Original Research

Respiratory diseases have varied clinical presentations and are classified as restrictive, obstructive, mixed, or normal. Restrictive lung diseases have reduced lung volumes, either due to an alteration in lung parenchyma or a disease of the pleura, chest wall, or neuromuscular apparatus. If caused by parenchymal lung disease, restrictive lung disorders are accompanied by reduced gas transfer, which may be portrayed clinically by desaturation after exercise. Based on anatomical structures, the causes of lung volume reduction may be intrinsic or extrinsic. Intrinsic causes correspond to diseases of the lung parenchyma, such as idiopathic fibrotic diseases, connective-tissue diseases, drug-induced lung diseases, and other primary diseases of the lungs. Extrinsic causes refer to disorders outside the lungs or extra-pulmonary diseases such as neuromuscular and nonmuscular diseases of the chest wall.¹ For example, obesity and myasthenia gravis can cause restrictive lung diseases, one through mechanical interference of lung expansion and the other through neuromuscular impedance of thoracic cage expansion. All these diseases eventually result in lung restriction, impaired lung function, and respiratory failure. This heterogenicity of disease makes establishing a single severity criterion difficult.

Laboratory testing, imaging studies, and examinations are important for determining the pulmonary disease and its course and progression. The pulmonary function test (PFT), which consists of multiple procedures that are performed depending on the information needed, has been an essential tool in practice for the pulmonologist. The PFT includes spirometry, lung volume measurement, respiratory muscle strength, diffusion capacity, and a broncho-provocation test. Each test has a particular role in assisting the diagnosis and/or follow-up of the patient. Spirometry is frequently used due to its range of dynamic physiological parameters, ease of use, and accessibility. It is used for the diagnosis of pulmonary symptoms, in the assessment of disability, and preoperatory evaluation, including lung resection surgery, assisting in the diagnosis, monitoring, and therapy response of pulmonary diseases.

A systematic approach to PFT interpretation is recommended by several societies, such as the American Thoracic Society (ATS) and the European Respiratory Society (ERS).² The pulmonary function test results must be reproducible and meet established standards to ensure reliable and consistent clinical outcomes. A restrictive respiratory disease is defined by a decrease in total lung capacity (TLC) (< 5% of predicted value) and a normal forced expiratory volume in 1 second (FEV₁)/forced vital capacity (FVC) ratio.² Although other findings—such as a decrease in vital capacity—should prompt an investigation into whether the patient has a possible restrictive respiratory disease, the sole presence of this parameter is not definitive or diagnostic of a restrictive impairment.^2-4 The assessment of severity is typically determined by TLC. Unfortunately, the severity of a restrictive respiratory disease and the degree of patient discomfort do not always correlate when utilizing just TLC. Pulmonary sarcoidosis, for example, is a granulomatous lung disease with a restrictive PFT pattern and a disease burden that may vary over time. Having a more consistent method of grading the severity of the restrictive lung disease may help guide treatment. The modified Medical Research Council (mMRC) scale, a 5-point dyspnea scale, is widely used in assessing the severity of dyspnea in various respiratory conditions, including chronic obstructive pulmonary disease (COPD), where its scores have been associated with patient mortality.^1,5 The goal of this study was to document the associations between objective parameters obtained through PFT and other variables, with an established measurement of dyspnea to assess the severity grade of restrictive lung diseases.

Methods

This retrospective record review at the Veterans Affairs Caribbean Healthcare System (VACHS) in San Juan, Puerto Rico, wasconducted using the Veterans Health Information Systems and Technology Architecture to identify patients with a PFT, including spirometry, that indicated a restrictive ventilator pattern based on the current ATS/ERS Task Force on Lung Function Testing.² Patients were included if they were aged ≥ 21 years, PFT with TLC ≤ 80% predicted, mMRC score documented on PFT, and documented diffusing capacity of the lung for carbon monoxide (DLCO). Patients were excluded if their FEV₁/vital capacity (VC) was < 70% predicted using the largest VC, or no mMRC score was available. All patients meeting the inclusion criteria were considered regardless of comorbidities.

The PFT results of all adult patients, including those performed between June 1, 2013, and January 6, 2016, were submitted to spirometry, and lung volume measurements were analyzed. Sociodemographic information was collected, including sex, ethnicity, age, height, weight, and basal metabolic index. Other data found in PFTs, such as smoking status, smoking in packs/year, mMRC score, predicted TLC value, imaging present (chest X-ray, computed tomography), and hospitalizations and exacerbations within 1 year were collected. In addition, we examined the predicted values for FEV₁, DLCO, and DLCO/VA (calculated using the Ayer equation), FVC (calculated using the Knudson equation), expiratory reserve volume, inspiratory VC, and slow VC. PaO₂, PaCO₂, and Alveolar-arterial gradients also were collected.^6-9 Information about heart failure status was gathered through medical evaluation of notes and cardiac studies. All categorical variables were correlated with Spearman analysis and quantitative variables with average percentages. P values were calculated with analysis of variance.

Of 6461 VACHS patient records reviewed, 415 met the inclusion criteria. Patients were divided according to their mMRC score: 65 had mMRC score of 0, 87 had an mMRC score of 1, 2 had an mMRC score of 2, 146 had an mMRC of 3, and 115 had an mMRC score of 4. The population was primarily male (98.6%) and of Hispanic ethnicity (96.4%), with a mean age of 72 years (Table 1). Most patients (n = 269, 64.0%) were prior smokers, while 135 patients (32.5%) had never smoked, and 11 (2.7%) were current smokers. At baseline, 169 patients (41.4%) had interstitial lung disease, 39 (9.6%) had chest wall disorders, 29 (7.1%) had occupational exposure, 25 (6.1%) had pneumonitis, and 14 (3.4%) had neuromuscular disorders.

There was a statistically significant relationship between mMRC score and hospitalization and FEV₁ but not TLC (Table 2). As mMRC increased, so did hospitalizations: a total of 168 patients (40.5%) were hospitalized; 24 patients (36.9%) had an mMRC score of 0, 30 patients (34.0%) had an mMRC score of 1, 2 patients (100%) had an mMRC score of 2, 54 patients (37.0%) had an mMRC score of 3, and 58 patients (50.0%) had an mMRC score of 4 (P = .04). Mean (SD) TLC values increased as mMRC scores increased. Mean (SD) TLC was 70.5% (33.0) for the entire population; 68.8% (7.2) for patients with an mMRC score of 0, 70.8% (5.8) for patients with an mMRC score of 1, 75.0% (1.4) for patients with an mMRC score of 2, 70.1% (7.2) for patients with an mMRC score of 3, and 71.5% (62.1) for patients with an mMRC score of 4 (P = .10) (Figure 1). There was an associated decrease in mean (SD) FEV₁ with mMRC. Mean (SD) FEV₁ was 76.2% (18.9) for the entire population; 81.7% (19.3) for patients with an mMRC score of 0, 80.9% (18) for patients with an mMRC score of 1, 93.5% (34.6) for patients with an mMRC score of 2, 76.2% (17.1) for patients with an mMRC score of 3, and 69.2% (19.4) for patients with an mMRC score of 4; (P < .001) (Figure 2).

The correlation between mMRC and FEV₁ (r = 0.25, P < .001) was stronger than the correlation between mMRC and TLC (r = 0.15, P < .001). The correlations for DLCO (P < .001), DLCO/VA (P < .001), hemoglobin (P < .02), and PaO₂ (P < .001) were all statistically significant (P < .005), but with no strong identifiable trend.

Discussion

The patient population of this study was primarily older males of Hispanic ethnicity with a history of smoking. There was no association between body mass index or smoking status with worsening dyspnea as measured with mMRC scores. We observed no significant correlation between mMRC scores and various factors such as comorbidities including heart conditions, and epidemiological factors like the etiology of lung disease, including both intrinsic and extrinsic causes. This lack of association was anticipated, as restrictive lung diseases in our study predominantly arose from intrinsic pulmonary etiologies, such as interstitial lung disease. A difference between more hospitalizations and worsening dyspnea was identified. There was a slightly higher correlation between FEV₁ and mMRC scores when compared with TLC and mMRC scores concerning worsening dyspnea, which could indicate that the use of FEV₁ should be preferred over previous recommendations to use TLC.¹⁰ Other guidelines have utilized exercise capacity via the 6-minute walk test as a marker of severity with spirometry values and found that DLCO was correlated with severity.¹¹

The latest ERS/ATS guidelines recommend z scores for grading the severity of obstructive lung diseases but do not recommend them for the diagnosis of restrictive lung diseases.¹² A z score encompasses diverse variables (eg, age, sex, and ethnicity) to provide more uniform and consistent results. Other studies have been done to relate z scores to other spirometry variables with restrictive lung disease. One such study indicates the potential benefit of using FVC alone to grade restrictive lung diseases.¹³ There continues to be great diversity in the interpretation of pulmonary function tests, and we believe the information gathered can provide valuable insight for managing patients with restrictive lung diseases.

Limitations

Only 2 patients reported an mMRC score of 2 in our study. This may have affected statistical outcomes. It also may reveal possible deficits in the efficacy of patient education on the mMRC scale. This study was also limited by its small sample size, single center location, and the distribution of patients that reported an mMRC favored either low or high values. The patients in this study, who were all veterans, may not be representative of other patient populations.

Conclusions

There continue to be few factors associated with the physiological severity of the defective oxygen delivery and reported dyspnea of a patient with restrictive lung disease that allows for an accurate, repeatable grading of severity. Using FEV₁ instead of TLC to determine the severity of a restrictive lung disease should be reconsidered. We could not find any other strong correlation among other factors studied. Further research should be conducted to continue looking for variables that more accurately depict patient dyspnea in restrictive lung disease.

Acknowledgments

This study is based upon work supported by the Veterans Affairs Caribbean Healthcare System in San Juan, Puerto Rico, and is the result of work supported by Pulmonary & Critical Care Medicine service, with resources and the use of its facilities.

Respiratory diseases have varied clinical presentations and are classified as restrictive, obstructive, mixed, or normal. Restrictive lung diseases have reduced lung volumes, either due to an alteration in lung parenchyma or a disease of the pleura, chest wall, or neuromuscular apparatus. If caused by parenchymal lung disease, restrictive lung disorders are accompanied by reduced gas transfer, which may be portrayed clinically by desaturation after exercise. Based on anatomical structures, the causes of lung volume reduction may be intrinsic or extrinsic. Intrinsic causes correspond to diseases of the lung parenchyma, such as idiopathic fibrotic diseases, connective-tissue diseases, drug-induced lung diseases, and other primary diseases of the lungs. Extrinsic causes refer to disorders outside the lungs or extra-pulmonary diseases such as neuromuscular and nonmuscular diseases of the chest wall.¹ For example, obesity and myasthenia gravis can cause restrictive lung diseases, one through mechanical interference of lung expansion and the other through neuromuscular impedance of thoracic cage expansion. All these diseases eventually result in lung restriction, impaired lung function, and respiratory failure. This heterogenicity of disease makes establishing a single severity criterion difficult.

Laboratory testing, imaging studies, and examinations are important for determining the pulmonary disease and its course and progression. The pulmonary function test (PFT), which consists of multiple procedures that are performed depending on the information needed, has been an essential tool in practice for the pulmonologist. The PFT includes spirometry, lung volume measurement, respiratory muscle strength, diffusion capacity, and a broncho-provocation test. Each test has a particular role in assisting the diagnosis and/or follow-up of the patient. Spirometry is frequently used due to its range of dynamic physiological parameters, ease of use, and accessibility. It is used for the diagnosis of pulmonary symptoms, in the assessment of disability, and preoperatory evaluation, including lung resection surgery, assisting in the diagnosis, monitoring, and therapy response of pulmonary diseases.

A systematic approach to PFT interpretation is recommended by several societies, such as the American Thoracic Society (ATS) and the European Respiratory Society (ERS).² The pulmonary function test results must be reproducible and meet established standards to ensure reliable and consistent clinical outcomes. A restrictive respiratory disease is defined by a decrease in total lung capacity (TLC) (< 5% of predicted value) and a normal forced expiratory volume in 1 second (FEV₁)/forced vital capacity (FVC) ratio.² Although other findings—such as a decrease in vital capacity—should prompt an investigation into whether the patient has a possible restrictive respiratory disease, the sole presence of this parameter is not definitive or diagnostic of a restrictive impairment.^2-4 The assessment of severity is typically determined by TLC. Unfortunately, the severity of a restrictive respiratory disease and the degree of patient discomfort do not always correlate when utilizing just TLC. Pulmonary sarcoidosis, for example, is a granulomatous lung disease with a restrictive PFT pattern and a disease burden that may vary over time. Having a more consistent method of grading the severity of the restrictive lung disease may help guide treatment. The modified Medical Research Council (mMRC) scale, a 5-point dyspnea scale, is widely used in assessing the severity of dyspnea in various respiratory conditions, including chronic obstructive pulmonary disease (COPD), where its scores have been associated with patient mortality.^1,5 The goal of this study was to document the associations between objective parameters obtained through PFT and other variables, with an established measurement of dyspnea to assess the severity grade of restrictive lung diseases.

Methods

This retrospective record review at the Veterans Affairs Caribbean Healthcare System (VACHS) in San Juan, Puerto Rico, wasconducted using the Veterans Health Information Systems and Technology Architecture to identify patients with a PFT, including spirometry, that indicated a restrictive ventilator pattern based on the current ATS/ERS Task Force on Lung Function Testing.² Patients were included if they were aged ≥ 21 years, PFT with TLC ≤ 80% predicted, mMRC score documented on PFT, and documented diffusing capacity of the lung for carbon monoxide (DLCO). Patients were excluded if their FEV₁/vital capacity (VC) was < 70% predicted using the largest VC, or no mMRC score was available. All patients meeting the inclusion criteria were considered regardless of comorbidities.

The PFT results of all adult patients, including those performed between June 1, 2013, and January 6, 2016, were submitted to spirometry, and lung volume measurements were analyzed. Sociodemographic information was collected, including sex, ethnicity, age, height, weight, and basal metabolic index. Other data found in PFTs, such as smoking status, smoking in packs/year, mMRC score, predicted TLC value, imaging present (chest X-ray, computed tomography), and hospitalizations and exacerbations within 1 year were collected. In addition, we examined the predicted values for FEV₁, DLCO, and DLCO/VA (calculated using the Ayer equation), FVC (calculated using the Knudson equation), expiratory reserve volume, inspiratory VC, and slow VC. PaO₂, PaCO₂, and Alveolar-arterial gradients also were collected.^6-9 Information about heart failure status was gathered through medical evaluation of notes and cardiac studies. All categorical variables were correlated with Spearman analysis and quantitative variables with average percentages. P values were calculated with analysis of variance.

Of 6461 VACHS patient records reviewed, 415 met the inclusion criteria. Patients were divided according to their mMRC score: 65 had mMRC score of 0, 87 had an mMRC score of 1, 2 had an mMRC score of 2, 146 had an mMRC of 3, and 115 had an mMRC score of 4. The population was primarily male (98.6%) and of Hispanic ethnicity (96.4%), with a mean age of 72 years (Table 1). Most patients (n = 269, 64.0%) were prior smokers, while 135 patients (32.5%) had never smoked, and 11 (2.7%) were current smokers. At baseline, 169 patients (41.4%) had interstitial lung disease, 39 (9.6%) had chest wall disorders, 29 (7.1%) had occupational exposure, 25 (6.1%) had pneumonitis, and 14 (3.4%) had neuromuscular disorders.

There was a statistically significant relationship between mMRC score and hospitalization and FEV₁ but not TLC (Table 2). As mMRC increased, so did hospitalizations: a total of 168 patients (40.5%) were hospitalized; 24 patients (36.9%) had an mMRC score of 0, 30 patients (34.0%) had an mMRC score of 1, 2 patients (100%) had an mMRC score of 2, 54 patients (37.0%) had an mMRC score of 3, and 58 patients (50.0%) had an mMRC score of 4 (P = .04). Mean (SD) TLC values increased as mMRC scores increased. Mean (SD) TLC was 70.5% (33.0) for the entire population; 68.8% (7.2) for patients with an mMRC score of 0, 70.8% (5.8) for patients with an mMRC score of 1, 75.0% (1.4) for patients with an mMRC score of 2, 70.1% (7.2) for patients with an mMRC score of 3, and 71.5% (62.1) for patients with an mMRC score of 4 (P = .10) (Figure 1). There was an associated decrease in mean (SD) FEV₁ with mMRC. Mean (SD) FEV₁ was 76.2% (18.9) for the entire population; 81.7% (19.3) for patients with an mMRC score of 0, 80.9% (18) for patients with an mMRC score of 1, 93.5% (34.6) for patients with an mMRC score of 2, 76.2% (17.1) for patients with an mMRC score of 3, and 69.2% (19.4) for patients with an mMRC score of 4; (P < .001) (Figure 2).

The correlation between mMRC and FEV₁ (r = 0.25, P < .001) was stronger than the correlation between mMRC and TLC (r = 0.15, P < .001). The correlations for DLCO (P < .001), DLCO/VA (P < .001), hemoglobin (P < .02), and PaO₂ (P < .001) were all statistically significant (P < .005), but with no strong identifiable trend.

Discussion

The patient population of this study was primarily older males of Hispanic ethnicity with a history of smoking. There was no association between body mass index or smoking status with worsening dyspnea as measured with mMRC scores. We observed no significant correlation between mMRC scores and various factors such as comorbidities including heart conditions, and epidemiological factors like the etiology of lung disease, including both intrinsic and extrinsic causes. This lack of association was anticipated, as restrictive lung diseases in our study predominantly arose from intrinsic pulmonary etiologies, such as interstitial lung disease. A difference between more hospitalizations and worsening dyspnea was identified. There was a slightly higher correlation between FEV₁ and mMRC scores when compared with TLC and mMRC scores concerning worsening dyspnea, which could indicate that the use of FEV₁ should be preferred over previous recommendations to use TLC.¹⁰ Other guidelines have utilized exercise capacity via the 6-minute walk test as a marker of severity with spirometry values and found that DLCO was correlated with severity.¹¹

The latest ERS/ATS guidelines recommend z scores for grading the severity of obstructive lung diseases but do not recommend them for the diagnosis of restrictive lung diseases.¹² A z score encompasses diverse variables (eg, age, sex, and ethnicity) to provide more uniform and consistent results. Other studies have been done to relate z scores to other spirometry variables with restrictive lung disease. One such study indicates the potential benefit of using FVC alone to grade restrictive lung diseases.¹³ There continues to be great diversity in the interpretation of pulmonary function tests, and we believe the information gathered can provide valuable insight for managing patients with restrictive lung diseases.

