Associations of Fire Smoke and Other Pollutants With Incident Rheumatoid Arthritis and Rheumatoid Arthritis–Associated Interstitial Lung Disease

Abstract

Objective

The aim of this study was to determine whether pollutants such as fire smoke–related particulate matter <2.5 μm (PM_2.5) are associated with incident rheumatoid arthritis (RA) and RA‐associated interstitial lung disease (RA‐ILD).

Methods

This patient–control study used Veterans Affairs (VA) data from October 1, 2009, to December 31, 2018. We identified patients with incident RA and RA‐ILD using validated algorithms, matching each patient to ≤10 controls on age, sex, and VA enrollment year. We obtained pollutants including fire smoke PM_2.5, carbon monoxide, nitrogen oxides (NOx), ozone, overall PM_2.5, PM₁₀, and sulfur dioxide (SO₂) at least one year before the index date. We fit conditional logistic regression models to estimate adjusted odds ratios (aORs) with 95% confidence intervals (CIs) for incident RA and RA‐ILD, adjusted for confounders.

Results

We identified 9,701 patients with incident RA (mean age 65 years, 86% male), including 531 patients with RA‐ILD (mean age 69 years, 91% male), and 68,852 matched controls. Fire smoke PM_2.5 was not associated with RA (aOR 1.07, 95% CI 0.92–1.23) but was associated with RA‐ILD (aOR 1.98, 95% CI 1.08–3.62, per 1 μg/m³). Increased levels of NOx were associated with RA (aOR 1.16, 95% CI 1.06–1.27, highest vs lowest quartile). The highest quartiles of ozone (aOR 1.19, 95% CI 1.06–1.34) and PM₁₀ (aOR 1.25, 95% CI 1.10–1.43) were associated with seronegative RA. Carbon monoxide, overall PM_2.5, and SO₂ were not, or negatively, associated with RA and RA‐ILD.

Conclusion

Increased fire smoke PM_2.5 was associated with RA‐ILD, whereas NOx, ozone, and PM₁₀ were associated with RA risk. Thus, air pollution may increase the risk of RA and RA‐ILD.

INTRODUCTION

Wildfires are reaching record and alarming levels in the United States. ¹ , ² Such changes could have large impacts on human health because increased levels of particulate matter <2.5 μm (PM_2.5), the main pollutant in fire smoke, are associated with the onset of diabetes, ³ multiple sclerosis, ⁴ and Parkinson disease ⁵ and all‐cause death. ⁶ Importantly, several respiratory irritants have been found to be associated with increased risk of rheumatoid arthritis (RA), including cigarette smoking, ⁷ occupational inhalant exposures, ⁸ , ⁹ and respiratory tract diseases, ¹⁰ , ¹¹ especially in the 5 to 10 years before RA onset. ¹² Thus, fire smoke may also be a risk factor for RA. Indeed, PM_2.5 appears to be associated with anti–citrullinated protein antibody (ACPA) production, a hallmark of RA. ¹³ , ¹⁴ , ¹⁵ Some studies of overall PM_2.5 have shown positive associations with RA ¹⁶ , ¹⁷ , ¹⁸ and others none. ¹⁹ , ²⁰ , ²¹ However, none studied fire smoke exposure specifically.

Another important and unresolved question is whether other pollutants like carbon monoxide (CO), nitrogen oxides (NOx), ozone, PM₁₀, and sulfur dioxide (SO₂) increase the risk of RA. For example, one study showed that NO₂ and SO₂ increase the risk of RA, ²² but others have shown no association of NOx ¹⁹ , ²⁰ , ²³ or SO₂. ¹⁹ , ²³ Although some studies have suggested CO, ⁹ , ²³ , ²⁴ ozone, ²⁰ , ²³ , ²⁴ and PM₁₀ might increase RA risk, these reports studied exposures less than three years before RA onset and/or did not adjust for key confounders like smoking. ¹⁷ Finally, although some studies suggest pollutants might increase the risk of interstitial lung abnormalities ²⁵ and idiopathic pulmonary fibrosis, ²⁶ none have studied the association between pollutants and RA‐associated interstitial lung disease (RA‐ILD).

To address these gaps, we leveraged national Veterans Affairs (VA) data, the largest integrated health care system in the United States, ²⁷ and nationwide pollutant monitoring data from the Environmental Protection Agency (EPA). ²⁸ We aimed to determine the associations between fire smoke and other pollutant exposures with the risk of RA and RA‐ILD and to identify the timing of any such associations. We hypothesized that increasing levels of fire smoke and other pollutant exposures would be associated with increased risk of both RA and RA‐ILD, especially five or more years before RA onset.

