met the broad eligibility criteria and those that did often had a small sample size. Besides highlighting a gap in the extant research, the limited studies available for meta-analysis means that caution must be taken when interpreting results. Fortunately, this issue will likely resolve once additional studies investigating the impact of exercise on PTSD symptoms in veterans are available for synthesis.

Relatedly, the included study interventions varied considerably, both in the types of exercise used and the characteristics of the exercises (eg, frequency, duration, and intensity), which is relevant as different exercise modalities are associated with differential physical effects.³³ Including such a mixture of exercises may have given an incomplete picture of their potential therapeutic effects. Also, none of the RCTs compared exercise against first-line treatments for PTSD, such as prolonged exposure or cognitive processing therapy, which would have provided further insight into the role exercise could play in clinical settings.⁷

Another limitation is the elevated risk of bias found in most studies, particularly present in the longitudinal single-arm studies, all of which were rated at serious risk. For instance, no single-arm study controlled for preexisting baseline trends: without such (and lacking a comparison control group like in RCTs), it is possible that the observed effects were due to extraneous factors, rather than the exercise intervention. Although not as severe, the multi-arm RCTs also displayed at least moderate risk of bias. Therefore, SMDs may have been overestimated for each group of studies.

Finally, the results of the single-arm meta-analysis displayed high statistical heterogeneity, reducing the generalizability of the results. One possible cause of this heterogeneity may have been the yoga interventions, as a separate analysis removing the only nonyoga study did not reduce heterogeneity. This result was surprising, as the included yoga interventions seemed similar across studies. While the presence of high heterogeneity does require some caution when applying these results to outside interventions, the present study made use of random-effects meta-analysis, a technique that incorporates study heterogeneity into the statistical model, thereby strengthening the findings compared with that of a traditional fixed-effects approach.¹⁰

Future Steps

Several future steps are warranted to improve knowledge of exercise as a treatment for PTSD in veterans and in the general population. With current meta-analyses limited to small numbers of studies, additional studies of the efficacy of exercise for treating PTSD could help in several ways. A larger pool of studies would enable future meta-analyses to explore related questions, such as those regarding the impact of exercise on quality of life or depressive symptom reduction among veterans with PTSD. A greater number of studies also would enable meta-analysts to explore potentially critical moderators. For example, the duration, frequency, or type of exercise may moderate the effect of exercise on PTSD symptom reduction. Moderators related to patient or study design characteristics also should be explored in future studies.

Future work also should evaluate the impact that specific features of exercise regimens have on PTSD. Knowing whether the type or structure of exercise affects its clinical use would be invaluable in developing and implementing efficient exercise-based interventions. For example, if facilitated exercise was found to be significantly more effective at reducing PTSD symptoms than exercise completed independently, the development of exercise intervention programs in the VA and other facilities that commonly treat PTSD may be warranted. Additionally, it may be useful to identify specific mechanisms through which exercise reduces PTSD symptoms. For example, in addition to its beneficial biological effects, exercise also promotes psychological health through behavioral activation and alterations within reinforcement/reward systems, suggesting that exercise regularity may be more important than intensity.^34,35 Understanding which mechanisms contribute most to change will aid in the development of more efficient interventions.

Given that veterans are demonstrating considerable interest in complementary and alternative PTSD treatments, it is critical that researchers focus on high-quality randomized tests of these interventions. Therefore, in addition to greater quality of exercise intervention studies, future efforts should be focused on RCTs that are designed in such a way as to limit potential introduction of bias. For example, assessment data should be completed by blinded assessors using standardized measures, and analyses should account for missing data and unequal participant attrition between groups. Ideally, pre-intervention trends across multiple baseline datapoints also would be collected in single-arm studies to avoid confounding related to regression to the mean. It is also recommended that future meta-analyses use risk of bias assessments and consider how the results of such assessments may impact the interpretation of results.

Conclusions

Findings from both single-arm studies and RCTs suggest possible benefit of exercise on PTSD symptom reduction, although confirmation of findings is needed. No study found increased symptoms following exercise intervention. Thus, it is reasonable to consider physical exercise, such as yoga, as an adjunct, whole-health consistent treatment. HCPs working with veterans with past traumatic experiences should consider incorporating exercise into patient care. Enhanced educational efforts emphasizing the psychotherapeutic impact of exercise may also have value for the veteran population. Furthermore, the current risk of bias assessments highlights the need for additional high-quality RCTs evaluating the specific impact of exercise on PTSD symptom reduction in veterans. In particular, this field of inquiry would benefit from larger samples and design characteristics to reduce bias (eg, blinding when possible, use of CAPS vs only self-report symptom measures, reducing problematic attrition, corrections for missing data, etc).

Acknowledgments

This research is the result of work supported with resources and the use of facilities at the VA Eastern Kansas Healthcare System (Dwight D. Eisenhower VA Medical Center). It was also supported by the Department of Veterans Affairs Office of Academic Affiliations Advanced Fellowship Program in Mental Illness Research and Treatment, as well as the Rocky Mountain Mental Illness Research, Education, and Clinical Center. Since Dr. Reis and Dr. Gaddy are employees of the US Government and contributed to this manuscript as part of their official duties, the work is not subject to US copyright. This study was preregistered on PROSPERO (https://www.crd.york.ac.uk/prospero/; ID: CRD42020153419).

Physical exercise offers preventative and therapeutic benefits for a range of chronic health conditions, including cardiovascular disease, type 2 diabetes mellitus, Alzheimer disease, and depression.^1,2 Exercise has been well studied for its antidepressant effects, its ability to reduce risk of aging-related dementia, and favorable effects on a range of cognitive functions.² Lesser evidence exists regarding the impact of exercise on other mental health concerns. Therefore, an accurate understanding of whether physical exercise may ameliorate other conditions is important.

A small meta-analysis by Rosenbaum and colleagues found that exercise interventions were superior to control conditions for symptom reduction in study participants with posttraumatic stress disorder (PTSD).³ This meta-analysis included 4 randomized clinical trials representing 200 cases. The trial included a variety of physical activities (eg, yoga, aerobic, and strength-building exercises) and control conditions, and participants recruited from online, community, inpatient, and outpatient settings. The standardized mean difference (SMD) produced by the analysis indicated a small-to-medium effect (Hedges g, -0.35), with the authors reporting no evidence of publication bias, although an assessment of potential bias associated with individual trial design characteristics was not conducted. Of note, a meta-analysis by Watts and colleagues found that effect sizes for PTSD treatments tend to be smaller in veteran populations.⁴ Therefore, how much the mean effect size estimate in the study is applicable to veterans with PTSD is unknown.³

Veterans represent a unique subpopulation in which PTSD is common, although no meta-analysis yet published has synthesized the effects of exercise interventions from trials of veterans with PTSD.⁵ A recent systematic review by Whitworth and Ciccolo concluded that exercise may be associated with reduced risk of PTSD, a briefer course of PTSD symptoms, and/or reduced sleep- and depression-related difficulties.⁶ However, that review primarily included observational, cross-sectional, and qualitative works. No trials included in our meta-analysis were included in that review.⁶

Evidence-based psychotherapies like cognitive processing therapy and prolonged exposure have been shown to be effective for treating PTSD in veterans; however, these modalities are accompanied by high rates of dropout (eg, 40-60%), thereby limiting their clinical utility.⁷ The use of complementary and alternative approaches for treatment in the United States has increased in recent years, and exercise represents an important complementary treatment option.⁸ In a study by Baldwin and colleagues, nearly 50% of veterans reported using complementary or alternative approaches, and veterans with PTSD were among those likely to use such approaches.⁹ However, current studies of the effects of exercise interventions on PTSD symptom reduction are mostly small and varied, making determinations difficult regarding the potential utility of exercise for treating this condition in veterans.

