Article

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).

To the Editor:

Atopic dermatitis (AD) affects an estimated 7.2% of adults and 10.7% of children in the United States; however, AD might affect different races at a varying rate.¹ Compared to their European American counterparts, Asian/Pacific Islanders and African Americans are 7 and 3 times more likely, respectively, to be given a diagnosis of AD.²

Despite being disproportionately affected by AD, minority groups might be underrepresented in clinical trials of AD treatments.³ One explanation for this imbalance might be that ethnoracial representation differs across regions in the United States, perhaps in regions where clinical trials are conducted. Price et al³ investigated racial representation in clinical trials of AD globally and found that patients of color are consistently underrepresented.

Research on racial representation in clinical trials within the United States—on national and regional scales—is lacking from the current AD literature. We conducted a study to compare racial and ethnic disparities in AD clinical trials across regions of the United States. 

Using the ClinicalTrials.gov database (www.clinicaltrials.gov) of the National Library of Medicine, we identified clinical trials of AD treatments (encompassing phases 1 through 4) in the United States that were completed before March 14, 2021, with the earliest data from 2013. Search terms included atopic dermatitis, with an advanced search for interventional (clinical trials) and with results.

In total, 95 completed clinical trials were identified, of which 26 (27.4%) reported ethnoracial demographic data. One trial was excluded due to misrepresentation regarding the classification of individuals who identified as more than 1 racial category. Clinical trials for systemic treatments (7 [28%]) and topical treatments (18 [72%]) were identified.

All ethnoracial data were self-reported by trial participants based on US Food and Drug Administration guidelines for racial and ethnic categorization.⁴ Trial participants who identified ethnically as Hispanic or Latino might have been a part of any racial group. Only 7 of the 25 included clinical trials (28%) provided ethnic demographic data (Hispanic [Latino] or non-Hispanic); 72% of trials failed to report ethnicity. Ethnic data included in our analysis came from only the 7 clinical trials that included these data. International multicenter trials that included a US site were excluded.

Ultimately, the number of trials included in our analysis was 25, comprised of 2443 participants. Data were further organized by US geographic region (Northeast, Midwest, South, West, and multiregion trials [ie, conducted in ≥2 regions]). No AD clinical trials were conducted solely in the Midwest; it was only included within multiregion trials.

Compared to their representation in the 2019 US Census, most minority groups were overrepresented in clinical trials, while White individuals were underrepresented (eTable). The percentages of our findings on representation for race are as follows (US Census data are listed in parentheses for comparison⁵):

• White: 56.8% (72.5%)
• Black/African American: 28.3% (12.7%)
• Asian: 10.3% (5.5%)
• Multiracial: 1.1% (3.3%)
• Native Hawaiian or other Pacific Islander: 0.3% (0.2%)
• American Indian or Alaska Native: 0.2% (0.8%)
• Other: 0.5% (4.9%).

Our findings on representation for ethnicity are as follows (US Census data is listed in parentheses for comparison⁵):

• Hispanic or Latino: 21.4% (18.0%)

Although representation of Black/African American and Asian participants in clinical trials was higher than their representation in US Census data and representation of White participants was lower in clinical trials than their representation in census data, equal representation among all racial and ethnic groups is still lacking. A potential explanation for this finding might be that requirements for trial inclusion selected for more minority patients, given the propensity for greater severity of AD among those racial groups.² Another explanation might be that efforts to include minority patients in clinical trials are improving.

There were great differences in ethnoracial representation in clinical trials when regions within the United States were compared. Based on census population data by region, the West had the highest percentage (29.9%) of Hispanic or Latino residents; however, this group represented only 11.7% of participants in AD clinical trials in that region.⁵

The South had the greatest number of participants in AD clinical trials of any region, which was consistent with research findings on an association between severity of AD and heat.⁶ With a warmer climate correlating with an increased incidence of AD, it is possible that more people are willing to participate in clinical trials in the South.

The Midwest was the only region in which region-specific clinical trials were not conducted. Recent studies have shown that individuals with AD who live in the Midwest have comparatively less access to health care associated with AD treatment and are more likely to visit an emergency department because of AD than individuals in any other US region.⁷ This discrepancy highlights the need for increased access to resources and clinical trials focused on the treatment of AD in the Midwest.

In 1993, the National Institutes of Health Revitalization Act established a federal legislative mandate to encourage inclusion of women and people of color in clinical trials.⁸ During the last 2 decades, there have been improvements in ethnoracial reporting. A 2020 global study found that 81.1% of randomized controlled trials (phases 2 and 3) of AD treatments reported ethnoracial data.³

Equal representation in clinical trials allows for further investigation of the connection between race, AD severity, and treatment efficacy. Clinical trials need to have equal representation of ethnoracial categories across all regions of the United States. If one group is notably overrepresented, ethnoracial associations related to the treatment of AD might go undetected.⁹ Similarly, if representation is unequal, relationships of treatment efficacy within ethnoracial groups also might go undetected. None of the clinical trials that we analyzed investigated treatment efficacy by race, suggesting that there is a need for future research in this area.

It also is important to note that broad classifications of race and ethnicity are limiting and therefore overlook differences within ethnoracial categories. Although representation of minority patients in clinical trials for AD treatments is improving, we conclude that there remains a need for greater and equal representation of minority groups in clinical trials of AD treatments in the United States.

To the Editor:

Atopic dermatitis (AD) affects an estimated 7.2% of adults and 10.7% of children in the United States; however, AD might affect different races at a varying rate.¹ Compared to their European American counterparts, Asian/Pacific Islanders and African Americans are 7 and 3 times more likely, respectively, to be given a diagnosis of AD.²

Despite being disproportionately affected by AD, minority groups might be underrepresented in clinical trials of AD treatments.³ One explanation for this imbalance might be that ethnoracial representation differs across regions in the United States, perhaps in regions where clinical trials are conducted. Price et al³ investigated racial representation in clinical trials of AD globally and found that patients of color are consistently underrepresented.

Research on racial representation in clinical trials within the United States—on national and regional scales—is lacking from the current AD literature. We conducted a study to compare racial and ethnic disparities in AD clinical trials across regions of the United States. 

Using the ClinicalTrials.gov database (www.clinicaltrials.gov) of the National Library of Medicine, we identified clinical trials of AD treatments (encompassing phases 1 through 4) in the United States that were completed before March 14, 2021, with the earliest data from 2013. Search terms included atopic dermatitis, with an advanced search for interventional (clinical trials) and with results.

In total, 95 completed clinical trials were identified, of which 26 (27.4%) reported ethnoracial demographic data. One trial was excluded due to misrepresentation regarding the classification of individuals who identified as more than 1 racial category. Clinical trials for systemic treatments (7 [28%]) and topical treatments (18 [72%]) were identified.

All ethnoracial data were self-reported by trial participants based on US Food and Drug Administration guidelines for racial and ethnic categorization.⁴ Trial participants who identified ethnically as Hispanic or Latino might have been a part of any racial group. Only 7 of the 25 included clinical trials (28%) provided ethnic demographic data (Hispanic [Latino] or non-Hispanic); 72% of trials failed to report ethnicity. Ethnic data included in our analysis came from only the 7 clinical trials that included these data. International multicenter trials that included a US site were excluded.

Ultimately, the number of trials included in our analysis was 25, comprised of 2443 participants. Data were further organized by US geographic region (Northeast, Midwest, South, West, and multiregion trials [ie, conducted in ≥2 regions]). No AD clinical trials were conducted solely in the Midwest; it was only included within multiregion trials.

Compared to their representation in the 2019 US Census, most minority groups were overrepresented in clinical trials, while White individuals were underrepresented (eTable). The percentages of our findings on representation for race are as follows (US Census data are listed in parentheses for comparison⁵):

• White: 56.8% (72.5%)
• Black/African American: 28.3% (12.7%)
• Asian: 10.3% (5.5%)
• Multiracial: 1.1% (3.3%)
• Native Hawaiian or other Pacific Islander: 0.3% (0.2%)
• American Indian or Alaska Native: 0.2% (0.8%)
• Other: 0.5% (4.9%).

Our findings on representation for ethnicity are as follows (US Census data is listed in parentheses for comparison⁵):

• Hispanic or Latino: 21.4% (18.0%)

Although representation of Black/African American and Asian participants in clinical trials was higher than their representation in US Census data and representation of White participants was lower in clinical trials than their representation in census data, equal representation among all racial and ethnic groups is still lacking. A potential explanation for this finding might be that requirements for trial inclusion selected for more minority patients, given the propensity for greater severity of AD among those racial groups.² Another explanation might be that efforts to include minority patients in clinical trials are improving.

There were great differences in ethnoracial representation in clinical trials when regions within the United States were compared. Based on census population data by region, the West had the highest percentage (29.9%) of Hispanic or Latino residents; however, this group represented only 11.7% of participants in AD clinical trials in that region.⁵

The South had the greatest number of participants in AD clinical trials of any region, which was consistent with research findings on an association between severity of AD and heat.⁶ With a warmer climate correlating with an increased incidence of AD, it is possible that more people are willing to participate in clinical trials in the South.

The Midwest was the only region in which region-specific clinical trials were not conducted. Recent studies have shown that individuals with AD who live in the Midwest have comparatively less access to health care associated with AD treatment and are more likely to visit an emergency department because of AD than individuals in any other US region.⁷ This discrepancy highlights the need for increased access to resources and clinical trials focused on the treatment of AD in the Midwest.

In 1993, the National Institutes of Health Revitalization Act established a federal legislative mandate to encourage inclusion of women and people of color in clinical trials.⁸ During the last 2 decades, there have been improvements in ethnoracial reporting. A 2020 global study found that 81.1% of randomized controlled trials (phases 2 and 3) of AD treatments reported ethnoracial data.³

Equal representation in clinical trials allows for further investigation of the connection between race, AD severity, and treatment efficacy. Clinical trials need to have equal representation of ethnoracial categories across all regions of the United States. If one group is notably overrepresented, ethnoracial associations related to the treatment of AD might go undetected.⁹ Similarly, if representation is unequal, relationships of treatment efficacy within ethnoracial groups also might go undetected. None of the clinical trials that we analyzed investigated treatment efficacy by race, suggesting that there is a need for future research in this area.

It also is important to note that broad classifications of race and ethnicity are limiting and therefore overlook differences within ethnoracial categories. Although representation of minority patients in clinical trials for AD treatments is improving, we conclude that there remains a need for greater and equal representation of minority groups in clinical trials of AD treatments in the United States.

To the Editor:

Atopic dermatitis (AD) affects an estimated 7.2% of adults and 10.7% of children in the United States; however, AD might affect different races at a varying rate.¹ Compared to their European American counterparts, Asian/Pacific Islanders and African Americans are 7 and 3 times more likely, respectively, to be given a diagnosis of AD.²

Despite being disproportionately affected by AD, minority groups might be underrepresented in clinical trials of AD treatments.³ One explanation for this imbalance might be that ethnoracial representation differs across regions in the United States, perhaps in regions where clinical trials are conducted. Price et al³ investigated racial representation in clinical trials of AD globally and found that patients of color are consistently underrepresented.

Research on racial representation in clinical trials within the United States—on national and regional scales—is lacking from the current AD literature. We conducted a study to compare racial and ethnic disparities in AD clinical trials across regions of the United States. 

Using the ClinicalTrials.gov database (www.clinicaltrials.gov) of the National Library of Medicine, we identified clinical trials of AD treatments (encompassing phases 1 through 4) in the United States that were completed before March 14, 2021, with the earliest data from 2013. Search terms included atopic dermatitis, with an advanced search for interventional (clinical trials) and with results.

In total, 95 completed clinical trials were identified, of which 26 (27.4%) reported ethnoracial demographic data. One trial was excluded due to misrepresentation regarding the classification of individuals who identified as more than 1 racial category. Clinical trials for systemic treatments (7 [28%]) and topical treatments (18 [72%]) were identified.

All ethnoracial data were self-reported by trial participants based on US Food and Drug Administration guidelines for racial and ethnic categorization.⁴ Trial participants who identified ethnically as Hispanic or Latino might have been a part of any racial group. Only 7 of the 25 included clinical trials (28%) provided ethnic demographic data (Hispanic [Latino] or non-Hispanic); 72% of trials failed to report ethnicity. Ethnic data included in our analysis came from only the 7 clinical trials that included these data. International multicenter trials that included a US site were excluded.

Ultimately, the number of trials included in our analysis was 25, comprised of 2443 participants. Data were further organized by US geographic region (Northeast, Midwest, South, West, and multiregion trials [ie, conducted in ≥2 regions]). No AD clinical trials were conducted solely in the Midwest; it was only included within multiregion trials.

Compared to their representation in the 2019 US Census, most minority groups were overrepresented in clinical trials, while White individuals were underrepresented (eTable). The percentages of our findings on representation for race are as follows (US Census data are listed in parentheses for comparison⁵):

• White: 56.8% (72.5%)
• Black/African American: 28.3% (12.7%)
• Asian: 10.3% (5.5%)
• Multiracial: 1.1% (3.3%)
• Native Hawaiian or other Pacific Islander: 0.3% (0.2%)
• American Indian or Alaska Native: 0.2% (0.8%)
• Other: 0.5% (4.9%).

Our findings on representation for ethnicity are as follows (US Census data is listed in parentheses for comparison⁵):

• Hispanic or Latino: 21.4% (18.0%)

Although representation of Black/African American and Asian participants in clinical trials was higher than their representation in US Census data and representation of White participants was lower in clinical trials than their representation in census data, equal representation among all racial and ethnic groups is still lacking. A potential explanation for this finding might be that requirements for trial inclusion selected for more minority patients, given the propensity for greater severity of AD among those racial groups.² Another explanation might be that efforts to include minority patients in clinical trials are improving.

There were great differences in ethnoracial representation in clinical trials when regions within the United States were compared. Based on census population data by region, the West had the highest percentage (29.9%) of Hispanic or Latino residents; however, this group represented only 11.7% of participants in AD clinical trials in that region.⁵

The South had the greatest number of participants in AD clinical trials of any region, which was consistent with research findings on an association between severity of AD and heat.⁶ With a warmer climate correlating with an increased incidence of AD, it is possible that more people are willing to participate in clinical trials in the South.

The Midwest was the only region in which region-specific clinical trials were not conducted. Recent studies have shown that individuals with AD who live in the Midwest have comparatively less access to health care associated with AD treatment and are more likely to visit an emergency department because of AD than individuals in any other US region.⁷ This discrepancy highlights the need for increased access to resources and clinical trials focused on the treatment of AD in the Midwest.

In 1993, the National Institutes of Health Revitalization Act established a federal legislative mandate to encourage inclusion of women and people of color in clinical trials.⁸ During the last 2 decades, there have been improvements in ethnoracial reporting. A 2020 global study found that 81.1% of randomized controlled trials (phases 2 and 3) of AD treatments reported ethnoracial data.³

Equal representation in clinical trials allows for further investigation of the connection between race, AD severity, and treatment efficacy. Clinical trials need to have equal representation of ethnoracial categories across all regions of the United States. If one group is notably overrepresented, ethnoracial associations related to the treatment of AD might go undetected.⁹ Similarly, if representation is unequal, relationships of treatment efficacy within ethnoracial groups also might go undetected. None of the clinical trials that we analyzed investigated treatment efficacy by race, suggesting that there is a need for future research in this area.

It also is important to note that broad classifications of race and ethnicity are limiting and therefore overlook differences within ethnoracial categories. Although representation of minority patients in clinical trials for AD treatments is improving, we conclude that there remains a need for greater and equal representation of minority groups in clinical trials of AD treatments in the United States.

There are limited clinical data concerning inpatient and emergency department (ED) dermatologic consultations. The indications for these consultations vary widely, but in one study (N=271), it was found that 21% of inpatient consultations were for contact dermatitis and 10% were for drug eruptions.¹ In the same study, 77% of patients who required a dermatology consultation eventually were given a different diagnosis or change in treatment after consultation. For example, of all consultations for suspected cellulitis, only 10% were confirmed after dermatology evaluation.¹

Hospitalists and emergency physicians continue to struggle with the assessment of dermatologic conditions, often consulting dermatology whenever a patient has a “rash” or skin concern. Dermatology is still not emphasized in medical education and often is taught to most medical students in an abbreviated fashion, which results in physicians feeling ill-equipped to deal with any dermatologic condition—either mundane or potentially life-threatening. A study in 2016 showed that a monthly lecture series given to hospitalists over the course of 5 years did not improve diagnostic accuracy in patients who were admitted with skin manifestations.² This further shows that there is a need for dermatologic experts in the hospital.

We need to develop better guidelines for physicians in the ED and on inpatient units to guide them on appropriate use of dermatologic consultation outside the ambulatory office and the clinic. A 2013 study showed that patients often were discharged immediately after a dermatologic consultation, furthering our hypothesis that many inpatient consultations can be delayed until after discharge.³

In an era in which medical costs are soaring and there is constant surveillance for ways to reduce costs without impairing quality of care, limiting unnecessary specialty consultations should be embraced. In 2009, $1.8 billion in Medicare claims was paid for dermatology-related admissions.³ A substantial savings to Medicare consulting fees for certain diagnoses, such as cellulitis or contact dermatitis, could be realized if patients were referred for outpatient assessment and treatment. In a study of 271 consultations, 54 patients also had a skin biopsy, which further increases dollars spent on inpatient care and is (usually) something that can be performed in the outpatient setting.¹ In another study, the more common recommended treatments were topical corticosteroids and supportive educational measures for patients and hospital staff,³ which further substantiates that most dermatology consultations are not truly emergent and can wait for outpatient consultation.

In addition, we are dealing with the COVID-19 pandemic in our hospitals and EDs. Many physicians, including dermatologists, would prefer to avoid unnecessary exposure to SARS-CoV-2 on inpatient units and in the ED. It certainly would be preferable to require consultants to come in to evaluate patients only when they truly need to be seen while in the hospital.

There also is limited dermatology training in other specialties, and the dermatology team can help fill this gap with educational programs and one-on-one teaching. Hospital teams have signaled this need, but there has been limited success with multiple teaching opportunities.⁴

We believe that this need for inpatient dermatology services can be filled with the newer subspecialty of hospital dermatology, which is not commonly present at most hospitals; a reason why the subspecialty has not been more popular is that there are few available data in the form of randomized clinical trials that can guide inpatient dermatologists with the care of rare hospital skin diseases.⁵ Having a dermatologic hospitalist available might allow for patients to be seen more readily, which ultimately will save lives and health care dollars and would increase real-time teaching and education for house staff, nursing, and attendings at the bedside.

In a 2018 article,⁶ it was postulated that quicker diagnosis of pseudocellulitis and initiation of antibiotics to treat this condition would save the US health care system $210 million annually. We believe that pseudocellulitis would be best evaluated by inpatient dermatology teams, thereby avoiding costly dermatologic consultations, at an average rate of $138.89.⁶

Morbilliform drug eruptions are among the most common skin conditions seen in the hospital; approximately 95% of cases are an uncomplicated reaction to a medication or virus. Data show that many of these consultations might be unnecessary.⁷

Our institution (Hackensack University Medical Center, New Jersey) is a tertiary hospital that also is connected with a major cancer center. Given this connection, skin eruptions due to chemotherapy and radiation are common. The treatment of drug eruptions, graft-vs-host disease, and other oncologic or drug-related eruptions should be within the scope of practice of our hospitalists, but these cases frequently involve dermatologic consultation.

We constructed a consultation flowchart (Figure) to help guide the triage of patients in need of dermatologic evaluation by inpatient teams and possibly to avoid unnecessary consultation fees. The manner in which this—or any flowchart or teaching aid—is conveyed to hospital personnel is critical so that these tools are not perceived as patronizing or confrontational. In our flowchart, we list emergent dermatologic conditions that we believe are appropriate for dermatology consultation including erythrodermic psoriasis, bullous pemphigoid flare, and Stevens-Johnson syndrome/toxic epidermal necrolysis.

We believe that the flowchart can educate inpatient medical teams about appropriate dermatology consultation. Use of the flowchart also may decrease unnecessary consultations, which ultimately will lower health care spending overall.

There are limited clinical data concerning inpatient and emergency department (ED) dermatologic consultations. The indications for these consultations vary widely, but in one study (N=271), it was found that 21% of inpatient consultations were for contact dermatitis and 10% were for drug eruptions.¹ In the same study, 77% of patients who required a dermatology consultation eventually were given a different diagnosis or change in treatment after consultation. For example, of all consultations for suspected cellulitis, only 10% were confirmed after dermatology evaluation.¹

Hospitalists and emergency physicians continue to struggle with the assessment of dermatologic conditions, often consulting dermatology whenever a patient has a “rash” or skin concern. Dermatology is still not emphasized in medical education and often is taught to most medical students in an abbreviated fashion, which results in physicians feeling ill-equipped to deal with any dermatologic condition—either mundane or potentially life-threatening. A study in 2016 showed that a monthly lecture series given to hospitalists over the course of 5 years did not improve diagnostic accuracy in patients who were admitted with skin manifestations.² This further shows that there is a need for dermatologic experts in the hospital.

We need to develop better guidelines for physicians in the ED and on inpatient units to guide them on appropriate use of dermatologic consultation outside the ambulatory office and the clinic. A 2013 study showed that patients often were discharged immediately after a dermatologic consultation, furthering our hypothesis that many inpatient consultations can be delayed until after discharge.³

In an era in which medical costs are soaring and there is constant surveillance for ways to reduce costs without impairing quality of care, limiting unnecessary specialty consultations should be embraced. In 2009, $1.8 billion in Medicare claims was paid for dermatology-related admissions.³ A substantial savings to Medicare consulting fees for certain diagnoses, such as cellulitis or contact dermatitis, could be realized if patients were referred for outpatient assessment and treatment. In a study of 271 consultations, 54 patients also had a skin biopsy, which further increases dollars spent on inpatient care and is (usually) something that can be performed in the outpatient setting.¹ In another study, the more common recommended treatments were topical corticosteroids and supportive educational measures for patients and hospital staff,³ which further substantiates that most dermatology consultations are not truly emergent and can wait for outpatient consultation.

In addition, we are dealing with the COVID-19 pandemic in our hospitals and EDs. Many physicians, including dermatologists, would prefer to avoid unnecessary exposure to SARS-CoV-2 on inpatient units and in the ED. It certainly would be preferable to require consultants to come in to evaluate patients only when they truly need to be seen while in the hospital.

