Abstract
Objective
Rheumatoid arthritis (RA) and other autoimmune (AI) conditions are associated with inorganic dust exposure. Many military activities are likely to entail inorganic dust exposures. We wished to identify associations between prior military dust exposure and RA and other AI conditions.
Methods
We studied persons from a roster of Army, Navy, Air Force, or Marine Corps personnel who had served in Operation Enduring Freedom and Operations Iraqi Freedom and New Dawn. We linked military occupational codes to a job exposure matrix assigning dust exposure likelihood. We used the Veterans Affairs Health Care System (VAHCS) electronic health care records to identify cases of RA, systemic lupus erythematosus (SLE), systemic sclerosis (SSc), vasculitis, and inflammatory myositis. Generalized estimating equations modeled risk of RA and other AI conditions associated with dust exposure, taking into account military service branch, age at first VAHCS encounter, sex, race/ethnicity, smoking status, and years of military service.
Results
Of 438 086 veterans (68% ever‐smokers), 44% were classified with likely or somewhat likely dust exposure. Cases included 1139 cases with RA, 467 cases with SLE, and 180 cases with other AI diseases (SSc, vasculitis, or inflammatory myositis). Military dust exposure was associated with increased odds of RA (odds ratio [OR] = 1.10; 95% confidence interval [CI] = 1.003‐1.20) and increased odds of SSc, vasculitis, or inflammatory myositis (OR = 1.23; 95% CI = 1.14‐1.34) but was protective for SLE (OR = 0.81; 95% CI = 0.76‐0.88).
Conclusion
Dust exposure during past military service comprises an occupational and environmental risk factor for RA and other AI diseases. This is potentially relevant for prevention activities.
Significance & Innovations.
In the first study of this kind, dusty military occupations were associated with increased odds of rheumatoid arthritis.
These were also associated increased odds of systemic sclerosis, vasculitis, or inflammatory myositis.
There were no similarly increased odds for systemic lupus erythematosus.
Military service includes occupations that may carry autoimmune disease risk from dust inhalation.
INTRODUCTION
There is mounting epidemiologic evidence that exposure to inorganic dust is associated with increased risk of rheumatologic and other autoimmune (AI) diseases. The links between dusty environments and rheumatoid arthritis (RA) are evident among miners, especially coal miners, and are strongly implicated in a number of other silica‐exposed vocations (1, 2). Epidemiological associations have also been shown between silica inhalation and systemic sclerosis (SSc) and systemic lupus erythematosus (SLE) (3, 4).
Military service and risk of AI disease have been considered only to a limited degree (5, 6). Nonetheless, when considered in the context of its associated vocational and environmental aspects, military service is likely to be associated with excess exposure to inorganic dusts through a variety of duty assignments. Occupational factors include welding, abrasive blasting, grinding and polishing metals, vehicle and other maintenance work, earth moving and other construction, and explosives detonation (7). In addition, ambient dusty conditions (from duties requiring substantial outdoor time, especially duties that typically disturb soil) can add to the burden of inhaled inorganic dust.
Identifying the presence and magnitude of risk for RA and related AI conditions (eg, SSc and SLE) associated with military service would be directly impactful to the health care of military service personnel and veterans, through both primary prevention and targeted follow‐up surveillance. We wished to determine the association between military assignments with likely job‐related inorganic dust inhalation and the subsequent development of RA and other AI conditions using an established roster of separated service men and women who served in Operation Enduring Freedom (OEF) or Operations Iraqi Freedom and New Dawn (OIF and OND, respectively) and who subsequently had received care through the Veterans Administration Health Care System (VAHCS).
PATIENTS AND METHODS
Study population
The sampling frame of our study consists of the rosters of OEF and OIF and OND veterans. OEF predominantly includes veterans who served in Afghanistan; the latter two include those who served in Iraq. The roster is a list of veterans who separated from OEF/OIF/OND military service and is provided to the Veterans Affairs (VA) by the Department of Defense Manpower Data Center. This roster has been previously used to link prior military service to disease outcomes through VA medical record linkages (8, 9, 10, 11). This study was approved by the San Francisco VA Health Care System and the University of California San Francisco internal review boards (#17‐23248); a waiver of informed consent was granted because of the nature of the study.
Data sources
Data fields related to military service such as branch of service (Army, Navy, Air Force, Coast Guard, or Marine Corps), component (National Guard, Reserve, or active duty), rank, dates of deployment, and military occupation codes were obtained from the OEF/OIF/OND roster. The roster does not contain clinical data. We obtained relevant clinical information from the VA Corporate Data Warehouse, which contains data elements extracted from the national VA electronic medical record. These include diagnostic and procedure codes associated with inpatient and outpatient encounters, laboratory results, and pharmacy data. Data were accessed through the VA Informatics and Computing Infrastructure (VINCI) platform. Our study included all clinical data from the time of post–military discharge VA enrollment through September 30, 2018.