Literature Search

No previous research has synthesized the literature on the effects of exercise on PTSD in the veteran population. The current meta-analysis aims to provide a synthesis of systematically selected studies on this topic to determine whether exercise-based interventions are effective at reducing veterans’ symptoms of PTSD. Our hypothesis was that, when used as a primary or adjuvant intervention for PTSD, physical exercise would be associated with a reduction of PTSD symptom scale scores. We planned a priori to produce separate estimates for single-arm and multi-arm trials. We also wanted to conduct a careful risk of bias assessment—or evaluation of study features that may have systematically influenced results—for included trials, not only to provide context for interpretation of results, but also to inform suggestions for research to advance this field of inquiry.¹⁰

Methods

This study was preregistered on PROSPERO and followed PRISMA guidelines for meta-analyses and systematic reviews.¹¹ Supplementary materials, such as the PRISMA checklist, study data, and funnel plots, are available online (doi.org/10.6084/m9.figshare.c.5618437.v1). Conference abstracts were omitted due to a lack of necessary information. We decided early in the planning process to include both randomized and single-arm trials, expecting the number of completed studies in the area of exercise for PTSD symptom reduction in veterans, and particularly randomized trials of such, would be relatively small.

Studies were included if they met the following criteria: (1) the study was a single- or multi-arm interventional trial; (2) participants were veterans; (3) participants had a current diagnosis of PTSD or exhibited subthreshold PTSD symptoms, as established by authors of the individual studies and supported by a structured clinical interview, semistructured interview, or elevated scores on PTSD symptom self-report measures; (4) the study included an intervention in which exercise (physical activity that is planned, structured, repetitive, and purposive in the sense that improvement or maintenance of physical fitness or health is an objective) was the primary component; (5) PTSD symptom severity was by a clinician-rated or self-report measure; and (6) the study was published in a peer-reviewed journal.¹² Studies were excluded if means, standard deviations, and sample sizes were not available or the full text of the study was not available in English.

Data Collection

Data were extracted from included studies using custom forms and included the following information based on PRISMA guidelines: (1) study design characteristics; (2) intervention details; and (3) PTSD outcome information.¹¹ PTSD symptom severity was the primary outcome of interest. Outcome data were included if they were derived from a measure of PTSD symptoms—equivalency across measures was assumed for meta-analyses. Potential study bias for each outcome was evaluated using the ROBINS-I and Cochrane Collaboration’s RoB 2 tools for single-arm and multi-arm trials, respectively.^13,14 These tools evaluate domains related to the design, conduct, and analysis of studies that are associated with bias (ie, systematic error in findings, such as under- or overestimation of results).¹⁰ Examples include how well authors performed and concealed randomization procedures, addressed missing data, and measured study outcomes.^13,14 The risk of bias (eg, low, moderate, serious) associated with each domain is rated and, based on the domain ratings, each study is then given an overall rating regarding how much risk influences bias.^13,14 Broadly, lower risk of bias corresponds to higher confidence in the validity of results.

Finally, 4 authors (associated with 2 single- and 2 multi-arm studies) were contacted and asked to provide further information. Data for 1 additional multi-arm study were obtained from these communications and included in the final study selection.¹⁵ These authors were also asked for information about any unpublished works of which they were aware, although no additional works were identified.

Statistical Analyses

Analyses were performed with R Studio R 3.6.0 software.¹⁶ An SMD (also known as Hedges g) was calculated for each study outcome: for single-arm trials, this was the SMD between pre- and postintervention scores, whereas for multi-arm trials, this was the SMD between postintervention outcome scores across groups. CIs for each SMD were calculated using a standard normal distribution. Combined SMDs were estimated separately for single- and multi-arm studies, using random-effects meta-analyses. In order to include multiple relevant outcomes from a single trial (ie, for studies using multiple PTSD symptom measures), robust variance estimation was used.¹⁷ Precision was used to weight SMDs.

Correlations between pre- and postintervention scores were not available for 1 single-arm study.¹⁸ A correlation coefficient of 0.8 was imputed to calculate the standard error of the of the SMDs for the Clinician-Administered PTSD Scale (CAPS) and the PTSD Checklist (PCL), as this value is consistent with past findings regarding the test-retest reliability of these measures.^19-22 A sensitivity analysis, using several alternative correlational values, revealed that the choice of correlation coefficient did not impact the overall results of the meta-analysis.

I² was used to evaluate between-study heterogeneity. Values of I² > 25%, 50%, and 75% were selected to reflect low, moderate, and high heterogeneity, respectively, in accordance with guidelines described by Higgins and colleagues.²³ Potential publication bias was assessed via funnel plot and Egger test.²⁴ Finally, although collection of depressive symptom scores was proposed as a secondary outcome in the study protocol, such data were available only for 1 multi-arm study. As a result, this outcome was not evaluated.

Results

Six studies with 101 total participants were included in the single-arm analyses (Table 1).^18,25-29 Participants consisted of veterans with chronic pain, post-9/11 veterans, female veterans of childbearing age, veterans with a history of trauma therapy, and other veterans. Types of exercise included moderate aerobic exercise and yoga. PTSD symptom measures included the CAPS and the PCL (PCL-5 or PCL-M versions). Reported financial sources for included studies included federal grant funding, nonprofit material support, outside organization support, use of US Department of Veterans Affairs (VA) resources, and no reported financial support.

With respect to individual studies, Shivakumar and colleagues found that completion of an aerobic exercise program was associated with reduced scores on 2 different PTSD symptom scales (PCL and CAPS) in 16 women veterans.¹⁸ A trauma-informed yoga intervention study with 18 participants by Cushing and colleagues demonstrated veteran participation to be associated with large reductions in PTSD, anxiety, and depression scale scores.²⁵ In a study with 34 veterans, Chopin and colleagues found that a trauma-informed yoga intervention was associated with a statistically significant reduction in PTSD symptoms, as did a study by Zaccari and colleagues with 17 veterans.^26,29 Justice and Brems also found some evidence that trauma-informed yoga interventions helped PTSD symptoms in a small sample of 4 veterans, although these results were not quantitatively analyzed.²⁷ In contrast, a small pilot study (n = 12) by Staples and colleagues testing a biweekly, 6-week yoga program did not show a significant effect on PTSD symptoms.²⁸

Three studies with 217 total veteran participants were included in the multi-arm analyses (Table 2).^15,30,31 As all multi-arm trials incorporated randomization, they will be referred to as randomized controlled trials (RCTs). On contact, Davis and colleagues provided veteran-specific results for their trial; as such, our data differ from those within the published article.¹⁵ Participants from all included studies were veterans currently experiencing symptoms of PTSD. Types of exercise included yoga and combined methods (eg, aerobic and strength training).^15,30,31 PTSD symptom measures included the CAPS or the PCL-5.^15,30,31 Reported financial sources for included studies included federal grant funding, as well as nonprofit support, private donations, and VA and Department of Defense resources.

