Original Research

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.

Wolff-Parkinson-White (WPW) syndrome is characterized by the presence of ≥ 1 accessory pathways and the development of both recurrent paroxysmal atrial fibrillation (AF) and supraventricular tachycardia that can lead to further malignant arrhythmias resulting in sudden cardiac death (SCD).^1-7 Historically, incidental, ventricular pre-excitation on electrocardiogram has conferred a relatively low SCD risk in adults; however, newer WPW syndrome data suggest the endpoint may not be as benign as previously thought.⁷ The current literature has defined atrioventricular reentrant tachycardia triggering AF, rather than symptoms, as an independent risk factor for malignant arrhythmias. Still, long-term data detailing the association of AF with serious cardiac events and death in patients with WPW syndrome are still limited.^1-7

While previous guidelines for the treatment of WPW syndrome only recommended routine electrophysiology testing (EPT) with liberal catheter ablation for symptomatic individuals, the 2015 American College of Cardiology/American Heart Association/Heart Rhythm Society guidelines now suggest its potential benefit for risk stratification in the asymptomatic population.^8-12 Given the limited existing data, more long-term studies are needed to corroborate the latest EPT recommendations before routinely applying them in practice. Furthermore, since concomitant AF can lead to adverse cardiac outcomes in patients with WPW syndrome, additional data evaluating this association are also necessary. In this study, we aimed to determine the impact of atrial fibrillation and/or flutter (AF/AFL) on adverse cardiac outcomes and mortality in patients with WPW syndrome.

METHODS

This study used data from the Military Health System (MHS) Database Repository. The MHS is one of the largest health care systems in the country and includes information on about 10 million active duty and retired military service members and their families (51% male; 49% female).^13,14 Data were fully anonymized and complied in accordance with federal and state laws, including the Health Insurance Portability and Accountability Act of 1996. The Naval Medical Center Portsmouth Institutional Review Board approved this study.

Study Design

This retrospective, observational cohort study identified MHS patients with WPW syndrome from January 1, 2014, to December 31, 2019. Patients were included if they had ≥ 2 International Classification of Diseases, Ninth Revision (ICD-9) or International Classification of Diseases, Tenth Revision (ICD-10) diagnosis codes for WPW syndrome (ICD-9, 426.7; ICD-10, I45.6) on separate dates; were aged ≥ 18 years at index date; and had ≥ 1 year of continuous eligibility prior to the index date (enrollment gaps ≤ 30 days were considered continuous). Patients were then divided into 2 subgroups by the presence or absence of AF/AFL using diagnostic codes. Patients were excluded if they had evidence of an implantable cardioverter-defibrillator, permanent pacemaker or were missing age or sex data. Patients were followed from index date until the first occurrence of the outcome of interest, MHS disenrollment, or the end of the study period.

Cardiac composite outcomes comprised of sudden cardiac arrest (SCA), ventricular fibrillation (VF), ventricular tachycardia and death, as well as death specifically, were the outcomes of interest and assessed after index date using ICD-9 and ICD-10 codes. Death was defined as all-cause mortality. Time to event was calculated based on the date of the initial component from the composite outcome and date of death specifically for mortality. Those not experiencing an outcome were followed until MHS disenrollment or the end of the study period.

Various patient characteristics were assessed at index including age, sex, military sponsor (the patient’s active or retired duty member through which their dependent receives TRICARE benefits) rank and branch, geographic region, type of US Department of Defense beneficiary, and index year. Clinical characteristics were assessed over a 1-year baseline period prior to index date and included the number of cardiologist and clinical visits for WPW syndrome, Charlson Comorbidity Index (CCI) scores calculated from diagnostic codes outlined in the Quan coding method, and preindex time.¹⁵ Comorbidities were assessed at baseline and defined as having ≥ 1 ICD-9 or ICD-10 code for a corresponding condition within 1 year prior to index.

Baseline characteristics were assessed and descriptive statistics for categorical and continuous variables were presented accordingly. To assess bivariate association with exposure, χ² tests were used to compare categorical variables, while t tests were used to compare continuous variables by exposure status. Incidence proportions and rates were reported for each outcome of interest. Kaplan-Meier curves were constructed to assess the bivariate association between exposure and study outcomes. Cox proportional hazard modeling was performed to estimate the association between AF/AFL and time to each of the outcomes. Multiple models were designed to assess cardiac and metabolic covariates, in addition to baseline characteristics. This included a base model adjusted for age, sex, military sponsor rank and branch, geographic region, and duty status.

Additional models adjusted for cardiac and metabolic confounders and CCI score. A comprehensive model included the base, cardiac, and metabolic covariates. Multicollinearity between covariates was assessed. Variables with a variance inflation factor > 4 or a tolerance level < 0.1 were added to the models. Cox proportional hazard models were used to estimate the unadjusted and adjusted hazard ratios (HRs) and 95% CIs of the association between AF/AFL and the study outcomes. Data were analyzed using SAS, version 9.4 for Windows.

RESULTS

From 2014 through 2019, 35,539 patients with WPW syndrome were identified in the MHS, 5291 had AF/AFL (14.9%); 19,961 were female (56.2%), the mean (SD) age was 62.9 (18.0) years, and 11,742 were aged ≥ 75 years (33.0%) (Table 1).

There were 4121 (11.6%), 322 (0.9%), and 848 (2.4%) patients with AF, AFL, and both arrhythmias, respectively. The mean (SD) number of cardiology visits was 3.9 (3.0). The mean (SD) baseline CCI score for the AF/AFL subgroup was 5.9 (3.5) vs 3.7 (2.2) for the non-AF/AFL subgroup (P < .001). The most prevalent comorbid conditions were hypertension, hyperlipidemia, chronic obstructive pulmonary disease, and diabetes (P < .001) (Figure 1).

Composite Outcomes

In the overall cohort, during a mean (SD) follow-up time of 3.4 (2.0) years comprising 119,682 total person-years, the components of the composite outcome occurred 6506 times with an incidence rate of 5.44 per 100 person-years. Ventricular tachycardia was the most common event, occurring 3281 times with an incidence rate of 2.74 per 100 person-years. SCA and VF occurred 841 and 135 times with incidence rates of 0.70 and 0.11 per 100 person-years, respectively. Death was the initial event 2249 times with an incidence rate of 1.88 per 100 person-years. Figure 2 shows the Kaplan-Meier curve of cardiac composite outcome by AF/AFL status.

The subgroup with AF/AFL comprised 17,412 total person-years and 1424 cardiac composite incidences compared with 102,270 person years and 5082 incidences in the no AF/AFL group (Table 2). Comparing AF/AFL vs no AF/AFL incidence rates were 8.18 vs 4.97 per 100 person-years, respectively (P < .001). SCA and VF occurred 233 and 38 times and respectively had incidence rates of 1.34 and 0.22 per 100 person-years in the AF/AFL group vs 0.59 and 0.09 per 100 person-years in the no AF/AFL group (P < .001). There were 549 deaths and a 3.15 per 100 person-years incidence rate in the AF/AFL group vs 1700 deaths and a 1.66 incidence rate in the no AF/AFL group (P < .001).

The HR for the composite outcome in the base model was 1.33 (95% CI, 1.26-1.42, P < .001) (Table 3). The association between AF/AFL and the composite outcome remained significant after adjusting for additional metabolic and cardiac covariates. The HRs for the metabolic and cardiac models were 1.30 (95% CI, 1.23-1.38, P < .001) and 1.11 (95% CI, 1.05-1.18, P < .001), respectively. After adjusting for the full model, the HR was 1.12 (95% CI, 1.05-1.19, P < .001).

Over the 6-year study period, there was a lower survival probability for patients with AF/AFL. In the overall cohort, during a mean (SD) follow-up time of 3.7 (1.9) years comprising 129,391 total person-years, there were 3130 (8.8%) deaths and an incidence rate of 2.42 per 100 person-years. Death occurred 786 times with a 4.09 incidence rate per 100 person-years in the AF/AFL vs 2344 deaths and a 2.13 incidence rate per 100 person-years in the no AF/AFL group (P < .001). In the non-AF/AFL subgroup, death occurred 2344 times during a mean (SD) follow-up of 3.7 (1.9) years comprising 110,151 total person-years. Figure 3 shows the Kaplan-Meier curve of mortality by AF/AFL status.

After adjusting for the base, metabolic and cardiac covariates, the HRs for mortality were 1.45 (95% CI, 1.33-1.57, P < .001), 1.40 (95% CI, 1.29-1.51, P < .001) and 1.15 (95% CI, 1.06-1.25, P = .001), respectively (Table 4). The HR after adjusting for the full model was 1.16 (95% CI, 1.07-1.26, P < .001).