Limitations

Only 2 patients reported an mMRC score of 2 in our study. This may have affected statistical outcomes. It also may reveal possible deficits in the efficacy of patient education on the mMRC scale. This study was also limited by its small sample size, single center location, and the distribution of patients that reported an mMRC favored either low or high values. The patients in this study, who were all veterans, may not be representative of other patient populations.

Conclusions

There continue to be few factors associated with the physiological severity of the defective oxygen delivery and reported dyspnea of a patient with restrictive lung disease that allows for an accurate, repeatable grading of severity. Using FEV₁ instead of TLC to determine the severity of a restrictive lung disease should be reconsidered. We could not find any other strong correlation among other factors studied. Further research should be conducted to continue looking for variables that more accurately depict patient dyspnea in restrictive lung disease.

Acknowledgments

This study is based upon work supported by the Veterans Affairs Caribbean Healthcare System in San Juan, Puerto Rico, and is the result of work supported by Pulmonary & Critical Care Medicine service, with resources and the use of its facilities.

Respiratory diseases have varied clinical presentations and are classified as restrictive, obstructive, mixed, or normal. Restrictive lung diseases have reduced lung volumes, either due to an alteration in lung parenchyma or a disease of the pleura, chest wall, or neuromuscular apparatus. If caused by parenchymal lung disease, restrictive lung disorders are accompanied by reduced gas transfer, which may be portrayed clinically by desaturation after exercise. Based on anatomical structures, the causes of lung volume reduction may be intrinsic or extrinsic. Intrinsic causes correspond to diseases of the lung parenchyma, such as idiopathic fibrotic diseases, connective-tissue diseases, drug-induced lung diseases, and other primary diseases of the lungs. Extrinsic causes refer to disorders outside the lungs or extra-pulmonary diseases such as neuromuscular and nonmuscular diseases of the chest wall.¹ For example, obesity and myasthenia gravis can cause restrictive lung diseases, one through mechanical interference of lung expansion and the other through neuromuscular impedance of thoracic cage expansion. All these diseases eventually result in lung restriction, impaired lung function, and respiratory failure. This heterogenicity of disease makes establishing a single severity criterion difficult.

Laboratory testing, imaging studies, and examinations are important for determining the pulmonary disease and its course and progression. The pulmonary function test (PFT), which consists of multiple procedures that are performed depending on the information needed, has been an essential tool in practice for the pulmonologist. The PFT includes spirometry, lung volume measurement, respiratory muscle strength, diffusion capacity, and a broncho-provocation test. Each test has a particular role in assisting the diagnosis and/or follow-up of the patient. Spirometry is frequently used due to its range of dynamic physiological parameters, ease of use, and accessibility. It is used for the diagnosis of pulmonary symptoms, in the assessment of disability, and preoperatory evaluation, including lung resection surgery, assisting in the diagnosis, monitoring, and therapy response of pulmonary diseases.

A systematic approach to PFT interpretation is recommended by several societies, such as the American Thoracic Society (ATS) and the European Respiratory Society (ERS).² The pulmonary function test results must be reproducible and meet established standards to ensure reliable and consistent clinical outcomes. A restrictive respiratory disease is defined by a decrease in total lung capacity (TLC) (< 5% of predicted value) and a normal forced expiratory volume in 1 second (FEV₁)/forced vital capacity (FVC) ratio.² Although other findings—such as a decrease in vital capacity—should prompt an investigation into whether the patient has a possible restrictive respiratory disease, the sole presence of this parameter is not definitive or diagnostic of a restrictive impairment.^2-4 The assessment of severity is typically determined by TLC. Unfortunately, the severity of a restrictive respiratory disease and the degree of patient discomfort do not always correlate when utilizing just TLC. Pulmonary sarcoidosis, for example, is a granulomatous lung disease with a restrictive PFT pattern and a disease burden that may vary over time. Having a more consistent method of grading the severity of the restrictive lung disease may help guide treatment. The modified Medical Research Council (mMRC) scale, a 5-point dyspnea scale, is widely used in assessing the severity of dyspnea in various respiratory conditions, including chronic obstructive pulmonary disease (COPD), where its scores have been associated with patient mortality.^1,5 The goal of this study was to document the associations between objective parameters obtained through PFT and other variables, with an established measurement of dyspnea to assess the severity grade of restrictive lung diseases.

Methods

This retrospective record review at the Veterans Affairs Caribbean Healthcare System (VACHS) in San Juan, Puerto Rico, wasconducted using the Veterans Health Information Systems and Technology Architecture to identify patients with a PFT, including spirometry, that indicated a restrictive ventilator pattern based on the current ATS/ERS Task Force on Lung Function Testing.² Patients were included if they were aged ≥ 21 years, PFT with TLC ≤ 80% predicted, mMRC score documented on PFT, and documented diffusing capacity of the lung for carbon monoxide (DLCO). Patients were excluded if their FEV₁/vital capacity (VC) was < 70% predicted using the largest VC, or no mMRC score was available. All patients meeting the inclusion criteria were considered regardless of comorbidities.

The PFT results of all adult patients, including those performed between June 1, 2013, and January 6, 2016, were submitted to spirometry, and lung volume measurements were analyzed. Sociodemographic information was collected, including sex, ethnicity, age, height, weight, and basal metabolic index. Other data found in PFTs, such as smoking status, smoking in packs/year, mMRC score, predicted TLC value, imaging present (chest X-ray, computed tomography), and hospitalizations and exacerbations within 1 year were collected. In addition, we examined the predicted values for FEV₁, DLCO, and DLCO/VA (calculated using the Ayer equation), FVC (calculated using the Knudson equation), expiratory reserve volume, inspiratory VC, and slow VC. PaO₂, PaCO₂, and Alveolar-arterial gradients also were collected.^6-9 Information about heart failure status was gathered through medical evaluation of notes and cardiac studies. All categorical variables were correlated with Spearman analysis and quantitative variables with average percentages. P values were calculated with analysis of variance.

Of 6461 VACHS patient records reviewed, 415 met the inclusion criteria. Patients were divided according to their mMRC score: 65 had mMRC score of 0, 87 had an mMRC score of 1, 2 had an mMRC score of 2, 146 had an mMRC of 3, and 115 had an mMRC score of 4. The population was primarily male (98.6%) and of Hispanic ethnicity (96.4%), with a mean age of 72 years (Table 1). Most patients (n = 269, 64.0%) were prior smokers, while 135 patients (32.5%) had never smoked, and 11 (2.7%) were current smokers. At baseline, 169 patients (41.4%) had interstitial lung disease, 39 (9.6%) had chest wall disorders, 29 (7.1%) had occupational exposure, 25 (6.1%) had pneumonitis, and 14 (3.4%) had neuromuscular disorders.

There was a statistically significant relationship between mMRC score and hospitalization and FEV₁ but not TLC (Table 2). As mMRC increased, so did hospitalizations: a total of 168 patients (40.5%) were hospitalized; 24 patients (36.9%) had an mMRC score of 0, 30 patients (34.0%) had an mMRC score of 1, 2 patients (100%) had an mMRC score of 2, 54 patients (37.0%) had an mMRC score of 3, and 58 patients (50.0%) had an mMRC score of 4 (P = .04). Mean (SD) TLC values increased as mMRC scores increased. Mean (SD) TLC was 70.5% (33.0) for the entire population; 68.8% (7.2) for patients with an mMRC score of 0, 70.8% (5.8) for patients with an mMRC score of 1, 75.0% (1.4) for patients with an mMRC score of 2, 70.1% (7.2) for patients with an mMRC score of 3, and 71.5% (62.1) for patients with an mMRC score of 4 (P = .10) (Figure 1). There was an associated decrease in mean (SD) FEV₁ with mMRC. Mean (SD) FEV₁ was 76.2% (18.9) for the entire population; 81.7% (19.3) for patients with an mMRC score of 0, 80.9% (18) for patients with an mMRC score of 1, 93.5% (34.6) for patients with an mMRC score of 2, 76.2% (17.1) for patients with an mMRC score of 3, and 69.2% (19.4) for patients with an mMRC score of 4; (P < .001) (Figure 2).

The correlation between mMRC and FEV₁ (r = 0.25, P < .001) was stronger than the correlation between mMRC and TLC (r = 0.15, P < .001). The correlations for DLCO (P < .001), DLCO/VA (P < .001), hemoglobin (P < .02), and PaO₂ (P < .001) were all statistically significant (P < .005), but with no strong identifiable trend.

Discussion

The patient population of this study was primarily older males of Hispanic ethnicity with a history of smoking. There was no association between body mass index or smoking status with worsening dyspnea as measured with mMRC scores. We observed no significant correlation between mMRC scores and various factors such as comorbidities including heart conditions, and epidemiological factors like the etiology of lung disease, including both intrinsic and extrinsic causes. This lack of association was anticipated, as restrictive lung diseases in our study predominantly arose from intrinsic pulmonary etiologies, such as interstitial lung disease. A difference between more hospitalizations and worsening dyspnea was identified. There was a slightly higher correlation between FEV₁ and mMRC scores when compared with TLC and mMRC scores concerning worsening dyspnea, which could indicate that the use of FEV₁ should be preferred over previous recommendations to use TLC.¹⁰ Other guidelines have utilized exercise capacity via the 6-minute walk test as a marker of severity with spirometry values and found that DLCO was correlated with severity.¹¹

The latest ERS/ATS guidelines recommend z scores for grading the severity of obstructive lung diseases but do not recommend them for the diagnosis of restrictive lung diseases.¹² A z score encompasses diverse variables (eg, age, sex, and ethnicity) to provide more uniform and consistent results. Other studies have been done to relate z scores to other spirometry variables with restrictive lung disease. One such study indicates the potential benefit of using FVC alone to grade restrictive lung diseases.¹³ There continues to be great diversity in the interpretation of pulmonary function tests, and we believe the information gathered can provide valuable insight for managing patients with restrictive lung diseases.

Limitations

Only 2 patients reported an mMRC score of 2 in our study. This may have affected statistical outcomes. It also may reveal possible deficits in the efficacy of patient education on the mMRC scale. This study was also limited by its small sample size, single center location, and the distribution of patients that reported an mMRC favored either low or high values. The patients in this study, who were all veterans, may not be representative of other patient populations.

Conclusions

There continue to be few factors associated with the physiological severity of the defective oxygen delivery and reported dyspnea of a patient with restrictive lung disease that allows for an accurate, repeatable grading of severity. Using FEV₁ instead of TLC to determine the severity of a restrictive lung disease should be reconsidered. We could not find any other strong correlation among other factors studied. Further research should be conducted to continue looking for variables that more accurately depict patient dyspnea in restrictive lung disease.

Acknowledgments

This study is based upon work supported by the Veterans Affairs Caribbean Healthcare System in San Juan, Puerto Rico, and is the result of work supported by Pulmonary & Critical Care Medicine service, with resources and the use of its facilities.

The Centers for Disease Control and Prevention (CDC) classifies individuals with a body mass index (BMI) of 25 to 29.9as overweight and those with a BMI > 30 as obese (obesity classes: I, BMI 30 to 34.9; II, BMI 35 to 39.9; and III, BMI ≥ 40).¹ In 2011, the CDC estimated that 27.4% of adults in the United States were obese; less than a decade later, that number increased to 31.9%.¹ In that same period, the percentage of adults in Indiana classified as obese increased from 30.8% to 36.8%.¹ About 1 in 14 individuals in the US have class III obesity and 86% of veterans are either overweight or obese.²

High medical expenses can likely be attributed to the long-term health consequences of obesity. Compared to those with a healthy weight, individuals who are overweight or obese are at an increased risk for high blood pressure, high low-density lipoprotein cholesterol levels, low high-density lipoprotein cholesterol levels, high triglyceride levels, type 2 diabetes mellitus (T2DM), coronary heart disease, stroke, gallbladder disease, osteoarthritis, sleep apnea, cancer, mental health disorders, body pain, low quality of life, and death.³ Many of these conditions lead to increased health care needs, medication needs, hospitalizations, and overall health care system use.

Guidelines for the prevention and treatment of obesity have been produced by the American Heart Association, American College of Cardiology, and The Obesity Society; the Endocrine Society; the American Diabetes Association; and the US Departments of Veterans Affairs (VA) and Defense. Each follows a general algorithm to manage and prevent adverse effects (AEs) related to obesity. General practice is to assess a patient for elevated BMI (> 25), implement intense lifestyle modifications including calorie restriction and exercise, reassess for a maintained 5% to 10% weight loss for cardiovascular benefits, and potentially assess for pharmacological or surgical intervention to assist in weight loss.^2,4-6

While some weight loss medications (eg, phentermine/topiramate, naltrexone/bupropion, orlistat, and lorcaserin) tend to have unfavorable AEs or mixed efficacy, glucagon-like peptide-1 receptor agonists (GLP-1RAs) have provided new options.^7-10 Lorcaserin, for example, was removed from the market in 2020 due to its association with cancer risks.¹¹ The GLP-1RAs liraglutide and semaglutide received US Food and Drug Administration (FDA) approval for weight loss in 2014 and 2021, respectively.^12,13 GLP-1RAs have shown the greatest efficacy and benefits in reducing hemoglobin A_1c (HbA_1c); they are the preferred agents for patients who qualify for pharmacologic intervention for weight loss, especially those with T2DM. However, these studies have not evaluated the long-term outcomes of using these medications for weight loss and may not reflect the veteran population.^14,15

At Veteran Health Indiana (VHI), clinicians may use several weight loss medications for patients to achieve 5% to 10% weight loss. The medications most often used include liraglutide, phentermine/topiramate, naltrexone/bupropion, orlistat, and phentermine alone. However, more research is needed to determine which weight loss medication is the most beneficial for veterans, particularly following FDA approval of GLP-1RAs. At VHI, phentermine/topiramate is the preferred first-line agent unless patients have contraindications for use, in which case naltrexone/bupropion is recommended. These are considered first-line due to their ease of use in pill form, lower cost, and comparable weight loss to the GLP-1 medication class.² However, for patients with prediabetes, T2DM, BMI > 40, or BMI > 35 with specific comorbid conditions, liraglutide is preferred because of its beneficial effects for both weight loss and blood glucose control.²

This study aimed to expand on the 2021 Hood and colleagues study that examined total weight loss and weight loss as a percentage of baseline weight in patients with obesity at 3, 6, 12, and > 12 months of pharmacologic therapy by extending the time frame to 48 months.¹⁶ This study excluded semaglutide because few patients were prescribed the medication for weight loss during the study.

METHODS

We conducted a single-center, retrospective chart review of patients prescribed weight loss medications at VHI. A patient list was generated based on prescription fills from June 1, 2017, to July 31, 2021. Data were obtained from the Computerized Patient Record System; patients were not contacted. This study was approved by the Indiana University Health Institutional Review Board and VHI Research and Development Committee.

At the time of this study, liraglutide, phentermine/topiramate, naltrexone/bupropion, orlistat, and phentermine alone were available at VHI for patients who met the clinical criteria for use. All patients must have been enrolled in dietary and lifestyle management programs, including the VA MOVE! program, to be approved for these medications. After the MOVE! orientation, patients could participate in group or individual 12-week programs that included weigh-ins, goal-setting strategies, meal planning, and habit modification support. If patients could not meet in person, phone and other telehealth opportunities were available.

Patients were included in the study if they were aged ≥ 18 years, received a prescription for any of the 5 available medications for weight loss during the enrollment period, and were on the medication for ≥ 6 consecutive months. Patients were excluded if they received a prescription, were treated outside the VA system, or were pregnant. The primary indication for the included medication was not weight loss; the primary indication for the GLP-1RA was T2DM, or the weight loss was attributed to another disease. Adherence was not a measured outcome of this study; if patients were filling the medication, it was assumed they were taking it. Data were collected for each instance of medication use; as a result, a few patients were included more than once. Data collection for a failed medication ended when failure was documented. New data points began when new medication was prescribed; all data were per medication, not per patient. This allowed us to account for medication failure and provide accurate weight loss results based on medication choice within VHI.

Primary outcomes included total weight loss and weight loss as a percentage ofbaseline weight during the study period at 3, 6, 12, 24, 36, and 48 months of therapy. Secondary outcomes included the percentage of patients who lost 5% to 10% of their body weight from baseline; the percentage of patients who maintained ≥ 5% weight loss from baseline to 12, 24, 36, and 48 months if maintained on medication for that duration; duration of medication treatment in weeks; medication discontinuation rate; reason for medication discontinuation; enrollment in the MOVE! clinic and the time enrolled; percentage of patients with a BMI of 18 to 24.9 at the end of the study; and change in HbA_1c at 3, 6, 12, 24, 36, and 48 months.

Demographic data included race, age, sex, baseline weight, height, baseline BMI, and comorbid conditions (collected based on the most recent primary care clinical note before initiating medication). Medication data collected included medications used to manage comorbidities. Data related to weight management medication included prescribing clinic, maintenance dose of medication, duration of medication during the study period, the reason for medication discontinuation, or bariatric surgery intervention if applicable.

Basic descriptive statistics were used to characterize study participants. For continuous data, analysis of variance tests were used; if those results were not normal, then nonparametric tests were used, followed by pairwise tests between medication groups if the overall test was significant using the Fisher significant differences test. For nominal data, χ² or Fisher exact tests were used. For comparisons of primary and secondary outcomes, if the analyses needed to include adjustment for confounding variables, analysis of covariance was used for continuous data. A 2-sided 5% significance level was used for all tests.

RESULTS

A total of 228 instances of medication use were identified based on prescription fills; 123 did not meet inclusion criteria (117 for < 6 consecutive months of medication use) (Figure). The study included 105 participants with a mean age of 56 years; 80 were male (76.2%), and 85 identified as White race (81.0%). Mean (SD) weight was 130.1 kg (26.8) and BMI was 41.6 (7.2). The most common comorbid disease states among patients included hypertension, dyslipidemia, obstructive sleep apnea, and T2DM (Table 1). The baseline characteristics were comparable to those of Hood and colleagues.¹⁶

Most patients at VHI started on liraglutide (63%) or phentermine/topiramate (28%). For primary and secondary outcomes, statistics were calculated to determine whether the results were statistically significant for comparing the liraglutide and phentermine/topiramate subgroups. Sample sizes were too small for statistical analysis for bupropion/naltrexone, phentermine, and orlistat.