Inclusion/exclusion criteria
On the basis of OEF/OIF/OND roster data, we considered veterans from the Army, Navy, Air Force, or Marine Corps (n = 722 352) as potentially eligible for inclusion, of whom 438 086 were included in the final analysis (Figure 1). The bulk of the 288 266 excluded fell into several categories, as follows: 57 362 officers or warrant officers (who would likely differ systematically in exposure and other characteristics), 70 472 who served for fewer than 2 years (minimal duration of potential exposure), and 78 177 with either no in‐person encounters in the VA health care system or less than 90 days of VA follow‐up. There were also 34 639 who were excluded because there was a record of a VA encounter prior to the final military discharge date, which could reflect a reenlistment after previously initiating VA care (thus introducing uncertainty in the timing of exposure). Lastly, we excluded 43 492 with missing smoking status from the main analyses.
Figure 1.
Study inclusion and exclusion criteria. VA, Veterans Affairs.
Definition of RA and related AI diseases
We used ICD9 and ICD10 codes associated with VA encounters to identify veterans with any of the following diseases: RA, SSc, SLE, inflammatory myositis, and vasculitis. For each condition, we required a minimum of two face to face encounters at least 30 days apart with the relevant International Classification of Diseases (ICD) code(s). For RA, we also created a more specific definition, requiring that the ICD codes be associated with two encounters at least 180 days apart, of which at least one encounter was classified as a rheumatology clinic, as well as at least one prescription for a disease‐modifying antirheumatic drug (DMARD). These definitions have been previously validated in administrative data–based research (including VA records) on rheumatoid arthritis and related Ai diseases (12, 13, 14, 15, 16). Patients with RA were categorized according to serologic status, based on positive results from either rheumatoid factor (RF) or cyclic citrullinated peptide (CCP) antibody assays versus negative results of both tests versus unknown results of both tests. We also created the following two combined diagnosis variables: any AI disease except RA and any AI disease except RA and SLE (ie, SSc, inflammatory myositis, or vasculitis).
Although it is possible for an individual to have more than one AI disease diagnosis, for the purposes of this analysis, we chose to hierarchically categorize all patients as only having a single diagnosis (from highest priority to lowest: SSc, vasculitis, inflammatory myositis, SLE, and RA). This hierarchy is based on a previously established diagnostic schema (17).
Exposure assessment
Inorganic dust exposure during service was estimated via military occupation codes (MOCs) contained within the OEF/OIF/OND roster. In these numeric codes (varying by branch of service), the first two characters indicate the general branch of activity, and additional digits provide further detail. We applied a job exposure matrix (JEM) developed for this analysis on the basis of these MOCs. We based this approach on previous experience with JEM exposure assignment in nonmilitary occupations, including JEM exposure in RA risk (18, 19). The JEM categorized military duties as likely, somewhat likely, or not likely to involve inorganic dust exposure. The higher level (likely) included occupations such as construction, vehicle repair, firefighting, explosives disposal, and utilities repair/installation. The moderate level (somewhat likely) included examples such as electricians, electronics repair workers, telecom technicians, and motor vehicle operators. This JEM schema did not characterize likelihood of exposure to other dusts (such as from organic materials), gases, fumes, or vapors, although occupational categories such as construction involve a range of potential inhaled exposures. All supervisors of either moderate‐ or higher‐level occupations were assigned intermediate‐level exposure. As a sensitivity analysis, we downgraded the exposure level of supervisors of moderate‐level occupations to the negligible exposure; the results did not differ from those of the main analysis and are not presented here.
Smoking status
We assigned smoking status (defined dichotomously as ever‐smoker: yes or no) using Health Factor data in the VA electronic medical record. These data are collected by VA providers in response to automated clinical reminders. Using Health Factor data in this way has been validated in comparison to patient questionnaire (20). For veterans not identified as ever‐smokers on this basis, additional information leading to ever‐smoker classification was based on any of the following being present in the medical record: a consult request for smoking cessation clinic, an encounter with a smoking cessation clinic, or a prescription for a nicotine replacement product and/or varenicline. As noted previously, indeterminate smoking status was a study exclusion criterion.
Other covariates
Demographic variables, including date of birth, sex, and race/ethnicity were obtained from the OEF/OIF/OND roster. If race was listed as “missing” or “other,” we substituted race/ethnicity data from the VA electronic medical record, using an algorithm to classify race/ethnicity that was employed in our previous work (21). This method resulted in classification of all but 2.6% of the sample, who were classified as “other” for these analyses. We used ICD9 and ICD10 codes associated with VA encounters 30 days or more apart to identify veterans with comorbid post‐traumatic stress disorder (PTSD) to include in an additional analysis, given that this diagnosis has been associated with RA (11).
Statistical analysis
We used generalized estimating equations (GEEs) to model the risk of RA and other AI conditions (separately) associated with inorganic dust exposure as measured by the JEM as a dichotomous variable (negligible versus intermediate/higher level). All models controlled for age at first VA encounter, race/ethnicity, sex, and smoking status. The GEE models accounted for correlation within members of the same branch of service.
To further explore the role of smoking in these analyses, we re‐estimated the main study model including a dust‐smoking interaction term and separately re‐estimated the model including patients with unknown smoking status (as a separate category). To explore the potential role of PTSD comorbidity on dust exposure risk in RA, we added PTSD as an independent predictor variable to the main study model.
