Burn Pit Smoke Exposure and Sleep Apnea in US Veterans

Abstract

Introduction:

Burn pit smoke exposure (BPSE) during military deployment has been linked to long-term cardiorespiratory conditions, but its relationship with sleep apnea (SA) remains unclear. This study examines the association between BPSE and SA using Veterans Health Administration (VHA) electronic medical records (EMR) and the Airborne Hazards and Open Burn Pit Registry (AHOBPR).

Methods:

We conducted a retrospective cohort study of veterans from AHOBPR with VHA sleep study data. BPSE was classified into quartiles based on the duration of exposure, and SA severity was measured using the Apnea-Hypopnea Index (AHI). Logistic regression models and Cox proportional hazards models were used to evaluate the association between BPSE and SA, adjusting for confounders such as age, body mass index, smoking status, post-traumatic stress disorder (PTSD), and comorbid disease burden.

Results:

The study included 17,064 veterans (mean age 40.2 y; 89.6% male; 58.3% with PTSD). Veterans in the highest BPSE quartile (≥245 d) had an unadjusted OR of 1.13 for SA, which became nonsignificant after adjustment (aOR: 1.10, P=0.058). The median time to SA diagnosis was 8.8 years in the highest BPSE group versus 11.1 years in the lowest. The adjusted Hazard Ratio for earlier SA diagnosis in the highest BPSE quartile was 1.16 (95% CI: 1.10, 1.22).

Discussion:

Although BPSE was not associated with SA prevalence, it was linked to earlier diagnosis. BPSE-related airway inflammation or increased health care use among exposed veterans may explain this pattern. Findings support early surveillance and screening for SA in highly exposed veterans.

The US military widely utilized burn pits during post-9/11 deployments in Iraq and Afghanistan to dispose of various types of waste, including electronics, medical waste, plastics, batteries, and human waste.^1,2 The burning of these materials releases a complex mixture of harmful pollutants, such as particulate matter (PM) 2.5, dioxins, furans, volatile organic compounds, lead, mercury, and polycyclic aromatic hydrocarbons.^3,4

This burn pit smoke exposure (BPSE) has been linked to numerous long-term health outcomes, particularly affecting the respiratory and cardiovascular systems, such as asthma, chronic obstructive pulmonary disease, and hypertension.^1,5,6 The STAMPEDE study investigated 50 military personnel who, despite having no pre-deployment history of lung or heart disease, reported new pulmonary symptoms post-deployment, and had higher objective airway hyperreactivity/asthma as well as reduced diffusing capacity of the lungs for carbon monoxide.⁷ A later cohort study of nearly 460,000 veterans with a median follow-up of 11 years post-deployment demonstrated that for every 100 days of burn pit exposure, the odds of developing asthma, COPD, hypertension, and ischemic stroke were higher.¹ These findings suggest that BPSE may be associated with mixed, heterogeneous lower airway syndromes. However, the risk may depend upon the duration and total amount of exposure to burn pit pollutants.⁴

Given the overall impact of these pollutants on the respiratory system, it is plausible that BPSE contributes to the development or exacerbation of sleep disorders involving the upper airways, such as sleep apnea (SA). SA is a significant public health issue in the US population due to its high prevalence and associated comorbidities. It is linked to adverse health outcomes such as hypertension, cardiovascular disease, diabetes, depression, and impaired quality of life.⁸ For veterans, the prevalence and impact of sleep apnea are even more pronounced. Studies indicate that SA is more than twice as prevalent among veterans compared with non-veterans.^9,10 Approximately 27% of post-9/11 veterans have been diagnosed with sleep-related breathing disorders, according to a 20-year prospective cohort study.¹¹ The economic burden of SA is substantial, with increased health care utilization, including outpatient visits, inpatient stays, and emergency department encounters.¹² The Veterans Health Administration (VHA) has implemented policies to address veterans’ sleep apnea needs through initiatives such as the TeleSleep Enterprise-Wide Initiative and the Remote Veteran Apnea Management Platform, which enhance access through telehealth and the Home Sleep Apnea Testing programs.