DISCUSSION

In this large retrospective cohort study, patients with WPW syndrome and comorbid AF/AFL had a significantly higher association with the cardiac composite outcome and death during a 3-year follow-up period when compared with patients without AF/AFL. After adjusting for confounding variables, the AF/AFL subgroup maintained a 12% and 16% higher association with the composite outcome and mortality, respectively. There was minimal difference in confounding effects between demographic data and metabolic profiles, suggesting one may serve as a proxy for the other.

To our knowledge, this is the largest WPW syndrome cohort study evaluating cardiac outcomes and mortality to date. Although previous research has shown the relatively low and mostly anecdotal SCD incidence within this population,our results demonstrate a higher association of adverse cardiac outcomes and death in an AF/AFL subgroup.^16-18 Notably, in this study the AF/AFL cohort was older and had higher CCI scores than their counterparts (P < .001), thus inferring an inherently greater degree of morbidity and 10-year mortality risk. Our study is also unique in that the mean patient age was significantly older than previously reported (63 vs 27 years), which may suggest a longer living history of both ventricular pre-excitation and the comorbidities outlined in Figure 1.¹⁹ Given these age discrepancies, it is possible that our overall study population was still relatively low risk and that not all reported deaths were necessarily related to WPW syndrome. Despite these assumptions, when comparing the WPW syndrome subgroups, we still found the AF/AFL cohort maintained a statistically significant higher association with the 2 study outcomes, even after adjusting for the greater presence of comorbidities. This suggests that the presence of AF/AFL may still portend a worse prognosis in patients with WPW syndrome.

Although the association of AF and development of VF in patients with WPW syndrome—due to rapid conduction over the accessory pathway(s)—was first reported > 40 years ago, there has still been few large, long-term data studies exploring mortality in this cohort.^19-25 Furthermore, even though the current literature attributes the development of AF with the electrophysiologic properties of the accessory pathway, as well as intrinsic atrial architecture and muscle vulnerability, there is still equivocal consensus regarding EPT screening and ablation indications for asymptomatic patients with WPW syndrome.^26-28 Notably, Pappone and colleagues demonstrated the potential benefit of liberal ablation indications for asymptomatic patients, arguing that the intrinsic electrophysiologic properties of the accessory pathway—ie, short accessory-pathway antegrade effective refractory period, inducibility of atrioventricular reentrant tachycardia triggering AF, and multiple accessory pathway—rather than symptoms, are independent predictors of developing malignant arrhythmia.^1-5

These findings contradict those reported by Obeyesekere and colleagues, who concluded that the low SCD incidence rates in patients with WPW syndrome precluded routine invasive screening.^19,28 They argued that Pappone and colleagues used malignant arrhythmia as a surrogate marker for death, and that the positive predictive value of a short accessory-pathway antegrade effective refractory period for developing malignant arrhythmia was lower than reported (15% vs 82%, respectively) and that its negative predictive value was 100%.^1,19,28 Given these conflicting recommendations, we hope our data elucidates the higher association of adverse outcomes and support considerations for more intensive EPT indications in patients with WPW syndrome.

While our study does not report SCD incidence, it does provide robust and reliable mortality data that suggests a greater association of death within an AF/AFL subgroup. Our findings would support more liberal EPT recommendations in patients with WPW syndrome.^1-5,8,9 In this study, the SCA incidence rate was more than double the rate in the AF/AFL cohort (P < .001) and is commonly the initial presenting event in WPW syndrome.⁹ Even though the reported SCD incidence rate is low in WPW syndrome, our data demonstrated an increased association of death within the AF/AFL cohort. Physicians should consider early risk stratification and ablation to prevent potential recurrent malignant arrhythmia leading to death.^{1-5,8,9,12,19,20}

Limitations

As a retrospective study and without access to the National Death Index, we were unable to determine the exact cause or events leading to death and instead utilized all-cause mortality data. Subsequently, our observations may only demonstrate association, rather than causality, between AF/AFL and death in patients with WPW syndrome. Additionally, we could not distinguish between AF and AFL as the arrhythmia leading to death. However, since overall survivability was the outcome of interest, our adjusted HR models were still able to demonstrate the increased association of the composite outcome and death within an AF/AFL cohort.

Although a large cohort was analyzed, due to the constraints of utilizing diagnostic codes to determine study outcomes, we could not distinguish between symptomatic and asymptomatic patients, nor how they were managed prior to the outcome event. However, as recent literature demonstrates, updated predictors of malignant arrhythmia and decisions for early EPT are similar for both symptomatic and asymptomatic patients and should be driven by the intrinsic electrophysiologic properties of the accessory pathway, rather than symptomatology;thus, our inability to discern this should have negligible consequence in determining when to perform risk stratification and ablation.¹

MHS eligible patients have direct access to care; the generalizability of our data may not necessarily correspond to a community population with lower socioeconomic status (we did adjust for military sponsor rank which has been used as a proxy), reduced access to care, or uninsured individuals. However, the prevalence of WPW syndrome within our cohort was comparable to the general population, 0.4% vs 0.1%-0.3%, respectively.^13,14,19 Similarly, the incidence of AF within our population was comparable to the general population, 15% vs 16%-26%, respectively.²³ These similar data points suggest our results may apply beyond MHS patients.

CONCLUSIONS

Patients with WPW syndrome and AF/AFL have a higher association with adverse cardiac outcomes and death. Despite previously reported low SCD incidence rates in this population, our study demonstrates the increased association of mortality in an AF/AFL cohort. The limitations of utilizing all-cause mortality data necessitate further investigation into the etiology behind the deaths in our study population. Since ventricular pre-excitation can predispose patients to AF and potentially lead to malignant arrhythmia and SCD, understanding the cause of mortality will allow physicians to determine the appropriate monitoring and intervention strategies to improve outcomes in this population. Our results suggest consideration for more aggressive EPT screening and ablation recommendations in patients with WPW syndrome may be warranted.

Wolff-Parkinson-White (WPW) syndrome is characterized by the presence of ≥ 1 accessory pathways and the development of both recurrent paroxysmal atrial fibrillation (AF) and supraventricular tachycardia that can lead to further malignant arrhythmias resulting in sudden cardiac death (SCD).^1-7 Historically, incidental, ventricular pre-excitation on electrocardiogram has conferred a relatively low SCD risk in adults; however, newer WPW syndrome data suggest the endpoint may not be as benign as previously thought.⁷ The current literature has defined atrioventricular reentrant tachycardia triggering AF, rather than symptoms, as an independent risk factor for malignant arrhythmias. Still, long-term data detailing the association of AF with serious cardiac events and death in patients with WPW syndrome are still limited.^1-7

While previous guidelines for the treatment of WPW syndrome only recommended routine electrophysiology testing (EPT) with liberal catheter ablation for symptomatic individuals, the 2015 American College of Cardiology/American Heart Association/Heart Rhythm Society guidelines now suggest its potential benefit for risk stratification in the asymptomatic population.^8-12 Given the limited existing data, more long-term studies are needed to corroborate the latest EPT recommendations before routinely applying them in practice. Furthermore, since concomitant AF can lead to adverse cardiac outcomes in patients with WPW syndrome, additional data evaluating this association are also necessary. In this study, we aimed to determine the impact of atrial fibrillation and/or flutter (AF/AFL) on adverse cardiac outcomes and mortality in patients with WPW syndrome.

METHODS

This study used data from the Military Health System (MHS) Database Repository. The MHS is one of the largest health care systems in the country and includes information on about 10 million active duty and retired military service members and their families (51% male; 49% female).^13,14 Data were fully anonymized and complied in accordance with federal and state laws, including the Health Insurance Portability and Accountability Act of 1996. The Naval Medical Center Portsmouth Institutional Review Board approved this study.

Study Design

This retrospective, observational cohort study identified MHS patients with WPW syndrome from January 1, 2014, to December 31, 2019. Patients were included if they had ≥ 2 International Classification of Diseases, Ninth Revision (ICD-9) or International Classification of Diseases, Tenth Revision (ICD-10) diagnosis codes for WPW syndrome (ICD-9, 426.7; ICD-10, I45.6) on separate dates; were aged ≥ 18 years at index date; and had ≥ 1 year of continuous eligibility prior to the index date (enrollment gaps ≤ 30 days were considered continuous). Patients were then divided into 2 subgroups by the presence or absence of AF/AFL using diagnostic codes. Patients were excluded if they had evidence of an implantable cardioverter-defibrillator, permanent pacemaker or were missing age or sex data. Patients were followed from index date until the first occurrence of the outcome of interest, MHS disenrollment, or the end of the study period.