Primary Outcomes

The mean (SD) weight of participants dropped 8.1% from 130.1 kg to 119.5 kg over the patient-specific duration of weight management medication therapy for an absolute difference of 10.6 kg (9.7). Duration of individual medication use varied from 6 to 48 months. Weight loss was recorded at 6, 12, 24, 36, and 48 months of weight management therapy. Patient weight was not recorded after the medication was discontinued.

When classified by medication choice, the mean change in weight over the duration of the study was −23.9 kg for 2 patients using orlistat, −10.2 kg for 46 patients using liraglutide, −11.0 kg for 25 patients using phentermine/topiramate, -7.4 kg for 1 patient using phentermine, and -13.0 kg for 4 patients using naltrexone/bupropion. Patients without a weight documented at the end of their therapy or at the conclusion of the data collection period were not included in the total weight loss at the end of therapy. There were 78 documented instances of weight loss at the end of therapy (Table 2).

Body weight loss percentage was recorded at 6, 12, 24, 36, and 48 months of weight management therapy. The mean (SD) body weight loss percentage over the duration of the study was 9.2% (11.2). When classified by medication choice, the mean percentage of body weight loss was 16.8% for 2 patients using orlistat, 9.4% for 46 patients using liraglutide, 8.2% for 25 patients using phentermine/topiramate, 6.0% for 1 patient using phentermine alone, and 10.6% for 4 patients using naltrexone/bupropion (Table 3).

Secondary Outcomes

While none of the secondary outcomes were statistically significant, the results of this study suggest that both medications may contribute to weight loss in many patients included in this study. Almost two-thirds of the included patients analyzed lost ≥ 5% of weight from baseline while taking weight management medication. Sixty-six patients (63%) lost ≥ 5% of body weight at any time during the data collection period. When stratified by liraglutide and phentermine/topiramate, 41 patients (63%) taking liraglutide and 20 patients (67%) taking phentermine/topiramate lost ≥ 5% of weight from baseline. Of the 66 patients who lost ≥ 5% of body weight from baseline, 36 (55%) lost ≥ 10% of body weight from baseline at any time during the data collection period.

The mean (SD) duration for weight management medication use was 23 months (14.9). Phentermine/topiramate was tolerated longer than liraglutide: 22.7 months vs 21.7 months, respectively (Table 4).

The average overall documented medication discontinuation rate was 35.2%. Reasons for discontinuation included 21 patient-elected discontinuations, 8 patients no longer met criteria for use, 4 medications were no longer indicated, and 4 patients experienced AEs. It is unknown whether weight management medication was discontinued or not in 18 patients (17.2%).

This study evaluated the use and outcomes of weight loss medications over a longer period (up to 48 months) than what was previously studied among patients at VHI (12 months). The study aimed to better understand the long-term effect of weight loss medications, determine which medication had better long-term outcomes, and examine the reasons for medication discontinuation.

The results of this study displayed some similarities and differences compared with the Hood and colleagues study.¹⁶ Both yielded similar results for 5% of body weight loss and 10% of body weight loss. The largest difference was mean weight loss over the study period. In this study, patients lost a mean 10.6 kg over the course of weight loss medication use compared to 15.8 kg found by Hood and colleagues.¹⁶ A reason patients in the current study lost less weight overall could be the difference in time frames. The current study encompassed the COVID-19 pandemic, meaning fewer overall in-person patient appointments, which led to patients being lost to follow-up, missing weigh-ins during the time period, and gaps in care. For some patients, the pandemic possibly contributed to depression, missed medication doses, and a more sedentary lifestyle, leading to more weight gain.¹⁷ Telemedicine services at VHI expanded during the pandemic in an attempt to increase patient monitoring and counseling. It is unclear whether this expansion was enough to replace the in-person contact necessary to promote a healthy lifestyle.

VA pharmacists now care for patients through telehealth and are more involved in weight loss management. Since the conclusion of the Hood and colleagues study and start of this research, 2 pharmacists at VHI have been assigned to follow patients for obesity management to help with adherence to medication and lifestyle changes, management of AEs, dispense logistics, interventions for medications that may cause weight gain, and case management of glycemic control and weight loss with GLP-1RAs. Care management by pharmacists at VHI helps improve the logistics of titratable orders and save money by improving the use of high-cost items like GLP-1RAs. VA clinical pharmacy practitioners already monitor GLP-1RAs for patients with T2DM, so they are prepared to educate and assist patients with these medications.

It is important to continue developing a standardized process for weight loss medication management across the VA to improve the quality of patient care and optimize prescription outcomes. VA facilities differ in how weight loss management care is delivered and the level at which pharmacists are involved. Given the high rate of obesity among patients at the VA, the advent of new prescription options for weight loss, and the high cost associated with these medications, there has been increased attention to obesity care. Some Veterans Integrated Service Networks are forming a weight management community of practice groups to create standard operating procedures and algorithms to standardize care. Developing consistent processes is necessary to improve weight loss and patient care for veterans regardless where they receive treatment.

Limitations

The data used in this study were dependent on clinician documentation. Because of a lack of documentation in many instances, it was difficult to determine the full efficacy of the medications studied due to missing weight recordings. The lack of documentation made it difficult to determine whether patients were enrolled and active in the MOVE! program. It is required that patients enroll in MOVE! to obtain medications, but many did not have any follow-up MOVE! visits after initially obtaining their weight loss medication.

In this study, differences in the outcomes of patients with and without T2DM were not compared. It is the VA standard of care to prefer liraglutide over phentermine/topiramate in patients with T2DM or prediabetes.² This makes it difficult to assess whether phentermine/topiramate or liraglutide is more effective for weight loss in patients with T2DM. Weight gain after the discontinuation of weight loss medications was not assessed. Collecting this data may help determine whether a certain weight loss medication is less likely to cause rebound weight gain when discontinued.

Other limitations to this study consisted of excluding patients who discontinued therapy within 6 months, small sample sizes on some medications, and lack of data on adherence. Adherence was based on medication refills, which means that if a patient refilled the medication, it was assumed they were taking it. This is not always the case, and while accurate data on adherence is difficult to gather, it can impact how results may be interpreted. These additional limitations make it difficult to accurately determine the efficacy of the medications in this study.

CONCLUSIONS

This study found similar outcomes to what has been observed in larger clinical trials regarding weight loss medications. Nevertheless, there was a lack of accurate clinical documentation for most patients, which limits the conclusions. This lack of documentation potentially led to inaccurate results. It revealed that many patients at VHI did not uniformly receive consistent follow-up after starting a weight loss medication during the study period. With more standardized processes implemented at VA facilities, increased pharmacist involvement in weight loss medication management, and increased use of established telehealth services, patients could have the opportunity for closer follow-up that may lead to better weight loss outcomes. With these changes, there is more reason for additional studies to be conducted to assess follow-up, medication management, and weight loss overall.

The Centers for Disease Control and Prevention (CDC) classifies individuals with a body mass index (BMI) of 25 to 29.9as overweight and those with a BMI > 30 as obese (obesity classes: I, BMI 30 to 34.9; II, BMI 35 to 39.9; and III, BMI ≥ 40).¹ In 2011, the CDC estimated that 27.4% of adults in the United States were obese; less than a decade later, that number increased to 31.9%.¹ In that same period, the percentage of adults in Indiana classified as obese increased from 30.8% to 36.8%.¹ About 1 in 14 individuals in the US have class III obesity and 86% of veterans are either overweight or obese.²

High medical expenses can likely be attributed to the long-term health consequences of obesity. Compared to those with a healthy weight, individuals who are overweight or obese are at an increased risk for high blood pressure, high low-density lipoprotein cholesterol levels, low high-density lipoprotein cholesterol levels, high triglyceride levels, type 2 diabetes mellitus (T2DM), coronary heart disease, stroke, gallbladder disease, osteoarthritis, sleep apnea, cancer, mental health disorders, body pain, low quality of life, and death.³ Many of these conditions lead to increased health care needs, medication needs, hospitalizations, and overall health care system use.

Guidelines for the prevention and treatment of obesity have been produced by the American Heart Association, American College of Cardiology, and The Obesity Society; the Endocrine Society; the American Diabetes Association; and the US Departments of Veterans Affairs (VA) and Defense. Each follows a general algorithm to manage and prevent adverse effects (AEs) related to obesity. General practice is to assess a patient for elevated BMI (> 25), implement intense lifestyle modifications including calorie restriction and exercise, reassess for a maintained 5% to 10% weight loss for cardiovascular benefits, and potentially assess for pharmacological or surgical intervention to assist in weight loss.^2,4-6

While some weight loss medications (eg, phentermine/topiramate, naltrexone/bupropion, orlistat, and lorcaserin) tend to have unfavorable AEs or mixed efficacy, glucagon-like peptide-1 receptor agonists (GLP-1RAs) have provided new options.^7-10 Lorcaserin, for example, was removed from the market in 2020 due to its association with cancer risks.¹¹ The GLP-1RAs liraglutide and semaglutide received US Food and Drug Administration (FDA) approval for weight loss in 2014 and 2021, respectively.^12,13 GLP-1RAs have shown the greatest efficacy and benefits in reducing hemoglobin A_1c (HbA_1c); they are the preferred agents for patients who qualify for pharmacologic intervention for weight loss, especially those with T2DM. However, these studies have not evaluated the long-term outcomes of using these medications for weight loss and may not reflect the veteran population.^14,15

At Veteran Health Indiana (VHI), clinicians may use several weight loss medications for patients to achieve 5% to 10% weight loss. The medications most often used include liraglutide, phentermine/topiramate, naltrexone/bupropion, orlistat, and phentermine alone. However, more research is needed to determine which weight loss medication is the most beneficial for veterans, particularly following FDA approval of GLP-1RAs. At VHI, phentermine/topiramate is the preferred first-line agent unless patients have contraindications for use, in which case naltrexone/bupropion is recommended. These are considered first-line due to their ease of use in pill form, lower cost, and comparable weight loss to the GLP-1 medication class.² However, for patients with prediabetes, T2DM, BMI > 40, or BMI > 35 with specific comorbid conditions, liraglutide is preferred because of its beneficial effects for both weight loss and blood glucose control.²

This study aimed to expand on the 2021 Hood and colleagues study that examined total weight loss and weight loss as a percentage of baseline weight in patients with obesity at 3, 6, 12, and > 12 months of pharmacologic therapy by extending the time frame to 48 months.¹⁶ This study excluded semaglutide because few patients were prescribed the medication for weight loss during the study.

METHODS

We conducted a single-center, retrospective chart review of patients prescribed weight loss medications at VHI. A patient list was generated based on prescription fills from June 1, 2017, to July 31, 2021. Data were obtained from the Computerized Patient Record System; patients were not contacted. This study was approved by the Indiana University Health Institutional Review Board and VHI Research and Development Committee.

At the time of this study, liraglutide, phentermine/topiramate, naltrexone/bupropion, orlistat, and phentermine alone were available at VHI for patients who met the clinical criteria for use. All patients must have been enrolled in dietary and lifestyle management programs, including the VA MOVE! program, to be approved for these medications. After the MOVE! orientation, patients could participate in group or individual 12-week programs that included weigh-ins, goal-setting strategies, meal planning, and habit modification support. If patients could not meet in person, phone and other telehealth opportunities were available.

Patients were included in the study if they were aged ≥ 18 years, received a prescription for any of the 5 available medications for weight loss during the enrollment period, and were on the medication for ≥ 6 consecutive months. Patients were excluded if they received a prescription, were treated outside the VA system, or were pregnant. The primary indication for the included medication was not weight loss; the primary indication for the GLP-1RA was T2DM, or the weight loss was attributed to another disease. Adherence was not a measured outcome of this study; if patients were filling the medication, it was assumed they were taking it. Data were collected for each instance of medication use; as a result, a few patients were included more than once. Data collection for a failed medication ended when failure was documented. New data points began when new medication was prescribed; all data were per medication, not per patient. This allowed us to account for medication failure and provide accurate weight loss results based on medication choice within VHI.

Primary outcomes included total weight loss and weight loss as a percentage ofbaseline weight during the study period at 3, 6, 12, 24, 36, and 48 months of therapy. Secondary outcomes included the percentage of patients who lost 5% to 10% of their body weight from baseline; the percentage of patients who maintained ≥ 5% weight loss from baseline to 12, 24, 36, and 48 months if maintained on medication for that duration; duration of medication treatment in weeks; medication discontinuation rate; reason for medication discontinuation; enrollment in the MOVE! clinic and the time enrolled; percentage of patients with a BMI of 18 to 24.9 at the end of the study; and change in HbA_1c at 3, 6, 12, 24, 36, and 48 months.

Demographic data included race, age, sex, baseline weight, height, baseline BMI, and comorbid conditions (collected based on the most recent primary care clinical note before initiating medication). Medication data collected included medications used to manage comorbidities. Data related to weight management medication included prescribing clinic, maintenance dose of medication, duration of medication during the study period, the reason for medication discontinuation, or bariatric surgery intervention if applicable.

Basic descriptive statistics were used to characterize study participants. For continuous data, analysis of variance tests were used; if those results were not normal, then nonparametric tests were used, followed by pairwise tests between medication groups if the overall test was significant using the Fisher significant differences test. For nominal data, χ² or Fisher exact tests were used. For comparisons of primary and secondary outcomes, if the analyses needed to include adjustment for confounding variables, analysis of covariance was used for continuous data. A 2-sided 5% significance level was used for all tests.

RESULTS

A total of 228 instances of medication use were identified based on prescription fills; 123 did not meet inclusion criteria (117 for < 6 consecutive months of medication use) (Figure). The study included 105 participants with a mean age of 56 years; 80 were male (76.2%), and 85 identified as White race (81.0%). Mean (SD) weight was 130.1 kg (26.8) and BMI was 41.6 (7.2). The most common comorbid disease states among patients included hypertension, dyslipidemia, obstructive sleep apnea, and T2DM (Table 1). The baseline characteristics were comparable to those of Hood and colleagues.¹⁶

Most patients at VHI started on liraglutide (63%) or phentermine/topiramate (28%). For primary and secondary outcomes, statistics were calculated to determine whether the results were statistically significant for comparing the liraglutide and phentermine/topiramate subgroups. Sample sizes were too small for statistical analysis for bupropion/naltrexone, phentermine, and orlistat.

Primary Outcomes

The mean (SD) weight of participants dropped 8.1% from 130.1 kg to 119.5 kg over the patient-specific duration of weight management medication therapy for an absolute difference of 10.6 kg (9.7). Duration of individual medication use varied from 6 to 48 months. Weight loss was recorded at 6, 12, 24, 36, and 48 months of weight management therapy. Patient weight was not recorded after the medication was discontinued.

When classified by medication choice, the mean change in weight over the duration of the study was −23.9 kg for 2 patients using orlistat, −10.2 kg for 46 patients using liraglutide, −11.0 kg for 25 patients using phentermine/topiramate, -7.4 kg for 1 patient using phentermine, and -13.0 kg for 4 patients using naltrexone/bupropion. Patients without a weight documented at the end of their therapy or at the conclusion of the data collection period were not included in the total weight loss at the end of therapy. There were 78 documented instances of weight loss at the end of therapy (Table 2).

Body weight loss percentage was recorded at 6, 12, 24, 36, and 48 months of weight management therapy. The mean (SD) body weight loss percentage over the duration of the study was 9.2% (11.2). When classified by medication choice, the mean percentage of body weight loss was 16.8% for 2 patients using orlistat, 9.4% for 46 patients using liraglutide, 8.2% for 25 patients using phentermine/topiramate, 6.0% for 1 patient using phentermine alone, and 10.6% for 4 patients using naltrexone/bupropion (Table 3).

Secondary Outcomes

While none of the secondary outcomes were statistically significant, the results of this study suggest that both medications may contribute to weight loss in many patients included in this study. Almost two-thirds of the included patients analyzed lost ≥ 5% of weight from baseline while taking weight management medication. Sixty-six patients (63%) lost ≥ 5% of body weight at any time during the data collection period. When stratified by liraglutide and phentermine/topiramate, 41 patients (63%) taking liraglutide and 20 patients (67%) taking phentermine/topiramate lost ≥ 5% of weight from baseline. Of the 66 patients who lost ≥ 5% of body weight from baseline, 36 (55%) lost ≥ 10% of body weight from baseline at any time during the data collection period.

The mean (SD) duration for weight management medication use was 23 months (14.9). Phentermine/topiramate was tolerated longer than liraglutide: 22.7 months vs 21.7 months, respectively (Table 4).

The average overall documented medication discontinuation rate was 35.2%. Reasons for discontinuation included 21 patient-elected discontinuations, 8 patients no longer met criteria for use, 4 medications were no longer indicated, and 4 patients experienced AEs. It is unknown whether weight management medication was discontinued or not in 18 patients (17.2%).

This study evaluated the use and outcomes of weight loss medications over a longer period (up to 48 months) than what was previously studied among patients at VHI (12 months). The study aimed to better understand the long-term effect of weight loss medications, determine which medication had better long-term outcomes, and examine the reasons for medication discontinuation.

The results of this study displayed some similarities and differences compared with the Hood and colleagues study.¹⁶ Both yielded similar results for 5% of body weight loss and 10% of body weight loss. The largest difference was mean weight loss over the study period. In this study, patients lost a mean 10.6 kg over the course of weight loss medication use compared to 15.8 kg found by Hood and colleagues.¹⁶ A reason patients in the current study lost less weight overall could be the difference in time frames. The current study encompassed the COVID-19 pandemic, meaning fewer overall in-person patient appointments, which led to patients being lost to follow-up, missing weigh-ins during the time period, and gaps in care. For some patients, the pandemic possibly contributed to depression, missed medication doses, and a more sedentary lifestyle, leading to more weight gain.¹⁷ Telemedicine services at VHI expanded during the pandemic in an attempt to increase patient monitoring and counseling. It is unclear whether this expansion was enough to replace the in-person contact necessary to promote a healthy lifestyle.

VA pharmacists now care for patients through telehealth and are more involved in weight loss management. Since the conclusion of the Hood and colleagues study and start of this research, 2 pharmacists at VHI have been assigned to follow patients for obesity management to help with adherence to medication and lifestyle changes, management of AEs, dispense logistics, interventions for medications that may cause weight gain, and case management of glycemic control and weight loss with GLP-1RAs. Care management by pharmacists at VHI helps improve the logistics of titratable orders and save money by improving the use of high-cost items like GLP-1RAs. VA clinical pharmacy practitioners already monitor GLP-1RAs for patients with T2DM, so they are prepared to educate and assist patients with these medications.