Several upper airway respiratory diseases, such as sinonasal disease, have been shown to be associated with BPSE.¹³ Liu et al¹⁴ found that long-term exposure to air pollutants can increase the occurrence of sleep disorders and can reduce sleep duration. Similarly, emerging evidence suggests that air pollution can influence body water distribution and contribute to the risk of sleep disorders such as low-arousal-threshold obstructive SA.¹⁵ Although the association between ambient air pollution and sleep disorders, including SA, is increasingly well-documented, the specific impact of BPSE on SA severity among veterans remains unknown.

The primary objective of this study is to evaluate the association between BPSE and SA prevalence and severity among US veterans, utilizing data from the VA’s Airborne Hazards and Open Burn Pit Registry (AHOBPR) and VHA electronic medical records (EMR). By leveraging a large, well-characterized cohort and applying advanced data extraction methods, this study aims to provide insights into the potential health risks associated with burn pit exposures, specifically focusing on their impact on sleep health.

METHODS

Study Design and Data Sources

This study used a retrospective cohort using data from 2 primary sources: the VA AHOBPR and the VHA EMR. The AHOBPR identified Veterans who were deployed to relevant combat theaters and their self-reported BPSE data, whereas the VHA EMR provided clinical information, including sleep study results, demographics, and comorbid diagnoses. This study was approved by the Institutional Review Board (IRB) of Baylor College of Medicine and Michael E. DeBakey VA Medical Center (H-47595).

Study Population

The study cohort consisted of US veterans who were enrolled in the AHOBPR and had sleep study data recorded in the VHA EMR. A total of 4,837,321 veterans with any sleep diagnosis or services up to September 2023 were initially identified from the VHA EMR. Veterans included in the study were those who met the following inclusion criteria: they completed the AHOBPR questionnaire, utilized VHA sleep clinic services with one or more documented sleep study results (either polysomnography (PSG) or home sleep apnea tests (HSAT), and had a recorded Apnea-Hypopnea Index (AHI) value derived from sleep study reports using a previously validated natural language processing (NLP) algorithm.¹⁶

After applying the inclusion criteria, the final study cohort comprised 17,217 veterans. (Fig. 1). Veterans with complete BPSE from the AHOBPR (n=330,469) or AHI from the EMR (n=423,087) were included from the analysis. The veteran that their data existed in both dataset (n=17,370) formed the final cohort. In addition, if a subject was diagnosed with sleep apnea before the last day of deployment, they were excluded (n=153).

FIGURE 1.

SSROBE diagram. The cohort is constructed by crossing 2 cohorts of data. The study cohort consisted of US veterans who were enrolled in the VA’s Airborne Hazards and Open Burn Pit Registry (AHOBPR) and had sleep study data recorded in the Veteran Health Administration (VHA) Electronic Medical Records (EMR).

Data Sources and Extraction

AHOBPR: Established in 2014, the AHOBPR collects data on veterans’ exposure to airborne hazards, including burn pit smoke, during deployment. The voluntary registry uses an online questionnaire to obtain self-reported data on deployment history, environmental exposures, symptoms, medical history, and lifestyle factors. The AHOBPR is a voluntary registry for Veterans and service members who served in specified military campaigns or locations. According to Department of Defense records, individuals who served in Operations Desert Shield, Desert Storm, Iraqi Freedom, Enduring Freedom, New Dawn, and in regions including Southwest Asia, Afghanistan, and associated airspaces between August 2, 1990, and August 31, 2021, are eligible to participate. Registry participants self-enrolled through an online portal from 25 May 2014 to 1 August 2022.

VHA EMR: The VHA is the largest integrated health care system in the United States, encompassing 1321 health care facilities, including 172 VA Medical Centers and 1138 outpatient sites.¹⁷ Clinical data were extracted from the VHA EMR, which includes detailed information from sleep studies, including PSG and HSAT results. The Corporate Data Warehouse (CDW) and VA Informatics and Computing Infrastructure (VINCI) were used to access and analyze the EMR data, ensuring data security and veterans’ privacy.