We further modeled risk stratification separately among seropositive RA and seronegative RA. In a sensitivity analysis, we examined the effects of assigning all patients with unknown serologic status first as seropositive and then as seronegative. In another sensitivity analysis, we used a more specific definition of RA that included prescription of a DMARD.
We also carried out a series of stratified analyses. Because of the potential interrelationship of disease onset and length of military service, the data were partitioned at 2 to 4 years, 4 to 8 years, and more than 8 years of service, roughly equivalent to the 25th, 50th, and 75th percentiles of the study population. We remodeled the risk of SLE stratified by sex, given the extreme sex imbalance in that disease and likely sex differences in exposures.
To reduce misclassification bias, models of each condition studied excluded those diagnosed with any of the other conditions, such that the comparison group always consisted of individuals with none of the AI diagnoses of interest. In analyses based on RA serologic status, we also excluded persons with positive serology who did not meet the study definition of RA. All analyses were conducted in SAS version 9.4.
RESULTS
Among 438 086 veterans included in this analysis, 88% were men and 63% white, non‐Hispanic (Table 1). Average age at first VAHCS encounter after military discharge was 30.4 years, ranging from 19 years to 66 years. Nearly half of the veterans had served in the Army. The median length of service for the entire sample was 4.3 years (interquartile range = 3.2‐8.2 years). Approximately two‐thirds were ever‐smokers. The majority of those studied (56%) were classified as having low or no likelihood of inorganic dust exposure through their military occupation. The proportions with moderate or high dust exposure employments were similar (22.2% and 21.6 %, respectively). The median duration of care through the VA was 6.8 years (interquartile range = 4.7‐9.5 years).
Table 1.
Subject characteristics among 438 086 military veterans included in the analysis
| Characteristic | Frequency |
|---|---|
| Sex, male, n (%) | 385 516 (88.0) |
| Age at first VA encounter, mean ± SD (range), years | 30.4±7.3 (19‐66) |
| Race/ethnicity, n (%) | |
| White | 276 481 (63.1) |
| Black | 84 576 (19.3) |
| Hispanic | 51 815 (11.8) |
| Asian/Pacific Islander | 13 782 (3.1) |
| Other | 11 432 (2.6) |
| Years of military service, median (IQR) | 4.3 (3.2‐8.2) |
| Service branch, n (%) | |
| Army | 213 182 (48.7) |
| Air Force | 53 848 (12.3) |
| Marine | 90 469 (20.7) |
| Navy | 80 587 (18.4) |
| Education level, n (%) | |
| Less than HS | 3359 (0.8) |
| HS diploma | 397 629 (90.8) |
| Post‐secondary education | 37 098 (8.5) |
| Smoking status, n (%) | |
| Ever | 296 796 (67.7) |
| Never | 141 290 (32.3) |
| Dust exposure in military occupation, n (%) | |
| Low/no exposure | 246 200 (56.2) |
| Any exposure | 191 886 (43.8) |
| Moderate exposure | 97 320 (22.2) |
| High exposure | 94 566 (21.6) |
| Years of VA observation, median (IQR) | 6.8 (4.7‐9.5) |
| Number of encounters, median (IQR) | 21 ( 9‐49) |
| Diagnoses of autoimmune disease, n (%) | |
| Rheumatoid arthritis, sensitive definition | 1139 (0.26) |
| Rheumatoid arthritis, specific definition | 660 (0.15) |
| Systemic lupus erythematosus | 467 (0.11) |
| Other autoimmune diseases | 180 (0.04) |
| Vasculitis | 76 (0.02) |
| Inflammatory myositis | 55 (0.01) |
| Systemic sclerosis | 49 (0.01) |
Abbreviation: HS, high school; IQR, interquartile range; VA, Veterans Affairs.
For sensitive and specific definitions of rheumatoid arthritis, see Methods.
Using the primary definition of RA, which was based on ICD‐linked encounters alone, we identified 1139 cases of RA (26 cases per 10 000), of whom 461 (40%) were known to be positive for either RF or anti‐CCP (171 [15%] had unknown serologic status). Using the more specific definition of RA, which included a longer observation period and a DMARD prescription, the number with RA was reduced to 660 (15 per 10 000), of whom 359 (54%) were seropositive, 271 (41%) were seronegative, and 30 (5%) had unknown serologic status. SLE was the next most common rheumatologic condition (467 cases, including 298 among women); other AI diseases (SSc, vasculitis, or inflammatory myositis) were present in the medical records of 180.
In multivariable general estimating modeling (Table 2), moderate‐ and higher‐level dust in military employment was associated with 10% increased odds of RA (odds ratio [OR] = 1.10; 95% confidence interval [CI] = 1.003‐1.20). Several of the covariates included in the model also were statistically associated with RA (Table 2), most notably, age at first VA encounter (OR = 1.07; 95% CI = 1.05‐1.08), female sex (OR = 3.03; 95% CI = 2.82‐3.25), ever smoking (OR = 1.25; 95% CI = 1.10‐1.41), and Hispanic ethnicity (OR = 1.26; 95% CI = 1.05‐1.52). A smoking × dust exposure term added to this model was not consistent with an interaction (P = 0.64). In a sensitivity analysis in which persons with unknown smoking status (n = 42 492) were added back into the study population, there was no substantive change in the estimated odds for RA associated with moderate to high dust exposure (OR = 1.09; 95% CI = 1.002‐1.19). We further retested the main model, adding PTSD as a covariate. This had minimal impact on the previously estimated OR for RA associated with moderate to higher‐level dust exposure (OR 1.11; 95% CI 1.02‐1.21).