Cardiac composite outcomes comprised of sudden cardiac arrest (SCA), ventricular fibrillation (VF), ventricular tachycardia and death, as well as death specifically, were the outcomes of interest and assessed after index date using ICD-9 and ICD-10 codes. Death was defined as all-cause mortality. Time to event was calculated based on the date of the initial component from the composite outcome and date of death specifically for mortality. Those not experiencing an outcome were followed until MHS disenrollment or the end of the study period.

Various patient characteristics were assessed at index including age, sex, military sponsor (the patient’s active or retired duty member through which their dependent receives TRICARE benefits) rank and branch, geographic region, type of US Department of Defense beneficiary, and index year. Clinical characteristics were assessed over a 1-year baseline period prior to index date and included the number of cardiologist and clinical visits for WPW syndrome, Charlson Comorbidity Index (CCI) scores calculated from diagnostic codes outlined in the Quan coding method, and preindex time.¹⁵ Comorbidities were assessed at baseline and defined as having ≥ 1 ICD-9 or ICD-10 code for a corresponding condition within 1 year prior to index.

Baseline characteristics were assessed and descriptive statistics for categorical and continuous variables were presented accordingly. To assess bivariate association with exposure, χ² tests were used to compare categorical variables, while t tests were used to compare continuous variables by exposure status. Incidence proportions and rates were reported for each outcome of interest. Kaplan-Meier curves were constructed to assess the bivariate association between exposure and study outcomes. Cox proportional hazard modeling was performed to estimate the association between AF/AFL and time to each of the outcomes. Multiple models were designed to assess cardiac and metabolic covariates, in addition to baseline characteristics. This included a base model adjusted for age, sex, military sponsor rank and branch, geographic region, and duty status.

Additional models adjusted for cardiac and metabolic confounders and CCI score. A comprehensive model included the base, cardiac, and metabolic covariates. Multicollinearity between covariates was assessed. Variables with a variance inflation factor > 4 or a tolerance level < 0.1 were added to the models. Cox proportional hazard models were used to estimate the unadjusted and adjusted hazard ratios (HRs) and 95% CIs of the association between AF/AFL and the study outcomes. Data were analyzed using SAS, version 9.4 for Windows.

RESULTS

From 2014 through 2019, 35,539 patients with WPW syndrome were identified in the MHS, 5291 had AF/AFL (14.9%); 19,961 were female (56.2%), the mean (SD) age was 62.9 (18.0) years, and 11,742 were aged ≥ 75 years (33.0%) (Table 1).

There were 4121 (11.6%), 322 (0.9%), and 848 (2.4%) patients with AF, AFL, and both arrhythmias, respectively. The mean (SD) number of cardiology visits was 3.9 (3.0). The mean (SD) baseline CCI score for the AF/AFL subgroup was 5.9 (3.5) vs 3.7 (2.2) for the non-AF/AFL subgroup (P < .001). The most prevalent comorbid conditions were hypertension, hyperlipidemia, chronic obstructive pulmonary disease, and diabetes (P < .001) (Figure 1).

Composite Outcomes

In the overall cohort, during a mean (SD) follow-up time of 3.4 (2.0) years comprising 119,682 total person-years, the components of the composite outcome occurred 6506 times with an incidence rate of 5.44 per 100 person-years. Ventricular tachycardia was the most common event, occurring 3281 times with an incidence rate of 2.74 per 100 person-years. SCA and VF occurred 841 and 135 times with incidence rates of 0.70 and 0.11 per 100 person-years, respectively. Death was the initial event 2249 times with an incidence rate of 1.88 per 100 person-years. Figure 2 shows the Kaplan-Meier curve of cardiac composite outcome by AF/AFL status.

The subgroup with AF/AFL comprised 17,412 total person-years and 1424 cardiac composite incidences compared with 102,270 person years and 5082 incidences in the no AF/AFL group (Table 2). Comparing AF/AFL vs no AF/AFL incidence rates were 8.18 vs 4.97 per 100 person-years, respectively (P < .001). SCA and VF occurred 233 and 38 times and respectively had incidence rates of 1.34 and 0.22 per 100 person-years in the AF/AFL group vs 0.59 and 0.09 per 100 person-years in the no AF/AFL group (P < .001). There were 549 deaths and a 3.15 per 100 person-years incidence rate in the AF/AFL group vs 1700 deaths and a 1.66 incidence rate in the no AF/AFL group (P < .001).

The HR for the composite outcome in the base model was 1.33 (95% CI, 1.26-1.42, P < .001) (Table 3). The association between AF/AFL and the composite outcome remained significant after adjusting for additional metabolic and cardiac covariates. The HRs for the metabolic and cardiac models were 1.30 (95% CI, 1.23-1.38, P < .001) and 1.11 (95% CI, 1.05-1.18, P < .001), respectively. After adjusting for the full model, the HR was 1.12 (95% CI, 1.05-1.19, P < .001).

Over the 6-year study period, there was a lower survival probability for patients with AF/AFL. In the overall cohort, during a mean (SD) follow-up time of 3.7 (1.9) years comprising 129,391 total person-years, there were 3130 (8.8%) deaths and an incidence rate of 2.42 per 100 person-years. Death occurred 786 times with a 4.09 incidence rate per 100 person-years in the AF/AFL vs 2344 deaths and a 2.13 incidence rate per 100 person-years in the no AF/AFL group (P < .001). In the non-AF/AFL subgroup, death occurred 2344 times during a mean (SD) follow-up of 3.7 (1.9) years comprising 110,151 total person-years. Figure 3 shows the Kaplan-Meier curve of mortality by AF/AFL status.

After adjusting for the base, metabolic and cardiac covariates, the HRs for mortality were 1.45 (95% CI, 1.33-1.57, P < .001), 1.40 (95% CI, 1.29-1.51, P < .001) and 1.15 (95% CI, 1.06-1.25, P = .001), respectively (Table 4). The HR after adjusting for the full model was 1.16 (95% CI, 1.07-1.26, P < .001).

DISCUSSION

In this large retrospective cohort study, patients with WPW syndrome and comorbid AF/AFL had a significantly higher association with the cardiac composite outcome and death during a 3-year follow-up period when compared with patients without AF/AFL. After adjusting for confounding variables, the AF/AFL subgroup maintained a 12% and 16% higher association with the composite outcome and mortality, respectively. There was minimal difference in confounding effects between demographic data and metabolic profiles, suggesting one may serve as a proxy for the other.

To our knowledge, this is the largest WPW syndrome cohort study evaluating cardiac outcomes and mortality to date. Although previous research has shown the relatively low and mostly anecdotal SCD incidence within this population,our results demonstrate a higher association of adverse cardiac outcomes and death in an AF/AFL subgroup.^16-18 Notably, in this study the AF/AFL cohort was older and had higher CCI scores than their counterparts (P < .001), thus inferring an inherently greater degree of morbidity and 10-year mortality risk. Our study is also unique in that the mean patient age was significantly older than previously reported (63 vs 27 years), which may suggest a longer living history of both ventricular pre-excitation and the comorbidities outlined in Figure 1.¹⁹ Given these age discrepancies, it is possible that our overall study population was still relatively low risk and that not all reported deaths were necessarily related to WPW syndrome. Despite these assumptions, when comparing the WPW syndrome subgroups, we still found the AF/AFL cohort maintained a statistically significant higher association with the 2 study outcomes, even after adjusting for the greater presence of comorbidities. This suggests that the presence of AF/AFL may still portend a worse prognosis in patients with WPW syndrome.

Although the association of AF and development of VF in patients with WPW syndrome—due to rapid conduction over the accessory pathway(s)—was first reported > 40 years ago, there has still been few large, long-term data studies exploring mortality in this cohort.^19-25 Furthermore, even though the current literature attributes the development of AF with the electrophysiologic properties of the accessory pathway, as well as intrinsic atrial architecture and muscle vulnerability, there is still equivocal consensus regarding EPT screening and ablation indications for asymptomatic patients with WPW syndrome.^26-28 Notably, Pappone and colleagues demonstrated the potential benefit of liberal ablation indications for asymptomatic patients, arguing that the intrinsic electrophysiologic properties of the accessory pathway—ie, short accessory-pathway antegrade effective refractory period, inducibility of atrioventricular reentrant tachycardia triggering AF, and multiple accessory pathway—rather than symptoms, are independent predictors of developing malignant arrhythmia.^1-5

These findings contradict those reported by Obeyesekere and colleagues, who concluded that the low SCD incidence rates in patients with WPW syndrome precluded routine invasive screening.^19,28 They argued that Pappone and colleagues used malignant arrhythmia as a surrogate marker for death, and that the positive predictive value of a short accessory-pathway antegrade effective refractory period for developing malignant arrhythmia was lower than reported (15% vs 82%, respectively) and that its negative predictive value was 100%.^1,19,28 Given these conflicting recommendations, we hope our data elucidates the higher association of adverse outcomes and support considerations for more intensive EPT indications in patients with WPW syndrome.