Davis and colleagues conducted a recently concluded RCT with > 130 veteran participants and found that a novel manualized yoga program was superior to an attention control in reducing PTSD symptom scale scores for veterans.¹⁵ Goldstein and colleagues found that a program consisting of both aerobic and resistance exercises reduced PTSD symptoms to a greater extent than a waitlist control condition, with 47 veterans randomized in this trial.³⁰ Likewise, Hall and colleagues conducted a pilot RCT in which an intervention that integrated exercise and cognitive behavioral techniques was compared to a waitlist control condition.³¹ For the 48 veterans included in the analyses, the authors reported greater PTSD symptom reduction associated with integrated exercise than that of the control condition; however, the study was not powered to detect statistically significant differences between groups.

Bias Assessment

Results for the risk of bias assessments can be viewed in Tables 3 and 4. For single-arm studies, overall risk of bias was serious for all included trials. Serious risk of bias was found in 2 domains: confounding, due to a lack of accounting for potential preexisting baseline trends (eg, regression to the mean) that could have impacted study results; and measurement, due to the use of a self-report symptom measure (PCL) or CAPS with unblinded assessors. Multiple studies also showed moderate risk in the missing data domain due to participant dropout without appropriate analytic methods to address potential bias.

For RCTs, overall risk of bias ranged from some concerns to high risk. High risk of bias was found in 1 domain, measurement of outcome, due to use of a self-report symptom measure (PCL) with unblinded groups.³¹ The other 2 studies all had some concern of bias in at least 1 of the following domains: randomization, missing data, and measurement of outcome.

Pooled Standardized Mean Differences

Meta-analytic results can be viewed in Figure 2. The pooled SMD for the 6 single-arm studies was -0.60 (df = 4.41, 95% CI, -1.08 to -0.12, P = .03), indicating a statistically significant reduction in PTSD symptoms over the course of an exercise intervention. Combining SMDs for the 3 included RCTs revealed a pooled SMD of -0.40 (df = 1.57, 95% CI, -0.86 to 0.06, P = .06), indicating that exercise did not result in a statistically significant reduction in PTSD symptoms compared with control conditions.

Publication Bias and Heterogeneity

Visual inspection funnel plots and Egger test did not suggest the presence of publication bias for RCTs (t = 1.21, df = 2, P = .35) or single-arm studies (t = -0.36, df = 5, P = .73).

Single-arm studies displayed a high degree of heterogeneity (I² = 81.5%). Including sample size or exercise duration as variables in meta-regressions did not reduce heterogeneity (I² = 85.2% and I² = 83.8%, respectively). Performing a subgroup analysis only on studies using yoga as an intervention also did not reduce heterogeneity (I² = 79.2%). Due to the small number of studies, no further exploration of heterogeneity was conducted on single-arm studies. RCTs did not display any heterogeneity (I² = 0%).

Discussion

Our report represents an early synthesis of the first prospective studies of physical exercise interventions for PTSD in veterans. Results from meta-analyses of 6 single-arm studies (101 participants) and 3 RCTs (217 participants) provide early evidence that exercise may reduce PTSD symptoms in veterans. Yoga was the most common form of exercise used in single-arm studies, whereas RCTs used a wider range of interventions. The pooled SMD of -0.60 for single-arm longitudinal studies suggest a medium decrease in PTSD symptoms for veterans who engage in exercise interventions. Analysis of the RCTs supported this finding, with a pooled SMD of -0.40 reflecting a small-to-medium effect of exercise on PTSD symptoms over control conditions, although this result did not achieve statistical significance. Of note, while the nonsignificant finding for RCTs may have been due to insufficient power caused by the limited number of included studies, possibly exercise was not more efficacious than were the control conditions.

Although RCTs represented a variety of exercise types, PTSD symptom measures, and veteran subgroups, statistical results were not indicative of heterogeneity. However, only the largest and most comprehensive study of exercise for PTSD in veterans to date by Davis and colleagues had a statistically significant SMD.¹⁵ Of note, one of the other 2 RCTs displayed an SMD of a similar magnitude, but this study had a much smaller sample size and was underpowered to detect significance.³⁰ Additionally, risk of bias assessments for single-arm studies and RCTs revealed study characteristics that suggest possible inflation of absolute effect sizes for individual studies. Therefore, the pooled SMDs we report are interpretable but may exceed the true effect of exercise for PTSD symptom reduction in veterans.

Limitations

The present study has several important limitations. First, few studies were found that met the broad eligibility criteria and those that did often had a small sample size. Besides highlighting a gap in the extant research, the limited studies available for meta-analysis means that caution must be taken when interpreting results. Fortunately, this issue will likely resolve once additional studies investigating the impact of exercise on PTSD symptoms in veterans are available for synthesis.

Relatedly, the included study interventions varied considerably, both in the types of exercise used and the characteristics of the exercises (eg, frequency, duration, and intensity), which is relevant as different exercise modalities are associated with differential physical effects.³³ Including such a mixture of exercises may have given an incomplete picture of their potential therapeutic effects. Also, none of the RCTs compared exercise against first-line treatments for PTSD, such as prolonged exposure or cognitive processing therapy, which would have provided further insight into the role exercise could play in clinical settings.⁷

Another limitation is the elevated risk of bias found in most studies, particularly present in the longitudinal single-arm studies, all of which were rated at serious risk. For instance, no single-arm study controlled for preexisting baseline trends: without such (and lacking a comparison control group like in RCTs), it is possible that the observed effects were due to extraneous factors, rather than the exercise intervention. Although not as severe, the multi-arm RCTs also displayed at least moderate risk of bias. Therefore, SMDs may have been overestimated for each group of studies.

Finally, the results of the single-arm meta-analysis displayed high statistical heterogeneity, reducing the generalizability of the results. One possible cause of this heterogeneity may have been the yoga interventions, as a separate analysis removing the only nonyoga study did not reduce heterogeneity. This result was surprising, as the included yoga interventions seemed similar across studies. While the presence of high heterogeneity does require some caution when applying these results to outside interventions, the present study made use of random-effects meta-analysis, a technique that incorporates study heterogeneity into the statistical model, thereby strengthening the findings compared with that of a traditional fixed-effects approach.¹⁰

Future Steps

Several future steps are warranted to improve knowledge of exercise as a treatment for PTSD in veterans and in the general population. With current meta-analyses limited to small numbers of studies, additional studies of the efficacy of exercise for treating PTSD could help in several ways. A larger pool of studies would enable future meta-analyses to explore related questions, such as those regarding the impact of exercise on quality of life or depressive symptom reduction among veterans with PTSD. A greater number of studies also would enable meta-analysts to explore potentially critical moderators. For example, the duration, frequency, or type of exercise may moderate the effect of exercise on PTSD symptom reduction. Moderators related to patient or study design characteristics also should be explored in future studies.