Table 2.
Risk of RA associated with dust exposure in military occupations, with and without adjustment for covariates
| Risk Factors | Odds (Risk) of RA | |
|---|---|---|
| Unadjusted OR (95% CI) | Adjusted OR (95% CI) | |
| Dust exposure (any versus none) | 1.02 (0.87‐1.19) | 1.10 (1.003‐1.20) |
| Age at first VA encounter, years | 1.06 (1.04‐1.09) | 1.07 (1.05‐1.08) |
| Female sex | 2.80 (2.44‐3.21) | 3.03 (2.82‐3.25) |
| Race (versus white) | ||
| White, non‐Hispanic (reference) | 1.0 (‐‐‐) | 1.0 (‐‐‐) |
| American Indian | 1.68 (0.98‐2.87) | 1.39 (0.74‐2.61) |
| Asian | 1.23 (0.81‐1.87) | 0.95 (0.66‐1.39) |
| Black | 1.50 (1.12‐2.01) | 1.09 (0.94‐1.27) |
| Hispanic | 1.26 (0.92‐1.72) | 1.26 (1.05‐1.52) |
| Hawaiian/Pacific Islander | 1.91 (1.22‐2.97) | 1.54 (0.92‐2.58) |
| Other | 1.16 (0.99‐1.35) | 0.97 (0.65‐1.45) |
| Ever‐smoker (versus never‐smoker) | 0.95 (0.80‐1.12) | 1.25 (1.10‐1.41) |
| Years of service | 1.06 (1.04‐1.09) | 1.00 (0.99‐1.02) |
Abbreviation: OR, odds ratio; RA, rheumatoid arthritis; VA, Veterans Affairs.
All estimates from general estimating equation models accounting for military service branch. Adjusted general estimating equation model includes all variables shown.
In further multivariable modeling that included two separate variables for moderate and high dust exposure (compared with no/low exposure), the ORs were 1.06 (95% CI = 0.97‐1.15) for moderate exposure and 1.14 (95% CI = 1.01‐1.28) for high exposure (see Supplemental Table S1).
When patients were disaggregated according to serologic status (Table 3), dust exposure was a statistically significant risk factor for seronegative RA (OR = 1.25; 95% = CI 1.14‐1.36) but not seropositive RA (OR = 1.02; 95% CI = 0.85‐1.21). Because 15% of patients with RA were missing serologic status, we examined the effect of assigning all those with unknown status first to seropositive and then to seronegative statuses. When the unknown cases were assigned to be seronegative, the OR for seronegative RA associated with dust exposure was reduced to 1.15 (95% CI = 1.09‐1.21), consistent with misclassification of some of the unknown cases. The dust‐associated risk estimates for RA were similar when the latter was defined by more specific DMARD‐based criterion (OR = 1.12; 95% CI 1.06‐1.20). Similarly, the risk estimates for seronegative (OR = 1.23; 95% CI = 1.07‐1.41) and seropositive RA (OR = 1.06; 95% CI = 0.88‐1.24) followed the same pattern as they did for the more sensitive RA definition (see Supplemental Table S2).
Table 3.
Risk of RA by seropositive and seronegative status associated with dust exposure intensity from military occupations
| Dust Exposure | Any RA | Seropositive RA | Seronegative RA | |||
|---|---|---|---|---|---|---|
| Percentage With Condition | Adjusted OR (CI) | Percentage With Condition | Adjusted OR (CI) | Percentage With Condition | Adjusted OR (CI) | |
| All | 0.260 | NA | 0.106 | NA | 0.117 | NA |
| No/low exposure | 0.258 | 1.0 (reference) | 0.109 | 1.0 (reference) | 0.108 | 1.0 (reference) |
| Moderate/high exposure | 0.264 | 1.10 (1.003‐1.20) | 0.102 | 1.01 (0.86‐1.20) | 0.128 | 1.25 (1.14‐1.36) |
Abbreviation: CI, confidence interval; NA, not applicable; OR, odds ratio; RA, rheumatoid arthritis.
All model estimates from generalized estimating equations accounting for military branch and controlling for age at first Veterans Affairs encounter, sex, race/ethnicity (seven categories), smoking status, and years of military service.
Models of serologic status also exclude 171 patients with unknown serostatus, 2416 seropositive patients without RA, and patients with the opposite serostatus (507 seronegative excluded from seropositive model; 451 seropositive excluded from seronegative model).