Data Extraction and NLP:¹⁶ Sleep study results, particularly the Apnea-Hypopnea Index (AHI), were extracted from clinical notes using an NLP algorithm. This algorithm was designed to identify and quantify AHI values from both PSG and HSAT reports. The NLP approach was validated using a random sample of 921 annotated sleep study notes to ensure high accuracy in identifying AHI values across different categories of SA severity: No Sleep Apnea (AHI<5), Mild to Moderate Sleep Apnea (5 ≤ AHI<30), and Severe Sleep Apnea (AHI ≥ 30). These cutoffs for AHI are recommended by the American Academy of Sleep Medicine (AASM).¹⁸ The NLP algorithm has high sensitivity (92%), specificity (88%), positive predictive value (PPV: 85%), and negative predictive value (NPV: 90%).¹⁶

Variables

Exposure Classification

Cumulative burn pit smoke exposure was calculated based on self-reported data from each deployment post–October 7, 2001. Participants reported average hours per day of exposure to burn pit smoke for each deployment segment. Deployment duration was calculated from start and end dates, and the reported average daily exposure was multiplied by the number of days in each segment to estimate cumulative hours of exposure. These values were then summed across deployment segments. Finally, cumulative hours were divided by 24 (hours/day).¹⁹ The total duration of BPSE was categorized into 5 levels to represent different levels of exposure as recommended in previous studies: Q0: Zero exposure; Q1: first quartile (0 to 46 d); Q2: second quartile (46 to 118 d); Q3: third quartile (118-245 d); Q4: fourth quartile (245+ days)^1,6,19,20. (Supplemental Fig. 1, Supplemental Digital Content 1, http://links.lww.com/MLR/D89)

Outcome Variables

The primary outcome was the prevalence of sleep apnea, classified by severity according to the AHI obtained from sleep studies. For the time to event analysis the time to AHI was estimated by subtracting the sleep study date minus the last date of deployment. The last date of deployment was obtained from AHOBPR data.

Other Variables

We defined the index date the date that the veterans completed the AHOBPR questionnaires. We extracted other variables such as sex, age at the index date of completing AHOBPR questionnaires, race, ethnicity, body mass index (BMI), Charlson comorbidity index (CCI) for a year before the index date, smoking status, and post-traumatic stress disorder (PTSD) status were extracted from VA EMR at the index date. BMI was defined as the most recent recorded BMI within 2 years before or up to 6 months after the index date. PTSD was defined using ICD-9 (“309.81”) and ICD-10 (“F43.1X”) codes, requiring at least one outpatient encounter or one inpatient diagnosis before the index date of the cohort. For analyses, age and BMI were modeled as continuous variables, whereas also categorized (age by decade; BMI using WHO cutoffs of <25, 25–29.9, ≥30) for descriptive summaries. Sex was coded as binary (male/female). Smoking status was categorized as never, former, or current. PTSD was treated as a binary variable (yes/no), and CCI was analyzed as both a continuous score and categorized as comorbid conditions with CCI≥2.

Statistical Analysis

Descriptive statistics were used to summarize the demographic and clinical characteristics of the study participants.

Logistic regression models: The association between BPSE and SA was examined by comparing the odds of having SA (defined as AHI ≥ 5) and severe SA (defined as AHI ≥ 30) across different levels of exposure. Adjusted odds ratios (aOR) were reported for each exposure group compared with the reference (Q0).

We constructed multivariable logistic regression models adjusting for potential confounders including age, sex, BMI, smoking status, PTSD, and CCI. We reported unadjusted ORs and aORs with 95% CI.

Time-to-Event Analysis

Cox proportional hazards models were used to analyze the time to diagnosis of SA across different levels of BPSE exposure. Hazard ratios were adjusted for age, sex, BMI, smoking status, PTSD, CCI, and the year when the deployment ended (as a continuous variable). Because of low sample size in the Q0 group, we only reported the analysis for Q1 to Q4, which represented the veterans with BPSE experiences.

RESULTS

The cohort consisted of 17,064 veterans with a mean age of 40.2 years, predominantly male (89.6%), and a high prevalence of PTSD (58.3%). BMI averaged 31.4 kg/m2, with 57% classified as obese (BMI ≥ 30). Most of the veterans had a relatively low burden of comorbid conditions (CCI<2). The demographic and clinical characteristics across different levels of BPSE were generally similar, with PTSD prevalence and smoking rates slightly higher in those reporting higher BPSE levels. (Table 1)

TABLE 1.