Future work also should evaluate the impact that specific features of exercise regimens have on PTSD. Knowing whether the type or structure of exercise affects its clinical use would be invaluable in developing and implementing efficient exercise-based interventions. For example, if facilitated exercise was found to be significantly more effective at reducing PTSD symptoms than exercise completed independently, the development of exercise intervention programs in the VA and other facilities that commonly treat PTSD may be warranted. Additionally, it may be useful to identify specific mechanisms through which exercise reduces PTSD symptoms. For example, in addition to its beneficial biological effects, exercise also promotes psychological health through behavioral activation and alterations within reinforcement/reward systems, suggesting that exercise regularity may be more important than intensity.^34,35 Understanding which mechanisms contribute most to change will aid in the development of more efficient interventions.

Given that veterans are demonstrating considerable interest in complementary and alternative PTSD treatments, it is critical that researchers focus on high-quality randomized tests of these interventions. Therefore, in addition to greater quality of exercise intervention studies, future efforts should be focused on RCTs that are designed in such a way as to limit potential introduction of bias. For example, assessment data should be completed by blinded assessors using standardized measures, and analyses should account for missing data and unequal participant attrition between groups. Ideally, pre-intervention trends across multiple baseline datapoints also would be collected in single-arm studies to avoid confounding related to regression to the mean. It is also recommended that future meta-analyses use risk of bias assessments and consider how the results of such assessments may impact the interpretation of results.

Conclusions

Findings from both single-arm studies and RCTs suggest possible benefit of exercise on PTSD symptom reduction, although confirmation of findings is needed. No study found increased symptoms following exercise intervention. Thus, it is reasonable to consider physical exercise, such as yoga, as an adjunct, whole-health consistent treatment. HCPs working with veterans with past traumatic experiences should consider incorporating exercise into patient care. Enhanced educational efforts emphasizing the psychotherapeutic impact of exercise may also have value for the veteran population. Furthermore, the current risk of bias assessments highlights the need for additional high-quality RCTs evaluating the specific impact of exercise on PTSD symptom reduction in veterans. In particular, this field of inquiry would benefit from larger samples and design characteristics to reduce bias (eg, blinding when possible, use of CAPS vs only self-report symptom measures, reducing problematic attrition, corrections for missing data, etc).

Acknowledgments

This research is the result of work supported with resources and the use of facilities at the VA Eastern Kansas Healthcare System (Dwight D. Eisenhower VA Medical Center). It was also supported by the Department of Veterans Affairs Office of Academic Affiliations Advanced Fellowship Program in Mental Illness Research and Treatment, as well as the Rocky Mountain Mental Illness Research, Education, and Clinical Center. Since Dr. Reis and Dr. Gaddy are employees of the US Government and contributed to this manuscript as part of their official duties, the work is not subject to US copyright. This study was preregistered on PROSPERO (https://www.crd.york.ac.uk/prospero/; ID: CRD42020153419).

Physical exercise offers preventative and therapeutic benefits for a range of chronic health conditions, including cardiovascular disease, type 2 diabetes mellitus, Alzheimer disease, and depression.^1,2 Exercise has been well studied for its antidepressant effects, its ability to reduce risk of aging-related dementia, and favorable effects on a range of cognitive functions.² Lesser evidence exists regarding the impact of exercise on other mental health concerns. Therefore, an accurate understanding of whether physical exercise may ameliorate other conditions is important.

A small meta-analysis by Rosenbaum and colleagues found that exercise interventions were superior to control conditions for symptom reduction in study participants with posttraumatic stress disorder (PTSD).³ This meta-analysis included 4 randomized clinical trials representing 200 cases. The trial included a variety of physical activities (eg, yoga, aerobic, and strength-building exercises) and control conditions, and participants recruited from online, community, inpatient, and outpatient settings. The standardized mean difference (SMD) produced by the analysis indicated a small-to-medium effect (Hedges g, -0.35), with the authors reporting no evidence of publication bias, although an assessment of potential bias associated with individual trial design characteristics was not conducted. Of note, a meta-analysis by Watts and colleagues found that effect sizes for PTSD treatments tend to be smaller in veteran populations.⁴ Therefore, how much the mean effect size estimate in the study is applicable to veterans with PTSD is unknown.³

Veterans represent a unique subpopulation in which PTSD is common, although no meta-analysis yet published has synthesized the effects of exercise interventions from trials of veterans with PTSD.⁵ A recent systematic review by Whitworth and Ciccolo concluded that exercise may be associated with reduced risk of PTSD, a briefer course of PTSD symptoms, and/or reduced sleep- and depression-related difficulties.⁶ However, that review primarily included observational, cross-sectional, and qualitative works. No trials included in our meta-analysis were included in that review.⁶

Evidence-based psychotherapies like cognitive processing therapy and prolonged exposure have been shown to be effective for treating PTSD in veterans; however, these modalities are accompanied by high rates of dropout (eg, 40-60%), thereby limiting their clinical utility.⁷ The use of complementary and alternative approaches for treatment in the United States has increased in recent years, and exercise represents an important complementary treatment option.⁸ In a study by Baldwin and colleagues, nearly 50% of veterans reported using complementary or alternative approaches, and veterans with PTSD were among those likely to use such approaches.⁹ However, current studies of the effects of exercise interventions on PTSD symptom reduction are mostly small and varied, making determinations difficult regarding the potential utility of exercise for treating this condition in veterans.

Literature Search

No previous research has synthesized the literature on the effects of exercise on PTSD in the veteran population. The current meta-analysis aims to provide a synthesis of systematically selected studies on this topic to determine whether exercise-based interventions are effective at reducing veterans’ symptoms of PTSD. Our hypothesis was that, when used as a primary or adjuvant intervention for PTSD, physical exercise would be associated with a reduction of PTSD symptom scale scores. We planned a priori to produce separate estimates for single-arm and multi-arm trials. We also wanted to conduct a careful risk of bias assessment—or evaluation of study features that may have systematically influenced results—for included trials, not only to provide context for interpretation of results, but also to inform suggestions for research to advance this field of inquiry.¹⁰

Methods

This study was preregistered on PROSPERO and followed PRISMA guidelines for meta-analyses and systematic reviews.¹¹ Supplementary materials, such as the PRISMA checklist, study data, and funnel plots, are available online (doi.org/10.6084/m9.figshare.c.5618437.v1). Conference abstracts were omitted due to a lack of necessary information. We decided early in the planning process to include both randomized and single-arm trials, expecting the number of completed studies in the area of exercise for PTSD symptom reduction in veterans, and particularly randomized trials of such, would be relatively small.

Studies were included if they met the following criteria: (1) the study was a single- or multi-arm interventional trial; (2) participants were veterans; (3) participants had a current diagnosis of PTSD or exhibited subthreshold PTSD symptoms, as established by authors of the individual studies and supported by a structured clinical interview, semistructured interview, or elevated scores on PTSD symptom self-report measures; (4) the study included an intervention in which exercise (physical activity that is planned, structured, repetitive, and purposive in the sense that improvement or maintenance of physical fitness or health is an objective) was the primary component; (5) PTSD symptom severity was by a clinician-rated or self-report measure; and (6) the study was published in a peer-reviewed journal.¹² Studies were excluded if means, standard deviations, and sample sizes were not available or the full text of the study was not available in English.