In the models stratified by years of service (Table 4), RA was most strongly associated with dust exposure in those with 4 years to 8 years of service (OR = 1.4; 95% CI = 1.3‐1.6); there was no statistical association among those with either fewer or more years of service. In contrast to RA, dust exposure was statistically protective for SLE (OR = 0.81; 95% CI = 0.76‐0.88), with this statistically significant protective association being present in those with 2 years to 4 years and 4 years to 8 years of service but not in the those with more than 8 years of service. The point estimate was even larger among women, with dust being statistically protective up to 8 years. In men, dust was not statistically protective. Disaggregating SSc, vasculitis, and inflammatory myositis (thus excluding SLE and RA) showed that these three conditions as a group were statistically associated with dust exposure in the full sample (OR = 1.23; 95% CI = 1.14‐1.34) and among the stratum of veterans with more than 8 years of service (OR = 1.57; 95% CI = 1.52‐1.62).
Table 4.
Risk of RA and other AI diseases associated with dust exposure in military occupations from generalized estimating equations accounting for military branch, with and without stratification by years of service
| Models | Cases/Total | Full Sample | 2‐4 Years of Service | >4‐8 Years of Service | >8 Years of Service | |||
|---|---|---|---|---|---|---|---|---|
| Adjusted OR (CI) for Dust Exposure | Cases/Total | Adjusted OR (CI) for Dust Exposure | Cases/Total | Adjusted OR (CI) for Dust Exposure | Cases/Total | Adjusted OR (CI) for Dust Exposure | ||
| RA | 1139/437 439 | 1.10 (1.003‐1.2) | 298/196 077 | 1.00 (0.68‐1.48) | 278/129 286 | 1.41 (1.27‐1.56) | 563/112 076 | 0.97 (0.83‐1.12) |
| SLE and other AI diseases | 647/436 947 | 0.92 (0.86‐0.99) | 246/196 025 | 0.87 (0.79‐0.96) | 166/129 174 | 0.77 (0.56‐1.06) | 235/111 748 | 1.13 (1.04‐1.23) |
| SLE only | 467/436 767 | 0.81 (0.76‐0.88) | 165/195 944 | 0.80 (0.75‐0.86) | 135/129 143 | 0.63 (0.44‐0.91) | 167/111 680 | 0.98 (0.85‐1.13) |
| SLE, women | 298/52 570 | 0.68 (0.65‐0.71) | 100/22 217 | 0.67 (0.54‐0.83) | 94/17 338 | 0.54 (0.32‐0.91) | 104/12 741 | 0.78 (0.48‐1.28) |
| SLE, men | 169/384 471 | 1.05 (0.81‐1.36) | 65/173 727 | 1.02 (0.70‐1.49) | 41/111 805 | 0.76 (0.57‐1.01) | 63/98 876 | 1.29 (0.93‐1.80) |
| SSc, myositis, and vasculitis only | 180/436 480 | 1.23 (1.14‐1.34) | 81/195 860 | 0.98 (0.80‐1.20) | 31/129 039 | 1.49 (0.83‐2.50) | 68/111 581 | 1.57 (1.52‐1.62) |
Abbreviation: AI, autoimmune; CI, confidence interval; OR, odds ratio; RA, rheumatoid arthritis; SLE, systemic lupus erythematosus; SSc, systemic sclerosis.
All models are generalized estimating equations accounting for military branch, with and without stratification by years of service.
Full sample models control for age at first Veterans Affairs encounter, sex, race/ethnicity (seven categories), smoking status, and years of service.
Stratified models control for age at first Veterans Affairs encounter, race/ethnicity (white versus all others), sex, and smoking status.
Other AI diseases include SLE, SSc, myositis, and vasculitis.
Models of RA omit patients with other autoimmune diseases.
Models of other AI disease omit patients with RA.
Models of SLE omit patients with other AI diseases.
DISCUSSION
In this large sample of veterans of the Afghanistan and Iraqi wars, we observed 10% elevated odds of RA associated with military service more likely to involve exposure to inorganic dust. The odds were higher for seronegative RA (a 25% elevation), whereas there was no statistical increase observed for seropositive RA. For those with moderate length of service (4‐8 years; roughly one‐third of those studied), the excess odds of RA increased to 40%, although this did not carry through to those with longer service. In contrast, there was no such relationship to SLE, the second most common AI disease we studied. For SSc, myositis, and vasculitis, however, likely exposure to inorganic dust was associated with 23% increased odds, and service for more than 8 years was associated with 57% increased odds.
The specific question of exposure to inorganic dusts through military service in relation to prospective risk of RA and other AI conditions has not been addressed previously, although several studies have considered these conditions in relation to military service. For example, an analysis of Millennium Cohort data investigated factors associated with self‐reported RA and SLE, including broad military job categories, without identifying any consistent association (22). Occupational duties and dust were not included in that analysis. Another Millennium Cohort Study analyzed RA and SLE in relation to burn pits, which might be relevant to particulate inhalation (23). In that study, there was no significant relationship to exposure, but there were also few self‐reported cases confirmed by medical records. A more recent analysis using the VA Rheumatoid Arthritis Registry also investigated self‐reported burn pit exposure and waste disposal, observing an increased risk in relation to anti‐CCP positivity among persons with disease (5). Dust exposure in military occupations also was not considered in that analysis. Investigations exploiting the OEF/OIF/OND roster have provided insights into chronic conditions (8, 9, 10, 11). Of note, one analysis using that cohort did study AI disease risk in association with PTSD, although military occupation was not considered (11). A more recent Millennium Cohort Study also has observed an association between PTSD and AI disease (6). There is no basis, however, to posit a role of dust exposure as an unmeasured confounder in such a relationship.