PATIENTS AND METHODS

Study population and design. This patient–control study used national VA data, which contain detailed administrative and electronic health data on US veterans enrolled in VA health care. ²⁷

The index date for this study was the time of RA diagnosis (or matched date), as indicated by the date fulfilling all components of the RA algorithm. This study received institutional review board approval (00012917; 1619487).

Incident RA and RA‐ILD

The primary outcomes for this study were incident RA and RA‐ILD. We identified patients with incident RA using a validated algorithm requiring at least two RA codes (International Classification of Diseases, Ninth Revision [ICD‐9] 714.0, 714.1, 714.2, and 714.81; and ICD‐10 M05.x, M06.x, excluding M06.1, and M06.4), a rheumatologist diagnosis of RA, and either a prescription of a disease‐modifying antirheumatic drug (DMARD) or positive rheumatoid factor (RF) or ACPA. ²⁹ To ensure incident diagnoses, we also required a one‐year period after VA enrollment with no RA diagnostic codes or DMARD fills. This algorithm has >90% positive predictive value (PPV) for RA by the American College of Rheumatology criteria ³⁰ and 85% PPV for classifying incident versus prevalent RA. ³¹ Patients with RA were categorized as being seropositive or seronegative based on RF and/or ACPA.

Among these patients with RA, we defined RA‐ILD as the presence of two RA‐ILD codes at least 30 days apart (ICD‐9 515.x, 516.3, 516.8, and 516.9; and ICD‐10 J84.1, J84.89, and J84.9) along with either a pulmonologist diagnosis of ILD or procedure codes for a chest computed tomography scan plus either pulmonary function tests or lung biopsy. RA‐ILD assessment included data after RA diagnosis. This algorithm has a PPV for RA‐ILD of 81%. ³²

To ensure all participants could have at least five years of preceding pollutant exposure data, we a priori restricted the analysis to patients with incident RA and RA‐ILD with at least five years of preceding address data available. Because address data began October 1, 2009, we included patients with incident RA from October 1, 2014, to the time of extraction of data on patients with RA on December 31, 2018. We also required them to have available PM_2.5 data because they were the primary exposure of interest for this study. Availability of the other pollutant exposures was not required.

Controls

We matched patients with incident RA to up to ≤10 controls without RA based on year of birth, sex, and VA enrollment year. We did not match on smoking status to avoid introducing selection bias. ³³ We required that controls have none of the billing codes for RA mentioned in the Incident RA and RA‐ILD section. Like patients, we required they have at least five years preceding address data and available PM_2.5 data for pollutant exposure assessment.

Fire smoke PM_{2 .5}, other pollutants, and covariates

Our primary exposure was 24‐hour average fire smoke PM_2.5 exposure (in micrograms per meter cubed). This fire smoke exposure was a fused product of ground pollutant monitoring from the EPA Air Quality System ²⁸ and satellite‐based smoke plume imagery from the National Oceanic and Atmospheric Administration Hazard Mapping System. ³⁴ By overlaying these two data sources, this fused product produces daily grids of smoke‐related PM_2.5, nonsmoke PM_2.5, and overall total PM_2.5 over the contiguous United States at a 15 × 15 km resolution, starting in 2006. ³⁴ We interpolated these grids with address data from the National Change of Address Database maintained by the US Postal Service, which are updated quarterly.

We obtained secondary pollutants of interest including 1‐hour maximum (max) CO, 1‐hour max NOx, 8‐hour max ozone, 24‐hour average PM₁₀, and 1‐hour max SO₂ interpolated from the EPA ground pollutant monitoring above ²⁸ using a previously validated method. ³⁴ , ³⁵ Specifically, we assigned daily concentrations of each pollutant to Zip Code Tabulation Areas (ZCTAs) using the median of the measurements at all monitors within the ZCTAs and within 50 km of the ZCTA centroid. ²⁸ We chose 50 km to minimize missing data. For all pollutants, we categorized mean exposure into quartiles for ease of interpretation, along with deciles in a post hoc sensitivity analysis. Due to the narrow range of exposures including PM_2.5, we also reported results for continuous values.