Data Collection

Data were extracted from included studies using custom forms and included the following information based on PRISMA guidelines: (1) study design characteristics; (2) intervention details; and (3) PTSD outcome information.¹¹ PTSD symptom severity was the primary outcome of interest. Outcome data were included if they were derived from a measure of PTSD symptoms—equivalency across measures was assumed for meta-analyses. Potential study bias for each outcome was evaluated using the ROBINS-I and Cochrane Collaboration’s RoB 2 tools for single-arm and multi-arm trials, respectively.^13,14 These tools evaluate domains related to the design, conduct, and analysis of studies that are associated with bias (ie, systematic error in findings, such as under- or overestimation of results).¹⁰ Examples include how well authors performed and concealed randomization procedures, addressed missing data, and measured study outcomes.^13,14 The risk of bias (eg, low, moderate, serious) associated with each domain is rated and, based on the domain ratings, each study is then given an overall rating regarding how much risk influences bias.^13,14 Broadly, lower risk of bias corresponds to higher confidence in the validity of results.

Finally, 4 authors (associated with 2 single- and 2 multi-arm studies) were contacted and asked to provide further information. Data for 1 additional multi-arm study were obtained from these communications and included in the final study selection.¹⁵ These authors were also asked for information about any unpublished works of which they were aware, although no additional works were identified.

Statistical Analyses

Analyses were performed with R Studio R 3.6.0 software.¹⁶ An SMD (also known as Hedges g) was calculated for each study outcome: for single-arm trials, this was the SMD between pre- and postintervention scores, whereas for multi-arm trials, this was the SMD between postintervention outcome scores across groups. CIs for each SMD were calculated using a standard normal distribution. Combined SMDs were estimated separately for single- and multi-arm studies, using random-effects meta-analyses. In order to include multiple relevant outcomes from a single trial (ie, for studies using multiple PTSD symptom measures), robust variance estimation was used.¹⁷ Precision was used to weight SMDs.

Correlations between pre- and postintervention scores were not available for 1 single-arm study.¹⁸ A correlation coefficient of 0.8 was imputed to calculate the standard error of the of the SMDs for the Clinician-Administered PTSD Scale (CAPS) and the PTSD Checklist (PCL), as this value is consistent with past findings regarding the test-retest reliability of these measures.^19-22 A sensitivity analysis, using several alternative correlational values, revealed that the choice of correlation coefficient did not impact the overall results of the meta-analysis.

I² was used to evaluate between-study heterogeneity. Values of I² > 25%, 50%, and 75% were selected to reflect low, moderate, and high heterogeneity, respectively, in accordance with guidelines described by Higgins and colleagues.²³ Potential publication bias was assessed via funnel plot and Egger test.²⁴ Finally, although collection of depressive symptom scores was proposed as a secondary outcome in the study protocol, such data were available only for 1 multi-arm study. As a result, this outcome was not evaluated.

Results

Six studies with 101 total participants were included in the single-arm analyses (Table 1).^18,25-29 Participants consisted of veterans with chronic pain, post-9/11 veterans, female veterans of childbearing age, veterans with a history of trauma therapy, and other veterans. Types of exercise included moderate aerobic exercise and yoga. PTSD symptom measures included the CAPS and the PCL (PCL-5 or PCL-M versions). Reported financial sources for included studies included federal grant funding, nonprofit material support, outside organization support, use of US Department of Veterans Affairs (VA) resources, and no reported financial support.

With respect to individual studies, Shivakumar and colleagues found that completion of an aerobic exercise program was associated with reduced scores on 2 different PTSD symptom scales (PCL and CAPS) in 16 women veterans.¹⁸ A trauma-informed yoga intervention study with 18 participants by Cushing and colleagues demonstrated veteran participation to be associated with large reductions in PTSD, anxiety, and depression scale scores.²⁵ In a study with 34 veterans, Chopin and colleagues found that a trauma-informed yoga intervention was associated with a statistically significant reduction in PTSD symptoms, as did a study by Zaccari and colleagues with 17 veterans.^26,29 Justice and Brems also found some evidence that trauma-informed yoga interventions helped PTSD symptoms in a small sample of 4 veterans, although these results were not quantitatively analyzed.²⁷ In contrast, a small pilot study (n = 12) by Staples and colleagues testing a biweekly, 6-week yoga program did not show a significant effect on PTSD symptoms.²⁸

Three studies with 217 total veteran participants were included in the multi-arm analyses (Table 2).^15,30,31 As all multi-arm trials incorporated randomization, they will be referred to as randomized controlled trials (RCTs). On contact, Davis and colleagues provided veteran-specific results for their trial; as such, our data differ from those within the published article.¹⁵ Participants from all included studies were veterans currently experiencing symptoms of PTSD. Types of exercise included yoga and combined methods (eg, aerobic and strength training).^15,30,31 PTSD symptom measures included the CAPS or the PCL-5.^15,30,31 Reported financial sources for included studies included federal grant funding, as well as nonprofit support, private donations, and VA and Department of Defense resources.

Davis and colleagues conducted a recently concluded RCT with > 130 veteran participants and found that a novel manualized yoga program was superior to an attention control in reducing PTSD symptom scale scores for veterans.¹⁵ Goldstein and colleagues found that a program consisting of both aerobic and resistance exercises reduced PTSD symptoms to a greater extent than a waitlist control condition, with 47 veterans randomized in this trial.³⁰ Likewise, Hall and colleagues conducted a pilot RCT in which an intervention that integrated exercise and cognitive behavioral techniques was compared to a waitlist control condition.³¹ For the 48 veterans included in the analyses, the authors reported greater PTSD symptom reduction associated with integrated exercise than that of the control condition; however, the study was not powered to detect statistically significant differences between groups.

Bias Assessment

Results for the risk of bias assessments can be viewed in Tables 3 and 4. For single-arm studies, overall risk of bias was serious for all included trials. Serious risk of bias was found in 2 domains: confounding, due to a lack of accounting for potential preexisting baseline trends (eg, regression to the mean) that could have impacted study results; and measurement, due to the use of a self-report symptom measure (PCL) or CAPS with unblinded assessors. Multiple studies also showed moderate risk in the missing data domain due to participant dropout without appropriate analytic methods to address potential bias.

For RCTs, overall risk of bias ranged from some concerns to high risk. High risk of bias was found in 1 domain, measurement of outcome, due to use of a self-report symptom measure (PCL) with unblinded groups.³¹ The other 2 studies all had some concern of bias in at least 1 of the following domains: randomization, missing data, and measurement of outcome.

Pooled Standardized Mean Differences

Meta-analytic results can be viewed in Figure 2. The pooled SMD for the 6 single-arm studies was -0.60 (df = 4.41, 95% CI, -1.08 to -0.12, P = .03), indicating a statistically significant reduction in PTSD symptoms over the course of an exercise intervention. Combining SMDs for the 3 included RCTs revealed a pooled SMD of -0.40 (df = 1.57, 95% CI, -0.86 to 0.06, P = .06), indicating that exercise did not result in a statistically significant reduction in PTSD symptoms compared with control conditions.