There also is limited dermatology training in other specialties, and the dermatology team can help fill this gap with educational programs and one-on-one teaching. Hospital teams have signaled this need, but there has been limited success with multiple teaching opportunities.⁴

We believe that this need for inpatient dermatology services can be filled with the newer subspecialty of hospital dermatology, which is not commonly present at most hospitals; a reason why the subspecialty has not been more popular is that there are few available data in the form of randomized clinical trials that can guide inpatient dermatologists with the care of rare hospital skin diseases.⁵ Having a dermatologic hospitalist available might allow for patients to be seen more readily, which ultimately will save lives and health care dollars and would increase real-time teaching and education for house staff, nursing, and attendings at the bedside.

In a 2018 article,⁶ it was postulated that quicker diagnosis of pseudocellulitis and initiation of antibiotics to treat this condition would save the US health care system $210 million annually. We believe that pseudocellulitis would be best evaluated by inpatient dermatology teams, thereby avoiding costly dermatologic consultations, at an average rate of $138.89.⁶

Morbilliform drug eruptions are among the most common skin conditions seen in the hospital; approximately 95% of cases are an uncomplicated reaction to a medication or virus. Data show that many of these consultations might be unnecessary.⁷

Our institution (Hackensack University Medical Center, New Jersey) is a tertiary hospital that also is connected with a major cancer center. Given this connection, skin eruptions due to chemotherapy and radiation are common. The treatment of drug eruptions, graft-vs-host disease, and other oncologic or drug-related eruptions should be within the scope of practice of our hospitalists, but these cases frequently involve dermatologic consultation.

We constructed a consultation flowchart (Figure) to help guide the triage of patients in need of dermatologic evaluation by inpatient teams and possibly to avoid unnecessary consultation fees. The manner in which this—or any flowchart or teaching aid—is conveyed to hospital personnel is critical so that these tools are not perceived as patronizing or confrontational. In our flowchart, we list emergent dermatologic conditions that we believe are appropriate for dermatology consultation including erythrodermic psoriasis, bullous pemphigoid flare, and Stevens-Johnson syndrome/toxic epidermal necrolysis.

We believe that the flowchart can educate inpatient medical teams about appropriate dermatology consultation. Use of the flowchart also may decrease unnecessary consultations, which ultimately will lower health care spending overall.

There are limited clinical data concerning inpatient and emergency department (ED) dermatologic consultations. The indications for these consultations vary widely, but in one study (N=271), it was found that 21% of inpatient consultations were for contact dermatitis and 10% were for drug eruptions.¹ In the same study, 77% of patients who required a dermatology consultation eventually were given a different diagnosis or change in treatment after consultation. For example, of all consultations for suspected cellulitis, only 10% were confirmed after dermatology evaluation.¹

Hospitalists and emergency physicians continue to struggle with the assessment of dermatologic conditions, often consulting dermatology whenever a patient has a “rash” or skin concern. Dermatology is still not emphasized in medical education and often is taught to most medical students in an abbreviated fashion, which results in physicians feeling ill-equipped to deal with any dermatologic condition—either mundane or potentially life-threatening. A study in 2016 showed that a monthly lecture series given to hospitalists over the course of 5 years did not improve diagnostic accuracy in patients who were admitted with skin manifestations.² This further shows that there is a need for dermatologic experts in the hospital.

We need to develop better guidelines for physicians in the ED and on inpatient units to guide them on appropriate use of dermatologic consultation outside the ambulatory office and the clinic. A 2013 study showed that patients often were discharged immediately after a dermatologic consultation, furthering our hypothesis that many inpatient consultations can be delayed until after discharge.³

In an era in which medical costs are soaring and there is constant surveillance for ways to reduce costs without impairing quality of care, limiting unnecessary specialty consultations should be embraced. In 2009, $1.8 billion in Medicare claims was paid for dermatology-related admissions.³ A substantial savings to Medicare consulting fees for certain diagnoses, such as cellulitis or contact dermatitis, could be realized if patients were referred for outpatient assessment and treatment. In a study of 271 consultations, 54 patients also had a skin biopsy, which further increases dollars spent on inpatient care and is (usually) something that can be performed in the outpatient setting.¹ In another study, the more common recommended treatments were topical corticosteroids and supportive educational measures for patients and hospital staff,³ which further substantiates that most dermatology consultations are not truly emergent and can wait for outpatient consultation.

In addition, we are dealing with the COVID-19 pandemic in our hospitals and EDs. Many physicians, including dermatologists, would prefer to avoid unnecessary exposure to SARS-CoV-2 on inpatient units and in the ED. It certainly would be preferable to require consultants to come in to evaluate patients only when they truly need to be seen while in the hospital.

There also is limited dermatology training in other specialties, and the dermatology team can help fill this gap with educational programs and one-on-one teaching. Hospital teams have signaled this need, but there has been limited success with multiple teaching opportunities.⁴

We believe that this need for inpatient dermatology services can be filled with the newer subspecialty of hospital dermatology, which is not commonly present at most hospitals; a reason why the subspecialty has not been more popular is that there are few available data in the form of randomized clinical trials that can guide inpatient dermatologists with the care of rare hospital skin diseases.⁵ Having a dermatologic hospitalist available might allow for patients to be seen more readily, which ultimately will save lives and health care dollars and would increase real-time teaching and education for house staff, nursing, and attendings at the bedside.

In a 2018 article,⁶ it was postulated that quicker diagnosis of pseudocellulitis and initiation of antibiotics to treat this condition would save the US health care system $210 million annually. We believe that pseudocellulitis would be best evaluated by inpatient dermatology teams, thereby avoiding costly dermatologic consultations, at an average rate of $138.89.⁶

Morbilliform drug eruptions are among the most common skin conditions seen in the hospital; approximately 95% of cases are an uncomplicated reaction to a medication or virus. Data show that many of these consultations might be unnecessary.⁷

Our institution (Hackensack University Medical Center, New Jersey) is a tertiary hospital that also is connected with a major cancer center. Given this connection, skin eruptions due to chemotherapy and radiation are common. The treatment of drug eruptions, graft-vs-host disease, and other oncologic or drug-related eruptions should be within the scope of practice of our hospitalists, but these cases frequently involve dermatologic consultation.

We constructed a consultation flowchart (Figure) to help guide the triage of patients in need of dermatologic evaluation by inpatient teams and possibly to avoid unnecessary consultation fees. The manner in which this—or any flowchart or teaching aid—is conveyed to hospital personnel is critical so that these tools are not perceived as patronizing or confrontational. In our flowchart, we list emergent dermatologic conditions that we believe are appropriate for dermatology consultation including erythrodermic psoriasis, bullous pemphigoid flare, and Stevens-Johnson syndrome/toxic epidermal necrolysis.

We believe that the flowchart can educate inpatient medical teams about appropriate dermatology consultation. Use of the flowchart also may decrease unnecessary consultations, which ultimately will lower health care spending overall.

Sunscreen formulations typically protect from UV radiation (290–400 nm), as this is a well-established cause of photodamage, photoaging, and skin cancer.¹ However, sunlight also consists of visible (400–700 nm) and infrared (>700 nm) radiation.² In fact, UV radiation only comprises 5% to 7% of the solar radiation that reaches the surface of the earth, while visible and infrared lights comprise 44% and 53%, respectively.³ Visible light (VL) is the only portion of the solar spectrum visible to the human eye; it penetrates the skin to a depth range of 90 to 750 µm compared to 1.5 to 90 µm for UV radiation.⁴ Visible light also may come from artificial sources such as light bulbs and digital screens. The rapidly increasing use of smartphones, tablets, laptops, and other digital screens that emit high levels of short-wavelength VL has increased concerns about the safety of these devices. Although blue light exposure from screens is small compared with the amount of exposure from the sun, there is concern about the long-term effects of excessive screen time. Recent studies have demonstrated that exposure to light emitted from electronic devices, even for as little as 1 hour, may cause reactive oxygen species generation, apoptosis, collagen degradation, and necrosis of skin cells.⁵ Visible light increases tyrosinase activity and induces immediate erythema in light-skinned individuals and long-lasting pigmentation in dark-skinned individuals.^4,6

Sunscreens consist of chemical and mineral active ingredients that contain UV filters designed to absorb, scatter, and reflect UV photons with wavelengths up to 380 nm. Historically, traditional options do not protect against the effects induced by VL, as these sunscreens use nanosized particles that help to reduce the white appearance and result in transparency of the product.⁷ To block VL, the topical agent must be visible. Tinted sunscreens (TSs) are products that combine UV and VL filters. They give a colored base coverage that is achieved by incorporating a blend of black, red, and yellow iron oxides (IOs) and/or pigmentary titanium dioxide (PTD)(ie, titanium dioxide [TD] that is not nanosized). Because TSs offer an instant glow and protect the skin from both sun and artificial light, they have become increasingly popular and have been incorporated into makeup and skin care products to facilitate daily convenient use.

The purpose of this analysis was to study current available options and product factors that may influence consumer preference when choosing a TS based on the reviewer characteristics.

Methods

The keyword sunscreen was searched in the broader category of skin care products on an online supplier of sunscreens (www.sephora.com). This supplier was chosen because, unlike other sources, specific reviewer characteristics regarding underlying skin tone also were available. The search produced 161 results. For the purpose of this analysis, only facial TSs containing IO and/or PTD were included. Each sunscreen was checked by the authors, and 58 sunscreens that met the inclusion criteria were identified and further reviewed. Descriptive data, including formulation, sun protection factor (SPF), ingredient type (chemical or physical), pigments used, shades available, additional benefits, price range, rating, and user reviews, were gathered. The authors extracted these data from the product information on the website, manufacturer claims, ratings, and reviewer comments on each of the listed sunscreens.

For each product, the content of the top 10 most helpful positive and negative reviews as voted by consumers (1160 total reviews, consisting of 1 or more comments) was analyzed. Two authors (H.D.L.G. and P.V.) coded consumer-reported comments for positive and negative descriptors into the categories of cosmetic elegance, performance, skin compatibility and tolerance, tone compatibility, and affordability. Cosmetic elegance was defined as any feature associated with skin sensation (eg, greasy), color (eg, white cast), scent, ability to blend, and overall appearance of the product on the skin. Product performance included SPF, effectiveness in preventing sunburn, coverage, and finish claims (ie, matte, glow, invisible). Skin compatibility and tolerance were represented in the reviewers’ comments and reflected how the product performed in association with underlying dermatologic conditions, skin type, and if there were any side effects such as irritation or allergic reactions. Tone compatibility referred to TS color similarity with users’ skin and shades available for individual products. Affordability reflected consumers’ perceptions of the product price. Comments may be included in multiple categories (eg, a product was noted to blend well on the skin but did not provide enough coverage). Of entries, 10% (116/1160 reviews) were coded by first author (H.D.L.G.) to ensure internal validity. Reviewer characteristics were consistently available and were used to determine the top 5 recommended products for light-, medium-, and dark-skinned individuals based on the number of 5-star ratings in each group. Porcelain, fair, and light were considered light skin tones. Medium, tan, and olive were considered medium skin tones. Deep, dark, and ebony were considered dark skin tones.

Results

Sunscreen Characteristics—Among the 161 screened products, 58 met the inclusion criteria. Four types of formulations were included: lotion, cream, liquid, and powder. Twenty-nine (50%) were creams, followed by lotions (19%), liquids (28%), and powders (3%). More than 79% (46/58) of products had a reported SPF of 30 or higher. Sunscreens with an active physical ingredient—the minerals TD and/or zinc oxide (ZO)—were most common (33/58 [57%]), followed by the chemical sunscreens avobenzone, octinoxate, oxybenzone, homosalate, octisalate, and/or octocrylene active ingredients (14/58 [24%]), and a combination of chemical and physical sunscreens (11/58 [19%]). Nearly all products (55/58 [95%]) contained pigmentary IO (red, CI 77491; yellow, CI 77492; black, CI 77499). Notably, only 38% (22/58) of products had more than 1 shade. All products had additional claims associated with being hydrating, having antiaging effects, smoothing texture, minimizing the appearance of pores, softening lines, and/or promoting even skin tone. Traditional physical sunscreens (those containing TD and/or ZO) were more expensive than chemical sunscreens, with a median price of $30. The median review rating was 4.5 of 5 stars, with a median of 2300 customer reviews per product. Findings are summarized in Table 1.

Positive Features of Sunscreens—Based on an analysis of total reviews (N=1160), cosmetic elegance was the most cited positive feature associated with TS products (31%), followed by product performance (10%). Skin compatibility and tolerance (7%), tone compatibility (7%), and affordability (7%) were cited less commonly as positive features. When negative features were cited, consumers mostly noted tone incompatibility (16%) and cosmetic elegance concerns (14%). Product performance (13%) was comparatively cited as a negative feature (Table 1). Exemplary positive comments categorized in cosmetic elegance included the subthemes of rubs in well and natural glow. Exemplary negative comments in cosmetic elegance and tone compatibility categories included the subthemes patchy/dry finish and color mismatch. Table 1 illustrates these findings.

Product Recommendations—The top 5 recommendations of the best TS for each skin tone are listed in Table 2. The mean price of the recommended products was $42 for 1 to 1.9 oz. Laura Mercier Tinted Moisturizer Oil Free Natural Skin Perfector broad spectrum SPF 20 (Laura Mercier) was the top product for all 3 groups. Similarly, of 58 products available, the same 5 products—Laura Mercier Tinted Moisturizer Oil Free Natural Skin Perfector broad spectrum SPF 20, IT Cosmetics CC+ Cream with SPF 50 (IT Cosmetics, LLC), Tarte Amazonian Clay BB Tinted Moisturizer Broad Spectrum SPF 20 (Tarte Cosmetics), NARS Pure Radiant Tinted Moisturizer Broad Spectrum SPF 30 (NARS Cosmetics), and Laura Mercier Tinted Moisturizer Natural Skin Perfector broad spectrum SPF 30—were considered the best among consumers of all skin tones, with the addition of 2 different products (bareMinerals Original Liquid Mineral Foundation Broad Spectrum SPF 20 [bareMinerals] and ILIA Super Serum Skin Tint SPF 40 Foundation [ILIA Beauty]) in the dark skin group. Notably, these products were the only ones on Sephora’s website that offered up to 30 (22 on average) different shades.

Comment

Tone Compatibility—Tinted sunscreens were created to extend the range of photoprotection into the VL spectrum. The goal of TSs is to incorporate pigments that blend in with the natural skin tone, produce a glow, and have an aesthetically pleasing appearance. To accommodate a variety of skin colors, different shades can be obtained by mixing different amounts of yellow, red, and black IO with or without PTD. The pigments and reflective compounds provide color, opacity, and a natural coverage. Our qualitative analysis provides information on the lack of diversity among shades available for TS, especially for darker skin tones. Of the 58 products evaluated, 62% (32/58) only had 1 shade. In our cohort, tone compatibility was the most commonly cited negative feature. Of note, 89% of these comments were from consumers with dark skin tones, and there was a disproportional number of reviews by darker-skinned individuals compared to users with light and medium skin tones. This is of particular importance, as TSs have been shown to protect against dermatoses that disproportionally affect individuals with skin of color. When comparing sunscreen formulations containing IO with regular mineral sunscreens, Dumbuya et al³ found that IO-containing formulations significantly protected against VL-induced pigmentation compared with untreated skin or mineral sunscreen with SPF 50 or higher in individuals with Fitzpatrick skin type IV (P<.001). Similarly, Bernstein et al⁸ found that exposing patients with Fitzpatrick skin types III and IV to blue-violet light resulted in marked hyperpigmentation that lasted up to 3 months. Visible light elicits immediate and persistent pigment darkening in individuals with Fitzpatrick skin phototype III and above via the photo-oxidation of pre-existing melanin and de novo melanogenesis.⁹ Tinted sunscreens formulated with IO have been shown to aid in the treatment of melasma and prevent hyperpigmentation in individuals with Fitzpatrick skin types IV to VI.¹⁰ Patients with darker skin tones with dermatoses aggravated or induced by VL, such as melasma and postinflammatory hyperpigmentation, may seek photoprotection provided by TS but find the lack of matching shades unappealing. The dearth of shade diversity that matches all skin tones can lead to inequities and disproportionally affect those with darker skin.

Performance—Tinted sunscreen formulations containing IO have been proven effective in protecting against high-energy VL, especially when combined synergistically with ZO.¹¹ Kaye et al¹² found that TSs containing IO and the inorganic filters TD or ZO reduced transmittance of VL more effectively than nontinted sunscreens containing TD or ZO alone or products containing organic filters. The decreased VL transmittance in the former is due to synergistic effects of the VL-scattering properties of the TD and the VL absorption properties of the IO. Similarly, Sayre et al¹³ demonstrated that IO was superior to TD and ZO in attenuating the transmission of VL. Bernstein et al¹⁴ found that darker shades containing higher percentages of IO increased the attenuation of VL to 98% compared with lighter shades attenuating 93%. This correlates with the results of prior studies highlighting the potential of TSs in protecting individuals with skin of color.³ In our cohort, comments regarding product performance and protection were mostly positive, claiming that consistent use reduced hyperpigmentation on the skin surface, giving the appearance of a more even skin tone.

Tolerability—Iron oxides are minerals known to be safe, gentle, and nontoxic on the surface of the skin.¹⁵ Two case reports of contact dermatitis due to IO have been reported.^16,17 Within our cohort, only a few of the comments (6%) described negative product tolerance or compatibility with their skin type. However, it is more likely that these incompatibilities were due to other ingredients in the product or the individuals’ underlying dermatologic conditions.

Cosmetic Elegance—Most of the sunscreens available on the market today contain micronized forms of TD and ZO particles because they have better cosmetic acceptability.¹⁸ However, their reduced size compromises the protection provided against VL whereby the addition of IO is of vital importance. According to the RealSelf Sun Safety Report, only 11% of Americans wear sunscreen daily, and 46% never wear sunscreen.¹⁹ The most common reasons consumers reported for not wearing sunscreen included not liking how it looks on the skin, forgetting to apply it, and/or believing that application is inconvenient and time-consuming. Currently, TSs have been incorporated into daily-life products such as makeup, moisturizers, and serums, making application for users easy and convenient, decreasing the necessity of using multiple products, and offering the opportunity to choose from different presentations to make decisions for convenience and/or diverse occasions. Products containing IO blend in with the natural skin tone and have an aesthetically pleasing cosmetic appearance. In our cohort, comments regarding cosmetic elegance were highly valued and were present in multiple reviews (45%), with 69% being positive.

Affordability—In our cohort, product price was not predominantly mentioned in consumers’ reviews. However, negative comments regarding affordability were slightly higher than the positive (56% vs 44%). Notably, the mean price of our top recommendations was $42. Higher price was associated with products with a wider range of shades available. Prior studies have found similar results demonstrating that websites with recommendations on sunscreens for patients with skin of color compared with sunscreens for white or fair skin were more likely to recommend more expensive products (median, $14/oz vs $11.3/oz) despite the lower SPF level.²⁰ According to Schneider,²¹ daily use of the cheapest sunscreen on the head/neck region recommended for white/pale skin ($2/oz) would lead to an annual cost of $61 compared to $182 for darker skin ($6/oz). This showcases the considerable variation in sunscreen prices for both populations that could potentiate disparities and vulnerability in the latter group.

Conclusion

Tinted sunscreens provide both functional and cosmetic benefits and are a safe, effective, and convenient way to protect against high-energy VL. This study suggests that patients with skin of color encounter difficulties in finding matching shades in TS products. These difficulties may stem from the lack of knowledge regarding dark complexions and undertones and the lack of representation of black and brown skin that has persisted in dermatology research journals and textbooks for decades.²² Our study provides important insights to help dermatologists improve their familiarity with the brands and characteristics of TSs geared to patients with all skin tones, including skin of color. Limitations include single-retailer information and inclusion of both highly and poorly rated comments with subjective data, limiting generalizability. The limited selection of shades for darker skin poses a roadblock to proper treatment and prevention. These data represent an area for improvement within the beauty industry and the dermatologic field to deliver culturally sensitive care by being knowledgeable about darker skin tones and TS formulations tailored to people with skin of color.

Sunscreen formulations typically protect from UV radiation (290–400 nm), as this is a well-established cause of photodamage, photoaging, and skin cancer.¹ However, sunlight also consists of visible (400–700 nm) and infrared (>700 nm) radiation.² In fact, UV radiation only comprises 5% to 7% of the solar radiation that reaches the surface of the earth, while visible and infrared lights comprise 44% and 53%, respectively.³ Visible light (VL) is the only portion of the solar spectrum visible to the human eye; it penetrates the skin to a depth range of 90 to 750 µm compared to 1.5 to 90 µm for UV radiation.⁴ Visible light also may come from artificial sources such as light bulbs and digital screens. The rapidly increasing use of smartphones, tablets, laptops, and other digital screens that emit high levels of short-wavelength VL has increased concerns about the safety of these devices. Although blue light exposure from screens is small compared with the amount of exposure from the sun, there is concern about the long-term effects of excessive screen time. Recent studies have demonstrated that exposure to light emitted from electronic devices, even for as little as 1 hour, may cause reactive oxygen species generation, apoptosis, collagen degradation, and necrosis of skin cells.⁵ Visible light increases tyrosinase activity and induces immediate erythema in light-skinned individuals and long-lasting pigmentation in dark-skinned individuals.^4,6

Sunscreens consist of chemical and mineral active ingredients that contain UV filters designed to absorb, scatter, and reflect UV photons with wavelengths up to 380 nm. Historically, traditional options do not protect against the effects induced by VL, as these sunscreens use nanosized particles that help to reduce the white appearance and result in transparency of the product.⁷ To block VL, the topical agent must be visible. Tinted sunscreens (TSs) are products that combine UV and VL filters. They give a colored base coverage that is achieved by incorporating a blend of black, red, and yellow iron oxides (IOs) and/or pigmentary titanium dioxide (PTD)(ie, titanium dioxide [TD] that is not nanosized). Because TSs offer an instant glow and protect the skin from both sun and artificial light, they have become increasingly popular and have been incorporated into makeup and skin care products to facilitate daily convenient use.

The purpose of this analysis was to study current available options and product factors that may influence consumer preference when choosing a TS based on the reviewer characteristics.