Patients’ Characteristics and Demographics

Exposure days	All	Q0, 0	Q1, >0 & ≤46;	Q2, <46 & ≤118;	Q3, >118 &≤245;	Q4, >245
n	17,217	129	4281	4298	4250	4259
Age, M (SD)	40.6 (9.1)	39 (8.9)	41.1 (9.5)	40.8 (9.1)	40.2 (9.2)	40.2 (8.5)
Age, <30, N (%)	1655 (9.6)	18 (14)	419 (9.8)	418 (9.7)	466 (11)	334 (7.8)
Age, 30–40, N (%)	7155 (41.6)	54 (41.9)	1684 (39.3)	1714 (39.9)	1791 (42.1)	1912 (44.9)
Age, 40–50, N (%)	5157 (30)	41 (31.8)	1237 (28.9)	1330 (30.9)	1214 (28.6)	1335 (31.3)
Age, 50–65, N (%)	3141 (18.2)	14 (10.9)	908 (21.2)	810 (18.8)	750 (17.6)	659 (15.5)
Age, ≥65, N (%)	109 (0.6)	2 (1.6)	33 (0.8)	26 (0.6)	29 (0.7)	19 (0.4)
BMI, M (SD)	31.4 (5.1)	31.4 (5.8)	31.4 (5.2)	31.4 (5.1)	31.5 (5.2)	31.2 (5)
BMI, ≥30, N (%)	9846 (57.2)	70 (54.3)	2444 (57.1)	2455 (57.1)	2491 (58.6)	2386 (56)
Sex, male, N (%)	15,326 (89)	107 (82.9)	3803 (88.8)	3806 (88.6)	3790 (89.2)	3820 (89.7)
Race, N (%)
White	11424 (66.4)	75 (58.1)	2806 (65.5)	2841 (66.1)	2841 (66.8)	2861 (67.2)
Black	3532 (20.5)	38 (29.5)	921 (21.5)	887 (20.6)	833 (19.6)	853 (20)
Others	2261 (13.1)	16 (12.4)	554 (12.9)	570 (13.3)	576 (13.6)	545 (12.8)
Ethnic Hispanic, N (%)	2761 (16)	23 (17.8)	674 (15.7)	698 (16.2)	694 (16.3)	672 (15.8)
Smoking status, N (%)
Never smoker	8531 (49.5)	73 (56.6)	2171 (50.7)	2182 (50.8)	2086 (49.1)	2019 (47.4)
Former smoker	4639 (26.9)	30 (23.3)	1162 (27.1)	1153 (26.8)	1123 (26.4)	1171 (27.5)
Active smoker	4047 (23.5)	26 (20.2)	948 (22.1)	963 (22.4)	1041 (24.5)	1069 (25.1)
AHI, M (SD)	16.2 (16.5)	13.6 (16.1)	16.5 (16.9)	16 (16.3)	16.5 (16.9)	16.1 (16.2)
AHI <5, N (%)	3812 (22.1)	39 (30.2)	951 (22.2)	965 (22.5)	957 (22.5)	900 (21.1)
AHI, 5–30, N (%)	10,391 (60.4)	71 (55)	2547 (59.5)	2610 (60.7)	2530 (59.5)	2633 (61.8)
AHI ≥ 30, N (%)	3014 (17.5)	19 (14.7)	783 (18.3)	723 (16.8)	763 (18)	726 (17)
CCI, M (SD)	0.3 (0.7)	0.2 (0.5)	0.3 (0.8)	0.3 (0.8)	0.3 (0.8)	0.3 (0.7)
CCI, ≥2, N (%)	885 (5.1)	3 (2.3)	212 (5)	243 (5.7)	212 (5)	215 (5)
PTSD, N (%)	10,560 (61.3)	55 (42.6)	2345 (54.8)	2528 (58.8)	2742 (64.5)	2890 (67.9)

AHI indicates apnea-hypopnea index; BMI, body mass index (kg/m2); CCI, Charlson comorbidity index; M (SD), mean and standard deviations.