Publication Bias and Heterogeneity

Visual inspection funnel plots and Egger test did not suggest the presence of publication bias for RCTs (t = 1.21, df = 2, P = .35) or single-arm studies (t = -0.36, df = 5, P = .73).

Single-arm studies displayed a high degree of heterogeneity (I² = 81.5%). Including sample size or exercise duration as variables in meta-regressions did not reduce heterogeneity (I² = 85.2% and I² = 83.8%, respectively). Performing a subgroup analysis only on studies using yoga as an intervention also did not reduce heterogeneity (I² = 79.2%). Due to the small number of studies, no further exploration of heterogeneity was conducted on single-arm studies. RCTs did not display any heterogeneity (I² = 0%).

Discussion

Our report represents an early synthesis of the first prospective studies of physical exercise interventions for PTSD in veterans. Results from meta-analyses of 6 single-arm studies (101 participants) and 3 RCTs (217 participants) provide early evidence that exercise may reduce PTSD symptoms in veterans. Yoga was the most common form of exercise used in single-arm studies, whereas RCTs used a wider range of interventions. The pooled SMD of -0.60 for single-arm longitudinal studies suggest a medium decrease in PTSD symptoms for veterans who engage in exercise interventions. Analysis of the RCTs supported this finding, with a pooled SMD of -0.40 reflecting a small-to-medium effect of exercise on PTSD symptoms over control conditions, although this result did not achieve statistical significance. Of note, while the nonsignificant finding for RCTs may have been due to insufficient power caused by the limited number of included studies, possibly exercise was not more efficacious than were the control conditions.

Although RCTs represented a variety of exercise types, PTSD symptom measures, and veteran subgroups, statistical results were not indicative of heterogeneity. However, only the largest and most comprehensive study of exercise for PTSD in veterans to date by Davis and colleagues had a statistically significant SMD.¹⁵ Of note, one of the other 2 RCTs displayed an SMD of a similar magnitude, but this study had a much smaller sample size and was underpowered to detect significance.³⁰ Additionally, risk of bias assessments for single-arm studies and RCTs revealed study characteristics that suggest possible inflation of absolute effect sizes for individual studies. Therefore, the pooled SMDs we report are interpretable but may exceed the true effect of exercise for PTSD symptom reduction in veterans.

Limitations

The present study has several important limitations. First, few studies were found that met the broad eligibility criteria and those that did often had a small sample size. Besides highlighting a gap in the extant research, the limited studies available for meta-analysis means that caution must be taken when interpreting results. Fortunately, this issue will likely resolve once additional studies investigating the impact of exercise on PTSD symptoms in veterans are available for synthesis.

Relatedly, the included study interventions varied considerably, both in the types of exercise used and the characteristics of the exercises (eg, frequency, duration, and intensity), which is relevant as different exercise modalities are associated with differential physical effects.³³ Including such a mixture of exercises may have given an incomplete picture of their potential therapeutic effects. Also, none of the RCTs compared exercise against first-line treatments for PTSD, such as prolonged exposure or cognitive processing therapy, which would have provided further insight into the role exercise could play in clinical settings.⁷

Another limitation is the elevated risk of bias found in most studies, particularly present in the longitudinal single-arm studies, all of which were rated at serious risk. For instance, no single-arm study controlled for preexisting baseline trends: without such (and lacking a comparison control group like in RCTs), it is possible that the observed effects were due to extraneous factors, rather than the exercise intervention. Although not as severe, the multi-arm RCTs also displayed at least moderate risk of bias. Therefore, SMDs may have been overestimated for each group of studies.

Finally, the results of the single-arm meta-analysis displayed high statistical heterogeneity, reducing the generalizability of the results. One possible cause of this heterogeneity may have been the yoga interventions, as a separate analysis removing the only nonyoga study did not reduce heterogeneity. This result was surprising, as the included yoga interventions seemed similar across studies. While the presence of high heterogeneity does require some caution when applying these results to outside interventions, the present study made use of random-effects meta-analysis, a technique that incorporates study heterogeneity into the statistical model, thereby strengthening the findings compared with that of a traditional fixed-effects approach.¹⁰

Future Steps

Several future steps are warranted to improve knowledge of exercise as a treatment for PTSD in veterans and in the general population. With current meta-analyses limited to small numbers of studies, additional studies of the efficacy of exercise for treating PTSD could help in several ways. A larger pool of studies would enable future meta-analyses to explore related questions, such as those regarding the impact of exercise on quality of life or depressive symptom reduction among veterans with PTSD. A greater number of studies also would enable meta-analysts to explore potentially critical moderators. For example, the duration, frequency, or type of exercise may moderate the effect of exercise on PTSD symptom reduction. Moderators related to patient or study design characteristics also should be explored in future studies.

Future work also should evaluate the impact that specific features of exercise regimens have on PTSD. Knowing whether the type or structure of exercise affects its clinical use would be invaluable in developing and implementing efficient exercise-based interventions. For example, if facilitated exercise was found to be significantly more effective at reducing PTSD symptoms than exercise completed independently, the development of exercise intervention programs in the VA and other facilities that commonly treat PTSD may be warranted. Additionally, it may be useful to identify specific mechanisms through which exercise reduces PTSD symptoms. For example, in addition to its beneficial biological effects, exercise also promotes psychological health through behavioral activation and alterations within reinforcement/reward systems, suggesting that exercise regularity may be more important than intensity.^34,35 Understanding which mechanisms contribute most to change will aid in the development of more efficient interventions.

Given that veterans are demonstrating considerable interest in complementary and alternative PTSD treatments, it is critical that researchers focus on high-quality randomized tests of these interventions. Therefore, in addition to greater quality of exercise intervention studies, future efforts should be focused on RCTs that are designed in such a way as to limit potential introduction of bias. For example, assessment data should be completed by blinded assessors using standardized measures, and analyses should account for missing data and unequal participant attrition between groups. Ideally, pre-intervention trends across multiple baseline datapoints also would be collected in single-arm studies to avoid confounding related to regression to the mean. It is also recommended that future meta-analyses use risk of bias assessments and consider how the results of such assessments may impact the interpretation of results.

Conclusions

Findings from both single-arm studies and RCTs suggest possible benefit of exercise on PTSD symptom reduction, although confirmation of findings is needed. No study found increased symptoms following exercise intervention. Thus, it is reasonable to consider physical exercise, such as yoga, as an adjunct, whole-health consistent treatment. HCPs working with veterans with past traumatic experiences should consider incorporating exercise into patient care. Enhanced educational efforts emphasizing the psychotherapeutic impact of exercise may also have value for the veteran population. Furthermore, the current risk of bias assessments highlights the need for additional high-quality RCTs evaluating the specific impact of exercise on PTSD symptom reduction in veterans. In particular, this field of inquiry would benefit from larger samples and design characteristics to reduce bias (eg, blinding when possible, use of CAPS vs only self-report symptom measures, reducing problematic attrition, corrections for missing data, etc).