Methods

The keyword sunscreen was searched in the broader category of skin care products on an online supplier of sunscreens (www.sephora.com). This supplier was chosen because, unlike other sources, specific reviewer characteristics regarding underlying skin tone also were available. The search produced 161 results. For the purpose of this analysis, only facial TSs containing IO and/or PTD were included. Each sunscreen was checked by the authors, and 58 sunscreens that met the inclusion criteria were identified and further reviewed. Descriptive data, including formulation, sun protection factor (SPF), ingredient type (chemical or physical), pigments used, shades available, additional benefits, price range, rating, and user reviews, were gathered. The authors extracted these data from the product information on the website, manufacturer claims, ratings, and reviewer comments on each of the listed sunscreens.

For each product, the content of the top 10 most helpful positive and negative reviews as voted by consumers (1160 total reviews, consisting of 1 or more comments) was analyzed. Two authors (H.D.L.G. and P.V.) coded consumer-reported comments for positive and negative descriptors into the categories of cosmetic elegance, performance, skin compatibility and tolerance, tone compatibility, and affordability. Cosmetic elegance was defined as any feature associated with skin sensation (eg, greasy), color (eg, white cast), scent, ability to blend, and overall appearance of the product on the skin. Product performance included SPF, effectiveness in preventing sunburn, coverage, and finish claims (ie, matte, glow, invisible). Skin compatibility and tolerance were represented in the reviewers’ comments and reflected how the product performed in association with underlying dermatologic conditions, skin type, and if there were any side effects such as irritation or allergic reactions. Tone compatibility referred to TS color similarity with users’ skin and shades available for individual products. Affordability reflected consumers’ perceptions of the product price. Comments may be included in multiple categories (eg, a product was noted to blend well on the skin but did not provide enough coverage). Of entries, 10% (116/1160 reviews) were coded by first author (H.D.L.G.) to ensure internal validity. Reviewer characteristics were consistently available and were used to determine the top 5 recommended products for light-, medium-, and dark-skinned individuals based on the number of 5-star ratings in each group. Porcelain, fair, and light were considered light skin tones. Medium, tan, and olive were considered medium skin tones. Deep, dark, and ebony were considered dark skin tones.

Results

Sunscreen Characteristics—Among the 161 screened products, 58 met the inclusion criteria. Four types of formulations were included: lotion, cream, liquid, and powder. Twenty-nine (50%) were creams, followed by lotions (19%), liquids (28%), and powders (3%). More than 79% (46/58) of products had a reported SPF of 30 or higher. Sunscreens with an active physical ingredient—the minerals TD and/or zinc oxide (ZO)—were most common (33/58 [57%]), followed by the chemical sunscreens avobenzone, octinoxate, oxybenzone, homosalate, octisalate, and/or octocrylene active ingredients (14/58 [24%]), and a combination of chemical and physical sunscreens (11/58 [19%]). Nearly all products (55/58 [95%]) contained pigmentary IO (red, CI 77491; yellow, CI 77492; black, CI 77499). Notably, only 38% (22/58) of products had more than 1 shade. All products had additional claims associated with being hydrating, having antiaging effects, smoothing texture, minimizing the appearance of pores, softening lines, and/or promoting even skin tone. Traditional physical sunscreens (those containing TD and/or ZO) were more expensive than chemical sunscreens, with a median price of $30. The median review rating was 4.5 of 5 stars, with a median of 2300 customer reviews per product. Findings are summarized in Table 1.

Positive Features of Sunscreens—Based on an analysis of total reviews (N=1160), cosmetic elegance was the most cited positive feature associated with TS products (31%), followed by product performance (10%). Skin compatibility and tolerance (7%), tone compatibility (7%), and affordability (7%) were cited less commonly as positive features. When negative features were cited, consumers mostly noted tone incompatibility (16%) and cosmetic elegance concerns (14%). Product performance (13%) was comparatively cited as a negative feature (Table 1). Exemplary positive comments categorized in cosmetic elegance included the subthemes of rubs in well and natural glow. Exemplary negative comments in cosmetic elegance and tone compatibility categories included the subthemes patchy/dry finish and color mismatch. Table 1 illustrates these findings.

Product Recommendations—The top 5 recommendations of the best TS for each skin tone are listed in Table 2. The mean price of the recommended products was $42 for 1 to 1.9 oz. Laura Mercier Tinted Moisturizer Oil Free Natural Skin Perfector broad spectrum SPF 20 (Laura Mercier) was the top product for all 3 groups. Similarly, of 58 products available, the same 5 products—Laura Mercier Tinted Moisturizer Oil Free Natural Skin Perfector broad spectrum SPF 20, IT Cosmetics CC+ Cream with SPF 50 (IT Cosmetics, LLC), Tarte Amazonian Clay BB Tinted Moisturizer Broad Spectrum SPF 20 (Tarte Cosmetics), NARS Pure Radiant Tinted Moisturizer Broad Spectrum SPF 30 (NARS Cosmetics), and Laura Mercier Tinted Moisturizer Natural Skin Perfector broad spectrum SPF 30—were considered the best among consumers of all skin tones, with the addition of 2 different products (bareMinerals Original Liquid Mineral Foundation Broad Spectrum SPF 20 [bareMinerals] and ILIA Super Serum Skin Tint SPF 40 Foundation [ILIA Beauty]) in the dark skin group. Notably, these products were the only ones on Sephora’s website that offered up to 30 (22 on average) different shades.

Comment

Tone Compatibility—Tinted sunscreens were created to extend the range of photoprotection into the VL spectrum. The goal of TSs is to incorporate pigments that blend in with the natural skin tone, produce a glow, and have an aesthetically pleasing appearance. To accommodate a variety of skin colors, different shades can be obtained by mixing different amounts of yellow, red, and black IO with or without PTD. The pigments and reflective compounds provide color, opacity, and a natural coverage. Our qualitative analysis provides information on the lack of diversity among shades available for TS, especially for darker skin tones. Of the 58 products evaluated, 62% (32/58) only had 1 shade. In our cohort, tone compatibility was the most commonly cited negative feature. Of note, 89% of these comments were from consumers with dark skin tones, and there was a disproportional number of reviews by darker-skinned individuals compared to users with light and medium skin tones. This is of particular importance, as TSs have been shown to protect against dermatoses that disproportionally affect individuals with skin of color. When comparing sunscreen formulations containing IO with regular mineral sunscreens, Dumbuya et al³ found that IO-containing formulations significantly protected against VL-induced pigmentation compared with untreated skin or mineral sunscreen with SPF 50 or higher in individuals with Fitzpatrick skin type IV (P<.001). Similarly, Bernstein et al⁸ found that exposing patients with Fitzpatrick skin types III and IV to blue-violet light resulted in marked hyperpigmentation that lasted up to 3 months. Visible light elicits immediate and persistent pigment darkening in individuals with Fitzpatrick skin phototype III and above via the photo-oxidation of pre-existing melanin and de novo melanogenesis.⁹ Tinted sunscreens formulated with IO have been shown to aid in the treatment of melasma and prevent hyperpigmentation in individuals with Fitzpatrick skin types IV to VI.¹⁰ Patients with darker skin tones with dermatoses aggravated or induced by VL, such as melasma and postinflammatory hyperpigmentation, may seek photoprotection provided by TS but find the lack of matching shades unappealing. The dearth of shade diversity that matches all skin tones can lead to inequities and disproportionally affect those with darker skin.

Performance—Tinted sunscreen formulations containing IO have been proven effective in protecting against high-energy VL, especially when combined synergistically with ZO.¹¹ Kaye et al¹² found that TSs containing IO and the inorganic filters TD or ZO reduced transmittance of VL more effectively than nontinted sunscreens containing TD or ZO alone or products containing organic filters. The decreased VL transmittance in the former is due to synergistic effects of the VL-scattering properties of the TD and the VL absorption properties of the IO. Similarly, Sayre et al¹³ demonstrated that IO was superior to TD and ZO in attenuating the transmission of VL. Bernstein et al¹⁴ found that darker shades containing higher percentages of IO increased the attenuation of VL to 98% compared with lighter shades attenuating 93%. This correlates with the results of prior studies highlighting the potential of TSs in protecting individuals with skin of color.³ In our cohort, comments regarding product performance and protection were mostly positive, claiming that consistent use reduced hyperpigmentation on the skin surface, giving the appearance of a more even skin tone.

Tolerability—Iron oxides are minerals known to be safe, gentle, and nontoxic on the surface of the skin.¹⁵ Two case reports of contact dermatitis due to IO have been reported.^16,17 Within our cohort, only a few of the comments (6%) described negative product tolerance or compatibility with their skin type. However, it is more likely that these incompatibilities were due to other ingredients in the product or the individuals’ underlying dermatologic conditions.

Cosmetic Elegance—Most of the sunscreens available on the market today contain micronized forms of TD and ZO particles because they have better cosmetic acceptability.¹⁸ However, their reduced size compromises the protection provided against VL whereby the addition of IO is of vital importance. According to the RealSelf Sun Safety Report, only 11% of Americans wear sunscreen daily, and 46% never wear sunscreen.¹⁹ The most common reasons consumers reported for not wearing sunscreen included not liking how it looks on the skin, forgetting to apply it, and/or believing that application is inconvenient and time-consuming. Currently, TSs have been incorporated into daily-life products such as makeup, moisturizers, and serums, making application for users easy and convenient, decreasing the necessity of using multiple products, and offering the opportunity to choose from different presentations to make decisions for convenience and/or diverse occasions. Products containing IO blend in with the natural skin tone and have an aesthetically pleasing cosmetic appearance. In our cohort, comments regarding cosmetic elegance were highly valued and were present in multiple reviews (45%), with 69% being positive.

Affordability—In our cohort, product price was not predominantly mentioned in consumers’ reviews. However, negative comments regarding affordability were slightly higher than the positive (56% vs 44%). Notably, the mean price of our top recommendations was $42. Higher price was associated with products with a wider range of shades available. Prior studies have found similar results demonstrating that websites with recommendations on sunscreens for patients with skin of color compared with sunscreens for white or fair skin were more likely to recommend more expensive products (median, $14/oz vs $11.3/oz) despite the lower SPF level.²⁰ According to Schneider,²¹ daily use of the cheapest sunscreen on the head/neck region recommended for white/pale skin ($2/oz) would lead to an annual cost of $61 compared to $182 for darker skin ($6/oz). This showcases the considerable variation in sunscreen prices for both populations that could potentiate disparities and vulnerability in the latter group.

Conclusion

Tinted sunscreens provide both functional and cosmetic benefits and are a safe, effective, and convenient way to protect against high-energy VL. This study suggests that patients with skin of color encounter difficulties in finding matching shades in TS products. These difficulties may stem from the lack of knowledge regarding dark complexions and undertones and the lack of representation of black and brown skin that has persisted in dermatology research journals and textbooks for decades.²² Our study provides important insights to help dermatologists improve their familiarity with the brands and characteristics of TSs geared to patients with all skin tones, including skin of color. Limitations include single-retailer information and inclusion of both highly and poorly rated comments with subjective data, limiting generalizability. The limited selection of shades for darker skin poses a roadblock to proper treatment and prevention. These data represent an area for improvement within the beauty industry and the dermatologic field to deliver culturally sensitive care by being knowledgeable about darker skin tones and TS formulations tailored to people with skin of color.

Sunscreen formulations typically protect from UV radiation (290–400 nm), as this is a well-established cause of photodamage, photoaging, and skin cancer.¹ However, sunlight also consists of visible (400–700 nm) and infrared (>700 nm) radiation.² In fact, UV radiation only comprises 5% to 7% of the solar radiation that reaches the surface of the earth, while visible and infrared lights comprise 44% and 53%, respectively.³ Visible light (VL) is the only portion of the solar spectrum visible to the human eye; it penetrates the skin to a depth range of 90 to 750 µm compared to 1.5 to 90 µm for UV radiation.⁴ Visible light also may come from artificial sources such as light bulbs and digital screens. The rapidly increasing use of smartphones, tablets, laptops, and other digital screens that emit high levels of short-wavelength VL has increased concerns about the safety of these devices. Although blue light exposure from screens is small compared with the amount of exposure from the sun, there is concern about the long-term effects of excessive screen time. Recent studies have demonstrated that exposure to light emitted from electronic devices, even for as little as 1 hour, may cause reactive oxygen species generation, apoptosis, collagen degradation, and necrosis of skin cells.⁵ Visible light increases tyrosinase activity and induces immediate erythema in light-skinned individuals and long-lasting pigmentation in dark-skinned individuals.^4,6

Sunscreens consist of chemical and mineral active ingredients that contain UV filters designed to absorb, scatter, and reflect UV photons with wavelengths up to 380 nm. Historically, traditional options do not protect against the effects induced by VL, as these sunscreens use nanosized particles that help to reduce the white appearance and result in transparency of the product.⁷ To block VL, the topical agent must be visible. Tinted sunscreens (TSs) are products that combine UV and VL filters. They give a colored base coverage that is achieved by incorporating a blend of black, red, and yellow iron oxides (IOs) and/or pigmentary titanium dioxide (PTD)(ie, titanium dioxide [TD] that is not nanosized). Because TSs offer an instant glow and protect the skin from both sun and artificial light, they have become increasingly popular and have been incorporated into makeup and skin care products to facilitate daily convenient use.

The purpose of this analysis was to study current available options and product factors that may influence consumer preference when choosing a TS based on the reviewer characteristics.

Methods

The keyword sunscreen was searched in the broader category of skin care products on an online supplier of sunscreens (www.sephora.com). This supplier was chosen because, unlike other sources, specific reviewer characteristics regarding underlying skin tone also were available. The search produced 161 results. For the purpose of this analysis, only facial TSs containing IO and/or PTD were included. Each sunscreen was checked by the authors, and 58 sunscreens that met the inclusion criteria were identified and further reviewed. Descriptive data, including formulation, sun protection factor (SPF), ingredient type (chemical or physical), pigments used, shades available, additional benefits, price range, rating, and user reviews, were gathered. The authors extracted these data from the product information on the website, manufacturer claims, ratings, and reviewer comments on each of the listed sunscreens.

For each product, the content of the top 10 most helpful positive and negative reviews as voted by consumers (1160 total reviews, consisting of 1 or more comments) was analyzed. Two authors (H.D.L.G. and P.V.) coded consumer-reported comments for positive and negative descriptors into the categories of cosmetic elegance, performance, skin compatibility and tolerance, tone compatibility, and affordability. Cosmetic elegance was defined as any feature associated with skin sensation (eg, greasy), color (eg, white cast), scent, ability to blend, and overall appearance of the product on the skin. Product performance included SPF, effectiveness in preventing sunburn, coverage, and finish claims (ie, matte, glow, invisible). Skin compatibility and tolerance were represented in the reviewers’ comments and reflected how the product performed in association with underlying dermatologic conditions, skin type, and if there were any side effects such as irritation or allergic reactions. Tone compatibility referred to TS color similarity with users’ skin and shades available for individual products. Affordability reflected consumers’ perceptions of the product price. Comments may be included in multiple categories (eg, a product was noted to blend well on the skin but did not provide enough coverage). Of entries, 10% (116/1160 reviews) were coded by first author (H.D.L.G.) to ensure internal validity. Reviewer characteristics were consistently available and were used to determine the top 5 recommended products for light-, medium-, and dark-skinned individuals based on the number of 5-star ratings in each group. Porcelain, fair, and light were considered light skin tones. Medium, tan, and olive were considered medium skin tones. Deep, dark, and ebony were considered dark skin tones.

Results

Sunscreen Characteristics—Among the 161 screened products, 58 met the inclusion criteria. Four types of formulations were included: lotion, cream, liquid, and powder. Twenty-nine (50%) were creams, followed by lotions (19%), liquids (28%), and powders (3%). More than 79% (46/58) of products had a reported SPF of 30 or higher. Sunscreens with an active physical ingredient—the minerals TD and/or zinc oxide (ZO)—were most common (33/58 [57%]), followed by the chemical sunscreens avobenzone, octinoxate, oxybenzone, homosalate, octisalate, and/or octocrylene active ingredients (14/58 [24%]), and a combination of chemical and physical sunscreens (11/58 [19%]). Nearly all products (55/58 [95%]) contained pigmentary IO (red, CI 77491; yellow, CI 77492; black, CI 77499). Notably, only 38% (22/58) of products had more than 1 shade. All products had additional claims associated with being hydrating, having antiaging effects, smoothing texture, minimizing the appearance of pores, softening lines, and/or promoting even skin tone. Traditional physical sunscreens (those containing TD and/or ZO) were more expensive than chemical sunscreens, with a median price of $30. The median review rating was 4.5 of 5 stars, with a median of 2300 customer reviews per product. Findings are summarized in Table 1.

Positive Features of Sunscreens—Based on an analysis of total reviews (N=1160), cosmetic elegance was the most cited positive feature associated with TS products (31%), followed by product performance (10%). Skin compatibility and tolerance (7%), tone compatibility (7%), and affordability (7%) were cited less commonly as positive features. When negative features were cited, consumers mostly noted tone incompatibility (16%) and cosmetic elegance concerns (14%). Product performance (13%) was comparatively cited as a negative feature (Table 1). Exemplary positive comments categorized in cosmetic elegance included the subthemes of rubs in well and natural glow. Exemplary negative comments in cosmetic elegance and tone compatibility categories included the subthemes patchy/dry finish and color mismatch. Table 1 illustrates these findings.

Product Recommendations—The top 5 recommendations of the best TS for each skin tone are listed in Table 2. The mean price of the recommended products was $42 for 1 to 1.9 oz. Laura Mercier Tinted Moisturizer Oil Free Natural Skin Perfector broad spectrum SPF 20 (Laura Mercier) was the top product for all 3 groups. Similarly, of 58 products available, the same 5 products—Laura Mercier Tinted Moisturizer Oil Free Natural Skin Perfector broad spectrum SPF 20, IT Cosmetics CC+ Cream with SPF 50 (IT Cosmetics, LLC), Tarte Amazonian Clay BB Tinted Moisturizer Broad Spectrum SPF 20 (Tarte Cosmetics), NARS Pure Radiant Tinted Moisturizer Broad Spectrum SPF 30 (NARS Cosmetics), and Laura Mercier Tinted Moisturizer Natural Skin Perfector broad spectrum SPF 30—were considered the best among consumers of all skin tones, with the addition of 2 different products (bareMinerals Original Liquid Mineral Foundation Broad Spectrum SPF 20 [bareMinerals] and ILIA Super Serum Skin Tint SPF 40 Foundation [ILIA Beauty]) in the dark skin group. Notably, these products were the only ones on Sephora’s website that offered up to 30 (22 on average) different shades.

Comment

Tone Compatibility—Tinted sunscreens were created to extend the range of photoprotection into the VL spectrum. The goal of TSs is to incorporate pigments that blend in with the natural skin tone, produce a glow, and have an aesthetically pleasing appearance. To accommodate a variety of skin colors, different shades can be obtained by mixing different amounts of yellow, red, and black IO with or without PTD. The pigments and reflective compounds provide color, opacity, and a natural coverage. Our qualitative analysis provides information on the lack of diversity among shades available for TS, especially for darker skin tones. Of the 58 products evaluated, 62% (32/58) only had 1 shade. In our cohort, tone compatibility was the most commonly cited negative feature. Of note, 89% of these comments were from consumers with dark skin tones, and there was a disproportional number of reviews by darker-skinned individuals compared to users with light and medium skin tones. This is of particular importance, as TSs have been shown to protect against dermatoses that disproportionally affect individuals with skin of color. When comparing sunscreen formulations containing IO with regular mineral sunscreens, Dumbuya et al³ found that IO-containing formulations significantly protected against VL-induced pigmentation compared with untreated skin or mineral sunscreen with SPF 50 or higher in individuals with Fitzpatrick skin type IV (P<.001). Similarly, Bernstein et al⁸ found that exposing patients with Fitzpatrick skin types III and IV to blue-violet light resulted in marked hyperpigmentation that lasted up to 3 months. Visible light elicits immediate and persistent pigment darkening in individuals with Fitzpatrick skin phototype III and above via the photo-oxidation of pre-existing melanin and de novo melanogenesis.⁹ Tinted sunscreens formulated with IO have been shown to aid in the treatment of melasma and prevent hyperpigmentation in individuals with Fitzpatrick skin types IV to VI.¹⁰ Patients with darker skin tones with dermatoses aggravated or induced by VL, such as melasma and postinflammatory hyperpigmentation, may seek photoprotection provided by TS but find the lack of matching shades unappealing. The dearth of shade diversity that matches all skin tones can lead to inequities and disproportionally affect those with darker skin.

Performance—Tinted sunscreen formulations containing IO have been proven effective in protecting against high-energy VL, especially when combined synergistically with ZO.¹¹ Kaye et al¹² found that TSs containing IO and the inorganic filters TD or ZO reduced transmittance of VL more effectively than nontinted sunscreens containing TD or ZO alone or products containing organic filters. The decreased VL transmittance in the former is due to synergistic effects of the VL-scattering properties of the TD and the VL absorption properties of the IO. Similarly, Sayre et al¹³ demonstrated that IO was superior to TD and ZO in attenuating the transmission of VL. Bernstein et al¹⁴ found that darker shades containing higher percentages of IO increased the attenuation of VL to 98% compared with lighter shades attenuating 93%. This correlates with the results of prior studies highlighting the potential of TSs in protecting individuals with skin of color.³ In our cohort, comments regarding product performance and protection were mostly positive, claiming that consistent use reduced hyperpigmentation on the skin surface, giving the appearance of a more even skin tone.

Tolerability—Iron oxides are minerals known to be safe, gentle, and nontoxic on the surface of the skin.¹⁵ Two case reports of contact dermatitis due to IO have been reported.^16,17 Within our cohort, only a few of the comments (6%) described negative product tolerance or compatibility with their skin type. However, it is more likely that these incompatibilities were due to other ingredients in the product or the individuals’ underlying dermatologic conditions.