To minimize the potential for reverse causation, in which people may alter air exposures between RA symptom onset and clinical diagnosis, we included pollutant exposure measurements occurring at least one year before index date. To identify the timing of any potential associations, we also divided this pollutant exposure into time windows (over one to three, over three to five, and over five years before the index date). We took the mean of all the daily pollutant reads that were available within each time window. We obtained the following covariates from the VA corporate data warehouse: age at index date, self‐reported sex (male vs female), VA enrollment year, duration of fire smoke pollutant exposure data before index date (continuous), race and ethnicity (Asian, Black, Hispanic White, White non‐Hispanic, or unknown), body mass index nearest preceding index date (<20, 20–24.9, 25–29.9, 30–34.9, 35–39.9, or ≥40 kg/m²), and smoking status at index date (current, former, or never).

Statistical analysis

We performed conditional logistic regression models for each pollutant exposure to obtain adjusted odds ratios (aORs) with 95% confidence intervals (CIs) for RA (with subgroup analysis by RA serostatus) and RA‐ILD, adjusting for the above covariates. We also performed stratified analyses by exposure time windows (more than one to three, more than three to five, and more than five years before index date). For sensitivity analyses, we stratified by geographic region (Midwest, Northeast, South, and West) and smoking status (never vs ever) using unconditional logistic regression because patients with RA and controls were no longer matched. Second, we performed multiplicative interactions between fire smoke PM_2.5 and age (>65 vs ≤65 years), sex, race and ethnicity, and smoking status (ever vs never) for risk of RA. Third, we performed a post hoc analysis categorizing PM_2.5 exposure data into deciles.

Throughout this study, any individuals missing exposure or covariate data were excluded from that model (ie, complete patient analysis). To examine the effects of missing covariates on our results, we performed a sensitivity analysis with multiple imputation for missing covariates using fully conditional specification with 10 imputed datasets. We used a significance threshold of two‐sided alpha of 0.05. We outlined all analyses in a prespecified protocol and performed analyses using Stata version 18 (StataCorp LLC).

RESULTS

Patient characteristics

We identified 30,146 patients with incident RA (Supplementary Figure 1). Of these, 9,701 patients with incident RA met study eligibility criteria (mean age 65 years, 86% male) and were matched to 68,851 controls. Within these, 531 also met criteria as patients with RA‐ILD (mean age 69 years, 91% male; Table 1). Median time between RA and RA‐ILD was 0.0 years (interquartile range −2.1 to 1.4). Of patients with RA with serological data, 63% were seropositive, whereas 80% of patients with RA‐ILD were seropositive.

Table 1.

Characteristics of patients with incident RA and matched controls^*

Characteristic	Patients with RA (n = 9,701), n (%)	Patients with RA‐ILD (n = 531), n (%)	Controls (n = 68,851), n (%)
Age at index, mean (±SD), y a	65 (±11)	69 (±9)	63 (±11)
Male a	8,292 (86)	484 (91)	57,043 (83)
VA enrollment period a
1997–2001	5,548 (57)	329 (62)	37,347 (54)
2002–2006	2,785 (29)	147 (28)	21,160 (31)
2007–2011	1,306 (14)	52 (10)	10,072 (15)
2012–2016	62 (0.6)	3 (0.6)	272 (0.4)
Time of PM2.5 data before index, mean (±SD), y	7.0 (±1.4)	6.8 (±1.4)	6.9 (±1.4)
Race and ethnicity
Asian/other	321 (3.3)	14 (3)	1930 (2.8)
Black	1,825 (19)	99 (19)	13,821 (20)
Hispanic White	335 (3.5)	25 (5)	1,865 (2.7)
White non‐Hispanic	6,735 (69)	374 (70)	41,008 (60)
Unknown/missing	485 (5)	19 (4)	10,227 (15)
BMI at index, kg/m2
<20	126 (1.3)	3 (0.6)	1,061 (1.5)
20 to <25	1,506 (16)	87 (16)	10,526 (15)
25 to <30	3,523 (36)	211 (40)	24,418 (36)
30 to <35	2,902 (30)	157 (30)	17,706 (26)
35 to <40	1,158 (12)	54 (10)	7,103 (10)
≥40	485 (5)	19 (4)	3,220 (4.7)
Missing	1 (<1)	–	4,817 (7)
Smoking status at index
Never	1,317 (14)	49 (9)	13,516 (20)
Former	2,858 (30)	143 (27)	16,169 (24)
Current	5,439 (56)	332 (63)	31,664 (46)
Unknown/missing	87 (0.9)	7 (1.3)	7,502 (11)
Geographic region
Midwest	1,975 (20)	104 (20)	13,747 (20)
Northeast	1,503 (16)	82 (15)	6,956 (10)
South	4,183 (43)	240 (45)	33,825 (49)
West	2,029 (21)	104 (20)	14,211 (21)
Unknown	10 (0.1)	1 (0.2)	104 (0.2)

^*BMI, body mass index; ILD, interstitial lung disease; PM_2.5, particulate matter <2.5 μm; RA, rheumatoid arthritis; VA, Veterans Affairs.