Acknowledgments

This research is the result of work supported with resources and the use of facilities at the VA Eastern Kansas Healthcare System (Dwight D. Eisenhower VA Medical Center). It was also supported by the Department of Veterans Affairs Office of Academic Affiliations Advanced Fellowship Program in Mental Illness Research and Treatment, as well as the Rocky Mountain Mental Illness Research, Education, and Clinical Center. Since Dr. Reis and Dr. Gaddy are employees of the US Government and contributed to this manuscript as part of their official duties, the work is not subject to US copyright. This study was preregistered on PROSPERO (https://www.crd.york.ac.uk/prospero/; ID: CRD42020153419).

A new consensus statement from the American Head and Neck Society Endocrine Surgery Section and International Thyroid Oncology Group focuses on a definition of advanced thyroid cancer and outlines strategies for mutation testing and targeted treatment.

Mutation testing should not be pursued if cancer burden and disease threat is low, since most thyroid cancers have a very good prognosis and are highly treatable. But 15% of differentiated thyroid cancer cases are locally advanced, and radioiodine refractory differentiated thyroid cancer has a 10-year survival below 50%.

More generally, advanced thyroid cancer has not been well defined clinically. Physicians with experience diagnosing advanced disease may recognize it, but there is no widely accepted definition. “This may be the first time that an expert group of physicians has attempted to define what advanced thyroid cancer is,” said David Shonka, MD, who is a coauthor of the consensus statement, which was published online in Head & Neck. He is an associate professor of otolaryngology/head and neck surgery at the University of Virginia, Charlottesville.

“All patients with advanced thyroid disease and most patients with incurable radioiodine refractory differentiated thyroid cancer should undergo somatic mutational testing,” the authors wrote. “Next-generation sequencing can reveal targetable mutations and potentially give patients affected by advanced thyroid carcinoma systemic treatment options that can prolong survival. These new innovative approaches are changing the landscape of clinical care for patients with advanced thyroid cancer.”

For differentiated thyroid cancer and medullary thyroid carcinoma, the authors created a definition that combines structural factors on imaging, along with surgical findings, and biochemical, histologic, and molecular factors. Anaplastic thyroid cancer should always be considered advanced, even after a complete resection and incidental pathological identification.

The statement also summarizes recent advances in thyroid cancer that have revealed molecular markers which contribute to oncogenesis. Initially, those approaches were applied to indeterminate fine needle biopsies to improve diagnosis. More recent studies used them to match patients to targeted therapies. There are Food and Drug Administration–approved therapies targeting the BRAF and RET mutations, but advanced thyroid cancer is also included in some “basket” trials that test targeted agents against driver mutations across multiple tumor types.

Radioiodine refractory differentiated thyroid cancer had few treatments as recently as 10 years ago. But recent research has shown that multikinase inhibitors improve outcomes, and a range of mutations have been found in this type of thyroid cancer, including BRAF V600E, RET, PIK3CA, and PTEN, and fusions involving RET, NTRK, and ALK. Other mutations have been linked to more aggressive disease. Efforts to personalize treatment also include microsatellite stability status, tumor mutational burden, and programmed death–ligand 1 status as indicators for immunotherapy. “With discovery of many other molecular targets, and emerging literature showcasing promise of matched targeted therapies, we recommend that all patients with advanced thyroid cancer have comprehensive genomic profiling on tumor tissue through (next generation sequencing),” the authors wrote.

These newer and novel therapies have presented physicians with options outside of surgery, chemotherapy, or radiotherapy, which have low efficacy against advanced thyroid cancer. “It is an area in which there has been substantial change. Even 5-7 years ago, patients with advanced thyroid cancer that was not responsive to radioactive iodine or surgery really didn’t have a lot of options. This is a really an exciting and growing field,” Dr. Shonka said.

He specifically cited anaplastic thyroid cancer, which like radioiodine refractory differentiated thyroid cancer has had few treatment options until recently. “Now, if you see a patient with anaplastic thyroid cancer, your knee-jerk reaction should be ‘let’s do molecular testing on this, this is definitely advanced disease.’ If they have a BRAF mutation, that’s targetable, and we can treat this patient with combination therapy that actually improves their survival. So, there’s some exciting stuff happening and probably more coming down the road as we develop new drugs that can target these mutations that we’re identifying.”

Dr. Shonka has no relevant financial disclosures.

A new consensus statement from the American Head and Neck Society Endocrine Surgery Section and International Thyroid Oncology Group focuses on a definition of advanced thyroid cancer and outlines strategies for mutation testing and targeted treatment.

Mutation testing should not be pursued if cancer burden and disease threat is low, since most thyroid cancers have a very good prognosis and are highly treatable. But 15% of differentiated thyroid cancer cases are locally advanced, and radioiodine refractory differentiated thyroid cancer has a 10-year survival below 50%.

More generally, advanced thyroid cancer has not been well defined clinically. Physicians with experience diagnosing advanced disease may recognize it, but there is no widely accepted definition. “This may be the first time that an expert group of physicians has attempted to define what advanced thyroid cancer is,” said David Shonka, MD, who is a coauthor of the consensus statement, which was published online in Head & Neck. He is an associate professor of otolaryngology/head and neck surgery at the University of Virginia, Charlottesville.

“All patients with advanced thyroid disease and most patients with incurable radioiodine refractory differentiated thyroid cancer should undergo somatic mutational testing,” the authors wrote. “Next-generation sequencing can reveal targetable mutations and potentially give patients affected by advanced thyroid carcinoma systemic treatment options that can prolong survival. These new innovative approaches are changing the landscape of clinical care for patients with advanced thyroid cancer.”

For differentiated thyroid cancer and medullary thyroid carcinoma, the authors created a definition that combines structural factors on imaging, along with surgical findings, and biochemical, histologic, and molecular factors. Anaplastic thyroid cancer should always be considered advanced, even after a complete resection and incidental pathological identification.

The statement also summarizes recent advances in thyroid cancer that have revealed molecular markers which contribute to oncogenesis. Initially, those approaches were applied to indeterminate fine needle biopsies to improve diagnosis. More recent studies used them to match patients to targeted therapies. There are Food and Drug Administration–approved therapies targeting the BRAF and RET mutations, but advanced thyroid cancer is also included in some “basket” trials that test targeted agents against driver mutations across multiple tumor types.

Radioiodine refractory differentiated thyroid cancer had few treatments as recently as 10 years ago. But recent research has shown that multikinase inhibitors improve outcomes, and a range of mutations have been found in this type of thyroid cancer, including BRAF V600E, RET, PIK3CA, and PTEN, and fusions involving RET, NTRK, and ALK. Other mutations have been linked to more aggressive disease. Efforts to personalize treatment also include microsatellite stability status, tumor mutational burden, and programmed death–ligand 1 status as indicators for immunotherapy. “With discovery of many other molecular targets, and emerging literature showcasing promise of matched targeted therapies, we recommend that all patients with advanced thyroid cancer have comprehensive genomic profiling on tumor tissue through (next generation sequencing),” the authors wrote.