Cosmetic Elegance—Most of the sunscreens available on the market today contain micronized forms of TD and ZO particles because they have better cosmetic acceptability.¹⁸ However, their reduced size compromises the protection provided against VL whereby the addition of IO is of vital importance. According to the RealSelf Sun Safety Report, only 11% of Americans wear sunscreen daily, and 46% never wear sunscreen.¹⁹ The most common reasons consumers reported for not wearing sunscreen included not liking how it looks on the skin, forgetting to apply it, and/or believing that application is inconvenient and time-consuming. Currently, TSs have been incorporated into daily-life products such as makeup, moisturizers, and serums, making application for users easy and convenient, decreasing the necessity of using multiple products, and offering the opportunity to choose from different presentations to make decisions for convenience and/or diverse occasions. Products containing IO blend in with the natural skin tone and have an aesthetically pleasing cosmetic appearance. In our cohort, comments regarding cosmetic elegance were highly valued and were present in multiple reviews (45%), with 69% being positive.

Affordability—In our cohort, product price was not predominantly mentioned in consumers’ reviews. However, negative comments regarding affordability were slightly higher than the positive (56% vs 44%). Notably, the mean price of our top recommendations was $42. Higher price was associated with products with a wider range of shades available. Prior studies have found similar results demonstrating that websites with recommendations on sunscreens for patients with skin of color compared with sunscreens for white or fair skin were more likely to recommend more expensive products (median, $14/oz vs $11.3/oz) despite the lower SPF level.²⁰ According to Schneider,²¹ daily use of the cheapest sunscreen on the head/neck region recommended for white/pale skin ($2/oz) would lead to an annual cost of $61 compared to $182 for darker skin ($6/oz). This showcases the considerable variation in sunscreen prices for both populations that could potentiate disparities and vulnerability in the latter group.

Conclusion

Tinted sunscreens provide both functional and cosmetic benefits and are a safe, effective, and convenient way to protect against high-energy VL. This study suggests that patients with skin of color encounter difficulties in finding matching shades in TS products. These difficulties may stem from the lack of knowledge regarding dark complexions and undertones and the lack of representation of black and brown skin that has persisted in dermatology research journals and textbooks for decades.²² Our study provides important insights to help dermatologists improve their familiarity with the brands and characteristics of TSs geared to patients with all skin tones, including skin of color. Limitations include single-retailer information and inclusion of both highly and poorly rated comments with subjective data, limiting generalizability. The limited selection of shades for darker skin poses a roadblock to proper treatment and prevention. These data represent an area for improvement within the beauty industry and the dermatologic field to deliver culturally sensitive care by being knowledgeable about darker skin tones and TS formulations tailored to people with skin of color.

Edema blisters are a common but often underreported entity most commonly seen on the lower extremities in the setting of acute edema. ¹ Reported risk factors and associations include chronic venous insufficiency, congestive heart failure, hereditary angioedema, and medications (eg, amlodipine). ^1,2 We report a newly described variant that we have termed anasarca-induced desquamation in which a patient sloughed the entire cutaneous surface of the body after gaining almost 40 pounds over 5 days.

Case Report

A 50-year-old man without a home was found minimally responsive in a yard. His core body temperature was 25.5 °C. He was profoundly acidotic (pH, <6.733 [reference range, 7.35–7.45]; lactic acid, 20.5 mmol/L [reference range, 0.5–2.2 mmol/L]) at admission. His medical history was notable for diabetes mellitus, hypertension, alcohol abuse, and pulmonary embolism. The patient was resuscitated with rewarming and intravenous fluids in the setting of acute renal insufficiency. By day 5 of the hospital stay, he had a net positive intake of 21.8 L and an 18-kg (39.7-lb) weight gain.

Dermatology was consulted for skin sloughing. Physical examination revealed nonpainful desquamation of the vermilion lip, periorbital skin, right shoulder, and hips without notable mucosal changes. Two 4-mm punch biopsies of the shoulder revealed an intracorneal split with desquamation of the stratum corneum and a mild dermal lymphocytic infiltrate, consistent with exfoliation secondary to edema or staphylococcal scalded skin syndrome (Figure 1). No staphylococcal growth was noted on blood, urine, nasal, wound, and ocular cultures throughout the hospital stay.

As the patient’s anasarca improved with diuretics and continuous renal replacement therapy, the entire cutaneous surface—head to toe—underwent desquamation, including the palms and soles. He was managed with supportive skin care. The anasarca healed completely with residual hypopigmentation (Figures 2 and 3).

Comment

Anasarca-induced desquamation represents a more diffuse form of a known entity: edema blisters. Occurring most commonly in the setting of acute exacerbation of chronic venous insufficiency, edema blisters can mimic other vesiculobullous conditions, such as bullous pemphigoid and herpes zoster.³

Pathogenesis of Edema Blisters—Edema develops in the skin when the capillary filtration rate, determined by the hydrostatic and oncotic pressures of the capillaries and interstitium, exceeds venous and lymphatic drainage. The appearance of edema blisters in the acute setting likely is related to the speed at which edema develops in skin.¹ Although edema blisters often are described as tense, there is a paucity of histologic data at the anatomical level of split in the skin.In our patient, desquamation was within the stratum corneum and likely multifactorial. His weight gain of nearly 40 lb, the result of intravenous instillation of fluids and low urine output, was undeniably a contributing factor. The anasarca was aggravated by hypoalbuminemia (2.1 g/dL) in the setting of known liver disease. Other possible contributing factors were hypotension, which required vasopressor therapy that led to hypoperfusion of the skin, and treatment of hypothermia, with resulting reactive vasodilation and capillary leak.

Management—Treatment of acute edema blisters is focused on the underlying cause of the edema. In a study of 13 patients with edema blisters, all had blisters on the legs that resolved with treatment, such as diuretics or compression therapy.¹

Anasarca-induced desquamation is an inherently benign condition that mimics potentially fatal disorders, such as Stevens-Johnson syndrome, staphylococcal scalded skin syndrome, and toxic shock syndrome. Therefore, patients presenting with diffuse superficial desquamation should be assessed for the mucosal changes of Stevens-Johnson syndrome and a history of acute edema in the affected areas to avoid potentially harmful empiric treatments, such as corticosteroids and intravenous antibiotics.

Conclusion

Anasarca-induced desquamation represents a more diffuse form of edema blisters. This desquamation can mimic a potentially fatal rash, such as Stevens-Johnson syndrome and staphylococcal scalded skin syndrome.

Edema blisters are a common but often underreported entity most commonly seen on the lower extremities in the setting of acute edema. ¹ Reported risk factors and associations include chronic venous insufficiency, congestive heart failure, hereditary angioedema, and medications (eg, amlodipine). ^1,2 We report a newly described variant that we have termed anasarca-induced desquamation in which a patient sloughed the entire cutaneous surface of the body after gaining almost 40 pounds over 5 days.

Case Report

A 50-year-old man without a home was found minimally responsive in a yard. His core body temperature was 25.5 °C. He was profoundly acidotic (pH, <6.733 [reference range, 7.35–7.45]; lactic acid, 20.5 mmol/L [reference range, 0.5–2.2 mmol/L]) at admission. His medical history was notable for diabetes mellitus, hypertension, alcohol abuse, and pulmonary embolism. The patient was resuscitated with rewarming and intravenous fluids in the setting of acute renal insufficiency. By day 5 of the hospital stay, he had a net positive intake of 21.8 L and an 18-kg (39.7-lb) weight gain.

Dermatology was consulted for skin sloughing. Physical examination revealed nonpainful desquamation of the vermilion lip, periorbital skin, right shoulder, and hips without notable mucosal changes. Two 4-mm punch biopsies of the shoulder revealed an intracorneal split with desquamation of the stratum corneum and a mild dermal lymphocytic infiltrate, consistent with exfoliation secondary to edema or staphylococcal scalded skin syndrome (Figure 1). No staphylococcal growth was noted on blood, urine, nasal, wound, and ocular cultures throughout the hospital stay.

As the patient’s anasarca improved with diuretics and continuous renal replacement therapy, the entire cutaneous surface—head to toe—underwent desquamation, including the palms and soles. He was managed with supportive skin care. The anasarca healed completely with residual hypopigmentation (Figures 2 and 3).

Comment

Anasarca-induced desquamation represents a more diffuse form of a known entity: edema blisters. Occurring most commonly in the setting of acute exacerbation of chronic venous insufficiency, edema blisters can mimic other vesiculobullous conditions, such as bullous pemphigoid and herpes zoster.³

Pathogenesis of Edema Blisters—Edema develops in the skin when the capillary filtration rate, determined by the hydrostatic and oncotic pressures of the capillaries and interstitium, exceeds venous and lymphatic drainage. The appearance of edema blisters in the acute setting likely is related to the speed at which edema develops in skin.¹ Although edema blisters often are described as tense, there is a paucity of histologic data at the anatomical level of split in the skin.In our patient, desquamation was within the stratum corneum and likely multifactorial. His weight gain of nearly 40 lb, the result of intravenous instillation of fluids and low urine output, was undeniably a contributing factor. The anasarca was aggravated by hypoalbuminemia (2.1 g/dL) in the setting of known liver disease. Other possible contributing factors were hypotension, which required vasopressor therapy that led to hypoperfusion of the skin, and treatment of hypothermia, with resulting reactive vasodilation and capillary leak.

Management—Treatment of acute edema blisters is focused on the underlying cause of the edema. In a study of 13 patients with edema blisters, all had blisters on the legs that resolved with treatment, such as diuretics or compression therapy.¹

Anasarca-induced desquamation is an inherently benign condition that mimics potentially fatal disorders, such as Stevens-Johnson syndrome, staphylococcal scalded skin syndrome, and toxic shock syndrome. Therefore, patients presenting with diffuse superficial desquamation should be assessed for the mucosal changes of Stevens-Johnson syndrome and a history of acute edema in the affected areas to avoid potentially harmful empiric treatments, such as corticosteroids and intravenous antibiotics.

Conclusion

Anasarca-induced desquamation represents a more diffuse form of edema blisters. This desquamation can mimic a potentially fatal rash, such as Stevens-Johnson syndrome and staphylococcal scalded skin syndrome.

Edema blisters are a common but often underreported entity most commonly seen on the lower extremities in the setting of acute edema. ¹ Reported risk factors and associations include chronic venous insufficiency, congestive heart failure, hereditary angioedema, and medications (eg, amlodipine). ^1,2 We report a newly described variant that we have termed anasarca-induced desquamation in which a patient sloughed the entire cutaneous surface of the body after gaining almost 40 pounds over 5 days.

Case Report

A 50-year-old man without a home was found minimally responsive in a yard. His core body temperature was 25.5 °C. He was profoundly acidotic (pH, <6.733 [reference range, 7.35–7.45]; lactic acid, 20.5 mmol/L [reference range, 0.5–2.2 mmol/L]) at admission. His medical history was notable for diabetes mellitus, hypertension, alcohol abuse, and pulmonary embolism. The patient was resuscitated with rewarming and intravenous fluids in the setting of acute renal insufficiency. By day 5 of the hospital stay, he had a net positive intake of 21.8 L and an 18-kg (39.7-lb) weight gain.

Dermatology was consulted for skin sloughing. Physical examination revealed nonpainful desquamation of the vermilion lip, periorbital skin, right shoulder, and hips without notable mucosal changes. Two 4-mm punch biopsies of the shoulder revealed an intracorneal split with desquamation of the stratum corneum and a mild dermal lymphocytic infiltrate, consistent with exfoliation secondary to edema or staphylococcal scalded skin syndrome (Figure 1). No staphylococcal growth was noted on blood, urine, nasal, wound, and ocular cultures throughout the hospital stay.

As the patient’s anasarca improved with diuretics and continuous renal replacement therapy, the entire cutaneous surface—head to toe—underwent desquamation, including the palms and soles. He was managed with supportive skin care. The anasarca healed completely with residual hypopigmentation (Figures 2 and 3).

Comment

Anasarca-induced desquamation represents a more diffuse form of a known entity: edema blisters. Occurring most commonly in the setting of acute exacerbation of chronic venous insufficiency, edema blisters can mimic other vesiculobullous conditions, such as bullous pemphigoid and herpes zoster.³

Pathogenesis of Edema Blisters—Edema develops in the skin when the capillary filtration rate, determined by the hydrostatic and oncotic pressures of the capillaries and interstitium, exceeds venous and lymphatic drainage. The appearance of edema blisters in the acute setting likely is related to the speed at which edema develops in skin.¹ Although edema blisters often are described as tense, there is a paucity of histologic data at the anatomical level of split in the skin.In our patient, desquamation was within the stratum corneum and likely multifactorial. His weight gain of nearly 40 lb, the result of intravenous instillation of fluids and low urine output, was undeniably a contributing factor. The anasarca was aggravated by hypoalbuminemia (2.1 g/dL) in the setting of known liver disease. Other possible contributing factors were hypotension, which required vasopressor therapy that led to hypoperfusion of the skin, and treatment of hypothermia, with resulting reactive vasodilation and capillary leak.

Management—Treatment of acute edema blisters is focused on the underlying cause of the edema. In a study of 13 patients with edema blisters, all had blisters on the legs that resolved with treatment, such as diuretics or compression therapy.¹

Anasarca-induced desquamation is an inherently benign condition that mimics potentially fatal disorders, such as Stevens-Johnson syndrome, staphylococcal scalded skin syndrome, and toxic shock syndrome. Therefore, patients presenting with diffuse superficial desquamation should be assessed for the mucosal changes of Stevens-Johnson syndrome and a history of acute edema in the affected areas to avoid potentially harmful empiric treatments, such as corticosteroids and intravenous antibiotics.

Conclusion

Anasarca-induced desquamation represents a more diffuse form of edema blisters. This desquamation can mimic a potentially fatal rash, such as Stevens-Johnson syndrome and staphylococcal scalded skin syndrome.

The medicated intrauterine devices (IUDs), including the levonorgestrel-releasing IUD (LNG-IUD) (Mirena, Kyleena, Skyla, and Liletta) and the copper IUD (Cu-IUD; Paragard), are remarkably effective contraceptives. For the 52-mg LNG-IUD (Mirena, Liletta) the pregnancy rate over 6 years of use averaged less than 0.2% per year.^1,2 For the Cu-IUD, the pregnancy rate over 10 years of use averaged 0.5% per year for the first 3 years of use and 0.2% per year over the following 7 years of use.³ IUD perforation of the uterus, expulsion, and malposition are recognized complications of IUD use. Our understanding of the prevalence and management of malpositioned IUDs is evolving and the main focus of this editorial.

Complete and partial uterus perforation

A complete uterine perforation occurs when the entire IUD is outside the walls of the uterus. A partial uterine perforation occurs when the IUD is outside the uterine cavity, but a portion of the IUD remains in the myometrium. When uterine perforation is suspected, ultrasound can determine if the IUD is properly sited within the uterus. If ultrasonography does not detect the IUD within the uterus, an x-ray of the pelvis and abdomen should be obtained to determine if the IUD is in the peritoneal cavity. If both an ultrasound and a pelvic-abdominal x-ray do not detect the IUD, the IUD was probably expelled from the patient.

Uterine perforation is uncommon and occurs once in every 500 to 1,000 insertions in non-breastfeeding women.^4-8 The most common symptoms reported by patients with a perforated IUD are pain and/or bleeding.⁸ Investigators in the European Active Surveillance Study on Intrauterine Devices (EURAS) enrolled more than 60,000 patients who had an IUD insertion and followed them for 12 months with more than 39,000 followed for up to 60 months.^7,8 The uterine perforation rate per 1,000 IUD insertions in non-breastfeeding women with 60 months of follow-up was 1.6 for the LNG-IUD and 0.8 for the Cu-IUD.⁸ The rate of uterine perforation was much higher in women who are breastfeeding or recently postpartum. In the EURAS study after 60 months of follow-up, the perforation rate per 1,000 insertions among breastfeeding women was 7.9 for the LNG-IUS and 4.7 for the Cu-IUD.⁸

Remarkably very few IUD perforations were detected at the time of insertion, including only 2% of the LNG-IUD insertions and 17% of the Cu-IUD insertions.⁸ Many perforations were not detected until more than 12 months following insertion, including 32% of the LNG-IUD insertions and 22% of the Cu-IUD insertions.⁸ Obviously, an IUD that has completely perforated the uterus and resides in the peritoneal cavity is not an effective contraceptive. For some patients, the IUD perforation was initially diagnosed after they became pregnant, and imaging studies to locate the IUD and assess the pregnancy were initiated. Complete perforation is usually treated with laparoscopy to remove the IUD and reduce the risk of injury to intra-abdominal organs.

Patients with an IUD partial perforation may present with pelvic pain or abnormal uterine bleeding.⁹ An ultrasound study to explore the cause of the presenting symptom may detect the partial perforation. It is estimated that approximately 20% of cases of IUD perforation are partial perforation.⁹ Over time, a partial perforation may progress to a complete perforation. In some cases of partial perforation, the IUD string may still be visible in the cervix, and the IUD may be removed by pulling on the strings.⁸ Hysteroscopy and/or laparoscopy may be needed to remove a partially perforated IUD. Following a partial or complete IUD perforation, if the patient desires to continue with IUD contraception, it would be wise to insert a new IUD under ultrasound guidance or assess proper placement with a postplacement ultrasound.

Continue to: Expulsion...

IUD expulsion occurs in approximately 3% to 11% of patients.^10-13 The age of the patient influences the rate of expulsion. In a study of 2,748 patients with a Cu-IUD, the rate of expulsion by age for patients <20 years, 20–24 years, 25–29 years, 30–34 years, and ≥35 years was 8.2%, 3.2%, 3.0%, 2.3%, and 1.8%, respectively.¹⁰ In this study, age did not influence the rate of IUD removal for pelvic pain or abnormal bleeding, which was 4% to 5% across all age groups.¹⁰ In a study of 5,403 patients with an IUD, the rate of IUD expulsion by age for patients <20 years, 20–29 years, and 30–45 years was 14.6%, 7.3%, and 7.2%, respectively.¹² In this study, the 3-year cumulative rate of expulsion was 10.2%.¹² There was no statistically significant difference in the 3-year cumulative rate of expulsion for the 52-mg LNG-IUD (10.1%) and Cu-IUD (10.7%).¹²

The majority of patients who have an IUD expulsion recognize the event and seek additional contraception care. A few patients first recognize the IUD expulsion when they become pregnant, and imaging studies detect no IUD in the uterus or the peritoneal cavity. In a study of more than 17,000 patients using an LNG-IUD, 108 pregnancies were reported. Seven pregnancies occurred in patients who did not realize their IUD was expelled.¹⁴ Patients who have had an IUD expulsion and receive a new IUD are at increased risk for re-expulsion. For these patients, reinsertion of an IUD could be performed under ultrasound guidance to ensure and document optimal initial IUD position within the uterus, or ultrasound can be obtained postinsertion to document appropriate IUD position.

Malposition—prevalence and management

Our understanding of the prevalence and management of a malpositioned IUD is evolving. For the purposes of this discussion a malpositioned IUD is defined as being in the uterus, but not properly positioned within the uterine cavity. Perforation into the peritoneal cavity and complete expulsion of an IUD are considered separate entities. However, a malpositioned IUD within the uterus may eventually perforate the uterus or be expelled from the body. For example, an IUD embedded in the uterine wall may eventually work its way through the wall and become perforated, residing in the peritoneal cavity. An IUD with the stem in the cervix below the internal os may eventually be expelled from the uterus and leave the body through the vagina.

High-quality ultrasonography, including 2-dimensional (2-D) ultrasound with videoclips or 3-dimensional (3-D) ultrasound with coronal views, has greatly advanced our understanding of the prevalence and characteristics of a malpositioned IUD.^15-18 Ultrasound features of an IUD correctly placed within the uterus include:

• the IUD is in the uterus
• the shaft is in the midline of the uterine cavity
• the shaft of the IUD is not in the endocervix
• the IUD arms are at a 90-degree angle from the shaft
• the top of the IUD is within 2 cm of the fundus
• the IUD is not rotated outside of the cornual plane, inverted or transverse.

Ultrasound imaging has identified multiple types of malpositioned IUDs, including:

• IUD embedded in the myometrium—a portion of the IUD is embedded in the uterine wall
• low-lying IUD—the IUD is low in the uterine cavity but not in the endocervix
• IUD in the endocervix—the stem is in the endocervical canal
• rotated—the IUD is rotated outside the cornual plane
• malpositioned arms—the arms are not at a 90-degree angle to the stem
• the IUD is inverted, transverse, or laterally displaced.

IUD malposition is highly prevalent and has been identified in 10% to 20% of convenience cohorts in which an ultrasound study was performed.^15-18

Benacerraf, Shipp, and Bromley were among the first experts to use ultrasound to detect the high prevalence of malpositioned IUDs among a convenience sample of 167 patients with an IUD undergoing ultrasound for a variety of indications. Using 3-D ultrasound, including reconstructed coronal views, they identified 28 patients (17%) with a malpositioned IUD based on the detection of the IUD “poking into the substance of the uterus or cervix.” Among the patients with a malpositioned IUD, the principal indication for the ultrasound study was pelvic pain (39%) or abnormal uterine bleeding (36%). Among women with a normally sited IUD, pelvic pain (19%) or abnormal uterine bleeding (15%) were less often the principal indication for the ultrasound.¹⁵ The malpositioned IUD was removed in 21 of the 28 cases and the symptoms of pelvic pain or abnormal bleeding resolved in 20 of the 21 patients.¹⁵

Other investigators have confirmed the observation that IUD malposition is common.^16-18 In a retrospective study of 1,748 pelvic ultrasounds performed for any indication where an IUD was present, after excluding 13 patients who were determined to have expelled their IUD (13) and 13 patients with a perforated IUD, 156 patients (8.9%) were diagnosed as having a malpositioned IUD.¹⁶ IUD malposition was diagnosed when the IUD was in the uterus but positioned in the lower uterine segment, cervix, rotated or embedded in the uterus. An IUD in the lower uterine segment or cervix was detected in 133 patients, representing 85% of cases. Among these cases, 29 IUDs were also embedded and/or rotated, indicating that some IUDs have multiple causes of the malposition. Twenty-one IUDs were near the fundus but embedded and/or rotated. Controls with a normally-sited IUD were selected for comparison to the case group. Among IUD users, the identification of suspected adenomyosis on the ultrasound was associated with an increased risk of IUD malposition (odds ratio [OR], 3.04; 95% confidence interval [CI], 1.08-8.52).¹⁶ In this study, removal of a malpositioned LNG-IUD, without initiating a highly reliable contraceptive was associated with an increased risk of pregnancy. It is important to initiate a highly reliable form of contraception if the plan is to remove a malpositioned IUD.^16,19

In a study of 1,253 pelvic ultrasounds performed for any indication where an IUD was identified in the uterus, 263 IUDs (19%) were determined to be malpositioned.¹⁷ In this study the location of the malpositioned IUDs included¹⁷:

• the lower uterine segment not extending into the cervix (38%)
• in the lower uterine segment extending into the cervix (22%)
• in the cervix (26%)
• rotated axis of the IUD (12%)
• other (2%).