^a The indicated factors were matched.

Pollutants and incident RA

The mean duration of PM_2.5 exposure data available for this study was 7.0 years in patients with RA and 6.9 years in controls (Table 1), with a range of 4.3 to 9.0 years. Mean pollutant levels did not show large absolute differences between patients with RA and controls (Supplementary Table 1). Fire smoke PM_2.5 was not associated with RA overall (aOR 1.07, 95% CI 0.92–1.23, per 1 μg/m³; Table 2). Post hoc analyses showed that the second and tenth deciles of exposure were associated with higher RA risk (Supplementary Table 2). Increased NOx levels were associated with an increased risk of RA (aOR 1.16, 95% CI 1.06–1.27 for highest vs lowest quartile; Table 2). In addition, the highest quartiles of ozone (aOR 1.19, 95% CI 1.06–1.34) and PM₁₀ (aOR 1.25, 95% CI 1.10–1.43) exposure were associated with patients with RA who were seronegative, but not seropositive, when compared to the lowest quartile. In contrast, CO (aOR 0.78, 95% CI 0.67–0.91, per 1 part per million), overall PM_2.5 (aOR 0.94, 95% CI 0.93–0.96, per 10 μg/m³), and SO₂ (aOR 0.98, 95% CI 0.97–0.98, per 1 part per billion) were negatively associated with RA risk.

Table 2.

Association between preceding fire smoke and other pollutant exposures and incident RA^*

Pollutant	Overall RA (n = 9,701), continuous exposure, adjusted OR (95% CI) a , b	Overall RA (n = 9,701), quartiled exposure, adjusted OR (95% CI) a	Seropositive RA (n = 5,588), quartiled exposure, adjusted OR (95% CI) a	Seronegative RA (n = 3,347), quartiled exposure, adjusted OR (95% CI) a
Fire smoke PM2.5, μg/m3	1.07 (0.92–1.23)	1.00 (ref)	1.00 (ref)	1.00 (ref)
Quartile 2	–	0.95 (0.89–1.02)	0.99 (0.90–1.08)	0.90 (0.80–1.00)
Quartile 3	–	0.91 (0.85–0.97)	0.98 (0.90–1.07)	0.82 (0.73–0.92)
Quartile 4	–	1.01 (0.95–1.08)	1.08 (0.99–1.18)	0.93 (0.83–1.04)
CO, ppm	0.78 (0.67–0.91)	1.00 (ref)	1.00 (ref)	1.00 (ref)
Quartile 2	–	0.92 (0.85–0.99)	0.91 (0.81–1.01)	0.88 (0.77–1.02)
Quartile 3	–	0.84 (0.77–0.91)	0.90 (0.80–0.99)	0.77 (0.66–0.89)
Quartile 4	–	0.81 (0.74–0.88)	0.78 (0.70–0.88)	0.85 (0.73–0.98)
NOx, ppb	1.02 (1.00–1.03)	1.00 (ref)	1.00 (ref)	1.00 (ref)
Quartile 2	–	1.18 (1.08–1.29)	1.17 (1.04–1.31)	1.17 (1.01–1.36)
Quartile 3	–	1.08 (0.99–1.18)	1.14 (1.01–1.28)	0.98 (0.84–1.14)
Quartile 4	–	1.16 (1.06–1.27)	1.16 (1.03–1.31)	1.15 (0.99–1.34)
O3, ppm	0.97 (0.91–1.03)	1.00 (ref)	1.00 (ref)	1.00 (ref)
Quartile 2	–	0.99 (0.93–1.06)	0.97 (0.89–1.06)	1.05 (0.94–1.18)
Quartile 3	–	0.89 (0.83–0.95)	0.85 (0.78–0.93)	0.92 (0.82–1.04)
Quartile 4	–	0.97 (0.91–1.04)	0.90 (0.83–0.99)	1.19 (1.06–1.34)
PM2.5, μg/m3	0.94 (0.93–0.96)	1.00 (ref)	1.00 (ref)	1.00 (ref)
Quartile 2	–	0.90 (0.85–0.96)	0.94 (0.86–1.02)	0.83 (0.75–0.92)
Quartile 3	–	0.81 (0.76–0.86)	0.81 (0.75–0.89)	0.77 (0.69–0.86)
Quartile 4	–	0.72 (0.68–0.77)	0.76 (0.70–0.83)	0.65 (0.58–0.73)
PM10, μg/m3	1.03 (0.98–1.08)	1.00 (ref)	1.00 (ref)	1.00 (ref)
Quartile 2	–	0.83 (0.77–0.89)	0.78 (0.71–0.87)	0.99 (0.87–1.14)
Quartile 3	–	0.88 (0.81–0.95)	0.83 (0.75–0.91)	1.05 (0.92–1.19)
Quartile 4	–	1.02 (0.94–1.10)	0.96 (0.87–1.06)	1.25 (1.10–1.43)
SO2, ppb	0.98 (0.97–0.98)	1.00 (ref)	1.00 (ref)	1.00 (ref)
Quartile 2	–	0.96 (0.89–1.04)	0.99 (0.90–1.10)	0.95 (0.83–1.08)
Quartile 3	–	1.01 (0.94–1.10)	1.02 (0.92–1.13)	1.01 (0.89–1.16)
Quartile 4	–	0.78 (0.72–0.85)	0.80 (0.71–0.89)	0.77 (0.67–0.88)