These newer and novel therapies have presented physicians with options outside of surgery, chemotherapy, or radiotherapy, which have low efficacy against advanced thyroid cancer. “It is an area in which there has been substantial change. Even 5-7 years ago, patients with advanced thyroid cancer that was not responsive to radioactive iodine or surgery really didn’t have a lot of options. This is a really an exciting and growing field,” Dr. Shonka said.

He specifically cited anaplastic thyroid cancer, which like radioiodine refractory differentiated thyroid cancer has had few treatment options until recently. “Now, if you see a patient with anaplastic thyroid cancer, your knee-jerk reaction should be ‘let’s do molecular testing on this, this is definitely advanced disease.’ If they have a BRAF mutation, that’s targetable, and we can treat this patient with combination therapy that actually improves their survival. So, there’s some exciting stuff happening and probably more coming down the road as we develop new drugs that can target these mutations that we’re identifying.”

Dr. Shonka has no relevant financial disclosures.

A new consensus statement from the American Head and Neck Society Endocrine Surgery Section and International Thyroid Oncology Group focuses on a definition of advanced thyroid cancer and outlines strategies for mutation testing and targeted treatment.

Mutation testing should not be pursued if cancer burden and disease threat is low, since most thyroid cancers have a very good prognosis and are highly treatable. But 15% of differentiated thyroid cancer cases are locally advanced, and radioiodine refractory differentiated thyroid cancer has a 10-year survival below 50%.

More generally, advanced thyroid cancer has not been well defined clinically. Physicians with experience diagnosing advanced disease may recognize it, but there is no widely accepted definition. “This may be the first time that an expert group of physicians has attempted to define what advanced thyroid cancer is,” said David Shonka, MD, who is a coauthor of the consensus statement, which was published online in Head & Neck. He is an associate professor of otolaryngology/head and neck surgery at the University of Virginia, Charlottesville.

“All patients with advanced thyroid disease and most patients with incurable radioiodine refractory differentiated thyroid cancer should undergo somatic mutational testing,” the authors wrote. “Next-generation sequencing can reveal targetable mutations and potentially give patients affected by advanced thyroid carcinoma systemic treatment options that can prolong survival. These new innovative approaches are changing the landscape of clinical care for patients with advanced thyroid cancer.”

For differentiated thyroid cancer and medullary thyroid carcinoma, the authors created a definition that combines structural factors on imaging, along with surgical findings, and biochemical, histologic, and molecular factors. Anaplastic thyroid cancer should always be considered advanced, even after a complete resection and incidental pathological identification.

The statement also summarizes recent advances in thyroid cancer that have revealed molecular markers which contribute to oncogenesis. Initially, those approaches were applied to indeterminate fine needle biopsies to improve diagnosis. More recent studies used them to match patients to targeted therapies. There are Food and Drug Administration–approved therapies targeting the BRAF and RET mutations, but advanced thyroid cancer is also included in some “basket” trials that test targeted agents against driver mutations across multiple tumor types.

Radioiodine refractory differentiated thyroid cancer had few treatments as recently as 10 years ago. But recent research has shown that multikinase inhibitors improve outcomes, and a range of mutations have been found in this type of thyroid cancer, including BRAF V600E, RET, PIK3CA, and PTEN, and fusions involving RET, NTRK, and ALK. Other mutations have been linked to more aggressive disease. Efforts to personalize treatment also include microsatellite stability status, tumor mutational burden, and programmed death–ligand 1 status as indicators for immunotherapy. “With discovery of many other molecular targets, and emerging literature showcasing promise of matched targeted therapies, we recommend that all patients with advanced thyroid cancer have comprehensive genomic profiling on tumor tissue through (next generation sequencing),” the authors wrote.

These newer and novel therapies have presented physicians with options outside of surgery, chemotherapy, or radiotherapy, which have low efficacy against advanced thyroid cancer. “It is an area in which there has been substantial change. Even 5-7 years ago, patients with advanced thyroid cancer that was not responsive to radioactive iodine or surgery really didn’t have a lot of options. This is a really an exciting and growing field,” Dr. Shonka said.

He specifically cited anaplastic thyroid cancer, which like radioiodine refractory differentiated thyroid cancer has had few treatment options until recently. “Now, if you see a patient with anaplastic thyroid cancer, your knee-jerk reaction should be ‘let’s do molecular testing on this, this is definitely advanced disease.’ If they have a BRAF mutation, that’s targetable, and we can treat this patient with combination therapy that actually improves their survival. So, there’s some exciting stuff happening and probably more coming down the road as we develop new drugs that can target these mutations that we’re identifying.”

Dr. Shonka has no relevant financial disclosures.

A ketogenic diet fed to mice with epithelial ovarian cancer led to significantly increased tumor growth and gut microbiome alterations, according to study recently presented at the annual meeting of the Society of Gynecologic Oncology.

“The keto diet is very popular, especially among patients who believe it may treat cancer by starving tumors of the fuel they need to grow, altering the immune system, and other anticancer effects,” said study leader Mariam AlHilli, MD, of the Cleveland Clinic.

The findings are surprising because in other studies the high-fat, zero-carb ketogenic diet has demonstrated tumor-suppressing effects. It has been under study as a possible adjuvant therapy for other cancers, such as glioblastoma, colon cancer, prostate cancer, and pancreatic cancer.

“While we don’t know yet whether these findings extend to patients, the results in animals indicate that instead of being protective, the keto diet appears to promote ovarian cancer growth and progression,” Dr. AlHilli said. In the present study, tumor bearing mice were fed a keto diet consisting of 10% protein, 0% carbohydrates, and 90% fat, while the high-fat diet was 10% protein, 15% carbohydrates, and 75% fat. The control diet consisted of 10% protein, 77% carbohydrates, and 13% fat. Epithelial ovarian cancer tumor growth was monitored weekly.

Over the 6- to 10-week course of study, a 9.1-fold increase from baseline in tumor growth was observed in the keto diet-fed mice (n = 20). Among mice fed a high-fat diet (n = 20) that included some carbohydrates, tumor growth increased 2.0-fold from baseline, and among control group mice (n = 20) fed a low-fat, high carbohydrate diet, tumor growth increased 3.1-fold.

The investigators observed several hallmarks of tumor progression: tumor associated macrophages were enriched significantly, activated lymphoid cells (natural killer cells) were significantly reduced (P < .001), and M2:M1 polarization trended higher. Also, in keto diet–fed mice, gene set enrichment analysis revealed that epithelial ovarian cancer tumors had increased angiogenesis and inflammatory responses, enhanced epithelial-to-mesenchymal transition phenotype, and altered lipid metabolism. Compared with high-fat diet–fed mice, the keto-fed mice had increases in lipid catalytic activity and catabolism, as well as decreases in lipid synthesis.

“The tumor increase could be mediated by the gut microbiome or by gene alterations or by metabolite levels that influence tumor growth. It’s possible that each cancer type is different. The composition of the diet may be a factor, as well as how tumors metabolize fat and ketones,” Dr. AlHilli said.

The results need to be confirmed in preclinical animal studies and in additional models, she added.

The study was funded by a K12 Grant and internal funding from Cleveland Clinic. Dr. AlHilli declared no relevant disclosures.