None of these prior studies of rheumatologic disease outcomes employed a JEM to assign risk linked to military duties. Even in studies of other health conditions linked to military service, this approach is used rarely. An early study of brain tumor risk among Air Force personnel marks one notable exception (24). The general concept of exploiting MOS codes as a surrogate for exposures was explored by Gadermann et al in 2014 (25). More recently, a 2019 investigation went farther by linking specific MOS codes to a set of specific inhalational exposures, essentially a JEM approach (26). Ours is the first application of an MOS‐linked JEM in the study of RA and other AI conditions.
A number of challenges beyond exposure assessment limit the study of health outcomes in military cohorts generally. Epidemiological investigations that compare those with military service records with others typically face a “healthy soldier effect,” in which all‐cause and specific mortality rates are reduced for a number chronic diseases, but this effect appears to erode over time among veterans from multiple armed forces (27, 28, 29). Follow‐up in VAHCS treatment is able to take advantage of this erosion to reveal elevated risk (28).
Shared immunological citrullinated targets have been described in the lungs and synovial tissue of patients with RA, which would imply that dust might be more likely to be associated with anti‐CCP seropositive RA (30). Consistent with this, a population‐based case‐referent investigation of RA in Sweden (the Epidemiological Investigation of Rheumatoid Arthritis [EIRA] study) observed a strong association between RA and self‐reported occupational exposure to silica, restricted to those with seropositive disease (RF in one analysis and anti‐CCP in another), whereas silica was not a risk factor for seronegative RA (31, 32). A preferential relationship of occupational exposure to seropositive disease, however, has not been consistently reported. Further follow‐up from EIRA, for example, found that, among male bricklayers and concrete workers and among electronics workers, risk was similarly elevated for seropositive and seronegative disease (ORs = seropositive, 2.9 [brick/concrete], 2.4 [electronics]; seronegative, 2.2 [brick/concrete], 2.6 [electronics]) (33). Another analysis, combining EIRA with additional Swedish national registry cases, estimated a silica exposure OR in men of 1.4 for seropositive disease versus 1.3 for seronegative disease (34). In a separate cohort study, also from Sweden, the estimated risks of seropositive and seronegative disease associated with silica and other inorganic dust were all similar (relative risks ranging from 1.27 to 1.46) (19). The number of persons with unknown serologic status in this analysis is a limitation that may arise from VA patients who have received specialty care in the community with testing that was not performed at a VA laboratory and whose results thus were not warehoused in the VINCI system. To the extent that such patients, when they report negative testing, are more likely to be retested at the VA, this could lead to systematic bias in missingness of data, yielding a case mix with a higher proportion of seronegative persons than would otherwise be observed.
Other potential limitations of our study should be kept in view. We could only study OEF/OIF/OND veterans who chose to receive their care through the VA system. Although our identification of cases used an approach validated in electronic medical record–based research, we did not carry out sample chart review and data extraction. The anticipated associations we observed between RA and its known risk factors, notably cigarette smoking and sex, support the validity of our case identification methods. Our JEM approach does not employ supporting industrial hygiene data and is likely prone to misclassification of exposure, although there is no reason to presume that this is systematic in any direction. Random exposure misclassification should bias toward the null and would not account for the associations we did observe. Moreover, as opposed to self‐reported exposure, which dominates military occupational epidemiology, the JEM approach is free of recall bias, an advantage underscored in a 2020 review of exposure assessment in the epidemiologic study of chronic disease in military populations (35). The JEM that we applied considered likely inorganic dust exposure but did not address other inhalants, such as work‐related organic dusts. Animal and textile dusts, for example, were associated with RA risk (seropositive and seronegative) in a recent study (36). Endotoxin (a key component of organic dusts) has been shown to potentiate silica‐induced lung injury in an experimental model (37). We did not assess combined inorganic and organic dust exposure in terms of RA risk. Our failure to observe a monotonic association between RA and duration of military service could reflect a “healthy worker survivor effect,” which is well‐recognized in occupational studies (38). In contradistinction, the exposure‐associated risk of SSc, myositis, and vasculitis was strongest among those with the longest service. Because the roster does not contain health condition information, we are precluded from taking earlier disease or symptom onset into account. Only 12% of our cohort was female, and, moreover, work exposures even within JEM categories likely differed systematically by sex. This complicates the interpretation of AI diseases for which sex is a powerful risk factor and may account for the statistically significant protective effect of exposure for SLE among women (but not men).
In conclusion, our study is notable for being the first large‐scale investigation exploring dusty exposures in past military service as a risk factor for the future diagnosis of RA and other AI conditions. Although the excess odds we identified were modest, the large numbers of persons exposed suggests that the absolute number of theoretically attributable and potentially preventable cases of disease, especially RA, is not trivial.
AUTHOR CONTRIBUTIONS
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Blanc had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study conception and design
Schmajuk, Trupin, Blanc.
Acquisition of data
Ying, Trupin, Blanc.
Analysis and interpretation of data
Ying, Schmajuk, Trupin, Blanc.