^*Bold values indicate P < 0.05. CI, confidence interval; CO, carbon monoxide; NOx, nitrogen oxides; O₃, ozone; OR, odds ratio; PM_2.5, particulate matter <2.5 μm; ppb, parts per billion; ppm, parts per million; RA, rheumatoid arthritis; ref, reference; SO₂, sulfur dioxide.

^a The ref group was individuals with the lowest quartile of pollutant exposure. The patients were matched with controls on age, sex, and enrollment while adjusting for duration of pollutant exposure data, race and ethnicity, body mass index, and smoking status.

^b Values are per 1 unit exposure for all except O₃ (0.01 units) and NOx and PM₁₀ (10 units).

Pollutants and RA‐ILD

Higher fire smoke PM_2.5 was significantly associated with RA‐ILD (aOR 1.98, 95% CI 1.08–3.62, per 1 μg/m³; Table 3). Post hoc analyses showed that fire smoke PM_2.5 concentrations of 0.28 μg/m³ and higher seemed to drive this elevated risk (Supplementary Table 2). Except for the highest quartile of CO being negatively associated, no other pollutants were associated with RA‐ILD.

Table 3.

Association between preceding fire smoke and other pollutant exposures and RA‐ILD^*

Pollutant	Patients with RA‐ILD (n = 531), mean concentration (±SD)	Controls (n = 3,605), mean concentration (±SD)	Continuous exposure in patients with RA‐ILD, adjusted OR (95% CI) a , b	Quartiled exposure in patients with RA‐ILD, adjusted OR (95% CI) a
Fire smoke PM2.5, μg/m3	0.27 (±0.15)	0.26 (±0.14)	1.98 (1.08–3.62)	1.00 (ref)
Quartile 2	–	–	–	1.16 (0.87–1.55)
Quartile 3	–	–	–	1.06 (0.79–1.43)
Quartile 4	–	–	–	1.45 (1.09–1.93)
CO, ppm	0.57 (±0.22)	0.59 (±0.22)	0.54 (0.27–1.09)	1.00 (ref)
Quartile 2	–	–	–	1.05 (0.72–1.52)
Quartile 3	–	–	–	0.94 (0.63–1.39)
Quartile 4	–	–	–	0.59 (0.39–0.89)
NOx, ppb	43.2 (±22.6)	43.2 (±24.7)	0.99 (0.93–1.05)	1.00 (ref)
Quartile 2	–	–	–	1.20 (0.92–1.77)
Quartile 3	–	–	–	0.97 (0.64–1.47)
Quartile 4	–	–	–	1.03 (0.70–1.52)
O3, ppm	0.04 (±0.00)	0.04 (±0.00)	0.90 (0.67–1.19)	1.00 (ref)
Quartile 2	–	–	–	0.95 (0.71–1.27)
Quartile 3	–	–	–	0.89 (0.66–1.22)
Quartile 4	–	–	–	0.89 (0.65–1.21)
PM2.5, μg/m3	8.96 (±1.62)	8.99 (±1.58)	0.99 (0.93–1.06)	1.00 (ref)
Quartile 2	–	–	–	1.12 (0.85–1.47)
Quartile 3	–	–	–	0.97 (0.72–1.30)
Quartile 4	–	–	–	0.94 (0.70–1.25)
PM10, μg/m3	18.7 (±5.0)	18.8 (±5.3)	1.04 (0.92–1.32)	1.00 (ref)
Quartile 2	–	–	–	0.78 (0.56–1.10)
Quartile 3	–	–	–	0.94 (0.67–1.32)
Quartile 4	–	–	–	1.07 (0.77–1.51)
SO2, ppb	3.48 (±3.52)	3.93 (±4.39)	0.97 (0.93–1.00)	1.00 (ref)
Quartile 2	–	–	–	1.11 (0.77–1.60)
Quartile 3	–	–	–	1.29 (0.90–1.84)
Quartile 4	–	–	–	0.86 (0.58–1.26)