While our study does not report SCD incidence, it does provide robust and reliable mortality data that suggests a greater association of death within an AF/AFL subgroup. Our findings would support more liberal EPT recommendations in patients with WPW syndrome.^1-5,8,9 In this study, the SCA incidence rate was more than double the rate in the AF/AFL cohort (P < .001) and is commonly the initial presenting event in WPW syndrome.⁹ Even though the reported SCD incidence rate is low in WPW syndrome, our data demonstrated an increased association of death within the AF/AFL cohort. Physicians should consider early risk stratification and ablation to prevent potential recurrent malignant arrhythmia leading to death.^{1-5,8,9,12,19,20}

Limitations

As a retrospective study and without access to the National Death Index, we were unable to determine the exact cause or events leading to death and instead utilized all-cause mortality data. Subsequently, our observations may only demonstrate association, rather than causality, between AF/AFL and death in patients with WPW syndrome. Additionally, we could not distinguish between AF and AFL as the arrhythmia leading to death. However, since overall survivability was the outcome of interest, our adjusted HR models were still able to demonstrate the increased association of the composite outcome and death within an AF/AFL cohort.

Although a large cohort was analyzed, due to the constraints of utilizing diagnostic codes to determine study outcomes, we could not distinguish between symptomatic and asymptomatic patients, nor how they were managed prior to the outcome event. However, as recent literature demonstrates, updated predictors of malignant arrhythmia and decisions for early EPT are similar for both symptomatic and asymptomatic patients and should be driven by the intrinsic electrophysiologic properties of the accessory pathway, rather than symptomatology;thus, our inability to discern this should have negligible consequence in determining when to perform risk stratification and ablation.¹

MHS eligible patients have direct access to care; the generalizability of our data may not necessarily correspond to a community population with lower socioeconomic status (we did adjust for military sponsor rank which has been used as a proxy), reduced access to care, or uninsured individuals. However, the prevalence of WPW syndrome within our cohort was comparable to the general population, 0.4% vs 0.1%-0.3%, respectively.^13,14,19 Similarly, the incidence of AF within our population was comparable to the general population, 15% vs 16%-26%, respectively.²³ These similar data points suggest our results may apply beyond MHS patients.

CONCLUSIONS

Patients with WPW syndrome and AF/AFL have a higher association with adverse cardiac outcomes and death. Despite previously reported low SCD incidence rates in this population, our study demonstrates the increased association of mortality in an AF/AFL cohort. The limitations of utilizing all-cause mortality data necessitate further investigation into the etiology behind the deaths in our study population. Since ventricular pre-excitation can predispose patients to AF and potentially lead to malignant arrhythmia and SCD, understanding the cause of mortality will allow physicians to determine the appropriate monitoring and intervention strategies to improve outcomes in this population. Our results suggest consideration for more aggressive EPT screening and ablation recommendations in patients with WPW syndrome may be warranted.

Wolff-Parkinson-White (WPW) syndrome is characterized by the presence of ≥ 1 accessory pathways and the development of both recurrent paroxysmal atrial fibrillation (AF) and supraventricular tachycardia that can lead to further malignant arrhythmias resulting in sudden cardiac death (SCD).^1-7 Historically, incidental, ventricular pre-excitation on electrocardiogram has conferred a relatively low SCD risk in adults; however, newer WPW syndrome data suggest the endpoint may not be as benign as previously thought.⁷ The current literature has defined atrioventricular reentrant tachycardia triggering AF, rather than symptoms, as an independent risk factor for malignant arrhythmias. Still, long-term data detailing the association of AF with serious cardiac events and death in patients with WPW syndrome are still limited.^1-7

While previous guidelines for the treatment of WPW syndrome only recommended routine electrophysiology testing (EPT) with liberal catheter ablation for symptomatic individuals, the 2015 American College of Cardiology/American Heart Association/Heart Rhythm Society guidelines now suggest its potential benefit for risk stratification in the asymptomatic population.^8-12 Given the limited existing data, more long-term studies are needed to corroborate the latest EPT recommendations before routinely applying them in practice. Furthermore, since concomitant AF can lead to adverse cardiac outcomes in patients with WPW syndrome, additional data evaluating this association are also necessary. In this study, we aimed to determine the impact of atrial fibrillation and/or flutter (AF/AFL) on adverse cardiac outcomes and mortality in patients with WPW syndrome.

METHODS

This study used data from the Military Health System (MHS) Database Repository. The MHS is one of the largest health care systems in the country and includes information on about 10 million active duty and retired military service members and their families (51% male; 49% female).^13,14 Data were fully anonymized and complied in accordance with federal and state laws, including the Health Insurance Portability and Accountability Act of 1996. The Naval Medical Center Portsmouth Institutional Review Board approved this study.

Study Design

This retrospective, observational cohort study identified MHS patients with WPW syndrome from January 1, 2014, to December 31, 2019. Patients were included if they had ≥ 2 International Classification of Diseases, Ninth Revision (ICD-9) or International Classification of Diseases, Tenth Revision (ICD-10) diagnosis codes for WPW syndrome (ICD-9, 426.7; ICD-10, I45.6) on separate dates; were aged ≥ 18 years at index date; and had ≥ 1 year of continuous eligibility prior to the index date (enrollment gaps ≤ 30 days were considered continuous). Patients were then divided into 2 subgroups by the presence or absence of AF/AFL using diagnostic codes. Patients were excluded if they had evidence of an implantable cardioverter-defibrillator, permanent pacemaker or were missing age or sex data. Patients were followed from index date until the first occurrence of the outcome of interest, MHS disenrollment, or the end of the study period.

Cardiac composite outcomes comprised of sudden cardiac arrest (SCA), ventricular fibrillation (VF), ventricular tachycardia and death, as well as death specifically, were the outcomes of interest and assessed after index date using ICD-9 and ICD-10 codes. Death was defined as all-cause mortality. Time to event was calculated based on the date of the initial component from the composite outcome and date of death specifically for mortality. Those not experiencing an outcome were followed until MHS disenrollment or the end of the study period.

Various patient characteristics were assessed at index including age, sex, military sponsor (the patient’s active or retired duty member through which their dependent receives TRICARE benefits) rank and branch, geographic region, type of US Department of Defense beneficiary, and index year. Clinical characteristics were assessed over a 1-year baseline period prior to index date and included the number of cardiologist and clinical visits for WPW syndrome, Charlson Comorbidity Index (CCI) scores calculated from diagnostic codes outlined in the Quan coding method, and preindex time.¹⁵ Comorbidities were assessed at baseline and defined as having ≥ 1 ICD-9 or ICD-10 code for a corresponding condition within 1 year prior to index.

Baseline characteristics were assessed and descriptive statistics for categorical and continuous variables were presented accordingly. To assess bivariate association with exposure, χ² tests were used to compare categorical variables, while t tests were used to compare continuous variables by exposure status. Incidence proportions and rates were reported for each outcome of interest. Kaplan-Meier curves were constructed to assess the bivariate association between exposure and study outcomes. Cox proportional hazard modeling was performed to estimate the association between AF/AFL and time to each of the outcomes. Multiple models were designed to assess cardiac and metabolic covariates, in addition to baseline characteristics. This included a base model adjusted for age, sex, military sponsor rank and branch, geographic region, and duty status.

Additional models adjusted for cardiac and metabolic confounders and CCI score. A comprehensive model included the base, cardiac, and metabolic covariates. Multicollinearity between covariates was assessed. Variables with a variance inflation factor > 4 or a tolerance level < 0.1 were added to the models. Cox proportional hazard models were used to estimate the unadjusted and adjusted hazard ratios (HRs) and 95% CIs of the association between AF/AFL and the study outcomes. Data were analyzed using SAS, version 9.4 for Windows.

RESULTS

From 2014 through 2019, 35,539 patients with WPW syndrome were identified in the MHS, 5291 had AF/AFL (14.9%); 19,961 were female (56.2%), the mean (SD) age was 62.9 (18.0) years, and 11,742 were aged ≥ 75 years (33.0%) (Table 1).