Supporting information
Table S1‐S2
Supported by the US Department of Veterans Affairs (Merit: CX001548‐01A2 to Dr. Schmajuk, Ms. Trupin, and Dr. Blanc), the National Institutes of Health (T32 5T32AR007304 to Dr. Ying); and the Russell Engleman Rheumatology Research Center (to Dr. Schmajuk).
No potential conflicts of interest relevant to this article were reported.
REFERENCES
- 1. Schmajuk G, Trupin L, Yelin E, Blanc PD. Prevalence of arthritis and rheumatoid arthritis in coal mining counties of the United States. Arthritis Care Res 2019;71:1209–15. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Ilar A, Klareskog L, Saevarsdottir S, Wiebert P, Askling J, Gustavsson P, et al. Occupational exposure to asbestos and silica and risk of developing rheumatoid arthritis: findings from a Swedish population‐based case‐control study. RMD Open 2019;5:e000978. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Prisco LC, Martin LW, Sparks JA. Inhalants other than personal cigarette smoking and risk for developing rheumatoid arthritis. Curr Opin Rheumatol 2020;32:279–88. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Blanco Pérez JJ, Arnalich Montiel V, Salgado‐Barreira Á, Alvarez Moure MA, Caldera Díaz AC, Melero Gonzalez R, et al. Prevalence and clinical impact of systemic autoimmune rheumatic disease in patients with silicosis. Arch Bronconeumol 2020. E‐pub ahead of print. [DOI] [PubMed] [Google Scholar]
- 5. Ebel AV, Lutt G, Poole JA, Thiele GM, Baker JF, Cannon GW, et al. Association of agricultural, occupational, and military inhalants with autoantibodies and disease features in US veterans with rheumatoid arthritis. Arthritis Rheumatol 2021;73:392–400. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Bookwalter DB, Roenfeldt KA, LeardMann CA, Kong SY, Riddle MS, Rull RP. Posttraumatic stress disorder and risk of selected autoimmune diseases among US military personnel. BMC Psychiatry 2020;20:23. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Garshick E, Abraham JH, Baird CP, Ciminera P, Downey GP, Falvo MJ, et al. Respiratory health after military service in Southwest Asia and Afghanistan: an official American Thoracic Society workshop report. Ann Am Thorac Soc 2019;16:e1–16. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Seal KH, Bertenthal D, Miner CR, Sen S, Marmar C. Bringing the war back home: mental health disorders among 103,788 US veterans returning from Iraq and Afghanistan seen at Department of Veterans Affairs facilities. Arch Intern Med 2007;167:476–82. [DOI] [PubMed] [Google Scholar]
- 9. Koo KH, Hebenstreit CL, Madden E, Seal KH, Maguen S. Race/ethnicity and gender differences in mental health diagnoses among Iraq and Afghanistan veterans. Psychiatry Res 2015;229:724–31. [DOI] [PubMed] [Google Scholar]
- 10. Maguen S, Madden E, Cohen BE, Bertenthal D, Neylan TC, Seal KH. Suicide risk in Iraq and Afghanistan veterans with mental health problems in VA care. J Psychiatr Res 2015;68:120–4. [DOI] [PubMed] [Google Scholar]
- 11. O'Donovan A, Cohen BE, Seal KH, Bertenthal D, Margaretten M, Nishimi K, et al. Elevated risk for autoimmune disorders in Iraq and Afghanistan veterans with posttraumatic stress disorder. Biol Psychiatry 2015;77:365–74. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Ng B, Aslam F, Petersen NJ, Yu HJ, Suarez‐Almazor ME. Identification of rheumatoid arthritis patients using an administrative database: a Veterans Affairs study. Arthritis Care Res 2012;64:1490–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Schmajuk G, Yazdany J. Underestimation of the reliability of codes for rheumatoid arthritis within administrative data: comment on the article by Ng et al. Arthritis Care Res 2013;65:835–6. [DOI] [PubMed] [Google Scholar]
- 14. Widdifield J, Bombardier C, Bernatsky S, Paterson JM, Green D, Young J, et al. An administrative data validation study of the accuracy of algorithms for identifying rheumatoid arthritis: the influence of the reference standard on algorithm performance. BMC Musculoskelet Disord 2014;15:216. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Moores KG, Sathe NA. A systematic review of validated methods for identifying systemic lupus erythematosus (SLE) using administrative or claims data. Vaccine 2013;31 Suppl 10:K62–73. [DOI] [PubMed] [Google Scholar]
- 16. Ying D, Gianfrancesco MA, Trupin L, Yazdany J, Greidinger EL, Schmajuk G. Increased risk of ischemic stroke in systemic sclerosis: a national cohort study of US veterans. J Rheumatol 2020;47:82–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Schmajuk G, Li J, Evans M, Anastasiou C, Izadi Z, Kay JL, et al. RISE registry reveals potential gaps in medication safety for new users of biologics and targeted synthetic DMARDs. Semin Arthritis Rheum 2020;50:1542–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Quinlan PJ, Earnest G, Eisner MD, Yelin EH, Katz PP, Balmes JR, et al. Performance of self‐reported occupational exposure compared to a job‐exposure matrix approach in asthma and chronic rhinitis. Occup Environ Med 2009;66:154–60. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Blanc PD, Järvholm B, Torén K. Prospective risk of rheumatologic disease associated with occupational exposure in a cohort of male construction workers. Am J Med 2015;128:1094–101. [DOI] [PubMed] [Google Scholar]