^*Bold values indicate P < 0.05. CI, confidence interval; CO, carbon monoxide; ILD, interstitial lung disease; NOx, nitrogen oxides; O₃, ozone; OR, odds ratio; PM_2.5, particulate matter <2.5 μm; ppb, parts per billion; ppm, parts per million; RA, rheumatoid arthritis; ref, reference; SO₂, sulfur dioxide.

^a The ref group was individuals with the lowest quartile of pollutant exposure. The patients were matched with controls on age, sex, and enrollment while adjusting for duration of pollutant exposure data, race and ethnicity, body mass index, and smoking status.

^b Values are per 1 unit exposure for all except O₃ (0.01 units) and NOx and PM₁₀ (10 units).

Pollutant timing and RA

When stratifying by pollutant timing, we found fire smoke PM_2.5 exposure to be most associated with RA in the one to three years (aOR 1.12, 95% CI 1.03–1.23, per 1 μg/m³) and three to five years (aOR 1.13, 95% CI 1.02–1.26, per 1 μg/m³) before RA index date (Figure 1; Table 4). Higher smoke PM_2.5 exposure more than five years before RA diagnosis was negatively associated with RA risk. In contrast, NOx showed the strongest point estimates in the three to five years and more than five years before RA onset. Finally, PM₁₀ exposure was most strongly associated with RA in the three to five years before RA onset (aOR 1.08, 95% CI 1.03–1.14 for highest vs lowest quartile).

Figure 1.

Forest plot depicting associations of fire smoke PM_2.5 with risk of RA by exposure time period. Models are per 1‐μg/m³ increase in PM_2.5 and adjusted for age, sex, enrollment, duration of pollutant exposure data, race and ethnicity, body mass index, and smoking status. PM_2.5, particulate matter <2.5 μm; RA, rheumatoid arthritis.

Table 4.

Association between the timing of preceding pollutant exposures and incident RA^*