When comparing the odds of association between SA and different levels of BPSE, the multivariate logistic regression analysis showed that the highest quartile of BPSE (245+ d vs. no exposure as reference) was significantly associated with increased odds of SA in the unadjusted analysis (OR: 1.13, 95% CI: 1.02, 1.24, P=0.018). However, after adjusting for confounding factors, this association lost statistical significance (aOR: 1.10, 95% CI: 1.00, 1.21, P=0.058). In the adjusted analysis we noted that traditional risk factors for SA, such as obesity, male sex, age, and higher comorbidity burden, were associated with a higher risk of SA. (Table 2).

TABLE 2.

Comparing the Odds of Association Between Sleep Apnea and Different Levels of Burn Pit Smoke Exposure Compared With No self-Reported Burn Pit Smoke Exposure

	OR (95%CI)		aOR (95%CI)
No exposure	Reference	P	Reference	P
Low exposure, 0–49 d	1.12 (1.01, 1.23)	0.029	1.08 (0.98, 1.19)	0.132
Mild exposure, 49–123 d	1.12 (1.02, 1.24)	0.023	1.09 (0.98, 1.20)	0.099
Moderate exposure, 123–255 d	1.10 (1.00, 1.22)	0.048	1.07 (0.97, 1.18)	0.176
Severe exposure, 255+ d	1.13 (1.02, 1.24)	0.018	1.10 (1.00, 1.21)	0.058
CCI, ≥ 2 vs. <2			1.06 (1.02, 1.09)	0.002
Race, Black vs. White			1.00 (0.98, 1.02)	0.913
Race, Others vs. White			1.03 (1.01, 1.06)	0.009
Hispanic vs. non-Hispanic			1.06 (1.04, 1.09)	0.000
Sex, Male vs. female			1.24 (1.20, 1.27)	0.000
BMI, 18.5- vs. 18.5–30			1.04 (0.88, 1.23)	0.644
BMI, 30+ vs 18.5–30			1.19 (1.17, 1.21)	0.000
PTSD vs. no PTSD			1.00 (0.99, 1.02)	0.609
Age, ≥ 50 vs. <50			1.18 (1.15, 1.20)	0.000
Former smoker vs. never smoker			1.03 (1.01, 1.05)	0.007
Active smoker vs. never smoker			0.99 (0.97, 1.01)	0.463

BMI indicates body mass index; CCI indicates Charlson comorbidity index; OR, odds ratio; PTSD, post-traumatic stress disorder.

Cox proportional hazards models demonstrated that the highest quartile of BPSE exposure (Q4) was associated with earlier SA diagnosis, with a median time to diagnosis of 8.8 years (interquartile range: 5.5, 12.0) compared with 11.1 years (interquartile range: 7.2, 15.8) in the lowest quartile (Q1) (Fig. 2). There was a dose-dependent effect where the duration of BPSE progressively increased the odds of earlier SA diagnosis, with the highest exposure group (>245+ d) having an unadjusted hazard ratio of 1.76 (95% CI: 1.68, 1.85) and an adjusted Hazard Ratio of 1.16 (95% CI: 1.10, 1.22), Figure 2.

FIGURE 2.

Time to diagnosis of sleep apnea by burn pit smoke exposure (BPSE) duration using the lowest quartile of exposure as the reference, that is, Q1.(Q1, >0 & ≤46; Q2, <46 & ≤118; Q3, >118 &≤245; Q4, >245).

DISCUSSION

Our findings suggest that higher BPSE is associated with earlier diagnosis of SA even after adjusting for many clinically relevant variables. However, BPSE was not associated with a higher prevalence of SA overall as measured by AHI from clinical sleep studies. Therefore, this study provides evidence that BPSE may be linked to earlier SA diagnosis in veterans.

Veterans with the highest levels of BPSE were diagnosed with SA ~2.5 years earlier than those with lower levels of exposure in a dose-dependent manner. This association remained robust even after adjusting for the deployment end date and other relevant comorbidities. There are several possible hypothetical explanations for these findings, namely (a) BPSE’s impact on accelerating time to SA development and/or increased health care engagement among veterans with higher BPSE. The shortened latency to SA may be due to prolonged exposure to airborne toxins from burn pit smoke triggering an inflammatory response in the airways, potentially accelerating the development of SA in predisposed individuals. The resulting inflammation, along with increased airway reactivity and reduced muscle tone, may reduce the time to onset of SA, as detected in the clinical sleep studies. This study, however, did not specifically investigate the pathophysiological factors contributing to the earlier onset of SA.