Among the 236 malpositioned IUDs, 24% appeared to be embedded in the uterine wall.¹⁷ Compared with patients with a normally-sited IUD on ultrasound, patients with a malpositioned IUD more frequently reported vaginal bleeding (30% vs 19%; P<.005) and pelvic pain (43% vs 30%; P<.002), similar to the findings in the Benacerraf et al. study.¹⁴

Connolly and Fox¹⁸ designed an innovative study to determine the rate of malpositioned IUDs using 2-D ultrasound to ensure proper IUD placement at the time of insertion with a follow-up 3-D ultrasound 8 weeks after insertion to assess IUD position within the uterus. At the 8-week 3-D ultrasound, among 763 women, 16.6% of the IUDs were malpositioned.¹⁸ In this study, IUD position was determined to be correct if all the following features were identified:

• the IUD shaft was in the midline of the uterine cavity
• the IUD arms were at 90 degrees from the stem
• the top of the IUD was within 3 to 4 mm of the fundus
• the IUD was not rotated, inverted or transverse.

IUD malpositions were categorized as:

• embedded in the uterine wall
• low in the uterine cavity
• in the endocervical canal
• misaligned
• perforated
• expulsed.

At the 8-week follow-up, 636 patients (83.4%) had an IUD that was correctly positioned.¹⁸ In 127 patients (16.6%) IUD malposition was identified, with some patients having more than one type of malposition. The types of malposition identified were:

• embedded in the myometrium (54%)
• misaligned, including rotated, laterally displaced, inverted, transverse or arms not deployed (47%)
• low in the uterine cavity (39%)
• in the endocervical canal (14%)
• perforated (3%)
• expulsion (0%).

Recall that all of these patients had a 2-D ultrasound at the time of insertion that identified the IUD as correctly placed. This suggests that during the 8 weeks following IUD placement there were changes in the location of the IUD or that 2-D ultrasound has lower sensitivity than 3-D ultrasound to detect malposition. Of note, at the 8-week follow-up, bleeding or pain was reported by 36% of the patients with a malpositioned IUD and 20% of patients with a correctly positioned IUD.¹⁷ Sixty-seven of the 127 malpositioned IUDs “required” removal, but the precise reasons for the removals were not delineated. The investigators concluded that 3-D ultrasonography is useful for the detection of IUD malposition and could be considered as part of ongoing IUD care, if symptoms of pain or bleeding occur.¹⁸

Continue to: IUD malposition following postplacental insertion...

IUD malposition following postplacental insertion

IUD malposition is common in patients who have had a postplacental insertion. Ultrasound imaging plays an important role in detecting IUD expulsion and malposition in these cases. Postplacental IUD insertion is defined as the placement of an IUD within 10 minutes following delivery of the placenta. Postplacental IUD insertion can be performed following a vaginal or cesarean birth and with a Cu-IUD or LNG-IUD. The good news is that postplacental IUD insertion reduces the risk of unplanned pregnancy in the years following birth. However, postplacental IUD insertion is associated with a high rate of IUD malposition.

In a study of 162 patients who had postplacental insertion of a Cu-IUD following a vaginal birth, ultrasound and physical examination at 6 months demonstrated complete IUD expulsion in 8%, partial expulsion in 16%, and malposition in 15%.²⁰ The IUD was correctly sited in 56% of patients. Seven patients (4%) had the IUD removed, and 1 patient had a perforated IUD. Among the 25 malpositioned IUDs, 14 were not within 1 cm of the fundus, and 11 were rotated outside of the axis of the cornuas. In this study partial expulsion was defined as an IUD protruding from the external cervical os on physical exam or demonstration of the distal tip of the IUD below the internal os of the cervix on ultrasound. Malposition was defined as an IUD that was >1 cm from the fundus or in an abnormal location or axis, but not partially expelled.

In a study of 69 patients who had postplacental insertion of a Cu-IUD following a cesarean birth, ultrasound and physical examination at 6 months demonstrated complete IUD expulsion in 3%, partial expulsion (stem in the cervix below the internal os) in 4% and malposition in 30%.²⁰ The IUD was correctly positioned in 59% of the patients.²¹ The IUD had been electively removed in 3%. Among the 21 patients with a malpositioned IUD, 10 were rotated within the uterine cavity, 6 were inverted (upside down), 3 were low-lying, and 2 were transverse.²¹ Given the relatively high rate of IUD malposition following postplacental insertion, it may be useful to perform a pelvic ultrasound at a postpartum visit to assess the location of the IUD, if ultrasonography is available.

Management of the malpositioned IUD

There are no consensus guidelines on how to care for a patient with a malpositioned IUD. Clinicians need to use their best judgment and engage the patient in joint decision making when managing a malpositioned IUD. When an IUD is malpositioned and the patient has bothersome symptoms of pelvic pain or abnormal bleeding that have not responded to standard interventions, consideration may be given to a remove and replace strategy. When the stem of the IUD is below the level of the internal os on ultrasound or visible at the external os on physical examination, consideration should be given to removing and replacing the IUD. However, if the IUD is removed without replacement or the initiation of a highly reliable contraceptive, the risk of unplanned pregnancy is considerable.^16,19

IUD totally or partially within the cervix or low-lying. When an IUD is in the cervix, the contraceptive efficacy of the IUD may be diminished, especially with a Cu-IUD.²² In these cases, removing and replacing the IUD is an option. In a survey of 20 expert clinicians, >80% recommended replacing an IUD that was totally or partially in the cervical canal.²³ But most of the experts would not replace an IUD that was incidentally noted on ultrasound to be low-lying, being positioned more than 2 cm below the fundus, with no portion of the IUD in the cervical canal. In the same survey, for patients with a low-lying IUD and pelvic pain or bleeding, the majority of experts reported that they would explore other causes of bleeding and pelvic pain not related to the IUD itself and not replace the IUD, but 30% of the experts reported that they would remove and replace the device.²³

IUD embedded in the myometrium with pelvic pain. Based on my clinical experience, when a patient has persistent pelvic pain following the insertion of an IUD and the pain does not resolve with standard measures including medication, an ultrasound study is warranted to assess the position of the IUD. If the ultrasound demonstrates that an arm of the IUD is embedded in the myometrium, removal of the IUD may be associated with resolution of the pain. Reinsertion of an IUD under ultrasound guidance may result in a correctly-sited IUD with no recurrence of pelvic pain.

IUD rotated within the uterus with no pain or abnormal bleeding. For an IUD that is near the fundus and rotated on its axis within the uterus, if the patient has no symptoms of pain or abnormal bleeding, my recommendation to the patient would be to leave the device in situ.

Without available guidelines, engage in clinician-patient discussion

It is clear that IUD malposition is common, occurring in 10% to 20% of patients with an IUD. High-quality ultrasound imaging is helpful in detecting IUD malposition, including 2-D ultrasound with videoclips and/or 3-D ultrasound with coronal reconstruction. More data are needed to identify the best options for managing various types of malpositioned IUDs in patients with and without bothersome symptoms such as pain and bleeding. Until consensus guidelines are developed, clinicians need to engage the patient in a discussion of how to best manage the malpositioned IUD. Medicated IUDs and progestin subdermal implants are our two most effective reversible contraceptives. They are among the most important advances in health care over the past half-century. ●

The medicated intrauterine devices (IUDs), including the levonorgestrel-releasing IUD (LNG-IUD) (Mirena, Kyleena, Skyla, and Liletta) and the copper IUD (Cu-IUD; Paragard), are remarkably effective contraceptives. For the 52-mg LNG-IUD (Mirena, Liletta) the pregnancy rate over 6 years of use averaged less than 0.2% per year.^1,2 For the Cu-IUD, the pregnancy rate over 10 years of use averaged 0.5% per year for the first 3 years of use and 0.2% per year over the following 7 years of use.³ IUD perforation of the uterus, expulsion, and malposition are recognized complications of IUD use. Our understanding of the prevalence and management of malpositioned IUDs is evolving and the main focus of this editorial.

Complete and partial uterus perforation

A complete uterine perforation occurs when the entire IUD is outside the walls of the uterus. A partial uterine perforation occurs when the IUD is outside the uterine cavity, but a portion of the IUD remains in the myometrium. When uterine perforation is suspected, ultrasound can determine if the IUD is properly sited within the uterus. If ultrasonography does not detect the IUD within the uterus, an x-ray of the pelvis and abdomen should be obtained to determine if the IUD is in the peritoneal cavity. If both an ultrasound and a pelvic-abdominal x-ray do not detect the IUD, the IUD was probably expelled from the patient.

Uterine perforation is uncommon and occurs once in every 500 to 1,000 insertions in non-breastfeeding women.^4-8 The most common symptoms reported by patients with a perforated IUD are pain and/or bleeding.⁸ Investigators in the European Active Surveillance Study on Intrauterine Devices (EURAS) enrolled more than 60,000 patients who had an IUD insertion and followed them for 12 months with more than 39,000 followed for up to 60 months.^7,8 The uterine perforation rate per 1,000 IUD insertions in non-breastfeeding women with 60 months of follow-up was 1.6 for the LNG-IUD and 0.8 for the Cu-IUD.⁸ The rate of uterine perforation was much higher in women who are breastfeeding or recently postpartum. In the EURAS study after 60 months of follow-up, the perforation rate per 1,000 insertions among breastfeeding women was 7.9 for the LNG-IUS and 4.7 for the Cu-IUD.⁸

Remarkably very few IUD perforations were detected at the time of insertion, including only 2% of the LNG-IUD insertions and 17% of the Cu-IUD insertions.⁸ Many perforations were not detected until more than 12 months following insertion, including 32% of the LNG-IUD insertions and 22% of the Cu-IUD insertions.⁸ Obviously, an IUD that has completely perforated the uterus and resides in the peritoneal cavity is not an effective contraceptive. For some patients, the IUD perforation was initially diagnosed after they became pregnant, and imaging studies to locate the IUD and assess the pregnancy were initiated. Complete perforation is usually treated with laparoscopy to remove the IUD and reduce the risk of injury to intra-abdominal organs.

Patients with an IUD partial perforation may present with pelvic pain or abnormal uterine bleeding.⁹ An ultrasound study to explore the cause of the presenting symptom may detect the partial perforation. It is estimated that approximately 20% of cases of IUD perforation are partial perforation.⁹ Over time, a partial perforation may progress to a complete perforation. In some cases of partial perforation, the IUD string may still be visible in the cervix, and the IUD may be removed by pulling on the strings.⁸ Hysteroscopy and/or laparoscopy may be needed to remove a partially perforated IUD. Following a partial or complete IUD perforation, if the patient desires to continue with IUD contraception, it would be wise to insert a new IUD under ultrasound guidance or assess proper placement with a postplacement ultrasound.

Continue to: Expulsion...

IUD expulsion occurs in approximately 3% to 11% of patients.^10-13 The age of the patient influences the rate of expulsion. In a study of 2,748 patients with a Cu-IUD, the rate of expulsion by age for patients <20 years, 20–24 years, 25–29 years, 30–34 years, and ≥35 years was 8.2%, 3.2%, 3.0%, 2.3%, and 1.8%, respectively.¹⁰ In this study, age did not influence the rate of IUD removal for pelvic pain or abnormal bleeding, which was 4% to 5% across all age groups.¹⁰ In a study of 5,403 patients with an IUD, the rate of IUD expulsion by age for patients <20 years, 20–29 years, and 30–45 years was 14.6%, 7.3%, and 7.2%, respectively.¹² In this study, the 3-year cumulative rate of expulsion was 10.2%.¹² There was no statistically significant difference in the 3-year cumulative rate of expulsion for the 52-mg LNG-IUD (10.1%) and Cu-IUD (10.7%).¹²

The majority of patients who have an IUD expulsion recognize the event and seek additional contraception care. A few patients first recognize the IUD expulsion when they become pregnant, and imaging studies detect no IUD in the uterus or the peritoneal cavity. In a study of more than 17,000 patients using an LNG-IUD, 108 pregnancies were reported. Seven pregnancies occurred in patients who did not realize their IUD was expelled.¹⁴ Patients who have had an IUD expulsion and receive a new IUD are at increased risk for re-expulsion. For these patients, reinsertion of an IUD could be performed under ultrasound guidance to ensure and document optimal initial IUD position within the uterus, or ultrasound can be obtained postinsertion to document appropriate IUD position.

Malposition—prevalence and management

Our understanding of the prevalence and management of a malpositioned IUD is evolving. For the purposes of this discussion a malpositioned IUD is defined as being in the uterus, but not properly positioned within the uterine cavity. Perforation into the peritoneal cavity and complete expulsion of an IUD are considered separate entities. However, a malpositioned IUD within the uterus may eventually perforate the uterus or be expelled from the body. For example, an IUD embedded in the uterine wall may eventually work its way through the wall and become perforated, residing in the peritoneal cavity. An IUD with the stem in the cervix below the internal os may eventually be expelled from the uterus and leave the body through the vagina.

High-quality ultrasonography, including 2-dimensional (2-D) ultrasound with videoclips or 3-dimensional (3-D) ultrasound with coronal views, has greatly advanced our understanding of the prevalence and characteristics of a malpositioned IUD.^15-18 Ultrasound features of an IUD correctly placed within the uterus include:

• the IUD is in the uterus
• the shaft is in the midline of the uterine cavity
• the shaft of the IUD is not in the endocervix
• the IUD arms are at a 90-degree angle from the shaft
• the top of the IUD is within 2 cm of the fundus
• the IUD is not rotated outside of the cornual plane, inverted or transverse.

Ultrasound imaging has identified multiple types of malpositioned IUDs, including:

• IUD embedded in the myometrium—a portion of the IUD is embedded in the uterine wall
• low-lying IUD—the IUD is low in the uterine cavity but not in the endocervix
• IUD in the endocervix—the stem is in the endocervical canal
• rotated—the IUD is rotated outside the cornual plane
• malpositioned arms—the arms are not at a 90-degree angle to the stem
• the IUD is inverted, transverse, or laterally displaced.

IUD malposition is highly prevalent and has been identified in 10% to 20% of convenience cohorts in which an ultrasound study was performed.^15-18

Benacerraf, Shipp, and Bromley were among the first experts to use ultrasound to detect the high prevalence of malpositioned IUDs among a convenience sample of 167 patients with an IUD undergoing ultrasound for a variety of indications. Using 3-D ultrasound, including reconstructed coronal views, they identified 28 patients (17%) with a malpositioned IUD based on the detection of the IUD “poking into the substance of the uterus or cervix.” Among the patients with a malpositioned IUD, the principal indication for the ultrasound study was pelvic pain (39%) or abnormal uterine bleeding (36%). Among women with a normally sited IUD, pelvic pain (19%) or abnormal uterine bleeding (15%) were less often the principal indication for the ultrasound.¹⁵ The malpositioned IUD was removed in 21 of the 28 cases and the symptoms of pelvic pain or abnormal bleeding resolved in 20 of the 21 patients.¹⁵

Other investigators have confirmed the observation that IUD malposition is common.^16-18 In a retrospective study of 1,748 pelvic ultrasounds performed for any indication where an IUD was present, after excluding 13 patients who were determined to have expelled their IUD (13) and 13 patients with a perforated IUD, 156 patients (8.9%) were diagnosed as having a malpositioned IUD.¹⁶ IUD malposition was diagnosed when the IUD was in the uterus but positioned in the lower uterine segment, cervix, rotated or embedded in the uterus. An IUD in the lower uterine segment or cervix was detected in 133 patients, representing 85% of cases. Among these cases, 29 IUDs were also embedded and/or rotated, indicating that some IUDs have multiple causes of the malposition. Twenty-one IUDs were near the fundus but embedded and/or rotated. Controls with a normally-sited IUD were selected for comparison to the case group. Among IUD users, the identification of suspected adenomyosis on the ultrasound was associated with an increased risk of IUD malposition (odds ratio [OR], 3.04; 95% confidence interval [CI], 1.08-8.52).¹⁶ In this study, removal of a malpositioned LNG-IUD, without initiating a highly reliable contraceptive was associated with an increased risk of pregnancy. It is important to initiate a highly reliable form of contraception if the plan is to remove a malpositioned IUD.^16,19

In a study of 1,253 pelvic ultrasounds performed for any indication where an IUD was identified in the uterus, 263 IUDs (19%) were determined to be malpositioned.¹⁷ In this study the location of the malpositioned IUDs included¹⁷:

• the lower uterine segment not extending into the cervix (38%)
• in the lower uterine segment extending into the cervix (22%)
• in the cervix (26%)
• rotated axis of the IUD (12%)
• other (2%).

Among the 236 malpositioned IUDs, 24% appeared to be embedded in the uterine wall.¹⁷ Compared with patients with a normally-sited IUD on ultrasound, patients with a malpositioned IUD more frequently reported vaginal bleeding (30% vs 19%; P<.005) and pelvic pain (43% vs 30%; P<.002), similar to the findings in the Benacerraf et al. study.¹⁴

Connolly and Fox¹⁸ designed an innovative study to determine the rate of malpositioned IUDs using 2-D ultrasound to ensure proper IUD placement at the time of insertion with a follow-up 3-D ultrasound 8 weeks after insertion to assess IUD position within the uterus. At the 8-week 3-D ultrasound, among 763 women, 16.6% of the IUDs were malpositioned.¹⁸ In this study, IUD position was determined to be correct if all the following features were identified:

• the IUD shaft was in the midline of the uterine cavity
• the IUD arms were at 90 degrees from the stem
• the top of the IUD was within 3 to 4 mm of the fundus
• the IUD was not rotated, inverted or transverse.

IUD malpositions were categorized as:

• embedded in the uterine wall
• low in the uterine cavity
• in the endocervical canal
• misaligned
• perforated
• expulsed.

At the 8-week follow-up, 636 patients (83.4%) had an IUD that was correctly positioned.¹⁸ In 127 patients (16.6%) IUD malposition was identified, with some patients having more than one type of malposition. The types of malposition identified were:

• embedded in the myometrium (54%)
• misaligned, including rotated, laterally displaced, inverted, transverse or arms not deployed (47%)
• low in the uterine cavity (39%)
• in the endocervical canal (14%)
• perforated (3%)
• expulsion (0%).

Recall that all of these patients had a 2-D ultrasound at the time of insertion that identified the IUD as correctly placed. This suggests that during the 8 weeks following IUD placement there were changes in the location of the IUD or that 2-D ultrasound has lower sensitivity than 3-D ultrasound to detect malposition. Of note, at the 8-week follow-up, bleeding or pain was reported by 36% of the patients with a malpositioned IUD and 20% of patients with a correctly positioned IUD.¹⁷ Sixty-seven of the 127 malpositioned IUDs “required” removal, but the precise reasons for the removals were not delineated. The investigators concluded that 3-D ultrasonography is useful for the detection of IUD malposition and could be considered as part of ongoing IUD care, if symptoms of pain or bleeding occur.¹⁸

Continue to: IUD malposition following postplacental insertion...

IUD malposition following postplacental insertion

IUD malposition is common in patients who have had a postplacental insertion. Ultrasound imaging plays an important role in detecting IUD expulsion and malposition in these cases. Postplacental IUD insertion is defined as the placement of an IUD within 10 minutes following delivery of the placenta. Postplacental IUD insertion can be performed following a vaginal or cesarean birth and with a Cu-IUD or LNG-IUD. The good news is that postplacental IUD insertion reduces the risk of unplanned pregnancy in the years following birth. However, postplacental IUD insertion is associated with a high rate of IUD malposition.

In a study of 162 patients who had postplacental insertion of a Cu-IUD following a vaginal birth, ultrasound and physical examination at 6 months demonstrated complete IUD expulsion in 8%, partial expulsion in 16%, and malposition in 15%.²⁰ The IUD was correctly sited in 56% of patients. Seven patients (4%) had the IUD removed, and 1 patient had a perforated IUD. Among the 25 malpositioned IUDs, 14 were not within 1 cm of the fundus, and 11 were rotated outside of the axis of the cornuas. In this study partial expulsion was defined as an IUD protruding from the external cervical os on physical exam or demonstration of the distal tip of the IUD below the internal os of the cervix on ultrasound. Malposition was defined as an IUD that was >1 cm from the fundus or in an abnormal location or axis, but not partially expelled.

In a study of 69 patients who had postplacental insertion of a Cu-IUD following a cesarean birth, ultrasound and physical examination at 6 months demonstrated complete IUD expulsion in 3%, partial expulsion (stem in the cervix below the internal os) in 4% and malposition in 30%.²⁰ The IUD was correctly positioned in 59% of the patients.²¹ The IUD had been electively removed in 3%. Among the 21 patients with a malpositioned IUD, 10 were rotated within the uterine cavity, 6 were inverted (upside down), 3 were low-lying, and 2 were transverse.²¹ Given the relatively high rate of IUD malposition following postplacental insertion, it may be useful to perform a pelvic ultrasound at a postpartum visit to assess the location of the IUD, if ultrasonography is available.

Management of the malpositioned IUD

There are no consensus guidelines on how to care for a patient with a malpositioned IUD. Clinicians need to use their best judgment and engage the patient in joint decision making when managing a malpositioned IUD. When an IUD is malpositioned and the patient has bothersome symptoms of pelvic pain or abnormal bleeding that have not responded to standard interventions, consideration may be given to a remove and replace strategy. When the stem of the IUD is below the level of the internal os on ultrasound or visible at the external os on physical examination, consideration should be given to removing and replacing the IUD. However, if the IUD is removed without replacement or the initiation of a highly reliable contraceptive, the risk of unplanned pregnancy is considerable.^16,19

IUD totally or partially within the cervix or low-lying. When an IUD is in the cervix, the contraceptive efficacy of the IUD may be diminished, especially with a Cu-IUD.²² In these cases, removing and replacing the IUD is an option. In a survey of 20 expert clinicians, >80% recommended replacing an IUD that was totally or partially in the cervical canal.²³ But most of the experts would not replace an IUD that was incidentally noted on ultrasound to be low-lying, being positioned more than 2 cm below the fundus, with no portion of the IUD in the cervical canal. In the same survey, for patients with a low-lying IUD and pelvic pain or bleeding, the majority of experts reported that they would explore other causes of bleeding and pelvic pain not related to the IUD itself and not replace the IUD, but 30% of the experts reported that they would remove and replace the device.²³

IUD embedded in the myometrium with pelvic pain. Based on my clinical experience, when a patient has persistent pelvic pain following the insertion of an IUD and the pain does not resolve with standard measures including medication, an ultrasound study is warranted to assess the position of the IUD. If the ultrasound demonstrates that an arm of the IUD is embedded in the myometrium, removal of the IUD may be associated with resolution of the pain. Reinsertion of an IUD under ultrasound guidance may result in a correctly-sited IUD with no recurrence of pelvic pain.