It is important to continue developing a standardized process for weight loss medication management across the VA to improve the quality of patient care and optimize prescription outcomes. VA facilities differ in how weight loss management care is delivered and the level at which pharmacists are involved. Given the high rate of obesity among patients at the VA, the advent of new prescription options for weight loss, and the high cost associated with these medications, there has been increased attention to obesity care. Some Veterans Integrated Service Networks are forming a weight management community of practice groups to create standard operating procedures and algorithms to standardize care. Developing consistent processes is necessary to improve weight loss and patient care for veterans regardless where they receive treatment.

Limitations

The data used in this study were dependent on clinician documentation. Because of a lack of documentation in many instances, it was difficult to determine the full efficacy of the medications studied due to missing weight recordings. The lack of documentation made it difficult to determine whether patients were enrolled and active in the MOVE! program. It is required that patients enroll in MOVE! to obtain medications, but many did not have any follow-up MOVE! visits after initially obtaining their weight loss medication.

In this study, differences in the outcomes of patients with and without T2DM were not compared. It is the VA standard of care to prefer liraglutide over phentermine/topiramate in patients with T2DM or prediabetes.² This makes it difficult to assess whether phentermine/topiramate or liraglutide is more effective for weight loss in patients with T2DM. Weight gain after the discontinuation of weight loss medications was not assessed. Collecting this data may help determine whether a certain weight loss medication is less likely to cause rebound weight gain when discontinued.

Other limitations to this study consisted of excluding patients who discontinued therapy within 6 months, small sample sizes on some medications, and lack of data on adherence. Adherence was based on medication refills, which means that if a patient refilled the medication, it was assumed they were taking it. This is not always the case, and while accurate data on adherence is difficult to gather, it can impact how results may be interpreted. These additional limitations make it difficult to accurately determine the efficacy of the medications in this study.

CONCLUSIONS

This study found similar outcomes to what has been observed in larger clinical trials regarding weight loss medications. Nevertheless, there was a lack of accurate clinical documentation for most patients, which limits the conclusions. This lack of documentation potentially led to inaccurate results. It revealed that many patients at VHI did not uniformly receive consistent follow-up after starting a weight loss medication during the study period. With more standardized processes implemented at VA facilities, increased pharmacist involvement in weight loss medication management, and increased use of established telehealth services, patients could have the opportunity for closer follow-up that may lead to better weight loss outcomes. With these changes, there is more reason for additional studies to be conducted to assess follow-up, medication management, and weight loss overall.

The Centers for Disease Control and Prevention (CDC) classifies individuals with a body mass index (BMI) of 25 to 29.9as overweight and those with a BMI > 30 as obese (obesity classes: I, BMI 30 to 34.9; II, BMI 35 to 39.9; and III, BMI ≥ 40).¹ In 2011, the CDC estimated that 27.4% of adults in the United States were obese; less than a decade later, that number increased to 31.9%.¹ In that same period, the percentage of adults in Indiana classified as obese increased from 30.8% to 36.8%.¹ About 1 in 14 individuals in the US have class III obesity and 86% of veterans are either overweight or obese.²

High medical expenses can likely be attributed to the long-term health consequences of obesity. Compared to those with a healthy weight, individuals who are overweight or obese are at an increased risk for high blood pressure, high low-density lipoprotein cholesterol levels, low high-density lipoprotein cholesterol levels, high triglyceride levels, type 2 diabetes mellitus (T2DM), coronary heart disease, stroke, gallbladder disease, osteoarthritis, sleep apnea, cancer, mental health disorders, body pain, low quality of life, and death.³ Many of these conditions lead to increased health care needs, medication needs, hospitalizations, and overall health care system use.

Guidelines for the prevention and treatment of obesity have been produced by the American Heart Association, American College of Cardiology, and The Obesity Society; the Endocrine Society; the American Diabetes Association; and the US Departments of Veterans Affairs (VA) and Defense. Each follows a general algorithm to manage and prevent adverse effects (AEs) related to obesity. General practice is to assess a patient for elevated BMI (> 25), implement intense lifestyle modifications including calorie restriction and exercise, reassess for a maintained 5% to 10% weight loss for cardiovascular benefits, and potentially assess for pharmacological or surgical intervention to assist in weight loss.^2,4-6

While some weight loss medications (eg, phentermine/topiramate, naltrexone/bupropion, orlistat, and lorcaserin) tend to have unfavorable AEs or mixed efficacy, glucagon-like peptide-1 receptor agonists (GLP-1RAs) have provided new options.^7-10 Lorcaserin, for example, was removed from the market in 2020 due to its association with cancer risks.¹¹ The GLP-1RAs liraglutide and semaglutide received US Food and Drug Administration (FDA) approval for weight loss in 2014 and 2021, respectively.^12,13 GLP-1RAs have shown the greatest efficacy and benefits in reducing hemoglobin A_1c (HbA_1c); they are the preferred agents for patients who qualify for pharmacologic intervention for weight loss, especially those with T2DM. However, these studies have not evaluated the long-term outcomes of using these medications for weight loss and may not reflect the veteran population.^14,15

At Veteran Health Indiana (VHI), clinicians may use several weight loss medications for patients to achieve 5% to 10% weight loss. The medications most often used include liraglutide, phentermine/topiramate, naltrexone/bupropion, orlistat, and phentermine alone. However, more research is needed to determine which weight loss medication is the most beneficial for veterans, particularly following FDA approval of GLP-1RAs. At VHI, phentermine/topiramate is the preferred first-line agent unless patients have contraindications for use, in which case naltrexone/bupropion is recommended. These are considered first-line due to their ease of use in pill form, lower cost, and comparable weight loss to the GLP-1 medication class.² However, for patients with prediabetes, T2DM, BMI > 40, or BMI > 35 with specific comorbid conditions, liraglutide is preferred because of its beneficial effects for both weight loss and blood glucose control.²

This study aimed to expand on the 2021 Hood and colleagues study that examined total weight loss and weight loss as a percentage of baseline weight in patients with obesity at 3, 6, 12, and > 12 months of pharmacologic therapy by extending the time frame to 48 months.¹⁶ This study excluded semaglutide because few patients were prescribed the medication for weight loss during the study.

METHODS

We conducted a single-center, retrospective chart review of patients prescribed weight loss medications at VHI. A patient list was generated based on prescription fills from June 1, 2017, to July 31, 2021. Data were obtained from the Computerized Patient Record System; patients were not contacted. This study was approved by the Indiana University Health Institutional Review Board and VHI Research and Development Committee.

At the time of this study, liraglutide, phentermine/topiramate, naltrexone/bupropion, orlistat, and phentermine alone were available at VHI for patients who met the clinical criteria for use. All patients must have been enrolled in dietary and lifestyle management programs, including the VA MOVE! program, to be approved for these medications. After the MOVE! orientation, patients could participate in group or individual 12-week programs that included weigh-ins, goal-setting strategies, meal planning, and habit modification support. If patients could not meet in person, phone and other telehealth opportunities were available.

Patients were included in the study if they were aged ≥ 18 years, received a prescription for any of the 5 available medications for weight loss during the enrollment period, and were on the medication for ≥ 6 consecutive months. Patients were excluded if they received a prescription, were treated outside the VA system, or were pregnant. The primary indication for the included medication was not weight loss; the primary indication for the GLP-1RA was T2DM, or the weight loss was attributed to another disease. Adherence was not a measured outcome of this study; if patients were filling the medication, it was assumed they were taking it. Data were collected for each instance of medication use; as a result, a few patients were included more than once. Data collection for a failed medication ended when failure was documented. New data points began when new medication was prescribed; all data were per medication, not per patient. This allowed us to account for medication failure and provide accurate weight loss results based on medication choice within VHI.

Primary outcomes included total weight loss and weight loss as a percentage ofbaseline weight during the study period at 3, 6, 12, 24, 36, and 48 months of therapy. Secondary outcomes included the percentage of patients who lost 5% to 10% of their body weight from baseline; the percentage of patients who maintained ≥ 5% weight loss from baseline to 12, 24, 36, and 48 months if maintained on medication for that duration; duration of medication treatment in weeks; medication discontinuation rate; reason for medication discontinuation; enrollment in the MOVE! clinic and the time enrolled; percentage of patients with a BMI of 18 to 24.9 at the end of the study; and change in HbA_1c at 3, 6, 12, 24, 36, and 48 months.

Demographic data included race, age, sex, baseline weight, height, baseline BMI, and comorbid conditions (collected based on the most recent primary care clinical note before initiating medication). Medication data collected included medications used to manage comorbidities. Data related to weight management medication included prescribing clinic, maintenance dose of medication, duration of medication during the study period, the reason for medication discontinuation, or bariatric surgery intervention if applicable.

Basic descriptive statistics were used to characterize study participants. For continuous data, analysis of variance tests were used; if those results were not normal, then nonparametric tests were used, followed by pairwise tests between medication groups if the overall test was significant using the Fisher significant differences test. For nominal data, χ² or Fisher exact tests were used. For comparisons of primary and secondary outcomes, if the analyses needed to include adjustment for confounding variables, analysis of covariance was used for continuous data. A 2-sided 5% significance level was used for all tests.

RESULTS

A total of 228 instances of medication use were identified based on prescription fills; 123 did not meet inclusion criteria (117 for < 6 consecutive months of medication use) (Figure). The study included 105 participants with a mean age of 56 years; 80 were male (76.2%), and 85 identified as White race (81.0%). Mean (SD) weight was 130.1 kg (26.8) and BMI was 41.6 (7.2). The most common comorbid disease states among patients included hypertension, dyslipidemia, obstructive sleep apnea, and T2DM (Table 1). The baseline characteristics were comparable to those of Hood and colleagues.¹⁶

Most patients at VHI started on liraglutide (63%) or phentermine/topiramate (28%). For primary and secondary outcomes, statistics were calculated to determine whether the results were statistically significant for comparing the liraglutide and phentermine/topiramate subgroups. Sample sizes were too small for statistical analysis for bupropion/naltrexone, phentermine, and orlistat.

Primary Outcomes

The mean (SD) weight of participants dropped 8.1% from 130.1 kg to 119.5 kg over the patient-specific duration of weight management medication therapy for an absolute difference of 10.6 kg (9.7). Duration of individual medication use varied from 6 to 48 months. Weight loss was recorded at 6, 12, 24, 36, and 48 months of weight management therapy. Patient weight was not recorded after the medication was discontinued.

When classified by medication choice, the mean change in weight over the duration of the study was −23.9 kg for 2 patients using orlistat, −10.2 kg for 46 patients using liraglutide, −11.0 kg for 25 patients using phentermine/topiramate, -7.4 kg for 1 patient using phentermine, and -13.0 kg for 4 patients using naltrexone/bupropion. Patients without a weight documented at the end of their therapy or at the conclusion of the data collection period were not included in the total weight loss at the end of therapy. There were 78 documented instances of weight loss at the end of therapy (Table 2).

Body weight loss percentage was recorded at 6, 12, 24, 36, and 48 months of weight management therapy. The mean (SD) body weight loss percentage over the duration of the study was 9.2% (11.2). When classified by medication choice, the mean percentage of body weight loss was 16.8% for 2 patients using orlistat, 9.4% for 46 patients using liraglutide, 8.2% for 25 patients using phentermine/topiramate, 6.0% for 1 patient using phentermine alone, and 10.6% for 4 patients using naltrexone/bupropion (Table 3).

Secondary Outcomes

While none of the secondary outcomes were statistically significant, the results of this study suggest that both medications may contribute to weight loss in many patients included in this study. Almost two-thirds of the included patients analyzed lost ≥ 5% of weight from baseline while taking weight management medication. Sixty-six patients (63%) lost ≥ 5% of body weight at any time during the data collection period. When stratified by liraglutide and phentermine/topiramate, 41 patients (63%) taking liraglutide and 20 patients (67%) taking phentermine/topiramate lost ≥ 5% of weight from baseline. Of the 66 patients who lost ≥ 5% of body weight from baseline, 36 (55%) lost ≥ 10% of body weight from baseline at any time during the data collection period.

The mean (SD) duration for weight management medication use was 23 months (14.9). Phentermine/topiramate was tolerated longer than liraglutide: 22.7 months vs 21.7 months, respectively (Table 4).

The average overall documented medication discontinuation rate was 35.2%. Reasons for discontinuation included 21 patient-elected discontinuations, 8 patients no longer met criteria for use, 4 medications were no longer indicated, and 4 patients experienced AEs. It is unknown whether weight management medication was discontinued or not in 18 patients (17.2%).

This study evaluated the use and outcomes of weight loss medications over a longer period (up to 48 months) than what was previously studied among patients at VHI (12 months). The study aimed to better understand the long-term effect of weight loss medications, determine which medication had better long-term outcomes, and examine the reasons for medication discontinuation.

The results of this study displayed some similarities and differences compared with the Hood and colleagues study.¹⁶ Both yielded similar results for 5% of body weight loss and 10% of body weight loss. The largest difference was mean weight loss over the study period. In this study, patients lost a mean 10.6 kg over the course of weight loss medication use compared to 15.8 kg found by Hood and colleagues.¹⁶ A reason patients in the current study lost less weight overall could be the difference in time frames. The current study encompassed the COVID-19 pandemic, meaning fewer overall in-person patient appointments, which led to patients being lost to follow-up, missing weigh-ins during the time period, and gaps in care. For some patients, the pandemic possibly contributed to depression, missed medication doses, and a more sedentary lifestyle, leading to more weight gain.¹⁷ Telemedicine services at VHI expanded during the pandemic in an attempt to increase patient monitoring and counseling. It is unclear whether this expansion was enough to replace the in-person contact necessary to promote a healthy lifestyle.

VA pharmacists now care for patients through telehealth and are more involved in weight loss management. Since the conclusion of the Hood and colleagues study and start of this research, 2 pharmacists at VHI have been assigned to follow patients for obesity management to help with adherence to medication and lifestyle changes, management of AEs, dispense logistics, interventions for medications that may cause weight gain, and case management of glycemic control and weight loss with GLP-1RAs. Care management by pharmacists at VHI helps improve the logistics of titratable orders and save money by improving the use of high-cost items like GLP-1RAs. VA clinical pharmacy practitioners already monitor GLP-1RAs for patients with T2DM, so they are prepared to educate and assist patients with these medications.

It is important to continue developing a standardized process for weight loss medication management across the VA to improve the quality of patient care and optimize prescription outcomes. VA facilities differ in how weight loss management care is delivered and the level at which pharmacists are involved. Given the high rate of obesity among patients at the VA, the advent of new prescription options for weight loss, and the high cost associated with these medications, there has been increased attention to obesity care. Some Veterans Integrated Service Networks are forming a weight management community of practice groups to create standard operating procedures and algorithms to standardize care. Developing consistent processes is necessary to improve weight loss and patient care for veterans regardless where they receive treatment.

Limitations

The data used in this study were dependent on clinician documentation. Because of a lack of documentation in many instances, it was difficult to determine the full efficacy of the medications studied due to missing weight recordings. The lack of documentation made it difficult to determine whether patients were enrolled and active in the MOVE! program. It is required that patients enroll in MOVE! to obtain medications, but many did not have any follow-up MOVE! visits after initially obtaining their weight loss medication.

In this study, differences in the outcomes of patients with and without T2DM were not compared. It is the VA standard of care to prefer liraglutide over phentermine/topiramate in patients with T2DM or prediabetes.² This makes it difficult to assess whether phentermine/topiramate or liraglutide is more effective for weight loss in patients with T2DM. Weight gain after the discontinuation of weight loss medications was not assessed. Collecting this data may help determine whether a certain weight loss medication is less likely to cause rebound weight gain when discontinued.

Other limitations to this study consisted of excluding patients who discontinued therapy within 6 months, small sample sizes on some medications, and lack of data on adherence. Adherence was based on medication refills, which means that if a patient refilled the medication, it was assumed they were taking it. This is not always the case, and while accurate data on adherence is difficult to gather, it can impact how results may be interpreted. These additional limitations make it difficult to accurately determine the efficacy of the medications in this study.

CONCLUSIONS

This study found similar outcomes to what has been observed in larger clinical trials regarding weight loss medications. Nevertheless, there was a lack of accurate clinical documentation for most patients, which limits the conclusions. This lack of documentation potentially led to inaccurate results. It revealed that many patients at VHI did not uniformly receive consistent follow-up after starting a weight loss medication during the study period. With more standardized processes implemented at VA facilities, increased pharmacist involvement in weight loss medication management, and increased use of established telehealth services, patients could have the opportunity for closer follow-up that may lead to better weight loss outcomes. With these changes, there is more reason for additional studies to be conducted to assess follow-up, medication management, and weight loss overall.

It is common practice to routinely measure postoperative hematocrit levels at US Department of Veterans Affairs (VA) hospitals for a wide range of elective general surgeries. While hematocrit measurement is a low-cost test, the high frequency with which these tests are performed may drastically increase overall costs.

Numerous studies have suggested that physicians overuse laboratory testing.^1-10 Kohli and colleagues recommended that the routine practice of obtaining postoperative hematocrit tests following elective gynecologic surgery be abandoned.¹ A similar recommendation was made by Olus and colleagues after studying uneventful, unplanned cesarean sections and by Wu and colleagues after investigating routine laboratory tests post total hip arthroplasty.^2,3

To our knowledge, a study assessing routine postoperative hematocrit testing in elective general surgery has not yet been conducted. Many laboratory tests ordered in the perioperative period are not indicated, including complete blood count (CBC), electrolytes, and coagulation studies.⁴ Based on the results of these studies, we expected that the routine measurement of postoperative hematocrit levels after elective general surgeries at VA medical centers would not be cost effective. A PubMed search for articles published from 1990 to 2023 using the search terms “hematocrit,” “hemoglobin,” “general,” “surgery,” “routine,” and “cost” or “cost-effectiveness,” suggests that the clinical usefulness of postoperative hematocrit testing has not been well studied in the general surgery setting. The purpose of this study was to determine the clinical utility and associated cost of measuring routine postoperative hematocrit levels in order to generate a guide as to when the practice is warranted following common elective general surgery.

Although gynecologic textbooks may describe recommendations of routine hematocrit checking after elective gynecologic operations, one has difficulty finding the same recommendations in general surgery textbooks.¹ However, it is common practice for surgical residents and attending surgeons to routinely order hematocrit on postoperative day-1 to ensure that the operation did not result in unsuspected anemia that then would need treatment (either with fluids or a blood transfusion). Many other surgeons rely on clinical factors such as tachycardia, oliguria, or hypotension to trigger a hematocrit (and other laboratory) tests. Our hypothesis is that the latter group has chosen the most cost-effective and prudent practice. One problem with checking the hematocrit routinely, as with any other screening test, is what to do with an abnormal result, assuming an asymptomatic patient? If the postoperative hematocrit is lower than expected given the estimated blood loss (EBL), what is one to do?