There were 4121 (11.6%), 322 (0.9%), and 848 (2.4%) patients with AF, AFL, and both arrhythmias, respectively. The mean (SD) number of cardiology visits was 3.9 (3.0). The mean (SD) baseline CCI score for the AF/AFL subgroup was 5.9 (3.5) vs 3.7 (2.2) for the non-AF/AFL subgroup (P < .001). The most prevalent comorbid conditions were hypertension, hyperlipidemia, chronic obstructive pulmonary disease, and diabetes (P < .001) (Figure 1).

Composite Outcomes

In the overall cohort, during a mean (SD) follow-up time of 3.4 (2.0) years comprising 119,682 total person-years, the components of the composite outcome occurred 6506 times with an incidence rate of 5.44 per 100 person-years. Ventricular tachycardia was the most common event, occurring 3281 times with an incidence rate of 2.74 per 100 person-years. SCA and VF occurred 841 and 135 times with incidence rates of 0.70 and 0.11 per 100 person-years, respectively. Death was the initial event 2249 times with an incidence rate of 1.88 per 100 person-years. Figure 2 shows the Kaplan-Meier curve of cardiac composite outcome by AF/AFL status.

The subgroup with AF/AFL comprised 17,412 total person-years and 1424 cardiac composite incidences compared with 102,270 person years and 5082 incidences in the no AF/AFL group (Table 2). Comparing AF/AFL vs no AF/AFL incidence rates were 8.18 vs 4.97 per 100 person-years, respectively (P < .001). SCA and VF occurred 233 and 38 times and respectively had incidence rates of 1.34 and 0.22 per 100 person-years in the AF/AFL group vs 0.59 and 0.09 per 100 person-years in the no AF/AFL group (P < .001). There were 549 deaths and a 3.15 per 100 person-years incidence rate in the AF/AFL group vs 1700 deaths and a 1.66 incidence rate in the no AF/AFL group (P < .001).

The HR for the composite outcome in the base model was 1.33 (95% CI, 1.26-1.42, P < .001) (Table 3). The association between AF/AFL and the composite outcome remained significant after adjusting for additional metabolic and cardiac covariates. The HRs for the metabolic and cardiac models were 1.30 (95% CI, 1.23-1.38, P < .001) and 1.11 (95% CI, 1.05-1.18, P < .001), respectively. After adjusting for the full model, the HR was 1.12 (95% CI, 1.05-1.19, P < .001).

Over the 6-year study period, there was a lower survival probability for patients with AF/AFL. In the overall cohort, during a mean (SD) follow-up time of 3.7 (1.9) years comprising 129,391 total person-years, there were 3130 (8.8%) deaths and an incidence rate of 2.42 per 100 person-years. Death occurred 786 times with a 4.09 incidence rate per 100 person-years in the AF/AFL vs 2344 deaths and a 2.13 incidence rate per 100 person-years in the no AF/AFL group (P < .001). In the non-AF/AFL subgroup, death occurred 2344 times during a mean (SD) follow-up of 3.7 (1.9) years comprising 110,151 total person-years. Figure 3 shows the Kaplan-Meier curve of mortality by AF/AFL status.

After adjusting for the base, metabolic and cardiac covariates, the HRs for mortality were 1.45 (95% CI, 1.33-1.57, P < .001), 1.40 (95% CI, 1.29-1.51, P < .001) and 1.15 (95% CI, 1.06-1.25, P = .001), respectively (Table 4). The HR after adjusting for the full model was 1.16 (95% CI, 1.07-1.26, P < .001).

DISCUSSION

In this large retrospective cohort study, patients with WPW syndrome and comorbid AF/AFL had a significantly higher association with the cardiac composite outcome and death during a 3-year follow-up period when compared with patients without AF/AFL. After adjusting for confounding variables, the AF/AFL subgroup maintained a 12% and 16% higher association with the composite outcome and mortality, respectively. There was minimal difference in confounding effects between demographic data and metabolic profiles, suggesting one may serve as a proxy for the other.

To our knowledge, this is the largest WPW syndrome cohort study evaluating cardiac outcomes and mortality to date. Although previous research has shown the relatively low and mostly anecdotal SCD incidence within this population,our results demonstrate a higher association of adverse cardiac outcomes and death in an AF/AFL subgroup.^16-18 Notably, in this study the AF/AFL cohort was older and had higher CCI scores than their counterparts (P < .001), thus inferring an inherently greater degree of morbidity and 10-year mortality risk. Our study is also unique in that the mean patient age was significantly older than previously reported (63 vs 27 years), which may suggest a longer living history of both ventricular pre-excitation and the comorbidities outlined in Figure 1.¹⁹ Given these age discrepancies, it is possible that our overall study population was still relatively low risk and that not all reported deaths were necessarily related to WPW syndrome. Despite these assumptions, when comparing the WPW syndrome subgroups, we still found the AF/AFL cohort maintained a statistically significant higher association with the 2 study outcomes, even after adjusting for the greater presence of comorbidities. This suggests that the presence of AF/AFL may still portend a worse prognosis in patients with WPW syndrome.

Although the association of AF and development of VF in patients with WPW syndrome—due to rapid conduction over the accessory pathway(s)—was first reported > 40 years ago, there has still been few large, long-term data studies exploring mortality in this cohort.^19-25 Furthermore, even though the current literature attributes the development of AF with the electrophysiologic properties of the accessory pathway, as well as intrinsic atrial architecture and muscle vulnerability, there is still equivocal consensus regarding EPT screening and ablation indications for asymptomatic patients with WPW syndrome.^26-28 Notably, Pappone and colleagues demonstrated the potential benefit of liberal ablation indications for asymptomatic patients, arguing that the intrinsic electrophysiologic properties of the accessory pathway—ie, short accessory-pathway antegrade effective refractory period, inducibility of atrioventricular reentrant tachycardia triggering AF, and multiple accessory pathway—rather than symptoms, are independent predictors of developing malignant arrhythmia.^1-5

These findings contradict those reported by Obeyesekere and colleagues, who concluded that the low SCD incidence rates in patients with WPW syndrome precluded routine invasive screening.^19,28 They argued that Pappone and colleagues used malignant arrhythmia as a surrogate marker for death, and that the positive predictive value of a short accessory-pathway antegrade effective refractory period for developing malignant arrhythmia was lower than reported (15% vs 82%, respectively) and that its negative predictive value was 100%.^1,19,28 Given these conflicting recommendations, we hope our data elucidates the higher association of adverse outcomes and support considerations for more intensive EPT indications in patients with WPW syndrome.

While our study does not report SCD incidence, it does provide robust and reliable mortality data that suggests a greater association of death within an AF/AFL subgroup. Our findings would support more liberal EPT recommendations in patients with WPW syndrome.^1-5,8,9 In this study, the SCA incidence rate was more than double the rate in the AF/AFL cohort (P < .001) and is commonly the initial presenting event in WPW syndrome.⁹ Even though the reported SCD incidence rate is low in WPW syndrome, our data demonstrated an increased association of death within the AF/AFL cohort. Physicians should consider early risk stratification and ablation to prevent potential recurrent malignant arrhythmia leading to death.^{1-5,8,9,12,19,20}

Limitations

As a retrospective study and without access to the National Death Index, we were unable to determine the exact cause or events leading to death and instead utilized all-cause mortality data. Subsequently, our observations may only demonstrate association, rather than causality, between AF/AFL and death in patients with WPW syndrome. Additionally, we could not distinguish between AF and AFL as the arrhythmia leading to death. However, since overall survivability was the outcome of interest, our adjusted HR models were still able to demonstrate the increased association of the composite outcome and death within an AF/AFL cohort.

Although a large cohort was analyzed, due to the constraints of utilizing diagnostic codes to determine study outcomes, we could not distinguish between symptomatic and asymptomatic patients, nor how they were managed prior to the outcome event. However, as recent literature demonstrates, updated predictors of malignant arrhythmia and decisions for early EPT are similar for both symptomatic and asymptomatic patients and should be driven by the intrinsic electrophysiologic properties of the accessory pathway, rather than symptomatology;thus, our inability to discern this should have negligible consequence in determining when to perform risk stratification and ablation.¹

MHS eligible patients have direct access to care; the generalizability of our data may not necessarily correspond to a community population with lower socioeconomic status (we did adjust for military sponsor rank which has been used as a proxy), reduced access to care, or uninsured individuals. However, the prevalence of WPW syndrome within our cohort was comparable to the general population, 0.4% vs 0.1%-0.3%, respectively.^13,14,19 Similarly, the incidence of AF within our population was comparable to the general population, 15% vs 16%-26%, respectively.²³ These similar data points suggest our results may apply beyond MHS patients.