IUD rotated within the uterus with no pain or abnormal bleeding. For an IUD that is near the fundus and rotated on its axis within the uterus, if the patient has no symptoms of pain or abnormal bleeding, my recommendation to the patient would be to leave the device in situ.

Without available guidelines, engage in clinician-patient discussion

It is clear that IUD malposition is common, occurring in 10% to 20% of patients with an IUD. High-quality ultrasound imaging is helpful in detecting IUD malposition, including 2-D ultrasound with videoclips and/or 3-D ultrasound with coronal reconstruction. More data are needed to identify the best options for managing various types of malpositioned IUDs in patients with and without bothersome symptoms such as pain and bleeding. Until consensus guidelines are developed, clinicians need to engage the patient in a discussion of how to best manage the malpositioned IUD. Medicated IUDs and progestin subdermal implants are our two most effective reversible contraceptives. They are among the most important advances in health care over the past half-century. ●

The medicated intrauterine devices (IUDs), including the levonorgestrel-releasing IUD (LNG-IUD) (Mirena, Kyleena, Skyla, and Liletta) and the copper IUD (Cu-IUD; Paragard), are remarkably effective contraceptives. For the 52-mg LNG-IUD (Mirena, Liletta) the pregnancy rate over 6 years of use averaged less than 0.2% per year.^1,2 For the Cu-IUD, the pregnancy rate over 10 years of use averaged 0.5% per year for the first 3 years of use and 0.2% per year over the following 7 years of use.³ IUD perforation of the uterus, expulsion, and malposition are recognized complications of IUD use. Our understanding of the prevalence and management of malpositioned IUDs is evolving and the main focus of this editorial.

Complete and partial uterus perforation

A complete uterine perforation occurs when the entire IUD is outside the walls of the uterus. A partial uterine perforation occurs when the IUD is outside the uterine cavity, but a portion of the IUD remains in the myometrium. When uterine perforation is suspected, ultrasound can determine if the IUD is properly sited within the uterus. If ultrasonography does not detect the IUD within the uterus, an x-ray of the pelvis and abdomen should be obtained to determine if the IUD is in the peritoneal cavity. If both an ultrasound and a pelvic-abdominal x-ray do not detect the IUD, the IUD was probably expelled from the patient.

Uterine perforation is uncommon and occurs once in every 500 to 1,000 insertions in non-breastfeeding women.^4-8 The most common symptoms reported by patients with a perforated IUD are pain and/or bleeding.⁸ Investigators in the European Active Surveillance Study on Intrauterine Devices (EURAS) enrolled more than 60,000 patients who had an IUD insertion and followed them for 12 months with more than 39,000 followed for up to 60 months.^7,8 The uterine perforation rate per 1,000 IUD insertions in non-breastfeeding women with 60 months of follow-up was 1.6 for the LNG-IUD and 0.8 for the Cu-IUD.⁸ The rate of uterine perforation was much higher in women who are breastfeeding or recently postpartum. In the EURAS study after 60 months of follow-up, the perforation rate per 1,000 insertions among breastfeeding women was 7.9 for the LNG-IUS and 4.7 for the Cu-IUD.⁸

Remarkably very few IUD perforations were detected at the time of insertion, including only 2% of the LNG-IUD insertions and 17% of the Cu-IUD insertions.⁸ Many perforations were not detected until more than 12 months following insertion, including 32% of the LNG-IUD insertions and 22% of the Cu-IUD insertions.⁸ Obviously, an IUD that has completely perforated the uterus and resides in the peritoneal cavity is not an effective contraceptive. For some patients, the IUD perforation was initially diagnosed after they became pregnant, and imaging studies to locate the IUD and assess the pregnancy were initiated. Complete perforation is usually treated with laparoscopy to remove the IUD and reduce the risk of injury to intra-abdominal organs.

Patients with an IUD partial perforation may present with pelvic pain or abnormal uterine bleeding.⁹ An ultrasound study to explore the cause of the presenting symptom may detect the partial perforation. It is estimated that approximately 20% of cases of IUD perforation are partial perforation.⁹ Over time, a partial perforation may progress to a complete perforation. In some cases of partial perforation, the IUD string may still be visible in the cervix, and the IUD may be removed by pulling on the strings.⁸ Hysteroscopy and/or laparoscopy may be needed to remove a partially perforated IUD. Following a partial or complete IUD perforation, if the patient desires to continue with IUD contraception, it would be wise to insert a new IUD under ultrasound guidance or assess proper placement with a postplacement ultrasound.

Continue to: Expulsion...

IUD expulsion occurs in approximately 3% to 11% of patients.^10-13 The age of the patient influences the rate of expulsion. In a study of 2,748 patients with a Cu-IUD, the rate of expulsion by age for patients <20 years, 20–24 years, 25–29 years, 30–34 years, and ≥35 years was 8.2%, 3.2%, 3.0%, 2.3%, and 1.8%, respectively.¹⁰ In this study, age did not influence the rate of IUD removal for pelvic pain or abnormal bleeding, which was 4% to 5% across all age groups.¹⁰ In a study of 5,403 patients with an IUD, the rate of IUD expulsion by age for patients <20 years, 20–29 years, and 30–45 years was 14.6%, 7.3%, and 7.2%, respectively.¹² In this study, the 3-year cumulative rate of expulsion was 10.2%.¹² There was no statistically significant difference in the 3-year cumulative rate of expulsion for the 52-mg LNG-IUD (10.1%) and Cu-IUD (10.7%).¹²

The majority of patients who have an IUD expulsion recognize the event and seek additional contraception care. A few patients first recognize the IUD expulsion when they become pregnant, and imaging studies detect no IUD in the uterus or the peritoneal cavity. In a study of more than 17,000 patients using an LNG-IUD, 108 pregnancies were reported. Seven pregnancies occurred in patients who did not realize their IUD was expelled.¹⁴ Patients who have had an IUD expulsion and receive a new IUD are at increased risk for re-expulsion. For these patients, reinsertion of an IUD could be performed under ultrasound guidance to ensure and document optimal initial IUD position within the uterus, or ultrasound can be obtained postinsertion to document appropriate IUD position.

Malposition—prevalence and management

Our understanding of the prevalence and management of a malpositioned IUD is evolving. For the purposes of this discussion a malpositioned IUD is defined as being in the uterus, but not properly positioned within the uterine cavity. Perforation into the peritoneal cavity and complete expulsion of an IUD are considered separate entities. However, a malpositioned IUD within the uterus may eventually perforate the uterus or be expelled from the body. For example, an IUD embedded in the uterine wall may eventually work its way through the wall and become perforated, residing in the peritoneal cavity. An IUD with the stem in the cervix below the internal os may eventually be expelled from the uterus and leave the body through the vagina.

High-quality ultrasonography, including 2-dimensional (2-D) ultrasound with videoclips or 3-dimensional (3-D) ultrasound with coronal views, has greatly advanced our understanding of the prevalence and characteristics of a malpositioned IUD.^15-18 Ultrasound features of an IUD correctly placed within the uterus include:

• the IUD is in the uterus
• the shaft is in the midline of the uterine cavity
• the shaft of the IUD is not in the endocervix
• the IUD arms are at a 90-degree angle from the shaft
• the top of the IUD is within 2 cm of the fundus
• the IUD is not rotated outside of the cornual plane, inverted or transverse.

Ultrasound imaging has identified multiple types of malpositioned IUDs, including:

• IUD embedded in the myometrium—a portion of the IUD is embedded in the uterine wall
• low-lying IUD—the IUD is low in the uterine cavity but not in the endocervix
• IUD in the endocervix—the stem is in the endocervical canal
• rotated—the IUD is rotated outside the cornual plane
• malpositioned arms—the arms are not at a 90-degree angle to the stem
• the IUD is inverted, transverse, or laterally displaced.

IUD malposition is highly prevalent and has been identified in 10% to 20% of convenience cohorts in which an ultrasound study was performed.^15-18

Benacerraf, Shipp, and Bromley were among the first experts to use ultrasound to detect the high prevalence of malpositioned IUDs among a convenience sample of 167 patients with an IUD undergoing ultrasound for a variety of indications. Using 3-D ultrasound, including reconstructed coronal views, they identified 28 patients (17%) with a malpositioned IUD based on the detection of the IUD “poking into the substance of the uterus or cervix.” Among the patients with a malpositioned IUD, the principal indication for the ultrasound study was pelvic pain (39%) or abnormal uterine bleeding (36%). Among women with a normally sited IUD, pelvic pain (19%) or abnormal uterine bleeding (15%) were less often the principal indication for the ultrasound.¹⁵ The malpositioned IUD was removed in 21 of the 28 cases and the symptoms of pelvic pain or abnormal bleeding resolved in 20 of the 21 patients.¹⁵

Other investigators have confirmed the observation that IUD malposition is common.^16-18 In a retrospective study of 1,748 pelvic ultrasounds performed for any indication where an IUD was present, after excluding 13 patients who were determined to have expelled their IUD (13) and 13 patients with a perforated IUD, 156 patients (8.9%) were diagnosed as having a malpositioned IUD.¹⁶ IUD malposition was diagnosed when the IUD was in the uterus but positioned in the lower uterine segment, cervix, rotated or embedded in the uterus. An IUD in the lower uterine segment or cervix was detected in 133 patients, representing 85% of cases. Among these cases, 29 IUDs were also embedded and/or rotated, indicating that some IUDs have multiple causes of the malposition. Twenty-one IUDs were near the fundus but embedded and/or rotated. Controls with a normally-sited IUD were selected for comparison to the case group. Among IUD users, the identification of suspected adenomyosis on the ultrasound was associated with an increased risk of IUD malposition (odds ratio [OR], 3.04; 95% confidence interval [CI], 1.08-8.52).¹⁶ In this study, removal of a malpositioned LNG-IUD, without initiating a highly reliable contraceptive was associated with an increased risk of pregnancy. It is important to initiate a highly reliable form of contraception if the plan is to remove a malpositioned IUD.^16,19

In a study of 1,253 pelvic ultrasounds performed for any indication where an IUD was identified in the uterus, 263 IUDs (19%) were determined to be malpositioned.¹⁷ In this study the location of the malpositioned IUDs included¹⁷:

• the lower uterine segment not extending into the cervix (38%)
• in the lower uterine segment extending into the cervix (22%)
• in the cervix (26%)
• rotated axis of the IUD (12%)
• other (2%).

Among the 236 malpositioned IUDs, 24% appeared to be embedded in the uterine wall.¹⁷ Compared with patients with a normally-sited IUD on ultrasound, patients with a malpositioned IUD more frequently reported vaginal bleeding (30% vs 19%; P<.005) and pelvic pain (43% vs 30%; P<.002), similar to the findings in the Benacerraf et al. study.¹⁴

Connolly and Fox¹⁸ designed an innovative study to determine the rate of malpositioned IUDs using 2-D ultrasound to ensure proper IUD placement at the time of insertion with a follow-up 3-D ultrasound 8 weeks after insertion to assess IUD position within the uterus. At the 8-week 3-D ultrasound, among 763 women, 16.6% of the IUDs were malpositioned.¹⁸ In this study, IUD position was determined to be correct if all the following features were identified:

• the IUD shaft was in the midline of the uterine cavity
• the IUD arms were at 90 degrees from the stem
• the top of the IUD was within 3 to 4 mm of the fundus
• the IUD was not rotated, inverted or transverse.

IUD malpositions were categorized as:

• embedded in the uterine wall
• low in the uterine cavity
• in the endocervical canal
• misaligned
• perforated
• expulsed.

At the 8-week follow-up, 636 patients (83.4%) had an IUD that was correctly positioned.¹⁸ In 127 patients (16.6%) IUD malposition was identified, with some patients having more than one type of malposition. The types of malposition identified were:

• embedded in the myometrium (54%)
• misaligned, including rotated, laterally displaced, inverted, transverse or arms not deployed (47%)
• low in the uterine cavity (39%)
• in the endocervical canal (14%)
• perforated (3%)
• expulsion (0%).

Recall that all of these patients had a 2-D ultrasound at the time of insertion that identified the IUD as correctly placed. This suggests that during the 8 weeks following IUD placement there were changes in the location of the IUD or that 2-D ultrasound has lower sensitivity than 3-D ultrasound to detect malposition. Of note, at the 8-week follow-up, bleeding or pain was reported by 36% of the patients with a malpositioned IUD and 20% of patients with a correctly positioned IUD.¹⁷ Sixty-seven of the 127 malpositioned IUDs “required” removal, but the precise reasons for the removals were not delineated. The investigators concluded that 3-D ultrasonography is useful for the detection of IUD malposition and could be considered as part of ongoing IUD care, if symptoms of pain or bleeding occur.¹⁸

Continue to: IUD malposition following postplacental insertion...

IUD malposition following postplacental insertion

IUD malposition is common in patients who have had a postplacental insertion. Ultrasound imaging plays an important role in detecting IUD expulsion and malposition in these cases. Postplacental IUD insertion is defined as the placement of an IUD within 10 minutes following delivery of the placenta. Postplacental IUD insertion can be performed following a vaginal or cesarean birth and with a Cu-IUD or LNG-IUD. The good news is that postplacental IUD insertion reduces the risk of unplanned pregnancy in the years following birth. However, postplacental IUD insertion is associated with a high rate of IUD malposition.

In a study of 162 patients who had postplacental insertion of a Cu-IUD following a vaginal birth, ultrasound and physical examination at 6 months demonstrated complete IUD expulsion in 8%, partial expulsion in 16%, and malposition in 15%.²⁰ The IUD was correctly sited in 56% of patients. Seven patients (4%) had the IUD removed, and 1 patient had a perforated IUD. Among the 25 malpositioned IUDs, 14 were not within 1 cm of the fundus, and 11 were rotated outside of the axis of the cornuas. In this study partial expulsion was defined as an IUD protruding from the external cervical os on physical exam or demonstration of the distal tip of the IUD below the internal os of the cervix on ultrasound. Malposition was defined as an IUD that was >1 cm from the fundus or in an abnormal location or axis, but not partially expelled.

In a study of 69 patients who had postplacental insertion of a Cu-IUD following a cesarean birth, ultrasound and physical examination at 6 months demonstrated complete IUD expulsion in 3%, partial expulsion (stem in the cervix below the internal os) in 4% and malposition in 30%.²⁰ The IUD was correctly positioned in 59% of the patients.²¹ The IUD had been electively removed in 3%. Among the 21 patients with a malpositioned IUD, 10 were rotated within the uterine cavity, 6 were inverted (upside down), 3 were low-lying, and 2 were transverse.²¹ Given the relatively high rate of IUD malposition following postplacental insertion, it may be useful to perform a pelvic ultrasound at a postpartum visit to assess the location of the IUD, if ultrasonography is available.

Management of the malpositioned IUD

There are no consensus guidelines on how to care for a patient with a malpositioned IUD. Clinicians need to use their best judgment and engage the patient in joint decision making when managing a malpositioned IUD. When an IUD is malpositioned and the patient has bothersome symptoms of pelvic pain or abnormal bleeding that have not responded to standard interventions, consideration may be given to a remove and replace strategy. When the stem of the IUD is below the level of the internal os on ultrasound or visible at the external os on physical examination, consideration should be given to removing and replacing the IUD. However, if the IUD is removed without replacement or the initiation of a highly reliable contraceptive, the risk of unplanned pregnancy is considerable.^16,19

IUD totally or partially within the cervix or low-lying. When an IUD is in the cervix, the contraceptive efficacy of the IUD may be diminished, especially with a Cu-IUD.²² In these cases, removing and replacing the IUD is an option. In a survey of 20 expert clinicians, >80% recommended replacing an IUD that was totally or partially in the cervical canal.²³ But most of the experts would not replace an IUD that was incidentally noted on ultrasound to be low-lying, being positioned more than 2 cm below the fundus, with no portion of the IUD in the cervical canal. In the same survey, for patients with a low-lying IUD and pelvic pain or bleeding, the majority of experts reported that they would explore other causes of bleeding and pelvic pain not related to the IUD itself and not replace the IUD, but 30% of the experts reported that they would remove and replace the device.²³

IUD embedded in the myometrium with pelvic pain. Based on my clinical experience, when a patient has persistent pelvic pain following the insertion of an IUD and the pain does not resolve with standard measures including medication, an ultrasound study is warranted to assess the position of the IUD. If the ultrasound demonstrates that an arm of the IUD is embedded in the myometrium, removal of the IUD may be associated with resolution of the pain. Reinsertion of an IUD under ultrasound guidance may result in a correctly-sited IUD with no recurrence of pelvic pain.

IUD rotated within the uterus with no pain or abnormal bleeding. For an IUD that is near the fundus and rotated on its axis within the uterus, if the patient has no symptoms of pain or abnormal bleeding, my recommendation to the patient would be to leave the device in situ.

Without available guidelines, engage in clinician-patient discussion

It is clear that IUD malposition is common, occurring in 10% to 20% of patients with an IUD. High-quality ultrasound imaging is helpful in detecting IUD malposition, including 2-D ultrasound with videoclips and/or 3-D ultrasound with coronal reconstruction. More data are needed to identify the best options for managing various types of malpositioned IUDs in patients with and without bothersome symptoms such as pain and bleeding. Until consensus guidelines are developed, clinicians need to engage the patient in a discussion of how to best manage the malpositioned IUD. Medicated IUDs and progestin subdermal implants are our two most effective reversible contraceptives. They are among the most important advances in health care over the past half-century. ●

UTIs IN PREGNANCY: MANAGING URETHRITIS, ASYMPTOMATIC BACTERIURIA, CYSTITIS, AND PYELONEPHRITIS

PATRICK DUFF, MD (JANUARY 2022)

Clarification on UTI issues

Regarding the article on urinary tract infections (UTIs) in pregnancy, I have 3 points of clarification. First, in 27 years of practice in which I universally performed screening urine cultures on prenatal patients plus all of those with symptoms, I have seen a total of 2 cultures with Staphylococcus saprophyticus. I see this organism listed in references as a major UTI causative, but is that the case? Second, the clinical case and symptoms discussed are accurate, but costovertebral angle tenderness or fever of 101 °F or higher indicate pyelonephritis and should be treated aggressively. Many of these patients will have nausea and vomiting and will be dehydrated. This decreases urine flow, allowing progressive bacterial growth in renal parenchyma. An initial bolus of intravenous fluids, at least 2 L wide open through a large-bore catheter, rapidly decreases fever, flushes the urinary tract, and improves nausea, headaches, and malaise. Finally, nitrofurantoin is excreted in the urine so rapidly that it does not achieve adequate tissue levels, and it should never be used to treat pyelonephritis or, for that matter, any infection other than uncomplicated cystitis/urethritis.

David Janowitz, MD

Houston, Texas

Dr. Duff responds

I appreciate Dr. Janowitz’s interest and thoughtful comments. The patient presented in the case study has acute cystitis, characterized by a low-grade fever, suprapubic pain, dysuria, frequency, and hesitancy. Patients with pyelonephritis typically have a higher fever and significant costovertebral angle pain and tenderness. I agree completely with Dr. Janowitz’s observations about the seriousness of pyelonephritis in pregnancy. Pyelonephritis is an important cause of preterm labor, bacteremia, and even septic shock. As I point out in the article, women with moderate to severe kidney infections should be hospitalized and treated with intravenous fluids, antipyretics, antiemetics, and intravenous antibiotics. My usual recommendation is ceftriaxone. Intravenous antibiotics should be continued until the patient has been afebrile and asymptomatic for 24 to 48 hours. Once patients improve, they can be transitioned to oral antibiotics to complete a 10-day course of therapy. Again, I agree with Dr. Janowitz’s statement that nitrofurantoin is not an appropriate drug for treatment of pyelonephritis because it does not reach acceptable concentrations in either the blood or the renal parenchyma. Rather, amoxicillin-clavulanate and trimethoprim-sulfamethoxazole are much better choices for oral therapy. However, once the infection is cleared, nitrofurantoin is an excellent agent for suppression of recurrent infection.

Finally, there is no doubt that the principal pathogens that cause UTIs in pregnant women are Escherichia coli, Klebsiella pneumoniae, and Proteus species. However, 3 aerobic Gram-positive cocci do, in fact, cause a small percentage of infections: group B streptococci, enterococci, and Staphylococcus saprophyticus. When the latter bacterium is identified as a single organism in high colony count, particularly in a catheterized urine specimen, it should be considered a true pathogen and not simply a contaminant.

CAN WE RETURN TO THE ABCs OF CRAFTING A MEDICAL RECORD NOTE?

ROBERT L. BARBIERI, MD (OCTOBER 2021)

Another suggestion for reducing note bloat in the EMR

Thank you for picking up a topic that is important for all physicians and one that has been annoying me since the introduction of electronic medical records (EMRs). I like the APSO approach, that works well. My idea for reducing “note bloat” is to eliminate all normal and routine findings and to hide them behind a hyperlink or behind a QR code. This would give you a truly short note and, should you need or want more details, you could always scan the QR code for access to the complete (and bloated) note. I would also recommend hiding all details that do not contribute to the immediate pressing issue at hand (for example, routine depression screening) behind a hyperlink or QR code. The same principle should apply to sending faxes to other physicians’ offices. I “love” receiving a chart an inch thick, only to discover that the whole pile of paper could be reduced to a single page of true information. Too few people speak up about this major time and productivity thief. Thank you!

Matthias Muenzer, MD

Rochester, New Hampshire

Dr. Barbieri responds

I thank Dr. Muenzer for his innovative suggestions for improving medical record notes. We spend many hours per week crafting notes in the medical record. Yet, very little attention is given to the development of best practices for improving the value and effectiveness of our notes for our patients and colleagues.