Methods

This retrospective case-control study conducted at the New Mexico VA Health Care System (NMVAHCS) in Albuquerque compared data for patients who received transfusion within 72 hours of elective surgeries vs patients who did not. Patients who underwent elective general surgery from January 2011 through December 2014 were included. An elective general surgery was defined as surgery performed following an outpatient preoperative anesthesia evaluation ≥ 30 days prior to operation. Patients who underwent emergency operations, and those with baseline anemia (preoperative hematocrit < 30%), and those transfused > 72 hours after their operation were excluded. The NMVAHCSInstitutional Review Board approved this study (No. 15-H184).

A detailed record review was conducted to collect data on demographics and other preoperative risk factors, including age, sex, body mass index (BMI), race and ethnicity, cardiac and pulmonary comorbidities, tobacco use, alcohol intake, diabetes, American Society of Anesthesiologists Physical Status Classification, metabolic equivalent of task, hematologic conditions, and renal disease.

For each procedure, we recorded the type of elective general surgery performed, the diagnosis/indication, pre- and postoperative hemoglobin/hematocrit, intraoperative EBL, length of operation, surgical wound class, length of hospital stay (LOS), intensive care unit (ICU) status, number of hematocrit tests, cardiovascular risk of operation (defined by anesthesia assessment), presence or absence of malignancy, preoperative platelet count, albumin level, preoperative prothrombin time/activated partial thromboplastin time (aPTT), international normalized ratio (INR), hemoglobin A_1c, and incidence of transfusion. Signs and symptoms of anemia were recorded as present if the postoperative vital signs suggested low intravascular volume (pulse > 120 beats/minute, systolic blood pressure < 90 mm Hg, or vasoactive medication requirement [per anesthesia postoperative note]) or if the patient reported or exhibited symptoms of dizziness or fatigue or evidence of clinically apparent bleeding (ie, hematoma formation). Laboratory charges for hematocrit tests and CBC at the NMAVAHCS were used to assess cost.¹¹

To stratify the transfusion risk, patients were distributed among 3 groups based on the following criteria: discharged home the same day as surgery; admitted but did not have postoperative hematocrit testing; and admitted and had postoperative hematocrit testing. We also stratified operations into low or high risk based on the risk for postoperative transfusion (Figure). Recognizing that the American College of Chest Physicians guidelines for perioperative management of antithrombotic therapy places bowel resection in a high-risk category, we designated a surgery as high risk when ≥ 2 patients in the transfusion group had that type of surgery over the 4 years of the study.¹² Otherwise, the operations were deemed low risk.

Statistical Analysis

Numeric analysis used t tests and Binary and categorical variables used Fisher exact tests. P value ≤ .05 was considered statistically significant. SAS software was used for all statistical analyses.

From 2011 through 2014, 1531 patients had elective general surgery at NMVAHCS. Twenty-two patients with preoperative anemia (hematocrit < 30%) and 1 patient who received a transfusion > 72 hours after the operation were excluded. Most elective operations (70%, n = 1075) were performed on an outpatient basis; none involved transfusion. Inguinal hernia repair was most common with 479 operations; 17 patients were treated inpatient of which 2 patients had routine postoperative hematocrit checks; (neither received transfusion). One patient with inguinal hernia surgery received transfusion without routine postoperative hematocrit monitoring.

Of 112 partial colon resections, 1 patient had a postoperative transfusion; and all but 3 received postoperative hematocrit monitoring. Nineteen patients undergoing partial colon resection had a clinical indication for postoperative hematocrit monitoring. None of the 5 patients with partial gastrectomy received a postoperative transfusion. Of 121 elective cholecystectomies, no patients had postoperative transfusion, whereas 34 had postoperative hematocrit monitoring; only 2 patients had a clinical reason for the hematocrit monitoring.

Of 430 elective inpatient operations, 12 received transfusions and 288 patients had ≥ 1 postoperative hematocrit test (67%). All hematocrit tests were requested by the attending surgeon, resident surgeon, or the surgical ICU team. Of the group that had postoperative hematocrit monitoring, there was an average of 4.4 postoperative hematocrit tests per patient (range, 1-44).

There were 12 transfusions for inpatients (2.8%), which is similar to the findings of a recent study of VA general surgery (2.3%).¹³ Five of the 12 patients received intraoperative transfusions while 7 were transfused within 72 hours postoperation. All but 1 patient receiving transfusion had EBL > 199 mL (range, 5-3000; mean, 950 mL; median, 500 mL) and/or signs or symptoms of anemia or other indications for measurement of the postoperative hematocrit. There were no statistically significant differences in patients’ age, sex, BMI, or race and ethnicity between groups receiving and not receiving transfusion (Table 1).

When comparing the transfusion vs the nontransfusion groups (after excluding those with clinical preoperative anemia) the risk factors for transfusion included: relatively low mean preoperative hematocrit (mean, 36.9% vs 42.7%, respectively; P = .003), low postoperative hematocrit (mean, 30.2% vs 37.1%, respectively; P < .001), high EBL (mean, 844 mL vs 109 mL, respectively; P = .005), large infusion of intraoperative fluids (mean, 4625 mL vs 2505 mL, respectively; P = .005), longer duration of operation (mean, 397 min vs 183 min, respectively; P < .001), and longer LOS (mean, 14.5 d vs 4.9 d, respectively; P < .001) (Table 2). Similarly, we found an increased risk for transfusion with high/intermediate cardiovascular risk (vs low), any wound not classified as clean, ICU stay, and postoperative symptoms of anemia.

We found no increased risk for transfusion with ethanol, tobacco, warfarin, or clopidogrel use; polycythemia; thrombocytopenia; preoperative INR; preoperative aPTT; preoperative albumin; Hemoglobin A_1c; or diabetes mellitus; or for operations performed for malignancy. Ten patients in the ICU received transfusion (5.8%) compared with 2 patients (0.8%) not admitted to the ICU.

Operations were deemed high risk when ≥ 2 of patients having that operation received transfusions within 72 hours of their operation. There were 15 abdominoperineal resections; 3 of these received transfusions (20%). There were 7 total abdominal colectomies; 3 of these received transfusions (43%). We therefore had 22 high-risk operations, 6 of which were transfused (27%).

Routine measurement of postoperative hematocrit levels after elective general surgery at NMVAHCS was not necessary. There were 12 transfusions for inpatients (2.8%), which is similar to the findings of a recent study of VA general surgery (2.3%).¹³ We found that routine postoperative hematocrit measurements to assess anemia had little or no effect on clinical decision-making or clinical outcomes.

According to our results, 88% of initial hematocrit tests after elective partial colectomies could have been eliminated; only 32 of 146 patients demonstrated a clinical reason for postoperative hematocrit testing. Similarly, 36 of 40 postcholecystectomy hematocrit tests (90%) could have been eliminated had the surgeons relied on clinical signs indicating possible postoperative anemia (none were transfused). Excluding patients with major intraoperative blood loss (> 300 mL), only 29 of 288 (10%) patients who had postoperative hematocrit tests had a clinical indication for a postoperative hematocrit test (ie, symptoms of anemia and/or active bleeding). One patient with inguinal hernia surgery who received transfusion was taking an anticoagulant and had a clinically indicated hematocrit test for a large hematoma that eventually required reoperation.

Our study found that routine hematocrit checks may actually increase the risk that a patient would receive an unnecessary transfusion. For instance, one elderly patient, after a right colectomy, had 6 hematocrit levels while on a heparin drip and received transfusion despite being asymptomatic. His lowest hematocrit level prior to transfusion was 23.7%. This patient had a total of 18 hematocrit tests. His EBL was 350 mL and his first postoperative HCT level was 33.1%. In another instance, a patient undergoing abdominoperineal resection had a transfusion on postoperative day 1, despite being hypertensive, with a hematocrit that ranged from 26% before transfusion to 31% after the transfusion. These 2 cases illustrate what has been shown in a recent study: A substantial number of patients with colorectal cancer receive unnecessary transfusions.¹⁴ On the other hand, one ileostomy closure patient had 33 hematocrit tests, yet his initial postoperative hematocrit was 37%, and he never received a transfusion. With low-risk surgeries, clinical judgment should dictate when a postoperative hematocrit level is needed. This strategy would have eliminated 206 unnecessary initial postoperative hematocrit tests (72%), could have decreased the number of unnecessary transfusions, and would have saved NMVAHCS about $1600 annually.

Abdominoperineal resections and total abdominal colectomies accounted for a high proportion of transfusions in our study. Inpatient elective operations can be risk stratified and have routine hematocrit tests ordered for patients at high risk. The probability of transfusion was greater in high-risk vs low-risk surgeries; 27% (6 of 22 patients) vs 2% (6 of 408 patients), respectively (P < .001). Since 14 of the 22 patients undergoing high-risk operation already had clinical reasons for a postoperative hematocrit test, we only need to add the remaining 8 patients with high-risk operations to the 74 who had a clinical reason for a hematocrit test and conclude that 82 of 430 patients (19%) had a clinical reason for a hematocrit test, either from signs or symptoms of blood loss or because they were in a high-risk group.

While our elective general surgery cases may not represent many general surgery programs in the US and VA health care systems, we can extrapolate cost savings using the same cost analyses outlined by Kohli and colleagues.¹ Assuming 1.9 million elective inpatient general surgeries per year in the United States with an average cost of $21 per CBC, the annual cost of universal postoperative hematocrit testing would be $40 million.^11,15 If postoperative hematocrit testing were 70% consistent with our findings, the annual cost for hematocrit tests on 51% of the inpatient general surgeries would be approximately $20.4 million. A reduction in routine hematocrit testing to 25% of all inpatient general surgeries (vs our finding that 19% were deemed necessary) results in an annual savings of $30 million. This conservative estimate could be even higher since there were 4.4 hematocrit tests per patient; therefore, we have about $132 million in savings.

Assuming 181,384 elective VA inpatient general surgeries each year, costing $7.14 per CBC (the NMVAHCS cost), the VA could save $1.3 million annually. If postoperative HCT testing were 70% consistent with our findings, the annual cost for hematocrit tests on 50.4% of inpatient general surgery operations would be about $653,000. A reduction in routine hematocrit testing to 25% of all inpatient general surgeries (vs our 19%) results in annual VA savings of $330,000. This conservative estimate could be even higher since there were on average 4.4 hematocrit levels per patient; therefore, we estimate that annual savings for the VA of about $1.45 million.

Limitations

The retrospective chart review nature of this study may have led to selection bias. Only a small number of patients received a transfusion, which may have skewed the data. This study population comes from a single VA medical center; this patient population may not be reflective of other VA medical centers or the US population as a whole. Given that NMVAHCS does not perform hepatic, esophageal, pancreas, or transplant operations, the potential savings to both the US and the VA may be overestimated, but this could be studied in the future by VA medical centers that perform more complex operations.

Conclusions

This study found that over a 4-year period routine postoperative hematocrit tests for patients undergoing elective general surgery at a VA medical center were not necessary. General surgeons routinely order various pre- and postoperative laboratory tests despite their limited utility. Reduction in unneeded routine tests could result in notable savings to the VA without compromising quality of care.

Only general surgery patients undergoing operations that carry a high risk for needing a blood transfusion should have a routine postoperative hematocrit testing. In our study population, the chance of an elective colectomy, cholecystectomy, or hernia patient needing a transfusion was rare. This strategy could eliminate a considerable number of unnecessary blood tests and would potentially yield significant savings.

It is common practice to routinely measure postoperative hematocrit levels at US Department of Veterans Affairs (VA) hospitals for a wide range of elective general surgeries. While hematocrit measurement is a low-cost test, the high frequency with which these tests are performed may drastically increase overall costs.

Numerous studies have suggested that physicians overuse laboratory testing.^1-10 Kohli and colleagues recommended that the routine practice of obtaining postoperative hematocrit tests following elective gynecologic surgery be abandoned.¹ A similar recommendation was made by Olus and colleagues after studying uneventful, unplanned cesarean sections and by Wu and colleagues after investigating routine laboratory tests post total hip arthroplasty.^2,3

To our knowledge, a study assessing routine postoperative hematocrit testing in elective general surgery has not yet been conducted. Many laboratory tests ordered in the perioperative period are not indicated, including complete blood count (CBC), electrolytes, and coagulation studies.⁴ Based on the results of these studies, we expected that the routine measurement of postoperative hematocrit levels after elective general surgeries at VA medical centers would not be cost effective. A PubMed search for articles published from 1990 to 2023 using the search terms “hematocrit,” “hemoglobin,” “general,” “surgery,” “routine,” and “cost” or “cost-effectiveness,” suggests that the clinical usefulness of postoperative hematocrit testing has not been well studied in the general surgery setting. The purpose of this study was to determine the clinical utility and associated cost of measuring routine postoperative hematocrit levels in order to generate a guide as to when the practice is warranted following common elective general surgery.

Although gynecologic textbooks may describe recommendations of routine hematocrit checking after elective gynecologic operations, one has difficulty finding the same recommendations in general surgery textbooks.¹ However, it is common practice for surgical residents and attending surgeons to routinely order hematocrit on postoperative day-1 to ensure that the operation did not result in unsuspected anemia that then would need treatment (either with fluids or a blood transfusion). Many other surgeons rely on clinical factors such as tachycardia, oliguria, or hypotension to trigger a hematocrit (and other laboratory) tests. Our hypothesis is that the latter group has chosen the most cost-effective and prudent practice. One problem with checking the hematocrit routinely, as with any other screening test, is what to do with an abnormal result, assuming an asymptomatic patient? If the postoperative hematocrit is lower than expected given the estimated blood loss (EBL), what is one to do?

Methods

This retrospective case-control study conducted at the New Mexico VA Health Care System (NMVAHCS) in Albuquerque compared data for patients who received transfusion within 72 hours of elective surgeries vs patients who did not. Patients who underwent elective general surgery from January 2011 through December 2014 were included. An elective general surgery was defined as surgery performed following an outpatient preoperative anesthesia evaluation ≥ 30 days prior to operation. Patients who underwent emergency operations, and those with baseline anemia (preoperative hematocrit < 30%), and those transfused > 72 hours after their operation were excluded. The NMVAHCSInstitutional Review Board approved this study (No. 15-H184).

A detailed record review was conducted to collect data on demographics and other preoperative risk factors, including age, sex, body mass index (BMI), race and ethnicity, cardiac and pulmonary comorbidities, tobacco use, alcohol intake, diabetes, American Society of Anesthesiologists Physical Status Classification, metabolic equivalent of task, hematologic conditions, and renal disease.

For each procedure, we recorded the type of elective general surgery performed, the diagnosis/indication, pre- and postoperative hemoglobin/hematocrit, intraoperative EBL, length of operation, surgical wound class, length of hospital stay (LOS), intensive care unit (ICU) status, number of hematocrit tests, cardiovascular risk of operation (defined by anesthesia assessment), presence or absence of malignancy, preoperative platelet count, albumin level, preoperative prothrombin time/activated partial thromboplastin time (aPTT), international normalized ratio (INR), hemoglobin A_1c, and incidence of transfusion. Signs and symptoms of anemia were recorded as present if the postoperative vital signs suggested low intravascular volume (pulse > 120 beats/minute, systolic blood pressure < 90 mm Hg, or vasoactive medication requirement [per anesthesia postoperative note]) or if the patient reported or exhibited symptoms of dizziness or fatigue or evidence of clinically apparent bleeding (ie, hematoma formation). Laboratory charges for hematocrit tests and CBC at the NMAVAHCS were used to assess cost.¹¹

To stratify the transfusion risk, patients were distributed among 3 groups based on the following criteria: discharged home the same day as surgery; admitted but did not have postoperative hematocrit testing; and admitted and had postoperative hematocrit testing. We also stratified operations into low or high risk based on the risk for postoperative transfusion (Figure). Recognizing that the American College of Chest Physicians guidelines for perioperative management of antithrombotic therapy places bowel resection in a high-risk category, we designated a surgery as high risk when ≥ 2 patients in the transfusion group had that type of surgery over the 4 years of the study.¹² Otherwise, the operations were deemed low risk.

Statistical Analysis

Numeric analysis used t tests and Binary and categorical variables used Fisher exact tests. P value ≤ .05 was considered statistically significant. SAS software was used for all statistical analyses.

From 2011 through 2014, 1531 patients had elective general surgery at NMVAHCS. Twenty-two patients with preoperative anemia (hematocrit < 30%) and 1 patient who received a transfusion > 72 hours after the operation were excluded. Most elective operations (70%, n = 1075) were performed on an outpatient basis; none involved transfusion. Inguinal hernia repair was most common with 479 operations; 17 patients were treated inpatient of which 2 patients had routine postoperative hematocrit checks; (neither received transfusion). One patient with inguinal hernia surgery received transfusion without routine postoperative hematocrit monitoring.

Of 112 partial colon resections, 1 patient had a postoperative transfusion; and all but 3 received postoperative hematocrit monitoring. Nineteen patients undergoing partial colon resection had a clinical indication for postoperative hematocrit monitoring. None of the 5 patients with partial gastrectomy received a postoperative transfusion. Of 121 elective cholecystectomies, no patients had postoperative transfusion, whereas 34 had postoperative hematocrit monitoring; only 2 patients had a clinical reason for the hematocrit monitoring.

Of 430 elective inpatient operations, 12 received transfusions and 288 patients had ≥ 1 postoperative hematocrit test (67%). All hematocrit tests were requested by the attending surgeon, resident surgeon, or the surgical ICU team. Of the group that had postoperative hematocrit monitoring, there was an average of 4.4 postoperative hematocrit tests per patient (range, 1-44).

There were 12 transfusions for inpatients (2.8%), which is similar to the findings of a recent study of VA general surgery (2.3%).¹³ Five of the 12 patients received intraoperative transfusions while 7 were transfused within 72 hours postoperation. All but 1 patient receiving transfusion had EBL > 199 mL (range, 5-3000; mean, 950 mL; median, 500 mL) and/or signs or symptoms of anemia or other indications for measurement of the postoperative hematocrit. There were no statistically significant differences in patients’ age, sex, BMI, or race and ethnicity between groups receiving and not receiving transfusion (Table 1).