CONCLUSIONS

Patients with WPW syndrome and AF/AFL have a higher association with adverse cardiac outcomes and death. Despite previously reported low SCD incidence rates in this population, our study demonstrates the increased association of mortality in an AF/AFL cohort. The limitations of utilizing all-cause mortality data necessitate further investigation into the etiology behind the deaths in our study population. Since ventricular pre-excitation can predispose patients to AF and potentially lead to malignant arrhythmia and SCD, understanding the cause of mortality will allow physicians to determine the appropriate monitoring and intervention strategies to improve outcomes in this population. Our results suggest consideration for more aggressive EPT screening and ablation recommendations in patients with WPW syndrome may be warranted.

The prevalence of chronic obstructive pulmonary disease (COPD) among male US veterans is higher than in the general population.¹ Veterans with COPD have higher rates of comorbidities and increased respiratory-related and all-cause health care use, including the use of long-term oxygen therapy (LTOT).^2-5 It has been well established that LTOT reduces all-cause mortality in patients with COPD and resting hypoxemic chronic respiratory failure (CRF) when used for ≥ 15 hours per day.^6-8

Delivery of domiciliary LTOT entails placing a nasal cannula into both nostrils and loosely securing it around both ears throughout the wake-sleep cycle. Several veterans with hypoxemic CRF due to COPD at the Jesse Brown Veterans Affairs Medical Center (JBVAMC) in Chicago, Illinois, who were receiving LTOT reported nasal cannula dislodgement (NCD) while they slept. However, the clinical significance and impact of these repeated episodes on respiratory-related health care utilization, such as frequent COPD exacerbations with hospitalization, were not recognized. Moreover, we found no published reports or clinical practice guidelines alluding to similar events reported by patients with hypoxemic CRF due to COPD receiving LTOT either at home or in an acute care setting.^8,9 Nonetheless, frequent COPD exacerbations are associated with increased hospital admissions and account for a large portion of health care costs attributed to COPD.^10-13

The purpose of this study was to determine whether veterans with hypoxemic CRF due to COPD and receiving 24-hour LTOT at JBVAMC were experiencing NCD during sleep and, if so, its impact on their hospitalizations for COPD exacerbations.

METHODS

We reviewed electronic health records (EHRs) of veterans with hypoxemic CRF from COPD who received 24-hour LTOT administered through nasal cannula and were followed in the JBVAMC pulmonary outpatient clinic between February 1, 2022, and December 31, 2022. In each case, LTOT was prescribed by a board-certified pulmonologist based on Veterans Health Administration clinical practice guidelines.¹⁴ A licensed durable medical equipment company contracted by the JBVAMC delivered and established home oxygen equipment at each veteran’s residence.

Pertinent patient demographics, clinical and physiologic variables, and hospitalizations with length of JBVAMC stay for each physician-diagnosed COPD exacerbation in the preceding year from the date last seen in the clinic were abstracted from EHRs. Overall hospital cost, defined as a veteran overnight stay in either the medical intensive care unit (MICU) or a general acute medicine bed in a US Department of Veterans Affairs (VA) facility, was calculated for each hospitalization for physician-diagnosed COPD exacerbation using VA Managerial Cost Accounting System National Cost Extracts for inpatient encounters.¹⁵ We then contacted each veteran by telephone and asked whether they had experienced NCD and, if so, its weekly frequency ranging from once to nightly.

Data Analysis

Data were reported as mean (SD) where appropriate. The t test and Fisher exact test were used as indicated. P < .05 was considered statistically significant. The study protocol was determined to be exempt by the JBVAMC Institutional Review Board (Protocol #1725748).

During the study period, 75 patients with hypoxemic CRF from COPD received LTOT and were followed at the JBVAMC. No patients were hospitalized outside the JBVAMC for COPD exacerbation during this time frame. We also found no documentation in the EHRs indicating that the clinicians managing these patients at the JBVAMC inquired about NCD during sleep.

Of the 75 patients, 66 (88%) responded to the telephone survey and 22 patients (33%) reported weekly episodes of NCD while they slept (median, 4 dislodgments per week). (Table 1). Eight patients (36%) reported nightly NCDs (Figure). All 66 respondents were male and 14 of 22 in the NCD group as well as 21 of 44 in the no NCD group were Black veterans. The mean age was similar in both groups: 71 years in the NCD group and 72 years in the no NCD group. There were no statistically significant differences in demographics, including prevalence of obstructive sleep apnea (OSA), supplemental oxygen flow rate, and duration of LTOT, or in pulmonary function test results between patients who did and did not experience NCD while sleeping (Table 2).

Ten of 22 patients (45%) with NCD and 9 of 44 patients (20%) without NCD were hospitalized at the JBVAMC for ≥ 1 COPD exacerbation in the preceding year that was diagnosed by a physician (P = .045). Three of 22 patients (14%) with NCD and no patients in the no NCD group were admitted to the MICU. No patients required intubation and mechanical ventilation during hospitalization, and no patients died. Overall hospital costs were 25% ($64,342) higher in NCD group compared with the no NCD group and were attributed to the MICU admissions in the NCD group (Table 3). Nine veterans did not respond to repeated telephone calls. One physician-diagnosed COPD exacerbation requiring hospitalization was documented in the nonresponder group; the patient was hospitalized for 2 days. One veteran died before being contacted.

DISCUSSION

There are 3 new findings in this study. First, health care practitioners at JBVAMC did not document the presence of NCD during sleep in patients with hypoxemic CRF due to COPD and receiving LTOT. Second, one-third of these patients reported frequent NCD during sleep when interviewed. Third, the nocturnal events were associated with a higher hospitalization rate for physician-diagnosed COPD exacerbation and higher overall hospital costs. These findings are unlikely to be explained by differences in COPD severity and/or known triggers that lead to COPD exacerbation and require hospitalization because baseline physiologic and LTOT parameters were similar in both groups. Conceivably, patients with untreated OSA could be restless while asleep, leading to NCD. However, this explanation seems unlikely because the frequency of OSA was similar in both groups.

Nocturnal arterial oxygen desaturation in patients with COPD without evidence of OSA may contribute to the frequency of exacerbations.¹⁶ Although the mechanism(s) underlying this phenomenon is uncertain, we posit that prolonged nocturnal airway wall hypoxia could amplify underlying chronic inflammation through local generation of reactive oxygen species, thereby predisposing patients to exacerbations. Frequent COPD exacerbations promote disease progression and health status decline and are associated with increased mortality.^11,13 Moreover, hospitalization of patients with COPD is the largest contributor to the annual direct cost of COPD per patient.^10,12 The higher hospitalization rate observed in the NCD group in our study suggests that interruption of supplemental oxygen delivery while asleep may be a risk factor for COPD exacerbation. Alternatively, an independent factor or factors may have contributed to both NCD during sleep and COPD exacerbation in these patients or an impending exacerbation resulted in sleep disturbances that led to NCD. Additional research is warranted on veterans with hypoxemic CRF from COPD who are receiving LTOT and report frequent NCD during sleep that may support or refute these hypotheses.

To the best of our knowledge, NCD during sleep has not been previously reported in patients with hypoxemic CRF due to COPD who are receiving LTOT at home or in an acute care setting.^17-20 Several layperson proposals to secure nasal cannulas to the face while sleeping are posted online. These include wearing a commercially available headband with 2 Velcro loops that fix the cannula tube, using fabric medical tape on both cheeks, and wearing a sleep mask. Conceivably, the efficacy and safety of these inexpensive interventions to mitigate NCD during sleep in patients receiving LTOT with hypoxemic CRF from COPD could be tested in clinical trials.

Limitations

This was a small, single-site study, comprised entirely of male patients who are predominantly Black veterans. The telephone interviews with veterans self-reporting NCD during their sleep are prone to recall bias. In addition, the validity and reproducibility of NCD during sleep were not addressed in this study. Missing data from 9 nonresponders may have introduced a nonresponse bias in data analysis and interpretation. The overall hospital cost for a COPD exacerbation at JBVAMC was derived from VA data; US Centers for Medicare & Medicaid Services or commercial carrier data may be different.^15,21 Lastly, access to LTOT for veterans with hypoxemic CRF from COPD is regulated and supervised at VA medical facilities.¹⁴ This process may be different for patients outside the VA. Taken together, it is difficult to generalize our initial observations to non-VA patients with hypoxemic CRF from COPD who are receiving LTOT. We suggest a large, prospective study of veterans be conducted to determine the prevalence of NCD during sleep and its relationship with COPD exacerbations in veterans receiving LTOT with hypoxemic CRF due to COPD.