UTIs IN PREGNANCY: MANAGING URETHRITIS, ASYMPTOMATIC BACTERIURIA, CYSTITIS, AND PYELONEPHRITIS

PATRICK DUFF, MD (JANUARY 2022)

Clarification on UTI issues

Regarding the article on urinary tract infections (UTIs) in pregnancy, I have 3 points of clarification. First, in 27 years of practice in which I universally performed screening urine cultures on prenatal patients plus all of those with symptoms, I have seen a total of 2 cultures with Staphylococcus saprophyticus. I see this organism listed in references as a major UTI causative, but is that the case? Second, the clinical case and symptoms discussed are accurate, but costovertebral angle tenderness or fever of 101 °F or higher indicate pyelonephritis and should be treated aggressively. Many of these patients will have nausea and vomiting and will be dehydrated. This decreases urine flow, allowing progressive bacterial growth in renal parenchyma. An initial bolus of intravenous fluids, at least 2 L wide open through a large-bore catheter, rapidly decreases fever, flushes the urinary tract, and improves nausea, headaches, and malaise. Finally, nitrofurantoin is excreted in the urine so rapidly that it does not achieve adequate tissue levels, and it should never be used to treat pyelonephritis or, for that matter, any infection other than uncomplicated cystitis/urethritis.

David Janowitz, MD

Houston, Texas

Dr. Duff responds

I appreciate Dr. Janowitz’s interest and thoughtful comments. The patient presented in the case study has acute cystitis, characterized by a low-grade fever, suprapubic pain, dysuria, frequency, and hesitancy. Patients with pyelonephritis typically have a higher fever and significant costovertebral angle pain and tenderness. I agree completely with Dr. Janowitz’s observations about the seriousness of pyelonephritis in pregnancy. Pyelonephritis is an important cause of preterm labor, bacteremia, and even septic shock. As I point out in the article, women with moderate to severe kidney infections should be hospitalized and treated with intravenous fluids, antipyretics, antiemetics, and intravenous antibiotics. My usual recommendation is ceftriaxone. Intravenous antibiotics should be continued until the patient has been afebrile and asymptomatic for 24 to 48 hours. Once patients improve, they can be transitioned to oral antibiotics to complete a 10-day course of therapy. Again, I agree with Dr. Janowitz’s statement that nitrofurantoin is not an appropriate drug for treatment of pyelonephritis because it does not reach acceptable concentrations in either the blood or the renal parenchyma. Rather, amoxicillin-clavulanate and trimethoprim-sulfamethoxazole are much better choices for oral therapy. However, once the infection is cleared, nitrofurantoin is an excellent agent for suppression of recurrent infection.

Finally, there is no doubt that the principal pathogens that cause UTIs in pregnant women are Escherichia coli, Klebsiella pneumoniae, and Proteus species. However, 3 aerobic Gram-positive cocci do, in fact, cause a small percentage of infections: group B streptococci, enterococci, and Staphylococcus saprophyticus. When the latter bacterium is identified as a single organism in high colony count, particularly in a catheterized urine specimen, it should be considered a true pathogen and not simply a contaminant.

CAN WE RETURN TO THE ABCs OF CRAFTING A MEDICAL RECORD NOTE?

ROBERT L. BARBIERI, MD (OCTOBER 2021)

Another suggestion for reducing note bloat in the EMR

Thank you for picking up a topic that is important for all physicians and one that has been annoying me since the introduction of electronic medical records (EMRs). I like the APSO approach, that works well. My idea for reducing “note bloat” is to eliminate all normal and routine findings and to hide them behind a hyperlink or behind a QR code. This would give you a truly short note and, should you need or want more details, you could always scan the QR code for access to the complete (and bloated) note. I would also recommend hiding all details that do not contribute to the immediate pressing issue at hand (for example, routine depression screening) behind a hyperlink or QR code. The same principle should apply to sending faxes to other physicians’ offices. I “love” receiving a chart an inch thick, only to discover that the whole pile of paper could be reduced to a single page of true information. Too few people speak up about this major time and productivity thief. Thank you!

Matthias Muenzer, MD

Rochester, New Hampshire

Dr. Barbieri responds

I thank Dr. Muenzer for his innovative suggestions for improving medical record notes. We spend many hours per week crafting notes in the medical record. Yet, very little attention is given to the development of best practices for improving the value and effectiveness of our notes for our patients and colleagues.

UTIs IN PREGNANCY: MANAGING URETHRITIS, ASYMPTOMATIC BACTERIURIA, CYSTITIS, AND PYELONEPHRITIS

PATRICK DUFF, MD (JANUARY 2022)

Clarification on UTI issues

Regarding the article on urinary tract infections (UTIs) in pregnancy, I have 3 points of clarification. First, in 27 years of practice in which I universally performed screening urine cultures on prenatal patients plus all of those with symptoms, I have seen a total of 2 cultures with Staphylococcus saprophyticus. I see this organism listed in references as a major UTI causative, but is that the case? Second, the clinical case and symptoms discussed are accurate, but costovertebral angle tenderness or fever of 101 °F or higher indicate pyelonephritis and should be treated aggressively. Many of these patients will have nausea and vomiting and will be dehydrated. This decreases urine flow, allowing progressive bacterial growth in renal parenchyma. An initial bolus of intravenous fluids, at least 2 L wide open through a large-bore catheter, rapidly decreases fever, flushes the urinary tract, and improves nausea, headaches, and malaise. Finally, nitrofurantoin is excreted in the urine so rapidly that it does not achieve adequate tissue levels, and it should never be used to treat pyelonephritis or, for that matter, any infection other than uncomplicated cystitis/urethritis.

David Janowitz, MD

Houston, Texas

Dr. Duff responds

I appreciate Dr. Janowitz’s interest and thoughtful comments. The patient presented in the case study has acute cystitis, characterized by a low-grade fever, suprapubic pain, dysuria, frequency, and hesitancy. Patients with pyelonephritis typically have a higher fever and significant costovertebral angle pain and tenderness. I agree completely with Dr. Janowitz’s observations about the seriousness of pyelonephritis in pregnancy. Pyelonephritis is an important cause of preterm labor, bacteremia, and even septic shock. As I point out in the article, women with moderate to severe kidney infections should be hospitalized and treated with intravenous fluids, antipyretics, antiemetics, and intravenous antibiotics. My usual recommendation is ceftriaxone. Intravenous antibiotics should be continued until the patient has been afebrile and asymptomatic for 24 to 48 hours. Once patients improve, they can be transitioned to oral antibiotics to complete a 10-day course of therapy. Again, I agree with Dr. Janowitz’s statement that nitrofurantoin is not an appropriate drug for treatment of pyelonephritis because it does not reach acceptable concentrations in either the blood or the renal parenchyma. Rather, amoxicillin-clavulanate and trimethoprim-sulfamethoxazole are much better choices for oral therapy. However, once the infection is cleared, nitrofurantoin is an excellent agent for suppression of recurrent infection.

Finally, there is no doubt that the principal pathogens that cause UTIs in pregnant women are Escherichia coli, Klebsiella pneumoniae, and Proteus species. However, 3 aerobic Gram-positive cocci do, in fact, cause a small percentage of infections: group B streptococci, enterococci, and Staphylococcus saprophyticus. When the latter bacterium is identified as a single organism in high colony count, particularly in a catheterized urine specimen, it should be considered a true pathogen and not simply a contaminant.

CAN WE RETURN TO THE ABCs OF CRAFTING A MEDICAL RECORD NOTE?

ROBERT L. BARBIERI, MD (OCTOBER 2021)

Another suggestion for reducing note bloat in the EMR

Thank you for picking up a topic that is important for all physicians and one that has been annoying me since the introduction of electronic medical records (EMRs). I like the APSO approach, that works well. My idea for reducing “note bloat” is to eliminate all normal and routine findings and to hide them behind a hyperlink or behind a QR code. This would give you a truly short note and, should you need or want more details, you could always scan the QR code for access to the complete (and bloated) note. I would also recommend hiding all details that do not contribute to the immediate pressing issue at hand (for example, routine depression screening) behind a hyperlink or QR code. The same principle should apply to sending faxes to other physicians’ offices. I “love” receiving a chart an inch thick, only to discover that the whole pile of paper could be reduced to a single page of true information. Too few people speak up about this major time and productivity thief. Thank you!

Matthias Muenzer, MD

Rochester, New Hampshire

Dr. Barbieri responds

I thank Dr. Muenzer for his innovative suggestions for improving medical record notes. We spend many hours per week crafting notes in the medical record. Yet, very little attention is given to the development of best practices for improving the value and effectiveness of our notes for our patients and colleagues.

ILLUSTRATION BY KIMBERLY MARTENS FOR OBG MANAGEMENT

While its etiology remains uncertain, cesarean scar niche (CSN) is well publicized, as are its pathological clinical manifestations. In a future pregnancy, they include cesarean scar pregnancy (CSP), which in turn can lead to placenta accreta spectrum, and possible uterine rupture/dehiscence of a residual thin myometrial layer. CSP refers to the implantation of an early pregnancy on the scar or in the niche at the site of a prior cesarean delivery (CD); it has an incidence of 1 per 1,000 pregnancies. An estimated 52% of CSPs occur after even just one CD.¹ CSP has been linked to placenta accreta spectrum and has been shown to be its precursor.² Both CSP and placenta accreta spectrum can be consequences of CD and share a common histology of villous or placental attachment/invasion into the cesarean scar.³ The incidence of placenta accreta spectrum has risen from about 1 in 4,000 live births in the 1970s to 1 in 2,500 in the 1980s; in 2016, the incidence of placenta accreta spectrum was reported as 1 per 272 live births.⁴

Placenta accreta spectrum denotes the attachment of the placenta into and through the myometrium,⁵ and it can result in severe complications, including hemorrhage, hysterectomy, and intensive care treatment. The increasing rate of placenta accreta spectrum parallels the increasing CD rate, which rose from 5.8% in 1970 to 31.9% in 2016.⁶ Multiple repeat CDs are increasing in frequency as well. At the beginning of the century, placenta accreta spectrum mainly occurred after manual removal of the placenta, uterine curettage, or endometritis. Recently, experts are in agreement that the main determinant of placenta accreta spectrum is the uterine scar and niche formation after a previous CD.⁵ Larger niches are associated with an increased incidence of uterine rupture or dehiscence in a subsequent pregnancy.⁷

In the nonpregnant state, such niches are associated with intermenstrual bleeding, pelvic pain, painful intercourse, painful menses, and subfertility, becoming increasingly more severe in women with greater numbers of CDs.^8-10 Conception rate with assisted reproductive treatment is notably reduced.¹¹

Understanding its etiology

Monteagudo and colleagues first described a “niche” in 100% of 44 women evaluated for postmenopausal bleeding who had a prior CD.¹² CSN has been the subject of well over 3,000 publications over the past 30 years. While the topic generates much interest among researchers, it is garnering little traction among practicing obstetricians. Such “niches,” also referred to as isthmocele, cesarean scar defect, or a diverticulum, was first described in 1961¹³ and later defined on ultrasonography as a hypoechoic triangular-shaped uterine defect outlined by saline instillation sonohysterogram (SIS), reflecting a discontinuation of the myometrium at the site of a previous CD.¹² In 2019, a European task force further defined a CSN as an “indentation at the site in the cesarean section scar with a depth of at least 2 mm” and extended the classification to include branches as extensions toward the anterior uterine serosa.¹⁴ Using this criterion, sonographic postoperative evaluation after one CD revealed a CSN in 68.9% of women with one single-layer uterine closure and in 73.6% of women after a double-layer closure.¹⁵ Larger niche sizes with thinner residual myometrial thickness appeared more frequently when a single-layer closure technique was used, without closure of the peritoneum. Its prevalence varies from 56% to 84%.^16,17

Etiology of CSN formation: Our hypotheses

The precise pathophysiology of CSN remains elusive. Speculations attributed niche formation to numerous factors: timing of surgery, cervical incision, incomplete closure of the uterine incision, adhesion formation between the CD scar and the abdominal wall, and inherent maternal conditions which may impair healing, such as smoking, obesity, diabetes, maternal age, and labor status.^18-20 Retroflexion of the uterus is reportedly associated with increased incidence and size of the niche, with CSN 50% more likely to develop in women with a retroflexed versus an anteverted uterus.²¹ We demonstrated the origin of niche formation in real-time from the start to the completion of uterine closure by a video capture of a single-layer closure followed by an immediate SIS of the ex vivo hysterectomized uterus, and histopathologic proof of the presence of endometrial cells defining the “niche.”²² This case exposes the misalignment of the uterine wall, while including the endometrium in the closure (FIGURE 1). Similarly, pathologic studies of hysteroscopy-resected isthmocele ridges of symptomatic women with niche-related subfertility revealed the tissue edges lined by endocervical, endometrial, or isthmic mucosa either combined or isolated in the scar.²³ The presence of endometrial/cervical tissue in the myometrial closure has been debated for over a century.^24,25

Continue to: Uterine closure techniques...

In 1882, Max Sanger introduced a vertical uterine closure of a classical cesarean operation in response to hysterectomy as the contemporaneous alternative to prevent infection, bleeding, and death.²⁴ Dr. Sanger emphasized layer approximation, suturing, and the avoidance of decidua in the first layer (FIGURE 2). This became the teaching of the classical CD until the 1970s. In 1926, Munro Kerr addressed uterine rupture with labor after a classical CD by introducing the lower uterine segment transverse incision. He cautioned to maintain the decidua inside the uterine 2-layer closure of the cavity.²⁵ These pioneers were joined by others to rally for endometrium exclusion while promoting layer approximation. These techniques became universally standard and were taught across teaching medical centers in the United States and abroad until about 50 years ago.

In the 1970s, newer developments brought significant changes to uterine closure techniques. Initiated by Joel-Cohen,²⁶ blunt dissection of the abdominal incision was adapted by Michael Stark, creating what came to be known as the Misgav-Ladach cesarean technique.²⁷ Stark emphasized blunt dissection and introduced single-layer closure. Thereby the exclusion of the endometrium, used for more than 70 years, was abandoned by the present-day single- or double-layer uterine closure in favor of cost and time savings. Systematic reviews and meta-analyses comparing the two contrasting techniques were inconclusive, noting that the niche prevalence and size were similar in both groups. These studies did not take into account the variety of individual techniques or the position of the endometrium in the final closures.²⁸

Endometrium and uterine closure

Our recent study examining uterine scar defect in women after one primary CD by SIS concluded that a specific endometrium-free closure technique (EFCT) (FIGURE 3) is associated with fewer and less severe defects and a thicker residual myometrial thickness when compared with closures with unknown or endometrium inclusion.²⁹ The study found non-specific closure techniques to be 6 times more likely to form a niche of 2-mm deep or greater than the EFCT.

Furthermore, we surveyed the diversity of uterine closures and the location of the endometrium among obstetricians in one institution.³⁰ Presence of endometrium on the surface of the final uterine closure was reported by 20% of respondents (see Figure 1). When asked for their opinion on the impact of CD techniques on placenta accreta spectrum, without available evidence 80% of the survey respondents reported no relationship to techniques, and only 20% suggested an association. This particular study demonstrates that the surgical techniques just described are random, unfettered, and applied without consideration of clinical outcomes.

Our recent retrospective study that spanned 30 years and examined the EFCT—per­formed anywhere between 3 to 9 consecutive CDs—revealed no abnormal placentation in any subsequent pregnancies.³¹ This was one of the few clinical studies of the long-term consequences of a uterine closure technique. In this study, the endometrium was excluded during the uterine closure, allowing its free edges to abut and heal. This step avoids scarring the endometrial-myometrial (EM) interface and unintentional inclusion of endometrium in the closed uterine wall. In this context, Jauniaux and colleagues cited the destruction of the EM interface as the main factor for placenta-adherent disorders.³² Sholapurkar and others highlight the need to further examine intrinsic details of uterine closure beyond single- and double-layer techniques to better understand the etiology of cesarean scar formation.¹⁹ The search for the pathophysiology of CSN continues to present significant challenges imposed by the variety of currently practiced uterine closures.

Continue to: Focus on prevention...

Our research aims to address the endometrium, a specific layer that was the topic of concern in nascent CD techniques, as a renewed and contemporary one. The presence of the endometrium in ectopic locations or its destruction from intrauterine surgeries or infections has been implicated in abnormal placentation.^13,24 Our approach, in theory, is to limit the position of the endometrium to its innermost location and avoid its iatrogenic suturing and inclusion into the uterine wall closure. The rationale of sparing the endometrium in a layer-by-layer approximation is to allow for a closer restoration to normal anatomy and physiology than a random “en masse” uterine wall closure would permit. For this reason, the EM junction, the perimetrium, and the serosa must be identified and realigned for a more effective closure that incorporates the entire myometrial thickness. As evidence supports technical impact on the development of uterine scar defect in women after one CD, future studies are needed to evaluate uterine integrity by saline infusion sonohysterography in multiparous women with a prior random closure technique or a prior EFCT.

The potential long-term risks of blunt dissection for opening the uterus have not been studied. There are no physiologic lines in the uterine wall to facilitate a regular-bordered uterine stretch. The tissue stretch, which depends on the individual surgeon’s strength applied during the procedure and patient’s labor status, may result in an irregular tear and a difficult repair. The EFCT technique shows a more optimized risk-benefit ratio for an anatomical repair and is replicable. The safety of uterine layer re-approximation has been demonstrated and can be studied in large populations using strict uniform criteria.

Current and future challenges

Residency training

Most recently, teachers of resident trainees are mostly familiar with blunt dissection, techniques of which are passed on unchallenged from resident to resident. The endometrium and peritoneum are neither identified nor treated as separate layers, thus becoming obsolete as surgical and anatomical landmarks.

Standardization of CD techniques

Front-line obstetricians are persuaded to practice a standardized approach that relies on the benefits of cost related to operating room turnover as well as surgeons’ time savings without consideration of outcomes in subsequent pregnancies. Sholapurkar has warned that “wrong standardization” is far worse than no standardization, worse for the training of junior obstetricians, as it can inhibit critical reasoning about safe surgical techniques that can optimize outcomes of the condition of the lower uterine segment.³³

Emergence of cost and time savings in clinical practice

A time-cost savings argument is relatively negligeable in an estimated 40-minute CD. By contrast, deliberate surgical technique and carrying out the appropriate steps for the particular condition at hand to achieve the best outcomes assume more weight.³² Furthermore, this short-term cost benefit is challenged by the comparatively larger costs associated with the diagnosis, the treatment of post-CD adverse consequences (outlined above), as well as the emotional impact on women and their families. Additionally, the emphasis on time savings creates a generation of surgeons fixated with total operative time without consideration of long-term risks and adverse maternal outcomes.

Physician autonomy has led to the unmonitored freedom of obstetricians to choose their own technique for a CD, with some employing the commonly practiced culture of fastest turnaround even in nonurgent circumstances.

Documentation and terminology

Current documenting systems are not detail-oriented enough to assist in a thorough correlation between surgical techniques and outcomes. The use of single- or double-layer closure terminology is insufficient and has proven to be flawed, without describing the handling of the endometrium in terms of its inclusion or exclusion in the closure.

Quality improvement feedback

Long-term post-CD complications are often not reported to the physician or institution involved in the prior CD. In our opinion, some sort of registry would be of value. Perhaps then subsequent CD outcomes could be traced back and reported to the prior institution and surgeon. Feedback is critical to understanding the correlation between techniques and outcomes and more specifically to gathering learning points and using data for quality improvement of future cases.

Patient education

While women continue to have complications following the presently used surgical techniques, they often have expectations not discussed with their obstetricians. Women should be educated and empowered to realize the different approaches to all aspects and consequences of CDs.

Conclusion

The technique of excluding the endometrium in closing the uterine incision appears to reduce subsequent abnormal placentation and diminish the frequency and size of post-CD scar defect. The revival of the endometrium-free closure technique may allow significant change in the postoperative results. Currently, standardization of CD technique is being promoted on the basis of time- and cost-savings rather than clinical outcomes. Simultaneously, inroads are being made to better understand the risks and consequences of CD.