When comparing the transfusion vs the nontransfusion groups (after excluding those with clinical preoperative anemia) the risk factors for transfusion included: relatively low mean preoperative hematocrit (mean, 36.9% vs 42.7%, respectively; P = .003), low postoperative hematocrit (mean, 30.2% vs 37.1%, respectively; P < .001), high EBL (mean, 844 mL vs 109 mL, respectively; P = .005), large infusion of intraoperative fluids (mean, 4625 mL vs 2505 mL, respectively; P = .005), longer duration of operation (mean, 397 min vs 183 min, respectively; P < .001), and longer LOS (mean, 14.5 d vs 4.9 d, respectively; P < .001) (Table 2). Similarly, we found an increased risk for transfusion with high/intermediate cardiovascular risk (vs low), any wound not classified as clean, ICU stay, and postoperative symptoms of anemia.

We found no increased risk for transfusion with ethanol, tobacco, warfarin, or clopidogrel use; polycythemia; thrombocytopenia; preoperative INR; preoperative aPTT; preoperative albumin; Hemoglobin A_1c; or diabetes mellitus; or for operations performed for malignancy. Ten patients in the ICU received transfusion (5.8%) compared with 2 patients (0.8%) not admitted to the ICU.

Operations were deemed high risk when ≥ 2 of patients having that operation received transfusions within 72 hours of their operation. There were 15 abdominoperineal resections; 3 of these received transfusions (20%). There were 7 total abdominal colectomies; 3 of these received transfusions (43%). We therefore had 22 high-risk operations, 6 of which were transfused (27%).

Routine measurement of postoperative hematocrit levels after elective general surgery at NMVAHCS was not necessary. There were 12 transfusions for inpatients (2.8%), which is similar to the findings of a recent study of VA general surgery (2.3%).¹³ We found that routine postoperative hematocrit measurements to assess anemia had little or no effect on clinical decision-making or clinical outcomes.

According to our results, 88% of initial hematocrit tests after elective partial colectomies could have been eliminated; only 32 of 146 patients demonstrated a clinical reason for postoperative hematocrit testing. Similarly, 36 of 40 postcholecystectomy hematocrit tests (90%) could have been eliminated had the surgeons relied on clinical signs indicating possible postoperative anemia (none were transfused). Excluding patients with major intraoperative blood loss (> 300 mL), only 29 of 288 (10%) patients who had postoperative hematocrit tests had a clinical indication for a postoperative hematocrit test (ie, symptoms of anemia and/or active bleeding). One patient with inguinal hernia surgery who received transfusion was taking an anticoagulant and had a clinically indicated hematocrit test for a large hematoma that eventually required reoperation.

Our study found that routine hematocrit checks may actually increase the risk that a patient would receive an unnecessary transfusion. For instance, one elderly patient, after a right colectomy, had 6 hematocrit levels while on a heparin drip and received transfusion despite being asymptomatic. His lowest hematocrit level prior to transfusion was 23.7%. This patient had a total of 18 hematocrit tests. His EBL was 350 mL and his first postoperative HCT level was 33.1%. In another instance, a patient undergoing abdominoperineal resection had a transfusion on postoperative day 1, despite being hypertensive, with a hematocrit that ranged from 26% before transfusion to 31% after the transfusion. These 2 cases illustrate what has been shown in a recent study: A substantial number of patients with colorectal cancer receive unnecessary transfusions.¹⁴ On the other hand, one ileostomy closure patient had 33 hematocrit tests, yet his initial postoperative hematocrit was 37%, and he never received a transfusion. With low-risk surgeries, clinical judgment should dictate when a postoperative hematocrit level is needed. This strategy would have eliminated 206 unnecessary initial postoperative hematocrit tests (72%), could have decreased the number of unnecessary transfusions, and would have saved NMVAHCS about $1600 annually.

Abdominoperineal resections and total abdominal colectomies accounted for a high proportion of transfusions in our study. Inpatient elective operations can be risk stratified and have routine hematocrit tests ordered for patients at high risk. The probability of transfusion was greater in high-risk vs low-risk surgeries; 27% (6 of 22 patients) vs 2% (6 of 408 patients), respectively (P < .001). Since 14 of the 22 patients undergoing high-risk operation already had clinical reasons for a postoperative hematocrit test, we only need to add the remaining 8 patients with high-risk operations to the 74 who had a clinical reason for a hematocrit test and conclude that 82 of 430 patients (19%) had a clinical reason for a hematocrit test, either from signs or symptoms of blood loss or because they were in a high-risk group.

While our elective general surgery cases may not represent many general surgery programs in the US and VA health care systems, we can extrapolate cost savings using the same cost analyses outlined by Kohli and colleagues.¹ Assuming 1.9 million elective inpatient general surgeries per year in the United States with an average cost of $21 per CBC, the annual cost of universal postoperative hematocrit testing would be $40 million.^11,15 If postoperative hematocrit testing were 70% consistent with our findings, the annual cost for hematocrit tests on 51% of the inpatient general surgeries would be approximately $20.4 million. A reduction in routine hematocrit testing to 25% of all inpatient general surgeries (vs our finding that 19% were deemed necessary) results in an annual savings of $30 million. This conservative estimate could be even higher since there were 4.4 hematocrit tests per patient; therefore, we have about $132 million in savings.

Assuming 181,384 elective VA inpatient general surgeries each year, costing $7.14 per CBC (the NMVAHCS cost), the VA could save $1.3 million annually. If postoperative HCT testing were 70% consistent with our findings, the annual cost for hematocrit tests on 50.4% of inpatient general surgery operations would be about $653,000. A reduction in routine hematocrit testing to 25% of all inpatient general surgeries (vs our 19%) results in annual VA savings of $330,000. This conservative estimate could be even higher since there were on average 4.4 hematocrit levels per patient; therefore, we estimate that annual savings for the VA of about $1.45 million.

Limitations

The retrospective chart review nature of this study may have led to selection bias. Only a small number of patients received a transfusion, which may have skewed the data. This study population comes from a single VA medical center; this patient population may not be reflective of other VA medical centers or the US population as a whole. Given that NMVAHCS does not perform hepatic, esophageal, pancreas, or transplant operations, the potential savings to both the US and the VA may be overestimated, but this could be studied in the future by VA medical centers that perform more complex operations.

Conclusions

This study found that over a 4-year period routine postoperative hematocrit tests for patients undergoing elective general surgery at a VA medical center were not necessary. General surgeons routinely order various pre- and postoperative laboratory tests despite their limited utility. Reduction in unneeded routine tests could result in notable savings to the VA without compromising quality of care.

Only general surgery patients undergoing operations that carry a high risk for needing a blood transfusion should have a routine postoperative hematocrit testing. In our study population, the chance of an elective colectomy, cholecystectomy, or hernia patient needing a transfusion was rare. This strategy could eliminate a considerable number of unnecessary blood tests and would potentially yield significant savings.

It is common practice to routinely measure postoperative hematocrit levels at US Department of Veterans Affairs (VA) hospitals for a wide range of elective general surgeries. While hematocrit measurement is a low-cost test, the high frequency with which these tests are performed may drastically increase overall costs.

Numerous studies have suggested that physicians overuse laboratory testing.^1-10 Kohli and colleagues recommended that the routine practice of obtaining postoperative hematocrit tests following elective gynecologic surgery be abandoned.¹ A similar recommendation was made by Olus and colleagues after studying uneventful, unplanned cesarean sections and by Wu and colleagues after investigating routine laboratory tests post total hip arthroplasty.^2,3

To our knowledge, a study assessing routine postoperative hematocrit testing in elective general surgery has not yet been conducted. Many laboratory tests ordered in the perioperative period are not indicated, including complete blood count (CBC), electrolytes, and coagulation studies.⁴ Based on the results of these studies, we expected that the routine measurement of postoperative hematocrit levels after elective general surgeries at VA medical centers would not be cost effective. A PubMed search for articles published from 1990 to 2023 using the search terms “hematocrit,” “hemoglobin,” “general,” “surgery,” “routine,” and “cost” or “cost-effectiveness,” suggests that the clinical usefulness of postoperative hematocrit testing has not been well studied in the general surgery setting. The purpose of this study was to determine the clinical utility and associated cost of measuring routine postoperative hematocrit levels in order to generate a guide as to when the practice is warranted following common elective general surgery.

Although gynecologic textbooks may describe recommendations of routine hematocrit checking after elective gynecologic operations, one has difficulty finding the same recommendations in general surgery textbooks.¹ However, it is common practice for surgical residents and attending surgeons to routinely order hematocrit on postoperative day-1 to ensure that the operation did not result in unsuspected anemia that then would need treatment (either with fluids or a blood transfusion). Many other surgeons rely on clinical factors such as tachycardia, oliguria, or hypotension to trigger a hematocrit (and other laboratory) tests. Our hypothesis is that the latter group has chosen the most cost-effective and prudent practice. One problem with checking the hematocrit routinely, as with any other screening test, is what to do with an abnormal result, assuming an asymptomatic patient? If the postoperative hematocrit is lower than expected given the estimated blood loss (EBL), what is one to do?

Methods

This retrospective case-control study conducted at the New Mexico VA Health Care System (NMVAHCS) in Albuquerque compared data for patients who received transfusion within 72 hours of elective surgeries vs patients who did not. Patients who underwent elective general surgery from January 2011 through December 2014 were included. An elective general surgery was defined as surgery performed following an outpatient preoperative anesthesia evaluation ≥ 30 days prior to operation. Patients who underwent emergency operations, and those with baseline anemia (preoperative hematocrit < 30%), and those transfused > 72 hours after their operation were excluded. The NMVAHCSInstitutional Review Board approved this study (No. 15-H184).

A detailed record review was conducted to collect data on demographics and other preoperative risk factors, including age, sex, body mass index (BMI), race and ethnicity, cardiac and pulmonary comorbidities, tobacco use, alcohol intake, diabetes, American Society of Anesthesiologists Physical Status Classification, metabolic equivalent of task, hematologic conditions, and renal disease.

For each procedure, we recorded the type of elective general surgery performed, the diagnosis/indication, pre- and postoperative hemoglobin/hematocrit, intraoperative EBL, length of operation, surgical wound class, length of hospital stay (LOS), intensive care unit (ICU) status, number of hematocrit tests, cardiovascular risk of operation (defined by anesthesia assessment), presence or absence of malignancy, preoperative platelet count, albumin level, preoperative prothrombin time/activated partial thromboplastin time (aPTT), international normalized ratio (INR), hemoglobin A_1c, and incidence of transfusion. Signs and symptoms of anemia were recorded as present if the postoperative vital signs suggested low intravascular volume (pulse > 120 beats/minute, systolic blood pressure < 90 mm Hg, or vasoactive medication requirement [per anesthesia postoperative note]) or if the patient reported or exhibited symptoms of dizziness or fatigue or evidence of clinically apparent bleeding (ie, hematoma formation). Laboratory charges for hematocrit tests and CBC at the NMAVAHCS were used to assess cost.¹¹

To stratify the transfusion risk, patients were distributed among 3 groups based on the following criteria: discharged home the same day as surgery; admitted but did not have postoperative hematocrit testing; and admitted and had postoperative hematocrit testing. We also stratified operations into low or high risk based on the risk for postoperative transfusion (Figure). Recognizing that the American College of Chest Physicians guidelines for perioperative management of antithrombotic therapy places bowel resection in a high-risk category, we designated a surgery as high risk when ≥ 2 patients in the transfusion group had that type of surgery over the 4 years of the study.¹² Otherwise, the operations were deemed low risk.

Statistical Analysis

Numeric analysis used t tests and Binary and categorical variables used Fisher exact tests. P value ≤ .05 was considered statistically significant. SAS software was used for all statistical analyses.

From 2011 through 2014, 1531 patients had elective general surgery at NMVAHCS. Twenty-two patients with preoperative anemia (hematocrit < 30%) and 1 patient who received a transfusion > 72 hours after the operation were excluded. Most elective operations (70%, n = 1075) were performed on an outpatient basis; none involved transfusion. Inguinal hernia repair was most common with 479 operations; 17 patients were treated inpatient of which 2 patients had routine postoperative hematocrit checks; (neither received transfusion). One patient with inguinal hernia surgery received transfusion without routine postoperative hematocrit monitoring.

Of 112 partial colon resections, 1 patient had a postoperative transfusion; and all but 3 received postoperative hematocrit monitoring. Nineteen patients undergoing partial colon resection had a clinical indication for postoperative hematocrit monitoring. None of the 5 patients with partial gastrectomy received a postoperative transfusion. Of 121 elective cholecystectomies, no patients had postoperative transfusion, whereas 34 had postoperative hematocrit monitoring; only 2 patients had a clinical reason for the hematocrit monitoring.

Of 430 elective inpatient operations, 12 received transfusions and 288 patients had ≥ 1 postoperative hematocrit test (67%). All hematocrit tests were requested by the attending surgeon, resident surgeon, or the surgical ICU team. Of the group that had postoperative hematocrit monitoring, there was an average of 4.4 postoperative hematocrit tests per patient (range, 1-44).

There were 12 transfusions for inpatients (2.8%), which is similar to the findings of a recent study of VA general surgery (2.3%).¹³ Five of the 12 patients received intraoperative transfusions while 7 were transfused within 72 hours postoperation. All but 1 patient receiving transfusion had EBL > 199 mL (range, 5-3000; mean, 950 mL; median, 500 mL) and/or signs or symptoms of anemia or other indications for measurement of the postoperative hematocrit. There were no statistically significant differences in patients’ age, sex, BMI, or race and ethnicity between groups receiving and not receiving transfusion (Table 1).

When comparing the transfusion vs the nontransfusion groups (after excluding those with clinical preoperative anemia) the risk factors for transfusion included: relatively low mean preoperative hematocrit (mean, 36.9% vs 42.7%, respectively; P = .003), low postoperative hematocrit (mean, 30.2% vs 37.1%, respectively; P < .001), high EBL (mean, 844 mL vs 109 mL, respectively; P = .005), large infusion of intraoperative fluids (mean, 4625 mL vs 2505 mL, respectively; P = .005), longer duration of operation (mean, 397 min vs 183 min, respectively; P < .001), and longer LOS (mean, 14.5 d vs 4.9 d, respectively; P < .001) (Table 2). Similarly, we found an increased risk for transfusion with high/intermediate cardiovascular risk (vs low), any wound not classified as clean, ICU stay, and postoperative symptoms of anemia.

We found no increased risk for transfusion with ethanol, tobacco, warfarin, or clopidogrel use; polycythemia; thrombocytopenia; preoperative INR; preoperative aPTT; preoperative albumin; Hemoglobin A_1c; or diabetes mellitus; or for operations performed for malignancy. Ten patients in the ICU received transfusion (5.8%) compared with 2 patients (0.8%) not admitted to the ICU.

Operations were deemed high risk when ≥ 2 of patients having that operation received transfusions within 72 hours of their operation. There were 15 abdominoperineal resections; 3 of these received transfusions (20%). There were 7 total abdominal colectomies; 3 of these received transfusions (43%). We therefore had 22 high-risk operations, 6 of which were transfused (27%).

Routine measurement of postoperative hematocrit levels after elective general surgery at NMVAHCS was not necessary. There were 12 transfusions for inpatients (2.8%), which is similar to the findings of a recent study of VA general surgery (2.3%).¹³ We found that routine postoperative hematocrit measurements to assess anemia had little or no effect on clinical decision-making or clinical outcomes.

According to our results, 88% of initial hematocrit tests after elective partial colectomies could have been eliminated; only 32 of 146 patients demonstrated a clinical reason for postoperative hematocrit testing. Similarly, 36 of 40 postcholecystectomy hematocrit tests (90%) could have been eliminated had the surgeons relied on clinical signs indicating possible postoperative anemia (none were transfused). Excluding patients with major intraoperative blood loss (> 300 mL), only 29 of 288 (10%) patients who had postoperative hematocrit tests had a clinical indication for a postoperative hematocrit test (ie, symptoms of anemia and/or active bleeding). One patient with inguinal hernia surgery who received transfusion was taking an anticoagulant and had a clinically indicated hematocrit test for a large hematoma that eventually required reoperation.

Our study found that routine hematocrit checks may actually increase the risk that a patient would receive an unnecessary transfusion. For instance, one elderly patient, after a right colectomy, had 6 hematocrit levels while on a heparin drip and received transfusion despite being asymptomatic. His lowest hematocrit level prior to transfusion was 23.7%. This patient had a total of 18 hematocrit tests. His EBL was 350 mL and his first postoperative HCT level was 33.1%. In another instance, a patient undergoing abdominoperineal resection had a transfusion on postoperative day 1, despite being hypertensive, with a hematocrit that ranged from 26% before transfusion to 31% after the transfusion. These 2 cases illustrate what has been shown in a recent study: A substantial number of patients with colorectal cancer receive unnecessary transfusions.¹⁴ On the other hand, one ileostomy closure patient had 33 hematocrit tests, yet his initial postoperative hematocrit was 37%, and he never received a transfusion. With low-risk surgeries, clinical judgment should dictate when a postoperative hematocrit level is needed. This strategy would have eliminated 206 unnecessary initial postoperative hematocrit tests (72%), could have decreased the number of unnecessary transfusions, and would have saved NMVAHCS about $1600 annually.

Abdominoperineal resections and total abdominal colectomies accounted for a high proportion of transfusions in our study. Inpatient elective operations can be risk stratified and have routine hematocrit tests ordered for patients at high risk. The probability of transfusion was greater in high-risk vs low-risk surgeries; 27% (6 of 22 patients) vs 2% (6 of 408 patients), respectively (P < .001). Since 14 of the 22 patients undergoing high-risk operation already had clinical reasons for a postoperative hematocrit test, we only need to add the remaining 8 patients with high-risk operations to the 74 who had a clinical reason for a hematocrit test and conclude that 82 of 430 patients (19%) had a clinical reason for a hematocrit test, either from signs or symptoms of blood loss or because they were in a high-risk group.

While our elective general surgery cases may not represent many general surgery programs in the US and VA health care systems, we can extrapolate cost savings using the same cost analyses outlined by Kohli and colleagues.¹ Assuming 1.9 million elective inpatient general surgeries per year in the United States with an average cost of $21 per CBC, the annual cost of universal postoperative hematocrit testing would be $40 million.^11,15 If postoperative hematocrit testing were 70% consistent with our findings, the annual cost for hematocrit tests on 51% of the inpatient general surgeries would be approximately $20.4 million. A reduction in routine hematocrit testing to 25% of all inpatient general surgeries (vs our finding that 19% were deemed necessary) results in an annual savings of $30 million. This conservative estimate could be even higher since there were 4.4 hematocrit tests per patient; therefore, we have about $132 million in savings.