CONCLUSIONS

Clinicians at the JBVAMC did not document the presence of NCD during sleep in patients with hypoxemic CRF from COPD who received LTOT. However, self-reported, weekly nocturnal NCD episodes were associated with a higher hospitalization rate for COPD exacerbation and higher hospital costs. Accordingly, user-friendly devices to mitigate NCD during sleep should be developed.

Acknowledgments

We thank Yolanda Davis, RRT, and George Adam for their assistance with this project.

The prevalence of chronic obstructive pulmonary disease (COPD) among male US veterans is higher than in the general population.¹ Veterans with COPD have higher rates of comorbidities and increased respiratory-related and all-cause health care use, including the use of long-term oxygen therapy (LTOT).^2-5 It has been well established that LTOT reduces all-cause mortality in patients with COPD and resting hypoxemic chronic respiratory failure (CRF) when used for ≥ 15 hours per day.^6-8

Delivery of domiciliary LTOT entails placing a nasal cannula into both nostrils and loosely securing it around both ears throughout the wake-sleep cycle. Several veterans with hypoxemic CRF due to COPD at the Jesse Brown Veterans Affairs Medical Center (JBVAMC) in Chicago, Illinois, who were receiving LTOT reported nasal cannula dislodgement (NCD) while they slept. However, the clinical significance and impact of these repeated episodes on respiratory-related health care utilization, such as frequent COPD exacerbations with hospitalization, were not recognized. Moreover, we found no published reports or clinical practice guidelines alluding to similar events reported by patients with hypoxemic CRF due to COPD receiving LTOT either at home or in an acute care setting.^8,9 Nonetheless, frequent COPD exacerbations are associated with increased hospital admissions and account for a large portion of health care costs attributed to COPD.^10-13

The purpose of this study was to determine whether veterans with hypoxemic CRF due to COPD and receiving 24-hour LTOT at JBVAMC were experiencing NCD during sleep and, if so, its impact on their hospitalizations for COPD exacerbations.

METHODS

We reviewed electronic health records (EHRs) of veterans with hypoxemic CRF from COPD who received 24-hour LTOT administered through nasal cannula and were followed in the JBVAMC pulmonary outpatient clinic between February 1, 2022, and December 31, 2022. In each case, LTOT was prescribed by a board-certified pulmonologist based on Veterans Health Administration clinical practice guidelines.¹⁴ A licensed durable medical equipment company contracted by the JBVAMC delivered and established home oxygen equipment at each veteran’s residence.

Pertinent patient demographics, clinical and physiologic variables, and hospitalizations with length of JBVAMC stay for each physician-diagnosed COPD exacerbation in the preceding year from the date last seen in the clinic were abstracted from EHRs. Overall hospital cost, defined as a veteran overnight stay in either the medical intensive care unit (MICU) or a general acute medicine bed in a US Department of Veterans Affairs (VA) facility, was calculated for each hospitalization for physician-diagnosed COPD exacerbation using VA Managerial Cost Accounting System National Cost Extracts for inpatient encounters.¹⁵ We then contacted each veteran by telephone and asked whether they had experienced NCD and, if so, its weekly frequency ranging from once to nightly.

Data Analysis

Data were reported as mean (SD) where appropriate. The t test and Fisher exact test were used as indicated. P < .05 was considered statistically significant. The study protocol was determined to be exempt by the JBVAMC Institutional Review Board (Protocol #1725748).

During the study period, 75 patients with hypoxemic CRF from COPD received LTOT and were followed at the JBVAMC. No patients were hospitalized outside the JBVAMC for COPD exacerbation during this time frame. We also found no documentation in the EHRs indicating that the clinicians managing these patients at the JBVAMC inquired about NCD during sleep.

Of the 75 patients, 66 (88%) responded to the telephone survey and 22 patients (33%) reported weekly episodes of NCD while they slept (median, 4 dislodgments per week). (Table 1). Eight patients (36%) reported nightly NCDs (Figure). All 66 respondents were male and 14 of 22 in the NCD group as well as 21 of 44 in the no NCD group were Black veterans. The mean age was similar in both groups: 71 years in the NCD group and 72 years in the no NCD group. There were no statistically significant differences in demographics, including prevalence of obstructive sleep apnea (OSA), supplemental oxygen flow rate, and duration of LTOT, or in pulmonary function test results between patients who did and did not experience NCD while sleeping (Table 2).

Ten of 22 patients (45%) with NCD and 9 of 44 patients (20%) without NCD were hospitalized at the JBVAMC for ≥ 1 COPD exacerbation in the preceding year that was diagnosed by a physician (P = .045). Three of 22 patients (14%) with NCD and no patients in the no NCD group were admitted to the MICU. No patients required intubation and mechanical ventilation during hospitalization, and no patients died. Overall hospital costs were 25% ($64,342) higher in NCD group compared with the no NCD group and were attributed to the MICU admissions in the NCD group (Table 3). Nine veterans did not respond to repeated telephone calls. One physician-diagnosed COPD exacerbation requiring hospitalization was documented in the nonresponder group; the patient was hospitalized for 2 days. One veteran died before being contacted.

DISCUSSION

There are 3 new findings in this study. First, health care practitioners at JBVAMC did not document the presence of NCD during sleep in patients with hypoxemic CRF due to COPD and receiving LTOT. Second, one-third of these patients reported frequent NCD during sleep when interviewed. Third, the nocturnal events were associated with a higher hospitalization rate for physician-diagnosed COPD exacerbation and higher overall hospital costs. These findings are unlikely to be explained by differences in COPD severity and/or known triggers that lead to COPD exacerbation and require hospitalization because baseline physiologic and LTOT parameters were similar in both groups. Conceivably, patients with untreated OSA could be restless while asleep, leading to NCD. However, this explanation seems unlikely because the frequency of OSA was similar in both groups.

Nocturnal arterial oxygen desaturation in patients with COPD without evidence of OSA may contribute to the frequency of exacerbations.¹⁶ Although the mechanism(s) underlying this phenomenon is uncertain, we posit that prolonged nocturnal airway wall hypoxia could amplify underlying chronic inflammation through local generation of reactive oxygen species, thereby predisposing patients to exacerbations. Frequent COPD exacerbations promote disease progression and health status decline and are associated with increased mortality.^11,13 Moreover, hospitalization of patients with COPD is the largest contributor to the annual direct cost of COPD per patient.^10,12 The higher hospitalization rate observed in the NCD group in our study suggests that interruption of supplemental oxygen delivery while asleep may be a risk factor for COPD exacerbation. Alternatively, an independent factor or factors may have contributed to both NCD during sleep and COPD exacerbation in these patients or an impending exacerbation resulted in sleep disturbances that led to NCD. Additional research is warranted on veterans with hypoxemic CRF from COPD who are receiving LTOT and report frequent NCD during sleep that may support or refute these hypotheses.

To the best of our knowledge, NCD during sleep has not been previously reported in patients with hypoxemic CRF due to COPD who are receiving LTOT at home or in an acute care setting.^17-20 Several layperson proposals to secure nasal cannulas to the face while sleeping are posted online. These include wearing a commercially available headband with 2 Velcro loops that fix the cannula tube, using fabric medical tape on both cheeks, and wearing a sleep mask. Conceivably, the efficacy and safety of these inexpensive interventions to mitigate NCD during sleep in patients receiving LTOT with hypoxemic CRF from COPD could be tested in clinical trials.

Limitations

This was a small, single-site study, comprised entirely of male patients who are predominantly Black veterans. The telephone interviews with veterans self-reporting NCD during their sleep are prone to recall bias. In addition, the validity and reproducibility of NCD during sleep were not addressed in this study. Missing data from 9 nonresponders may have introduced a nonresponse bias in data analysis and interpretation. The overall hospital cost for a COPD exacerbation at JBVAMC was derived from VA data; US Centers for Medicare & Medicaid Services or commercial carrier data may be different.^15,21 Lastly, access to LTOT for veterans with hypoxemic CRF from COPD is regulated and supervised at VA medical facilities.¹⁴ This process may be different for patients outside the VA. Taken together, it is difficult to generalize our initial observations to non-VA patients with hypoxemic CRF from COPD who are receiving LTOT. We suggest a large, prospective study of veterans be conducted to determine the prevalence of NCD during sleep and its relationship with COPD exacerbations in veterans receiving LTOT with hypoxemic CRF due to COPD.

CONCLUSIONS

Clinicians at the JBVAMC did not document the presence of NCD during sleep in patients with hypoxemic CRF from COPD who received LTOT. However, self-reported, weekly nocturnal NCD episodes were associated with a higher hospitalization rate for COPD exacerbation and higher hospital costs. Accordingly, user-friendly devices to mitigate NCD during sleep should be developed.