Emerging evidence suggests that a post-CD niche is the result of poor layer approximation as well as inclusion of the endometrium, which prevent healing of the uterine wall and often enables faulty implantation of the fertilized oocyte in the next pregnancy, potentially giving rise to placenta accreta spectrum. The prevalence and size of the defect can be minimized by techniques aimed at restoring the anatomy of the uterine wall and the physiology of the endometrium. Specialized training and education are necessary to stress the importance of anatomical assessment and decision making at the time of uterine closure. ●

ILLUSTRATION BY KIMBERLY MARTENS FOR OBG MANAGEMENT

While its etiology remains uncertain, cesarean scar niche (CSN) is well publicized, as are its pathological clinical manifestations. In a future pregnancy, they include cesarean scar pregnancy (CSP), which in turn can lead to placenta accreta spectrum, and possible uterine rupture/dehiscence of a residual thin myometrial layer. CSP refers to the implantation of an early pregnancy on the scar or in the niche at the site of a prior cesarean delivery (CD); it has an incidence of 1 per 1,000 pregnancies. An estimated 52% of CSPs occur after even just one CD.¹ CSP has been linked to placenta accreta spectrum and has been shown to be its precursor.² Both CSP and placenta accreta spectrum can be consequences of CD and share a common histology of villous or placental attachment/invasion into the cesarean scar.³ The incidence of placenta accreta spectrum has risen from about 1 in 4,000 live births in the 1970s to 1 in 2,500 in the 1980s; in 2016, the incidence of placenta accreta spectrum was reported as 1 per 272 live births.⁴

Placenta accreta spectrum denotes the attachment of the placenta into and through the myometrium,⁵ and it can result in severe complications, including hemorrhage, hysterectomy, and intensive care treatment. The increasing rate of placenta accreta spectrum parallels the increasing CD rate, which rose from 5.8% in 1970 to 31.9% in 2016.⁶ Multiple repeat CDs are increasing in frequency as well. At the beginning of the century, placenta accreta spectrum mainly occurred after manual removal of the placenta, uterine curettage, or endometritis. Recently, experts are in agreement that the main determinant of placenta accreta spectrum is the uterine scar and niche formation after a previous CD.⁵ Larger niches are associated with an increased incidence of uterine rupture or dehiscence in a subsequent pregnancy.⁷

In the nonpregnant state, such niches are associated with intermenstrual bleeding, pelvic pain, painful intercourse, painful menses, and subfertility, becoming increasingly more severe in women with greater numbers of CDs.^8-10 Conception rate with assisted reproductive treatment is notably reduced.¹¹

Understanding its etiology

Monteagudo and colleagues first described a “niche” in 100% of 44 women evaluated for postmenopausal bleeding who had a prior CD.¹² CSN has been the subject of well over 3,000 publications over the past 30 years. While the topic generates much interest among researchers, it is garnering little traction among practicing obstetricians. Such “niches,” also referred to as isthmocele, cesarean scar defect, or a diverticulum, was first described in 1961¹³ and later defined on ultrasonography as a hypoechoic triangular-shaped uterine defect outlined by saline instillation sonohysterogram (SIS), reflecting a discontinuation of the myometrium at the site of a previous CD.¹² In 2019, a European task force further defined a CSN as an “indentation at the site in the cesarean section scar with a depth of at least 2 mm” and extended the classification to include branches as extensions toward the anterior uterine serosa.¹⁴ Using this criterion, sonographic postoperative evaluation after one CD revealed a CSN in 68.9% of women with one single-layer uterine closure and in 73.6% of women after a double-layer closure.¹⁵ Larger niche sizes with thinner residual myometrial thickness appeared more frequently when a single-layer closure technique was used, without closure of the peritoneum. Its prevalence varies from 56% to 84%.^16,17

Etiology of CSN formation: Our hypotheses

The precise pathophysiology of CSN remains elusive. Speculations attributed niche formation to numerous factors: timing of surgery, cervical incision, incomplete closure of the uterine incision, adhesion formation between the CD scar and the abdominal wall, and inherent maternal conditions which may impair healing, such as smoking, obesity, diabetes, maternal age, and labor status.^18-20 Retroflexion of the uterus is reportedly associated with increased incidence and size of the niche, with CSN 50% more likely to develop in women with a retroflexed versus an anteverted uterus.²¹ We demonstrated the origin of niche formation in real-time from the start to the completion of uterine closure by a video capture of a single-layer closure followed by an immediate SIS of the ex vivo hysterectomized uterus, and histopathologic proof of the presence of endometrial cells defining the “niche.”²² This case exposes the misalignment of the uterine wall, while including the endometrium in the closure (FIGURE 1). Similarly, pathologic studies of hysteroscopy-resected isthmocele ridges of symptomatic women with niche-related subfertility revealed the tissue edges lined by endocervical, endometrial, or isthmic mucosa either combined or isolated in the scar.²³ The presence of endometrial/cervical tissue in the myometrial closure has been debated for over a century.^24,25

Continue to: Uterine closure techniques...

In 1882, Max Sanger introduced a vertical uterine closure of a classical cesarean operation in response to hysterectomy as the contemporaneous alternative to prevent infection, bleeding, and death.²⁴ Dr. Sanger emphasized layer approximation, suturing, and the avoidance of decidua in the first layer (FIGURE 2). This became the teaching of the classical CD until the 1970s. In 1926, Munro Kerr addressed uterine rupture with labor after a classical CD by introducing the lower uterine segment transverse incision. He cautioned to maintain the decidua inside the uterine 2-layer closure of the cavity.²⁵ These pioneers were joined by others to rally for endometrium exclusion while promoting layer approximation. These techniques became universally standard and were taught across teaching medical centers in the United States and abroad until about 50 years ago.

In the 1970s, newer developments brought significant changes to uterine closure techniques. Initiated by Joel-Cohen,²⁶ blunt dissection of the abdominal incision was adapted by Michael Stark, creating what came to be known as the Misgav-Ladach cesarean technique.²⁷ Stark emphasized blunt dissection and introduced single-layer closure. Thereby the exclusion of the endometrium, used for more than 70 years, was abandoned by the present-day single- or double-layer uterine closure in favor of cost and time savings. Systematic reviews and meta-analyses comparing the two contrasting techniques were inconclusive, noting that the niche prevalence and size were similar in both groups. These studies did not take into account the variety of individual techniques or the position of the endometrium in the final closures.²⁸

Endometrium and uterine closure

Our recent study examining uterine scar defect in women after one primary CD by SIS concluded that a specific endometrium-free closure technique (EFCT) (FIGURE 3) is associated with fewer and less severe defects and a thicker residual myometrial thickness when compared with closures with unknown or endometrium inclusion.²⁹ The study found non-specific closure techniques to be 6 times more likely to form a niche of 2-mm deep or greater than the EFCT.

Furthermore, we surveyed the diversity of uterine closures and the location of the endometrium among obstetricians in one institution.³⁰ Presence of endometrium on the surface of the final uterine closure was reported by 20% of respondents (see Figure 1). When asked for their opinion on the impact of CD techniques on placenta accreta spectrum, without available evidence 80% of the survey respondents reported no relationship to techniques, and only 20% suggested an association. This particular study demonstrates that the surgical techniques just described are random, unfettered, and applied without consideration of clinical outcomes.

Our recent retrospective study that spanned 30 years and examined the EFCT—per­formed anywhere between 3 to 9 consecutive CDs—revealed no abnormal placentation in any subsequent pregnancies.³¹ This was one of the few clinical studies of the long-term consequences of a uterine closure technique. In this study, the endometrium was excluded during the uterine closure, allowing its free edges to abut and heal. This step avoids scarring the endometrial-myometrial (EM) interface and unintentional inclusion of endometrium in the closed uterine wall. In this context, Jauniaux and colleagues cited the destruction of the EM interface as the main factor for placenta-adherent disorders.³² Sholapurkar and others highlight the need to further examine intrinsic details of uterine closure beyond single- and double-layer techniques to better understand the etiology of cesarean scar formation.¹⁹ The search for the pathophysiology of CSN continues to present significant challenges imposed by the variety of currently practiced uterine closures.

Continue to: Focus on prevention...

Our research aims to address the endometrium, a specific layer that was the topic of concern in nascent CD techniques, as a renewed and contemporary one. The presence of the endometrium in ectopic locations or its destruction from intrauterine surgeries or infections has been implicated in abnormal placentation.^13,24 Our approach, in theory, is to limit the position of the endometrium to its innermost location and avoid its iatrogenic suturing and inclusion into the uterine wall closure. The rationale of sparing the endometrium in a layer-by-layer approximation is to allow for a closer restoration to normal anatomy and physiology than a random “en masse” uterine wall closure would permit. For this reason, the EM junction, the perimetrium, and the serosa must be identified and realigned for a more effective closure that incorporates the entire myometrial thickness. As evidence supports technical impact on the development of uterine scar defect in women after one CD, future studies are needed to evaluate uterine integrity by saline infusion sonohysterography in multiparous women with a prior random closure technique or a prior EFCT.

The potential long-term risks of blunt dissection for opening the uterus have not been studied. There are no physiologic lines in the uterine wall to facilitate a regular-bordered uterine stretch. The tissue stretch, which depends on the individual surgeon’s strength applied during the procedure and patient’s labor status, may result in an irregular tear and a difficult repair. The EFCT technique shows a more optimized risk-benefit ratio for an anatomical repair and is replicable. The safety of uterine layer re-approximation has been demonstrated and can be studied in large populations using strict uniform criteria.

Current and future challenges

Residency training

Most recently, teachers of resident trainees are mostly familiar with blunt dissection, techniques of which are passed on unchallenged from resident to resident. The endometrium and peritoneum are neither identified nor treated as separate layers, thus becoming obsolete as surgical and anatomical landmarks.

Standardization of CD techniques

Front-line obstetricians are persuaded to practice a standardized approach that relies on the benefits of cost related to operating room turnover as well as surgeons’ time savings without consideration of outcomes in subsequent pregnancies. Sholapurkar has warned that “wrong standardization” is far worse than no standardization, worse for the training of junior obstetricians, as it can inhibit critical reasoning about safe surgical techniques that can optimize outcomes of the condition of the lower uterine segment.³³

Emergence of cost and time savings in clinical practice

A time-cost savings argument is relatively negligeable in an estimated 40-minute CD. By contrast, deliberate surgical technique and carrying out the appropriate steps for the particular condition at hand to achieve the best outcomes assume more weight.³² Furthermore, this short-term cost benefit is challenged by the comparatively larger costs associated with the diagnosis, the treatment of post-CD adverse consequences (outlined above), as well as the emotional impact on women and their families. Additionally, the emphasis on time savings creates a generation of surgeons fixated with total operative time without consideration of long-term risks and adverse maternal outcomes.

Physician autonomy has led to the unmonitored freedom of obstetricians to choose their own technique for a CD, with some employing the commonly practiced culture of fastest turnaround even in nonurgent circumstances.

Documentation and terminology

Current documenting systems are not detail-oriented enough to assist in a thorough correlation between surgical techniques and outcomes. The use of single- or double-layer closure terminology is insufficient and has proven to be flawed, without describing the handling of the endometrium in terms of its inclusion or exclusion in the closure.

Quality improvement feedback

Long-term post-CD complications are often not reported to the physician or institution involved in the prior CD. In our opinion, some sort of registry would be of value. Perhaps then subsequent CD outcomes could be traced back and reported to the prior institution and surgeon. Feedback is critical to understanding the correlation between techniques and outcomes and more specifically to gathering learning points and using data for quality improvement of future cases.

Patient education

While women continue to have complications following the presently used surgical techniques, they often have expectations not discussed with their obstetricians. Women should be educated and empowered to realize the different approaches to all aspects and consequences of CDs.

Conclusion

The technique of excluding the endometrium in closing the uterine incision appears to reduce subsequent abnormal placentation and diminish the frequency and size of post-CD scar defect. The revival of the endometrium-free closure technique may allow significant change in the postoperative results. Currently, standardization of CD technique is being promoted on the basis of time- and cost-savings rather than clinical outcomes. Simultaneously, inroads are being made to better understand the risks and consequences of CD.

Emerging evidence suggests that a post-CD niche is the result of poor layer approximation as well as inclusion of the endometrium, which prevent healing of the uterine wall and often enables faulty implantation of the fertilized oocyte in the next pregnancy, potentially giving rise to placenta accreta spectrum. The prevalence and size of the defect can be minimized by techniques aimed at restoring the anatomy of the uterine wall and the physiology of the endometrium. Specialized training and education are necessary to stress the importance of anatomical assessment and decision making at the time of uterine closure. ●

ILLUSTRATION BY KIMBERLY MARTENS FOR OBG MANAGEMENT

While its etiology remains uncertain, cesarean scar niche (CSN) is well publicized, as are its pathological clinical manifestations. In a future pregnancy, they include cesarean scar pregnancy (CSP), which in turn can lead to placenta accreta spectrum, and possible uterine rupture/dehiscence of a residual thin myometrial layer. CSP refers to the implantation of an early pregnancy on the scar or in the niche at the site of a prior cesarean delivery (CD); it has an incidence of 1 per 1,000 pregnancies. An estimated 52% of CSPs occur after even just one CD.¹ CSP has been linked to placenta accreta spectrum and has been shown to be its precursor.² Both CSP and placenta accreta spectrum can be consequences of CD and share a common histology of villous or placental attachment/invasion into the cesarean scar.³ The incidence of placenta accreta spectrum has risen from about 1 in 4,000 live births in the 1970s to 1 in 2,500 in the 1980s; in 2016, the incidence of placenta accreta spectrum was reported as 1 per 272 live births.⁴

Placenta accreta spectrum denotes the attachment of the placenta into and through the myometrium,⁵ and it can result in severe complications, including hemorrhage, hysterectomy, and intensive care treatment. The increasing rate of placenta accreta spectrum parallels the increasing CD rate, which rose from 5.8% in 1970 to 31.9% in 2016.⁶ Multiple repeat CDs are increasing in frequency as well. At the beginning of the century, placenta accreta spectrum mainly occurred after manual removal of the placenta, uterine curettage, or endometritis. Recently, experts are in agreement that the main determinant of placenta accreta spectrum is the uterine scar and niche formation after a previous CD.⁵ Larger niches are associated with an increased incidence of uterine rupture or dehiscence in a subsequent pregnancy.⁷

In the nonpregnant state, such niches are associated with intermenstrual bleeding, pelvic pain, painful intercourse, painful menses, and subfertility, becoming increasingly more severe in women with greater numbers of CDs.^8-10 Conception rate with assisted reproductive treatment is notably reduced.¹¹

Understanding its etiology

Monteagudo and colleagues first described a “niche” in 100% of 44 women evaluated for postmenopausal bleeding who had a prior CD.¹² CSN has been the subject of well over 3,000 publications over the past 30 years. While the topic generates much interest among researchers, it is garnering little traction among practicing obstetricians. Such “niches,” also referred to as isthmocele, cesarean scar defect, or a diverticulum, was first described in 1961¹³ and later defined on ultrasonography as a hypoechoic triangular-shaped uterine defect outlined by saline instillation sonohysterogram (SIS), reflecting a discontinuation of the myometrium at the site of a previous CD.¹² In 2019, a European task force further defined a CSN as an “indentation at the site in the cesarean section scar with a depth of at least 2 mm” and extended the classification to include branches as extensions toward the anterior uterine serosa.¹⁴ Using this criterion, sonographic postoperative evaluation after one CD revealed a CSN in 68.9% of women with one single-layer uterine closure and in 73.6% of women after a double-layer closure.¹⁵ Larger niche sizes with thinner residual myometrial thickness appeared more frequently when a single-layer closure technique was used, without closure of the peritoneum. Its prevalence varies from 56% to 84%.^16,17

Etiology of CSN formation: Our hypotheses

The precise pathophysiology of CSN remains elusive. Speculations attributed niche formation to numerous factors: timing of surgery, cervical incision, incomplete closure of the uterine incision, adhesion formation between the CD scar and the abdominal wall, and inherent maternal conditions which may impair healing, such as smoking, obesity, diabetes, maternal age, and labor status.^18-20 Retroflexion of the uterus is reportedly associated with increased incidence and size of the niche, with CSN 50% more likely to develop in women with a retroflexed versus an anteverted uterus.²¹ We demonstrated the origin of niche formation in real-time from the start to the completion of uterine closure by a video capture of a single-layer closure followed by an immediate SIS of the ex vivo hysterectomized uterus, and histopathologic proof of the presence of endometrial cells defining the “niche.”²² This case exposes the misalignment of the uterine wall, while including the endometrium in the closure (FIGURE 1). Similarly, pathologic studies of hysteroscopy-resected isthmocele ridges of symptomatic women with niche-related subfertility revealed the tissue edges lined by endocervical, endometrial, or isthmic mucosa either combined or isolated in the scar.²³ The presence of endometrial/cervical tissue in the myometrial closure has been debated for over a century.^24,25

Continue to: Uterine closure techniques...

In 1882, Max Sanger introduced a vertical uterine closure of a classical cesarean operation in response to hysterectomy as the contemporaneous alternative to prevent infection, bleeding, and death.²⁴ Dr. Sanger emphasized layer approximation, suturing, and the avoidance of decidua in the first layer (FIGURE 2). This became the teaching of the classical CD until the 1970s. In 1926, Munro Kerr addressed uterine rupture with labor after a classical CD by introducing the lower uterine segment transverse incision. He cautioned to maintain the decidua inside the uterine 2-layer closure of the cavity.²⁵ These pioneers were joined by others to rally for endometrium exclusion while promoting layer approximation. These techniques became universally standard and were taught across teaching medical centers in the United States and abroad until about 50 years ago.

In the 1970s, newer developments brought significant changes to uterine closure techniques. Initiated by Joel-Cohen,²⁶ blunt dissection of the abdominal incision was adapted by Michael Stark, creating what came to be known as the Misgav-Ladach cesarean technique.²⁷ Stark emphasized blunt dissection and introduced single-layer closure. Thereby the exclusion of the endometrium, used for more than 70 years, was abandoned by the present-day single- or double-layer uterine closure in favor of cost and time savings. Systematic reviews and meta-analyses comparing the two contrasting techniques were inconclusive, noting that the niche prevalence and size were similar in both groups. These studies did not take into account the variety of individual techniques or the position of the endometrium in the final closures.²⁸

Endometrium and uterine closure

Our recent study examining uterine scar defect in women after one primary CD by SIS concluded that a specific endometrium-free closure technique (EFCT) (FIGURE 3) is associated with fewer and less severe defects and a thicker residual myometrial thickness when compared with closures with unknown or endometrium inclusion.²⁹ The study found non-specific closure techniques to be 6 times more likely to form a niche of 2-mm deep or greater than the EFCT.

Furthermore, we surveyed the diversity of uterine closures and the location of the endometrium among obstetricians in one institution.³⁰ Presence of endometrium on the surface of the final uterine closure was reported by 20% of respondents (see Figure 1). When asked for their opinion on the impact of CD techniques on placenta accreta spectrum, without available evidence 80% of the survey respondents reported no relationship to techniques, and only 20% suggested an association. This particular study demonstrates that the surgical techniques just described are random, unfettered, and applied without consideration of clinical outcomes.

Our recent retrospective study that spanned 30 years and examined the EFCT—per­formed anywhere between 3 to 9 consecutive CDs—revealed no abnormal placentation in any subsequent pregnancies.³¹ This was one of the few clinical studies of the long-term consequences of a uterine closure technique. In this study, the endometrium was excluded during the uterine closure, allowing its free edges to abut and heal. This step avoids scarring the endometrial-myometrial (EM) interface and unintentional inclusion of endometrium in the closed uterine wall. In this context, Jauniaux and colleagues cited the destruction of the EM interface as the main factor for placenta-adherent disorders.³² Sholapurkar and others highlight the need to further examine intrinsic details of uterine closure beyond single- and double-layer techniques to better understand the etiology of cesarean scar formation.¹⁹ The search for the pathophysiology of CSN continues to present significant challenges imposed by the variety of currently practiced uterine closures.

Continue to: Focus on prevention...

Our research aims to address the endometrium, a specific layer that was the topic of concern in nascent CD techniques, as a renewed and contemporary one. The presence of the endometrium in ectopic locations or its destruction from intrauterine surgeries or infections has been implicated in abnormal placentation.^13,24 Our approach, in theory, is to limit the position of the endometrium to its innermost location and avoid its iatrogenic suturing and inclusion into the uterine wall closure. The rationale of sparing the endometrium in a layer-by-layer approximation is to allow for a closer restoration to normal anatomy and physiology than a random “en masse” uterine wall closure would permit. For this reason, the EM junction, the perimetrium, and the serosa must be identified and realigned for a more effective closure that incorporates the entire myometrial thickness. As evidence supports technical impact on the development of uterine scar defect in women after one CD, future studies are needed to evaluate uterine integrity by saline infusion sonohysterography in multiparous women with a prior random closure technique or a prior EFCT.

The potential long-term risks of blunt dissection for opening the uterus have not been studied. There are no physiologic lines in the uterine wall to facilitate a regular-bordered uterine stretch. The tissue stretch, which depends on the individual surgeon’s strength applied during the procedure and patient’s labor status, may result in an irregular tear and a difficult repair. The EFCT technique shows a more optimized risk-benefit ratio for an anatomical repair and is replicable. The safety of uterine layer re-approximation has been demonstrated and can be studied in large populations using strict uniform criteria.

Current and future challenges

Residency training

Most recently, teachers of resident trainees are mostly familiar with blunt dissection, techniques of which are passed on unchallenged from resident to resident. The endometrium and peritoneum are neither identified nor treated as separate layers, thus becoming obsolete as surgical and anatomical landmarks.

Standardization of CD techniques

Front-line obstetricians are persuaded to practice a standardized approach that relies on the benefits of cost related to operating room turnover as well as surgeons’ time savings without consideration of outcomes in subsequent pregnancies. Sholapurkar has warned that “wrong standardization” is far worse than no standardization, worse for the training of junior obstetricians, as it can inhibit critical reasoning about safe surgical techniques that can optimize outcomes of the condition of the lower uterine segment.³³

Emergence of cost and time savings in clinical practice

A time-cost savings argument is relatively negligeable in an estimated 40-minute CD. By contrast, deliberate surgical technique and carrying out the appropriate steps for the particular condition at hand to achieve the best outcomes assume more weight.³² Furthermore, this short-term cost benefit is challenged by the comparatively larger costs associated with the diagnosis, the treatment of post-CD adverse consequences (outlined above), as well as the emotional impact on women and their families. Additionally, the emphasis on time savings creates a generation of surgeons fixated with total operative time without consideration of long-term risks and adverse maternal outcomes.

Physician autonomy has led to the unmonitored freedom of obstetricians to choose their own technique for a CD, with some employing the commonly practiced culture of fastest turnaround even in nonurgent circumstances.

Documentation and terminology

Current documenting systems are not detail-oriented enough to assist in a thorough correlation between surgical techniques and outcomes. The use of single- or double-layer closure terminology is insufficient and has proven to be flawed, without describing the handling of the endometrium in terms of its inclusion or exclusion in the closure.

Quality improvement feedback

Long-term post-CD complications are often not reported to the physician or institution involved in the prior CD. In our opinion, some sort of registry would be of value. Perhaps then subsequent CD outcomes could be traced back and reported to the prior institution and surgeon. Feedback is critical to understanding the correlation between techniques and outcomes and more specifically to gathering learning points and using data for quality improvement of future cases.

Patient education

While women continue to have complications following the presently used surgical techniques, they often have expectations not discussed with their obstetricians. Women should be educated and empowered to realize the different approaches to all aspects and consequences of CDs.

Conclusion

The technique of excluding the endometrium in closing the uterine incision appears to reduce subsequent abnormal placentation and diminish the frequency and size of post-CD scar defect. The revival of the endometrium-free closure technique may allow significant change in the postoperative results. Currently, standardization of CD technique is being promoted on the basis of time- and cost-savings rather than clinical outcomes. Simultaneously, inroads are being made to better understand the risks and consequences of CD.

Emerging evidence suggests that a post-CD niche is the result of poor layer approximation as well as inclusion of the endometrium, which prevent healing of the uterine wall and often enables faulty implantation of the fertilized oocyte in the next pregnancy, potentially giving rise to placenta accreta spectrum. The prevalence and size of the defect can be minimized by techniques aimed at restoring the anatomy of the uterine wall and the physiology of the endometrium. Specialized training and education are necessary to stress the importance of anatomical assessment and decision making at the time of uterine closure. ●