Assuming 181,384 elective VA inpatient general surgeries each year, costing $7.14 per CBC (the NMVAHCS cost), the VA could save $1.3 million annually. If postoperative HCT testing were 70% consistent with our findings, the annual cost for hematocrit tests on 50.4% of inpatient general surgery operations would be about $653,000. A reduction in routine hematocrit testing to 25% of all inpatient general surgeries (vs our 19%) results in annual VA savings of $330,000. This conservative estimate could be even higher since there were on average 4.4 hematocrit levels per patient; therefore, we estimate that annual savings for the VA of about $1.45 million.

Limitations

The retrospective chart review nature of this study may have led to selection bias. Only a small number of patients received a transfusion, which may have skewed the data. This study population comes from a single VA medical center; this patient population may not be reflective of other VA medical centers or the US population as a whole. Given that NMVAHCS does not perform hepatic, esophageal, pancreas, or transplant operations, the potential savings to both the US and the VA may be overestimated, but this could be studied in the future by VA medical centers that perform more complex operations.

Conclusions

This study found that over a 4-year period routine postoperative hematocrit tests for patients undergoing elective general surgery at a VA medical center were not necessary. General surgeons routinely order various pre- and postoperative laboratory tests despite their limited utility. Reduction in unneeded routine tests could result in notable savings to the VA without compromising quality of care.

Only general surgery patients undergoing operations that carry a high risk for needing a blood transfusion should have a routine postoperative hematocrit testing. In our study population, the chance of an elective colectomy, cholecystectomy, or hernia patient needing a transfusion was rare. This strategy could eliminate a considerable number of unnecessary blood tests and would potentially yield significant savings.

Following deployment to operations Desert Shield and Desert Storm (Gulf War) in 1990 and 1991, many Gulf War veterans (GWVs) developed chronic, complex symptoms, including pain, dyscognition, and fatigue, with gastrointestinal, skin, and respiratory manifestations. This Gulf War Illness (GWI) is reported to affect about 30% of those deployed. More than 30 years later, there is no consensus as to the etiology of GWI, although some deployment-related exposures have been implicated.¹

Accepted research definitions for GWI include the Centers for Disease Control and Prevention and Kansas definitions.² The US Department of Veterans Affairs (VA) uses the terminology chronic multisymptom illness (CMI), which is an overarching diagnosis under which GWI falls. Although there is no consensus case definition for CMI, there is overlap with conditions such as fibromyalgia, myalgic encephalomyelitis/chronic fatigue syndrome, and irritable bowel syndrome; the VA considers these as qualifying clinical diagnoses.³ The pathophysiology of GWI is also unknown, though a frequently reported unifying feature is that of autonomic nervous system (ANS) dysfunction. Studies have demonstrated differences between veterans with GWI and those without GWI in both the reporting of symptoms attributable to ANS dysfunction and in physiologic evaluations of the ANS.^4-10

Small fiber neuropathy (SFN), a condition with damage to the A-δ and C small nerve fibers, has been proposed as a potential mechanism for the pain and ANS dysfunction experienced in GWI.^11-13 Symptoms of SFN are similar to those of GWI, with pain and ANS symptoms commonly reported.^14,15 There are multiple diagnostic criteria for SFN, the most commonly used requiring the presence of appropriate symptoms in the absence of large fiber neuropathy and a skin biopsy demonstrating reduced intraepidermal nerve fiber density.^16-19 Several conditions reportedly cause SFN, most notably diabetes/prediabetes. Autoimmune disease, vitamin B12 deficiency, monoclonal gammopathies, celiac disease, paraneoplastic syndromes, and sodium channel gene mutations may also contribute to SFN.²⁰ Hyperlipidemia has been identified as a contributor, although it has been variably reported.^21,22

Idiopathic neuropathies, SFN included, may be secondary to neurotoxicant exposures. Agents whose exposure or consumption have been associated with SFN include alcohol most prominently, but also the organic solvent n-hexane, heavy metals, and excess vitamin B₆.^20,23-25 Agents associated with large fiber neuropathy may also have relevance for SFN, as small fibers have been likened to the “canary in the coal mine” in that they may be more susceptible to neurotoxicants and are affected earlier in the disease process.²⁶ In this way, SFN may be the harbinger of large fiber neuropathy in some cases. Of specific relevance for GWVs, organophosphates and carbamates are known to produce a delayed onset large fiber neuropathy.^27-30 Exposure to petrochemical solvents has also been associated with large fiber neuropathies.^31,32

The War Related Illness and Injury Study Center (WRIISC) is a clinical, research, and education center established by Congress in 2001. Its primary focus is on military exposures and postdeployment health of veterans. It is located at 3 sites: East Orange, New Jersey; Washington, DC; and Palo Alto, California. The New Jersey WRIISC began a program to evaluate GWVs with characteristic symptoms for possible SFN with use of a skin biopsy.

We hypothesize that SFN may underly much of GWI symptomatology and may not be accounted for by the putative etiologies detailed in review of the medical literature. This retrospective review of clinical evaluations for SFN in GWVs who sought care at the New Jersey WRIISC explored and addressed the following questions: (1) how common is biopsy-confirmed SFN in veterans with GWI; (2) do veterans with GWI and SFN report more symptoms attributable to ANS dysfunction when compared with veterans with GWI and no SFN; and (3) can SFN in veterans with GWI and SFN be explained by conditions and substances commonly associated with SFN? Institutional review board approval and waiver of consent was obtained from the Veterans Affairs New Jersey Health Care Center for the study.

Methods

A retrospective chart review was conducted on veterans evaluated at the WRIISC from March 1, 2015, to January 31, 2019. Inclusion criteria were: deployment to operations Desert Shield and Desert Storm between August 2, 1990, and February 28, 1991, and skin biopsy conducted at the WRIISC. Skin biopsies were obtained at the discretion of an examining clinician based on clinical indications, including neuropathic pain, ANS symptoms, and/or a fibromyalgia/chronic pain–type presentation.

Electronic health record review explicitly abstracted GWI status, results of the skin biopsy, and ANS symptom burden as determined by the Composite Autonomic Symptom Scale 31 (COMPASS 31) completed at the time of the WRIISC evaluation. Determination of GWI was established as per the clinical opinion of the WRIISC lead clinician or environmental exposure clinician as evidenced by a diagnosis of fibromyalgia or chronic fatigue syndrome, or explicit statement of CMI/GWI in the patient assessment. A diagnosis of SFN was established if clinical signs were present and an intraepidermal nerve fiber density below the lower limits, as compared to normative data from the clinical diagnostic laboratory (Therapath Neuropathology), was documented.

COMPASS 31 assesses symptoms across 6 domains (orthostatic, vasomotor, secretomotor, gastrointestinal, bladder, andpupillomotor). Patients are asked about symptom frequency (rarely to almost always), severity (mild to severe), and improvement (much worse to completely gone). Individual domain scores and a total weighted score (0-100) have demonstrated good validity, reliability, and consistency in SFN.^33,34

In veterans with GWI and documented SFN, a health record review was performed to identify potential etiologies for SFN (Appendix).

Statistical Analysis

Microsoft Excel and IBM SPSS 12.0.1 for Windows were used for data collection and statistical analysis. Fisher exact test was used for comparing the prevalence of SFN in veterans with GWI vs without GWI. The independent samples t test was used for comparing COMPASS 31 scores for veterans with GWI by SFN status. α < .05 was used for determining statistical significance. For those GWVs documented with SFN and GWI, potential explanations were documented in total and by condition.

Results

From March 1, 2015, to January 31, 2019, 141 GWVs received a comprehensive in person clinical evaluation at the WRIISC and 51 veterans (36%) received a skin biopsy and were included in this retrospective observational study (Figure). The mean age was 48.6 years, and the majority were male and served in the US Army. Skin biopsies met clinical criteria for GWI for 42 (82%) and 24 of 42 (57%) were determined to have SFN. Four of 9 (44%) veterans without GWI had positive SFN biopsies, though this difference was not statistically significant (Table 1). Veterans with SFN but no GWI were not included in the further analysis.

Thirty-five veterans with GWI—18 with SFN and 17 without SFN—completed the COMPASS 31 (Table 2). COMPASS 31 data were not analyzed for veterans without GWI. Individual domain scores and the difference in COMPASS 31 scores for veterans with GWI and SFN vs GWI and no SFN (38.3 vs 37.8, respectively) were not statistically significant.

Sixteen of 24 veterans with GWI and SFN (67%) had ≥ 1 conditions that could potentially be responsible for SFN (Table 3), including 11 veterans (46%) with prediabetes/diabetes. Hyperlipidemia is only variably reported as a cause of SFN; when included, 19 of 24 (79%) SFN cases were accounted for. We could not identify a medical explanation for SFN in 5 of 24 veterans (21%) with GWI, which were deemed to be idiopathic.

Discussion

Biopsy-confirmed SFN was present in more than half of our sample of veterans with GWI, which is broadly consistent with what has been reported in the literature.^13,35-38 In this clinical observation study, SFN was similarly prevalent in veterans with and without GWI; although it should be noted that biopsies only were obtained when there was a strong clinical suspicion for SFN. Almost half of patients with GWI did not have SFN, so our study does not support SFN as the underlying explanation for all GWI. Although our data cannot provide clinical guidance as to when skin biopsy may be indicated in GWI, work done in fibromyalgia found symptoms of dysautonomia and paresthesias are more specific for SFN and may be useful to help guide medical decision making.³⁹

Veterans with GWI in our clinical sample reported a high burden of clinical symptoms conceivably attributable to ANS dysfunction. This symptom reporting is consistent with that seen in other GWI studies, as well as in other studies of SFN.^{4,5,7-9,14,15,34,38,40} Our clinical sample of veterans with GWI found no differences in the ANS symptom reporting between those with and without SFN. Therefore, our study cannot support SFN alone as accounting for ANS symptom burden in patients with GWI.

Two-thirds of biopsy-confirmed SFN in our clinical sample of veterans with GWI could potentially be explained by established medical conditions. As in other studies of SFN, prediabetes and diabetes represented a plurality (46%). Even after considering hyperlipidemia as a potential explanation, about 21% of SFN cases in veterans with GWI still were deemed idiopathic.

Evidence supports certain environmental agents as causal factors for GWI. Neurotoxicants reportedly related to GWI include pesticides (particularly organophosphates and carbamates), pyridostigmine bromide (used during the Gulf War as a prophylactic agent against the use of chemical weapons), and low levels of the nerve agent sarin from environmental contamination due to chemical weapons detonations.¹ Some of these agents have been implicated in neuropathy as well.^1,28-30 It is biologically plausible that deployment-related exposures could trigger SFN, though the traditional consensus has been that remote exposure to neurotoxic substances is unlikely to produce neuropathy that presents many years after the exposure.⁴¹ In the WRIISC clinical experience, however, veterans often report that their neuropathic symptoms predate the diagnosis of the associated medical conditions, sometimes by decades. It is conceivable that remote exposures may trigger the condition that is then potentiated by ongoing exposures, metabolic factors, and/or other medical conditions. These may perpetuate neuropathic symptoms and the illness experience of affected veterans. Our clinical observation study cannot clarify the extent to which this may be the case. Despite these findings and arguments, an environmental contribution to SFN cannot be discounted, and further research is needed to explore a potential relationship.

Limitations

This study’s conclusions are limited by its observational/retrospective design in a relatively small clinical sample of veterans evaluated at a tertiary referral center for postdeployment exposure-related health concerns. The WRIISC clinical sample is not representative of all GWVs or even of all veterans with GWI, as there is inherent selection bias as to who gets referred to and evaluated at the WRIISC. As with studies based on retrospective chart review, data are reliant on clinical documentation andaccuracy/consistency of the reviewer. Evaluation for SFN with skin biopsy is an invasive procedure and was performed when a high index of clinical suspicion for this condition existed, possibly representing confirmation bias. Therefore, the relatively high prevalence ofbiopsy-confirmed SFN seen in our clinical sample cannot be generalized to GWVs as a whole or even to veterans with GWI.

Assessment of autonomic dysfunction was based on COMPASS 31 symptom reporting by an small subset of the clinical cohort. Symptom reporting may not be reflective of true abnormality in ANS function. Physiologic tests of the ANS were not performed; such studies could more objectively establish whether ANS dysfunction is more prevalent in GWI veterans with SFN.

Evaluation for all potential etiologic/contributory conditions to SFN was not exhaustive. For example, sodium channel gene mutations have been documented to account for up to one-third of all cases of idiopathic SFN.⁴² For those cases in which no compelling etiology was identified, it is plausible that medical explanations for SFN may be found on further investigation.

Clinical assessments at the WRIISC were performed on GWVs ≥ 26 years after their deployment-related exposures. Other conditions/exposures may have occurred in the interim. What is not clear is whether the SFN predated the onset of any of these medical conditions or other putative contributors. This observational study is not able to tease out a temporal association to make a cause-and-effect assessment.

Conclusions

Retrospective analysis of clinical data of veterans evaluated at a specialized center for postdeployment health demonstrated that skin biopsy–confirmed SFN was prevalent, but not ubiquitous, in veterans with GWI. Symptom that may be attributed to ANS dysfunction in this clinical sample was consistent with literature on SFN and with GWI, but we could not definitively attribute ANS symptoms to SFN. Our study does not support the hypothesis that GWI symptoms are solely due to SFN, though it may still be relevant in a subset of veterans with GWI with strongly suggestive clinical features. We were able to identify a potential etiology for SFN in most veterans with GWI. Further investigations are recommended to explore any potential relationship between Gulf War exposures and SFN.

Following deployment to operations Desert Shield and Desert Storm (Gulf War) in 1990 and 1991, many Gulf War veterans (GWVs) developed chronic, complex symptoms, including pain, dyscognition, and fatigue, with gastrointestinal, skin, and respiratory manifestations. This Gulf War Illness (GWI) is reported to affect about 30% of those deployed. More than 30 years later, there is no consensus as to the etiology of GWI, although some deployment-related exposures have been implicated.¹

Accepted research definitions for GWI include the Centers for Disease Control and Prevention and Kansas definitions.² The US Department of Veterans Affairs (VA) uses the terminology chronic multisymptom illness (CMI), which is an overarching diagnosis under which GWI falls. Although there is no consensus case definition for CMI, there is overlap with conditions such as fibromyalgia, myalgic encephalomyelitis/chronic fatigue syndrome, and irritable bowel syndrome; the VA considers these as qualifying clinical diagnoses.³ The pathophysiology of GWI is also unknown, though a frequently reported unifying feature is that of autonomic nervous system (ANS) dysfunction. Studies have demonstrated differences between veterans with GWI and those without GWI in both the reporting of symptoms attributable to ANS dysfunction and in physiologic evaluations of the ANS.^4-10

Small fiber neuropathy (SFN), a condition with damage to the A-δ and C small nerve fibers, has been proposed as a potential mechanism for the pain and ANS dysfunction experienced in GWI.^11-13 Symptoms of SFN are similar to those of GWI, with pain and ANS symptoms commonly reported.^14,15 There are multiple diagnostic criteria for SFN, the most commonly used requiring the presence of appropriate symptoms in the absence of large fiber neuropathy and a skin biopsy demonstrating reduced intraepidermal nerve fiber density.^16-19 Several conditions reportedly cause SFN, most notably diabetes/prediabetes. Autoimmune disease, vitamin B12 deficiency, monoclonal gammopathies, celiac disease, paraneoplastic syndromes, and sodium channel gene mutations may also contribute to SFN.²⁰ Hyperlipidemia has been identified as a contributor, although it has been variably reported.^21,22

Idiopathic neuropathies, SFN included, may be secondary to neurotoxicant exposures. Agents whose exposure or consumption have been associated with SFN include alcohol most prominently, but also the organic solvent n-hexane, heavy metals, and excess vitamin B₆.^20,23-25 Agents associated with large fiber neuropathy may also have relevance for SFN, as small fibers have been likened to the “canary in the coal mine” in that they may be more susceptible to neurotoxicants and are affected earlier in the disease process.²⁶ In this way, SFN may be the harbinger of large fiber neuropathy in some cases. Of specific relevance for GWVs, organophosphates and carbamates are known to produce a delayed onset large fiber neuropathy.^27-30 Exposure to petrochemical solvents has also been associated with large fiber neuropathies.^31,32

The War Related Illness and Injury Study Center (WRIISC) is a clinical, research, and education center established by Congress in 2001. Its primary focus is on military exposures and postdeployment health of veterans. It is located at 3 sites: East Orange, New Jersey; Washington, DC; and Palo Alto, California. The New Jersey WRIISC began a program to evaluate GWVs with characteristic symptoms for possible SFN with use of a skin biopsy.

We hypothesize that SFN may underly much of GWI symptomatology and may not be accounted for by the putative etiologies detailed in review of the medical literature. This retrospective review of clinical evaluations for SFN in GWVs who sought care at the New Jersey WRIISC explored and addressed the following questions: (1) how common is biopsy-confirmed SFN in veterans with GWI; (2) do veterans with GWI and SFN report more symptoms attributable to ANS dysfunction when compared with veterans with GWI and no SFN; and (3) can SFN in veterans with GWI and SFN be explained by conditions and substances commonly associated with SFN? Institutional review board approval and waiver of consent was obtained from the Veterans Affairs New Jersey Health Care Center for the study.

Methods

A retrospective chart review was conducted on veterans evaluated at the WRIISC from March 1, 2015, to January 31, 2019. Inclusion criteria were: deployment to operations Desert Shield and Desert Storm between August 2, 1990, and February 28, 1991, and skin biopsy conducted at the WRIISC. Skin biopsies were obtained at the discretion of an examining clinician based on clinical indications, including neuropathic pain, ANS symptoms, and/or a fibromyalgia/chronic pain–type presentation.

Electronic health record review explicitly abstracted GWI status, results of the skin biopsy, and ANS symptom burden as determined by the Composite Autonomic Symptom Scale 31 (COMPASS 31) completed at the time of the WRIISC evaluation. Determination of GWI was established as per the clinical opinion of the WRIISC lead clinician or environmental exposure clinician as evidenced by a diagnosis of fibromyalgia or chronic fatigue syndrome, or explicit statement of CMI/GWI in the patient assessment. A diagnosis of SFN was established if clinical signs were present and an intraepidermal nerve fiber density below the lower limits, as compared to normative data from the clinical diagnostic laboratory (Therapath Neuropathology), was documented.

COMPASS 31 assesses symptoms across 6 domains (orthostatic, vasomotor, secretomotor, gastrointestinal, bladder, andpupillomotor). Patients are asked about symptom frequency (rarely to almost always), severity (mild to severe), and improvement (much worse to completely gone). Individual domain scores and a total weighted score (0-100) have demonstrated good validity, reliability, and consistency in SFN.^33,34

In veterans with GWI and documented SFN, a health record review was performed to identify potential etiologies for SFN (Appendix).