Increased health care engagement is another possible explanation. Veterans may receive earlier, more frequent, or more intensive medical evaluations due to their exposure concerns or other medical conditions brought about or exacerbated by deployment. Similar patterns of higher health care utilization have been observed in other chronic disease conditions^21–24 and is called “confounding by indication”. Regardless of the underlying etiology, these results suggest that clinical and public health surveillance may be important for accelerating the detection and treatment of SA, especially in this deployed cohort.

The limited impact of BPSE on SA severity could be due to the complex and multifactorial nature of sleep apnea.^25–27 Known risk factors such as male sex, obesity, anatomical variations, and age may overshadow the effects of environmental exposures^28–30 The expected associations with SA in our study strengthen its internal validity. Our study population was predominantly male, with a high prevalence of obesity. Subjects were otherwise relatively healthy and had lower rates of comorbid conditions. These characteristics may dilute the isolated effect of BPSE, thus reflecting the complexity of sleep apnea pathogenesis.

Prior studies have also suggested an association between environmental pollution and SA.^14,15 In addition, current cigarette smokers, particularly heavy smokers, have a higher risk for sleep-disordered breathing, whereas former smokers do not show an increased risk compared with never-smokers.³¹ Although there are parallels between smoking exposure or environmental pollution and BPSE, there are also significant distinctions. Environmental pollution tends to be chronic, with long-term exposure affecting residents continuously. In contrast, BPSE is typically limited to the deployment period,³² with exposure ceasing once veterans return home. This potentially creates a latent period for the development of SA. Although this study was not specifically designed to evaluate the potential effects of this post-deployment period, it may offer a plausible explanation for the relatively modest impact of BPSE on SA.

This study has several strengths. The large sample size provides robust statistical power, and the use of a validated natural language processing algorithm for extracting the AHI from clinical text significantly enhances the precision of our outcome measures compared with traditional ICD code-based diagnoses. The comprehensive adjustment for confounders and the integration of multiple data sources (AHOBPR and VHA EMR) also strengthen the validity and relevance of our findings. Limitations of the study include reliance on self-reported exposure data from the AHOBPR, which introduces the potential for recall bias.³³ Although the NLP algorithm demonstrated high recall and specificity, its overall accuracy was in the range of 86%–87%. The lack of information on specific pollutants or intensity of exposure also restricts our ability to identify the most harmful components of BPSE. Although we adjusted for key confounders such as BMI, and smoking status, residual confounding from other factors, such as socioeconomic status or sleep-related behaviors, cannot be ruled out. Finally, the population in this study was predominantly male sex, older age, and had higher BMI, thus generalizability to other populations may be limited.

Despite these limitations, our findings carry important clinical and policy implications. If confirmed, our findings could have significant implications for the VHA. These results underscore the importance of targeted screening and enhanced clinical surveillance for SA among veterans with high BPSE to facilitate earlier diagnosis and intervention, perhaps using the mandatory Toxic Exposure Screen already asked of more than 5 million VHA users to initiate follow-up screening for SA. Integrating these findings into existing VHA initiatives, such as the TeleSleep Enterprise-Wide Initiative and Home Sleep Apnea Testing, could optimize resource allocation by prioritizing Veterans with risk factors who are more likely to benefit from timely treatment. For non-VA health care systems, our findings highlight the importance of considering environmental exposures in the clinical assessment of SA. The results can have implications for populations exposed to occupational or environmental pollutants, such as firefighters or other exposed workers.

Future research should aim to clarify the mechanisms through which BPSE influences sleep health and investigate whether specific subpopulations of veterans (eg, those with pre-existing respiratory conditions or comorbid PTSD) are more vulnerable to the effects of BPSE on sleep. Longitudinal studies that track changes in SA severity over time in relation to BPSE may also provide further insights into the long-term impacts of these exposures on sleep health.

In conclusion, this study found that higher BPSE is associated with earlier diagnosis of SA after adjusting for clinically relevant variables. These findings underscore the importance of continued research into the respiratory effects of BPSE and highlight the need for comprehensive monitoring of veterans’ health outcomes, including sleep disorders, after deployment.

Supplementary Material

mlr-64-s161-s001.docx ^{(684.8KB, docx)}