Acknowledgments

We thank Yolanda Davis, RRT, and George Adam for their assistance with this project.

The prevalence of chronic obstructive pulmonary disease (COPD) among male US veterans is higher than in the general population.¹ Veterans with COPD have higher rates of comorbidities and increased respiratory-related and all-cause health care use, including the use of long-term oxygen therapy (LTOT).^2-5 It has been well established that LTOT reduces all-cause mortality in patients with COPD and resting hypoxemic chronic respiratory failure (CRF) when used for ≥ 15 hours per day.^6-8

Delivery of domiciliary LTOT entails placing a nasal cannula into both nostrils and loosely securing it around both ears throughout the wake-sleep cycle. Several veterans with hypoxemic CRF due to COPD at the Jesse Brown Veterans Affairs Medical Center (JBVAMC) in Chicago, Illinois, who were receiving LTOT reported nasal cannula dislodgement (NCD) while they slept. However, the clinical significance and impact of these repeated episodes on respiratory-related health care utilization, such as frequent COPD exacerbations with hospitalization, were not recognized. Moreover, we found no published reports or clinical practice guidelines alluding to similar events reported by patients with hypoxemic CRF due to COPD receiving LTOT either at home or in an acute care setting.^8,9 Nonetheless, frequent COPD exacerbations are associated with increased hospital admissions and account for a large portion of health care costs attributed to COPD.^10-13

The purpose of this study was to determine whether veterans with hypoxemic CRF due to COPD and receiving 24-hour LTOT at JBVAMC were experiencing NCD during sleep and, if so, its impact on their hospitalizations for COPD exacerbations.

METHODS

We reviewed electronic health records (EHRs) of veterans with hypoxemic CRF from COPD who received 24-hour LTOT administered through nasal cannula and were followed in the JBVAMC pulmonary outpatient clinic between February 1, 2022, and December 31, 2022. In each case, LTOT was prescribed by a board-certified pulmonologist based on Veterans Health Administration clinical practice guidelines.¹⁴ A licensed durable medical equipment company contracted by the JBVAMC delivered and established home oxygen equipment at each veteran’s residence.

Pertinent patient demographics, clinical and physiologic variables, and hospitalizations with length of JBVAMC stay for each physician-diagnosed COPD exacerbation in the preceding year from the date last seen in the clinic were abstracted from EHRs. Overall hospital cost, defined as a veteran overnight stay in either the medical intensive care unit (MICU) or a general acute medicine bed in a US Department of Veterans Affairs (VA) facility, was calculated for each hospitalization for physician-diagnosed COPD exacerbation using VA Managerial Cost Accounting System National Cost Extracts for inpatient encounters.¹⁵ We then contacted each veteran by telephone and asked whether they had experienced NCD and, if so, its weekly frequency ranging from once to nightly.

Data Analysis

Data were reported as mean (SD) where appropriate. The t test and Fisher exact test were used as indicated. P < .05 was considered statistically significant. The study protocol was determined to be exempt by the JBVAMC Institutional Review Board (Protocol #1725748).

During the study period, 75 patients with hypoxemic CRF from COPD received LTOT and were followed at the JBVAMC. No patients were hospitalized outside the JBVAMC for COPD exacerbation during this time frame. We also found no documentation in the EHRs indicating that the clinicians managing these patients at the JBVAMC inquired about NCD during sleep.

Of the 75 patients, 66 (88%) responded to the telephone survey and 22 patients (33%) reported weekly episodes of NCD while they slept (median, 4 dislodgments per week). (Table 1). Eight patients (36%) reported nightly NCDs (Figure). All 66 respondents were male and 14 of 22 in the NCD group as well as 21 of 44 in the no NCD group were Black veterans. The mean age was similar in both groups: 71 years in the NCD group and 72 years in the no NCD group. There were no statistically significant differences in demographics, including prevalence of obstructive sleep apnea (OSA), supplemental oxygen flow rate, and duration of LTOT, or in pulmonary function test results between patients who did and did not experience NCD while sleeping (Table 2).

Ten of 22 patients (45%) with NCD and 9 of 44 patients (20%) without NCD were hospitalized at the JBVAMC for ≥ 1 COPD exacerbation in the preceding year that was diagnosed by a physician (P = .045). Three of 22 patients (14%) with NCD and no patients in the no NCD group were admitted to the MICU. No patients required intubation and mechanical ventilation during hospitalization, and no patients died. Overall hospital costs were 25% ($64,342) higher in NCD group compared with the no NCD group and were attributed to the MICU admissions in the NCD group (Table 3). Nine veterans did not respond to repeated telephone calls. One physician-diagnosed COPD exacerbation requiring hospitalization was documented in the nonresponder group; the patient was hospitalized for 2 days. One veteran died before being contacted.

DISCUSSION

There are 3 new findings in this study. First, health care practitioners at JBVAMC did not document the presence of NCD during sleep in patients with hypoxemic CRF due to COPD and receiving LTOT. Second, one-third of these patients reported frequent NCD during sleep when interviewed. Third, the nocturnal events were associated with a higher hospitalization rate for physician-diagnosed COPD exacerbation and higher overall hospital costs. These findings are unlikely to be explained by differences in COPD severity and/or known triggers that lead to COPD exacerbation and require hospitalization because baseline physiologic and LTOT parameters were similar in both groups. Conceivably, patients with untreated OSA could be restless while asleep, leading to NCD. However, this explanation seems unlikely because the frequency of OSA was similar in both groups.

Nocturnal arterial oxygen desaturation in patients with COPD without evidence of OSA may contribute to the frequency of exacerbations.¹⁶ Although the mechanism(s) underlying this phenomenon is uncertain, we posit that prolonged nocturnal airway wall hypoxia could amplify underlying chronic inflammation through local generation of reactive oxygen species, thereby predisposing patients to exacerbations. Frequent COPD exacerbations promote disease progression and health status decline and are associated with increased mortality.^11,13 Moreover, hospitalization of patients with COPD is the largest contributor to the annual direct cost of COPD per patient.^10,12 The higher hospitalization rate observed in the NCD group in our study suggests that interruption of supplemental oxygen delivery while asleep may be a risk factor for COPD exacerbation. Alternatively, an independent factor or factors may have contributed to both NCD during sleep and COPD exacerbation in these patients or an impending exacerbation resulted in sleep disturbances that led to NCD. Additional research is warranted on veterans with hypoxemic CRF from COPD who are receiving LTOT and report frequent NCD during sleep that may support or refute these hypotheses.

To the best of our knowledge, NCD during sleep has not been previously reported in patients with hypoxemic CRF due to COPD who are receiving LTOT at home or in an acute care setting.^17-20 Several layperson proposals to secure nasal cannulas to the face while sleeping are posted online. These include wearing a commercially available headband with 2 Velcro loops that fix the cannula tube, using fabric medical tape on both cheeks, and wearing a sleep mask. Conceivably, the efficacy and safety of these inexpensive interventions to mitigate NCD during sleep in patients receiving LTOT with hypoxemic CRF from COPD could be tested in clinical trials.

Limitations

This was a small, single-site study, comprised entirely of male patients who are predominantly Black veterans. The telephone interviews with veterans self-reporting NCD during their sleep are prone to recall bias. In addition, the validity and reproducibility of NCD during sleep were not addressed in this study. Missing data from 9 nonresponders may have introduced a nonresponse bias in data analysis and interpretation. The overall hospital cost for a COPD exacerbation at JBVAMC was derived from VA data; US Centers for Medicare & Medicaid Services or commercial carrier data may be different.^15,21 Lastly, access to LTOT for veterans with hypoxemic CRF from COPD is regulated and supervised at VA medical facilities.¹⁴ This process may be different for patients outside the VA. Taken together, it is difficult to generalize our initial observations to non-VA patients with hypoxemic CRF from COPD who are receiving LTOT. We suggest a large, prospective study of veterans be conducted to determine the prevalence of NCD during sleep and its relationship with COPD exacerbations in veterans receiving LTOT with hypoxemic CRF due to COPD.

CONCLUSIONS

Clinicians at the JBVAMC did not document the presence of NCD during sleep in patients with hypoxemic CRF from COPD who received LTOT. However, self-reported, weekly nocturnal NCD episodes were associated with a higher hospitalization rate for COPD exacerbation and higher hospital costs. Accordingly, user-friendly devices to mitigate NCD during sleep should be developed.

Acknowledgments

We thank Yolanda Davis, RRT, and George Adam for their assistance with this project.