Statistical Analysis

Microsoft Excel and IBM SPSS 12.0.1 for Windows were used for data collection and statistical analysis. Fisher exact test was used for comparing the prevalence of SFN in veterans with GWI vs without GWI. The independent samples t test was used for comparing COMPASS 31 scores for veterans with GWI by SFN status. α < .05 was used for determining statistical significance. For those GWVs documented with SFN and GWI, potential explanations were documented in total and by condition.

Results

From March 1, 2015, to January 31, 2019, 141 GWVs received a comprehensive in person clinical evaluation at the WRIISC and 51 veterans (36%) received a skin biopsy and were included in this retrospective observational study (Figure). The mean age was 48.6 years, and the majority were male and served in the US Army. Skin biopsies met clinical criteria for GWI for 42 (82%) and 24 of 42 (57%) were determined to have SFN. Four of 9 (44%) veterans without GWI had positive SFN biopsies, though this difference was not statistically significant (Table 1). Veterans with SFN but no GWI were not included in the further analysis.

Thirty-five veterans with GWI—18 with SFN and 17 without SFN—completed the COMPASS 31 (Table 2). COMPASS 31 data were not analyzed for veterans without GWI. Individual domain scores and the difference in COMPASS 31 scores for veterans with GWI and SFN vs GWI and no SFN (38.3 vs 37.8, respectively) were not statistically significant.

Sixteen of 24 veterans with GWI and SFN (67%) had ≥ 1 conditions that could potentially be responsible for SFN (Table 3), including 11 veterans (46%) with prediabetes/diabetes. Hyperlipidemia is only variably reported as a cause of SFN; when included, 19 of 24 (79%) SFN cases were accounted for. We could not identify a medical explanation for SFN in 5 of 24 veterans (21%) with GWI, which were deemed to be idiopathic.

Discussion

Biopsy-confirmed SFN was present in more than half of our sample of veterans with GWI, which is broadly consistent with what has been reported in the literature.^13,35-38 In this clinical observation study, SFN was similarly prevalent in veterans with and without GWI; although it should be noted that biopsies only were obtained when there was a strong clinical suspicion for SFN. Almost half of patients with GWI did not have SFN, so our study does not support SFN as the underlying explanation for all GWI. Although our data cannot provide clinical guidance as to when skin biopsy may be indicated in GWI, work done in fibromyalgia found symptoms of dysautonomia and paresthesias are more specific for SFN and may be useful to help guide medical decision making.³⁹

Veterans with GWI in our clinical sample reported a high burden of clinical symptoms conceivably attributable to ANS dysfunction. This symptom reporting is consistent with that seen in other GWI studies, as well as in other studies of SFN.^{4,5,7-9,14,15,34,38,40} Our clinical sample of veterans with GWI found no differences in the ANS symptom reporting between those with and without SFN. Therefore, our study cannot support SFN alone as accounting for ANS symptom burden in patients with GWI.

Two-thirds of biopsy-confirmed SFN in our clinical sample of veterans with GWI could potentially be explained by established medical conditions. As in other studies of SFN, prediabetes and diabetes represented a plurality (46%). Even after considering hyperlipidemia as a potential explanation, about 21% of SFN cases in veterans with GWI still were deemed idiopathic.

Evidence supports certain environmental agents as causal factors for GWI. Neurotoxicants reportedly related to GWI include pesticides (particularly organophosphates and carbamates), pyridostigmine bromide (used during the Gulf War as a prophylactic agent against the use of chemical weapons), and low levels of the nerve agent sarin from environmental contamination due to chemical weapons detonations.¹ Some of these agents have been implicated in neuropathy as well.^1,28-30 It is biologically plausible that deployment-related exposures could trigger SFN, though the traditional consensus has been that remote exposure to neurotoxic substances is unlikely to produce neuropathy that presents many years after the exposure.⁴¹ In the WRIISC clinical experience, however, veterans often report that their neuropathic symptoms predate the diagnosis of the associated medical conditions, sometimes by decades. It is conceivable that remote exposures may trigger the condition that is then potentiated by ongoing exposures, metabolic factors, and/or other medical conditions. These may perpetuate neuropathic symptoms and the illness experience of affected veterans. Our clinical observation study cannot clarify the extent to which this may be the case. Despite these findings and arguments, an environmental contribution to SFN cannot be discounted, and further research is needed to explore a potential relationship.

Limitations

This study’s conclusions are limited by its observational/retrospective design in a relatively small clinical sample of veterans evaluated at a tertiary referral center for postdeployment exposure-related health concerns. The WRIISC clinical sample is not representative of all GWVs or even of all veterans with GWI, as there is inherent selection bias as to who gets referred to and evaluated at the WRIISC. As with studies based on retrospective chart review, data are reliant on clinical documentation andaccuracy/consistency of the reviewer. Evaluation for SFN with skin biopsy is an invasive procedure and was performed when a high index of clinical suspicion for this condition existed, possibly representing confirmation bias. Therefore, the relatively high prevalence ofbiopsy-confirmed SFN seen in our clinical sample cannot be generalized to GWVs as a whole or even to veterans with GWI.

Assessment of autonomic dysfunction was based on COMPASS 31 symptom reporting by an small subset of the clinical cohort. Symptom reporting may not be reflective of true abnormality in ANS function. Physiologic tests of the ANS were not performed; such studies could more objectively establish whether ANS dysfunction is more prevalent in GWI veterans with SFN.

Evaluation for all potential etiologic/contributory conditions to SFN was not exhaustive. For example, sodium channel gene mutations have been documented to account for up to one-third of all cases of idiopathic SFN.⁴² For those cases in which no compelling etiology was identified, it is plausible that medical explanations for SFN may be found on further investigation.

Clinical assessments at the WRIISC were performed on GWVs ≥ 26 years after their deployment-related exposures. Other conditions/exposures may have occurred in the interim. What is not clear is whether the SFN predated the onset of any of these medical conditions or other putative contributors. This observational study is not able to tease out a temporal association to make a cause-and-effect assessment.

Conclusions

Retrospective analysis of clinical data of veterans evaluated at a specialized center for postdeployment health demonstrated that skin biopsy–confirmed SFN was prevalent, but not ubiquitous, in veterans with GWI. Symptom that may be attributed to ANS dysfunction in this clinical sample was consistent with literature on SFN and with GWI, but we could not definitively attribute ANS symptoms to SFN. Our study does not support the hypothesis that GWI symptoms are solely due to SFN, though it may still be relevant in a subset of veterans with GWI with strongly suggestive clinical features. We were able to identify a potential etiology for SFN in most veterans with GWI. Further investigations are recommended to explore any potential relationship between Gulf War exposures and SFN.

Following deployment to operations Desert Shield and Desert Storm (Gulf War) in 1990 and 1991, many Gulf War veterans (GWVs) developed chronic, complex symptoms, including pain, dyscognition, and fatigue, with gastrointestinal, skin, and respiratory manifestations. This Gulf War Illness (GWI) is reported to affect about 30% of those deployed. More than 30 years later, there is no consensus as to the etiology of GWI, although some deployment-related exposures have been implicated.¹

Accepted research definitions for GWI include the Centers for Disease Control and Prevention and Kansas definitions.² The US Department of Veterans Affairs (VA) uses the terminology chronic multisymptom illness (CMI), which is an overarching diagnosis under which GWI falls. Although there is no consensus case definition for CMI, there is overlap with conditions such as fibromyalgia, myalgic encephalomyelitis/chronic fatigue syndrome, and irritable bowel syndrome; the VA considers these as qualifying clinical diagnoses.³ The pathophysiology of GWI is also unknown, though a frequently reported unifying feature is that of autonomic nervous system (ANS) dysfunction. Studies have demonstrated differences between veterans with GWI and those without GWI in both the reporting of symptoms attributable to ANS dysfunction and in physiologic evaluations of the ANS.^4-10

Small fiber neuropathy (SFN), a condition with damage to the A-δ and C small nerve fibers, has been proposed as a potential mechanism for the pain and ANS dysfunction experienced in GWI.^11-13 Symptoms of SFN are similar to those of GWI, with pain and ANS symptoms commonly reported.^14,15 There are multiple diagnostic criteria for SFN, the most commonly used requiring the presence of appropriate symptoms in the absence of large fiber neuropathy and a skin biopsy demonstrating reduced intraepidermal nerve fiber density.^16-19 Several conditions reportedly cause SFN, most notably diabetes/prediabetes. Autoimmune disease, vitamin B12 deficiency, monoclonal gammopathies, celiac disease, paraneoplastic syndromes, and sodium channel gene mutations may also contribute to SFN.²⁰ Hyperlipidemia has been identified as a contributor, although it has been variably reported.^21,22

Idiopathic neuropathies, SFN included, may be secondary to neurotoxicant exposures. Agents whose exposure or consumption have been associated with SFN include alcohol most prominently, but also the organic solvent n-hexane, heavy metals, and excess vitamin B₆.^20,23-25 Agents associated with large fiber neuropathy may also have relevance for SFN, as small fibers have been likened to the “canary in the coal mine” in that they may be more susceptible to neurotoxicants and are affected earlier in the disease process.²⁶ In this way, SFN may be the harbinger of large fiber neuropathy in some cases. Of specific relevance for GWVs, organophosphates and carbamates are known to produce a delayed onset large fiber neuropathy.^27-30 Exposure to petrochemical solvents has also been associated with large fiber neuropathies.^31,32

The War Related Illness and Injury Study Center (WRIISC) is a clinical, research, and education center established by Congress in 2001. Its primary focus is on military exposures and postdeployment health of veterans. It is located at 3 sites: East Orange, New Jersey; Washington, DC; and Palo Alto, California. The New Jersey WRIISC began a program to evaluate GWVs with characteristic symptoms for possible SFN with use of a skin biopsy.

We hypothesize that SFN may underly much of GWI symptomatology and may not be accounted for by the putative etiologies detailed in review of the medical literature. This retrospective review of clinical evaluations for SFN in GWVs who sought care at the New Jersey WRIISC explored and addressed the following questions: (1) how common is biopsy-confirmed SFN in veterans with GWI; (2) do veterans with GWI and SFN report more symptoms attributable to ANS dysfunction when compared with veterans with GWI and no SFN; and (3) can SFN in veterans with GWI and SFN be explained by conditions and substances commonly associated with SFN? Institutional review board approval and waiver of consent was obtained from the Veterans Affairs New Jersey Health Care Center for the study.

Methods

A retrospective chart review was conducted on veterans evaluated at the WRIISC from March 1, 2015, to January 31, 2019. Inclusion criteria were: deployment to operations Desert Shield and Desert Storm between August 2, 1990, and February 28, 1991, and skin biopsy conducted at the WRIISC. Skin biopsies were obtained at the discretion of an examining clinician based on clinical indications, including neuropathic pain, ANS symptoms, and/or a fibromyalgia/chronic pain–type presentation.

Electronic health record review explicitly abstracted GWI status, results of the skin biopsy, and ANS symptom burden as determined by the Composite Autonomic Symptom Scale 31 (COMPASS 31) completed at the time of the WRIISC evaluation. Determination of GWI was established as per the clinical opinion of the WRIISC lead clinician or environmental exposure clinician as evidenced by a diagnosis of fibromyalgia or chronic fatigue syndrome, or explicit statement of CMI/GWI in the patient assessment. A diagnosis of SFN was established if clinical signs were present and an intraepidermal nerve fiber density below the lower limits, as compared to normative data from the clinical diagnostic laboratory (Therapath Neuropathology), was documented.

COMPASS 31 assesses symptoms across 6 domains (orthostatic, vasomotor, secretomotor, gastrointestinal, bladder, andpupillomotor). Patients are asked about symptom frequency (rarely to almost always), severity (mild to severe), and improvement (much worse to completely gone). Individual domain scores and a total weighted score (0-100) have demonstrated good validity, reliability, and consistency in SFN.^33,34

In veterans with GWI and documented SFN, a health record review was performed to identify potential etiologies for SFN (Appendix).

Statistical Analysis

Microsoft Excel and IBM SPSS 12.0.1 for Windows were used for data collection and statistical analysis. Fisher exact test was used for comparing the prevalence of SFN in veterans with GWI vs without GWI. The independent samples t test was used for comparing COMPASS 31 scores for veterans with GWI by SFN status. α < .05 was used for determining statistical significance. For those GWVs documented with SFN and GWI, potential explanations were documented in total and by condition.

Results

From March 1, 2015, to January 31, 2019, 141 GWVs received a comprehensive in person clinical evaluation at the WRIISC and 51 veterans (36%) received a skin biopsy and were included in this retrospective observational study (Figure). The mean age was 48.6 years, and the majority were male and served in the US Army. Skin biopsies met clinical criteria for GWI for 42 (82%) and 24 of 42 (57%) were determined to have SFN. Four of 9 (44%) veterans without GWI had positive SFN biopsies, though this difference was not statistically significant (Table 1). Veterans with SFN but no GWI were not included in the further analysis.

Thirty-five veterans with GWI—18 with SFN and 17 without SFN—completed the COMPASS 31 (Table 2). COMPASS 31 data were not analyzed for veterans without GWI. Individual domain scores and the difference in COMPASS 31 scores for veterans with GWI and SFN vs GWI and no SFN (38.3 vs 37.8, respectively) were not statistically significant.

Sixteen of 24 veterans with GWI and SFN (67%) had ≥ 1 conditions that could potentially be responsible for SFN (Table 3), including 11 veterans (46%) with prediabetes/diabetes. Hyperlipidemia is only variably reported as a cause of SFN; when included, 19 of 24 (79%) SFN cases were accounted for. We could not identify a medical explanation for SFN in 5 of 24 veterans (21%) with GWI, which were deemed to be idiopathic.

Discussion

Biopsy-confirmed SFN was present in more than half of our sample of veterans with GWI, which is broadly consistent with what has been reported in the literature.^13,35-38 In this clinical observation study, SFN was similarly prevalent in veterans with and without GWI; although it should be noted that biopsies only were obtained when there was a strong clinical suspicion for SFN. Almost half of patients with GWI did not have SFN, so our study does not support SFN as the underlying explanation for all GWI. Although our data cannot provide clinical guidance as to when skin biopsy may be indicated in GWI, work done in fibromyalgia found symptoms of dysautonomia and paresthesias are more specific for SFN and may be useful to help guide medical decision making.³⁹

Veterans with GWI in our clinical sample reported a high burden of clinical symptoms conceivably attributable to ANS dysfunction. This symptom reporting is consistent with that seen in other GWI studies, as well as in other studies of SFN.^{4,5,7-9,14,15,34,38,40} Our clinical sample of veterans with GWI found no differences in the ANS symptom reporting between those with and without SFN. Therefore, our study cannot support SFN alone as accounting for ANS symptom burden in patients with GWI.

Two-thirds of biopsy-confirmed SFN in our clinical sample of veterans with GWI could potentially be explained by established medical conditions. As in other studies of SFN, prediabetes and diabetes represented a plurality (46%). Even after considering hyperlipidemia as a potential explanation, about 21% of SFN cases in veterans with GWI still were deemed idiopathic.

Evidence supports certain environmental agents as causal factors for GWI. Neurotoxicants reportedly related to GWI include pesticides (particularly organophosphates and carbamates), pyridostigmine bromide (used during the Gulf War as a prophylactic agent against the use of chemical weapons), and low levels of the nerve agent sarin from environmental contamination due to chemical weapons detonations.¹ Some of these agents have been implicated in neuropathy as well.^1,28-30 It is biologically plausible that deployment-related exposures could trigger SFN, though the traditional consensus has been that remote exposure to neurotoxic substances is unlikely to produce neuropathy that presents many years after the exposure.⁴¹ In the WRIISC clinical experience, however, veterans often report that their neuropathic symptoms predate the diagnosis of the associated medical conditions, sometimes by decades. It is conceivable that remote exposures may trigger the condition that is then potentiated by ongoing exposures, metabolic factors, and/or other medical conditions. These may perpetuate neuropathic symptoms and the illness experience of affected veterans. Our clinical observation study cannot clarify the extent to which this may be the case. Despite these findings and arguments, an environmental contribution to SFN cannot be discounted, and further research is needed to explore a potential relationship.

Limitations

This study’s conclusions are limited by its observational/retrospective design in a relatively small clinical sample of veterans evaluated at a tertiary referral center for postdeployment exposure-related health concerns. The WRIISC clinical sample is not representative of all GWVs or even of all veterans with GWI, as there is inherent selection bias as to who gets referred to and evaluated at the WRIISC. As with studies based on retrospective chart review, data are reliant on clinical documentation andaccuracy/consistency of the reviewer. Evaluation for SFN with skin biopsy is an invasive procedure and was performed when a high index of clinical suspicion for this condition existed, possibly representing confirmation bias. Therefore, the relatively high prevalence ofbiopsy-confirmed SFN seen in our clinical sample cannot be generalized to GWVs as a whole or even to veterans with GWI.

Assessment of autonomic dysfunction was based on COMPASS 31 symptom reporting by an small subset of the clinical cohort. Symptom reporting may not be reflective of true abnormality in ANS function. Physiologic tests of the ANS were not performed; such studies could more objectively establish whether ANS dysfunction is more prevalent in GWI veterans with SFN.

Evaluation for all potential etiologic/contributory conditions to SFN was not exhaustive. For example, sodium channel gene mutations have been documented to account for up to one-third of all cases of idiopathic SFN.⁴² For those cases in which no compelling etiology was identified, it is plausible that medical explanations for SFN may be found on further investigation.

Clinical assessments at the WRIISC were performed on GWVs ≥ 26 years after their deployment-related exposures. Other conditions/exposures may have occurred in the interim. What is not clear is whether the SFN predated the onset of any of these medical conditions or other putative contributors. This observational study is not able to tease out a temporal association to make a cause-and-effect assessment.

Conclusions

Retrospective analysis of clinical data of veterans evaluated at a specialized center for postdeployment health demonstrated that skin biopsy–confirmed SFN was prevalent, but not ubiquitous, in veterans with GWI. Symptom that may be attributed to ANS dysfunction in this clinical sample was consistent with literature on SFN and with GWI, but we could not definitively attribute ANS symptoms to SFN. Our study does not support the hypothesis that GWI symptoms are solely due to SFN, though it may still be relevant in a subset of veterans with GWI with strongly suggestive clinical features. We were able to identify a potential etiology for SFN in most veterans with GWI. Further investigations are recommended to explore any potential relationship between Gulf War exposures and SFN.