Pollutant/years before RA	Continuous exposure for patients with RA, adjusted OR (95% CI) a , b			Quartiled exposure for patients with RA, adjusted OR (95% CI) a
	More than one to three years	More than three to five years	More than five years	More than one to three years	More than three to five years	More than five years
Fire smoke PM2.5, μg/m3	1.12 (1.03–1.23)	1.13 (1.02–1.26)	0.79 (0.67–0.93)	1.00 (ref)	1.00 (ref)	1.00 (ref)
Quartile 2	–	–	–	1.26 (1.18–1.35)	1.19 (1.12–1.28)	1.06 (0.98–1.15)
Quartile 3	–	–	–	1.21 (1.13–1.30)	1.11 (1.03–1.19)	1.06 (0.97–1.16)
Quartile 4	–	–	–	1.18 (1.10–1.27)	1.12 (1.04–1.20)	0.92 (0.84–1.00)
CO, ppm	0.72 (0.60–0.86)	0.73 (0.62–0.86)	0.80 (0.69–0.93)	1.00 (ref)	1.00 (ref)	1.00 (ref)
Quartile 2	–	–	–	0.99 (0.90–1.08)	0.96 (0.88–1.05)	1.00 (0.91–1.10)
Quartile 3	–	–	–	0.86 (0.78–0.94)	0.85 (0.78–0.94)	0.80 (0.72–0.88)
Quartile 4	–	–	–	0.81 (0.74–0.89)	0.83 (0.76–0.91)	0.82 (0.74–0.90)
NOx, ppb	1.01 (0.99–1.02)	1.02 (1.00–1.04)	1.02 (1.01–1.04)	1.00 (ref)	1.00 (ref)	1.00 (ref)
Quartile 2	–	–	–	1.13 (1.03–1.24)	1.17 (1.06–1.28)	1.31 (1.18–1.46)
Quartile 3	–	–	–	1.11 (1.01–1.21)	1.03 (0.93–1.14)	1.08 (0.97–1.20)
Quartile 4	–	–	–	1.16 (1.06–1.28)	1.17 (1.06–1.29)	1.27 (1.14–1.41)
O3, ppm	1.02 (0.96–1.08)	0.99 (0.94–1.05)	0.94 (0.89–1.00)	1.00 (ref)	1.00 (ref)	1.00 (ref)
Quartile 2	–	–	–	1.04 (0.97–1.12)	1.01 (0.95–1.09)	0.96 (0.89–1.03)
Quartile 3	–	–	–	0.98 (0.92–1.05)	0.96 (0.90–1.03)	0.93 (0.85–1.00)
Quartile 4	–	–	–	1.01 (0.94–1.09)	1.00 (0.93–1.08)	0.92 (0.85–1.00)
PM2.5, μg/m3	0.94 (0.92–0.95)	0.96 (0.94–0.97)	0.95 (0.94–0.97)	1.00 (ref)	1.00 (ref)	1.00 (ref)
Quartile 2	–	–	–	0.87 (0.81–0.92)	0.94 (0.89–1.00)	0.94 (0.88–1.00)
Quartile 3	–	–	–	0.75 (0.70–0.80)	0.81 (0.76–0.87)	0.84 (0.79–0.90)
Quartile 4	–	–	–	0.72 (0.67–0.77)	0.81 (0.75–0.86)	0.72 (0.67–0.78)
PM10, μg/m3	1.04 (0.98–1.09)	1.08 (1.03–1.14)	1.00 (0.94–1.06)	1.00 (ref)	1.00 (ref)	1.00 (ref)
Quartile 2	–	–	–	0.84 (0.78–0.92)	0.84 (0.77–0.91)	0.88 (0.81–0.95)
Quartile 3	–	–	–	0.98 (0.90–1.06)	0.97 (0.90–1.05)	0.78 (0.71–0.85)
Quartile 4	–	–	–	1.03 (0.95–1.12)	1.11 (1.03–1.21)	0.98 (0.90–1.07)
SO2, ppb	0.97 (0.96–0.98)	0.98 (0.97–0.98)	0.99 (0.98–0.99)	1.00 (ref)	1.00 (ref)	1.00 (ref)
Quartile 2	–	–	–	0.94 (0.87–1.02)	0.96 (0.88–1.05)	1.02 (0.93–1.11)
Quartile 3	–	–	–	1.01 (0.93–1.10)	1.11 (1.02–1.20)	1.00 (0.92–1.10)
Quartile 4	–	–	–	0.80 (0.73–0.87)	0.83 (0.76–0.90)	0.81 (0.73–0.89)

^*Bolded values are statistically significant. CI, confidence interval; CO, carbon monoxide; NOx, nitrogen oxides; O₃, ozone; OR, odds ratio; PM_2.5, particulate matter <2.5 μm; ppb, parts per billion; ppm, parts per million; RA, rheumatoid arthritis; ref, reference; SO₂, sulfur dioxide.

^a The ref group was individuals with the lowest quartile of pollutant exposure. The patients were matched with controls on age, sex, and enrollment while adjusting for duration of pollutant exposure data, race and ethnicity, body mass index, and smoking status.

^b Values are per 1 unit exposure for all except O₃ (0.01 units) and NOx and PM₁₀ (10 units).

Sensitivity analyses

Sensitivity analyses stratifying by geographic region showed NOx had the highest OR for RA in the West (OR 1.50, 95% CI 1.21–1.85 for highest vs lowest quartile; Supplementary Table 3). Restricting analyses to never‐smokers showed no meaningful differences in point estimates (Supplementary Table 3). We also found no evidence of interactions between fire smoke PM_2.5 and age, sex, race and ethnicity, and smoking status on risk of RA (P > 0.3 for each, data not shown). Regarding missing covariate data, 576 patients with RA (6%) and 13,716 controls (20%) were missing at least one covariate. However, results for each of the seven pollutants did not differ after performing multiple imputation for missing covariates (data not shown). Of note, 7,282 of study participants (9%) were missing all fire smoke PM_2.5 exposure data, and 25,984 (33%) were missing at least one of the other five pollutant exposures in this study. Participants missing all fire smoke PM_2.5 exposure data were more likely to be controls and live in the West (Supplementary Table 4).