- 20. Calhoun PS, Wilson SM, Hertzberg JS, Kirby AC, McDonald SD, Dennis PA, et al. Validation of Veterans Affairs electronic medical record smoking data among Iraq‐ and Afghanistan‐era veterans. J Gen Intern Med 2017;32:1228–34. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Ying D, Gianfrancesco MA, Trupin L, Yazdany J, Greidinger EL, Schmajuk G. Increased risk of ischemic stroke in systemic sclerosis: a national cohort study of US veterans. J Rheumatol 2020;47:82–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Jones KA, Granado NS, Smith B, Slymen DJ, Ryan MA, Boyko EJ, et al. A prospective study of lupus and rheumatoid arthritis in relation to deployment in support of Iraq and Afghanistan: the Millennium Cohort study. Autoimmune Dis 2011;2011:741267. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Jones KA, Smith B, Granado NS, Boyko EJ, Gackstetter GD, Ryan MA, et al. Newly reported lupus and rheumatoid arthritis in relation to deployment within proximity to a documented open‐air burn pit in Iraq. J Occup Environ Med 2012;54:698–707. [DOI] [PubMed] [Google Scholar]
- 24. Grayson JK. Radiation exposure, socioeconomic status, and brain tumor risk in the US Air Force: a nested case‐control study. Am J Epidem 1996;143:480–6. [DOI] [PubMed] [Google Scholar]
- 25. Gadermann AM, Heeringa SG, Stein MB, Ursano RJ, Colpe LJ, Fullerton CS, Classifying US, et al. Army military occupational specialties using the Occupational Information Network. Mil Med 2014;179:752–61. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Zell‐Baran LM, Meehan R, Wolff J, Strand M, Krefft SD, Gottschall EB, et al. Military occupational specialty codes: utility in predicting inhalation exposures in post‐9/11 deployers. J Occup Environ Med 2019;61:1036–40. [DOI] [PubMed] [Google Scholar]
- 27. Strand LA, Martinsen JI, Fadum EA, Borud EK. Temporal trends in the healthy soldier effect in a cohort of Royal Norwegian Navy servicemen followed for 67 years. Occup Environ Med 2020;77:775–81. [DOI] [PubMed] [Google Scholar]
- 28. Bollinger MJ, Schmidt S, Pugh JA, Parsons HM, Copeland LA, Pugh MJ. Erosion of the healthy soldier effect in veterans of US military service in Iraq and Afghanistan. Popul Health Metr 2015;13:8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Cox B, McBride D, Broughton J, Tong D. Health conditions in a cohort of New Zealand Vietnam veterans: hospital admissions between 1988 and 2009. BMJ Open 2015;5:e008409. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Chatzidionisyou A, Catrina AI. The lung in rheumatoid arthritis, cause or consequence? Curr Opin Rheumatol 2016;28:76–82. [DOI] [PubMed] [Google Scholar]
- 31. Stolt P, Källberg H, Lundberg I, Sjögren B, Klareskog L, Alfredsson L, et al. Silica exposure is associated with increased risk of developing rheumatoid arthritis: results from the Swedish EIRA study. Ann Rheum Dis 2005;64:582–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Stolt P, Yahya A, Bengtsson C, Källberg H, Rönnelid J, Lundberg I, et al. Silica exposure among male current smokers is associated with a high risk of developing ACPA‐positive rheumatoid arthritis. Ann Rheum Dis 2010;69:1072–6. [DOI] [PubMed] [Google Scholar]
- 33. Ilar A, Alfredsson L, Wiebert P, Klareskog L, Bengtsson C. Occupation and risk of developing rheumatoid arthritis; results from a population‐based case‐control study. Arthritis Care Res 2018;70:499–509. [DOI] [PubMed] [Google Scholar]
- 34. Ilar A, Klareskog L, Saevarsdottir S, Wiebert P, Askling J, Gustavsson P, et al. Occupational exposure to asbestos and silica and risk of developing rheumatoid arthritis: findings from a Swedish‐based case‐control study. RMD Open 2019;5:e000978. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Hall AL, MacLean MB, VanTil L, McBride DI, Glass DC. Considering exposure assessment in epidemiological studies of chronic health in military populations. Front Public Health 2020;8:577601. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Ilar A, Gustavsson P, Wiebert P, Alfredsson L. Occupational exposure to organic dusts and risk of developing rheumatoid arthritis: findings from a Swedish‐based case‐control study. RMD Open 2019;5:e001049. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. Ko J‐W, Lee H‐J, Shin N‐R, Seo Y‐S, Kim S‐H, Shin I‐S, et al. Silicon dioxide nanoparticles enhance endotoxin‐induced lung injury in mice. Molecules 2018;23:2247. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. Brown DM, Picciotto S, Costello S, Neophytou AM, Izano MA, Ferguson JM, et al. The healthy worker survivor effect: target parameters and target populations. Curr Environ Health Rep 2017;4:364–72. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Table S1‐S2