DISCUSSION

This large patient–control study using national VA data and nationwide pollutant monitoring found that high fire smoke PM_2.5 exposure was associated with RA‐ILD, along with incident RA one to five years later. NOx, a pollutant generated largely from fossil fuel combustion, ³⁶ was also associated with incident RA, and ozone and PM₁₀ were associated with incident seronegative RA. Several other pollutants including CO, overall PM_2.5, and SO₂ were not, or negatively, associated with incident RA and RA‐ILD. Overall, these findings add to the growing body of literature on inhalant exposures and RA risk and suggest that reducing fire smoke and fossil fuel emissions may have benefits for RA and RA‐ILD prevention. They also highlight the need for more robust infrastructure for studying the impact of pollutants on health in the United States.

The first key finding from this study was that high fire smoke PM_2.5 exposure was associated with incident RA‐ILD. No prior studies have examined fire smoke PM_2.5 and RA or RA‐ILD risk, though others have found an association between overall PM_2.5 and RA. ¹⁶ , ¹⁷ , ¹⁸ Our study did not show a significant association between fire smoke PM_2.5 and overall RA except in stratified analyses in the one to five years before RA onset. The association between PM_2.5 and RA both with and without ILD has strong biologic rationale because both PM_2.5 ¹³ , ¹⁴ , ¹⁵ and wood‐fire smoke ³⁷ have previously been associated with ACPA development to a higher degree than cigarette smoke. Furthermore, elemental carbon from PM_2.5 is associated with increased risk of interstitial lung abnormalities. ²⁵ Importantly, the observed association between fire smoke PM_2.5 and RA‐ILD occurred even despite the study period not including recent record‐breaking wildfire years. ³⁸ , ³⁹ Therefore, future studies are needed to determine whether higher levels of fire smoke PM_2.5 exposure have an even stronger effect.

The second key finding from this study was that high levels of other pollutants such as NOx were associated with incident RA overall, whereas ozone and PM₁₀ were associated with incident seronegative RA. Prior literature on the association between NOx and RA has mostly shown a positive association. ¹⁸ , ²¹ , ²² , ²⁴ Studies showing no association with RA had lower NOx levels than ours and these other studies, ¹⁹ , ²⁰ , ²³ perhaps explaining this discrepancy. Indeed, even the third quartile of NOx exposure was not associated with RA in our study. The three prior studies of ozone and RA all showed a positive association as well. ²⁰ , ²³ , ²⁴ Only one study has shown an association between PM₁₀ and RA, ¹⁷ whereas most show no association. ¹⁶ , ¹⁸ , ¹⁹ , ²² , ²³ , ²⁴ Importantly, the associations we observed for ozone and PM₁₀ were only present for seronegative RA. However, this association of ozone and PM₁₀ with seronegative RA risk is consistent with prior links between respiratory tract diseases and seronegative RA onset. ¹² The association of so many respiratory irritants with RA onset suggests that respiratory irritation itself, not specific irritants, leads to RA onset. Thus, many diverse forms of pollution may be harmful.

Although the above associations between fire smoke and RA‐ILD and NOx and RA were modest in size, small effects can translate into large health effects on a population level. For example, taking into account the incidence of RA‐ILD in the United States ⁴⁰ and the effect sizes observed in this study, halving fire smoke PM_2.5 exposure would correspond to 876 fewer patients with RA‐ILD in the United States each year. Similarly, halving NOx exposure would be associated with 1,385 fewer patients with RA in the United States each year. ⁴¹ Thus, small pollutant reductions can have massive health implications.

Regarding the timing of pollutant exposures, we observed that fire smoke PM_2.5 was most strongly associated with RA in the one to five years before the date of RA diagnosis, whereas NOx exposure more than three years and PM₁₀ exposure three to five years before RA diagnosis demonstrated the highest risk. The longer time window for NOx was also previously observed in a population‐based study in Sweden. ²² The association with more recent fire smoke exposures contrasted with our hypothesis that associations would be strong more than five years before diagnosis. This may be a result of reverse causation or detection bias. Alternatively, fire smoke may be an environmental trigger that has a more immediate effect on increasing risk. For example, previous studies with short follow‐up also showed PM_2.5 to be associated with RA ¹⁶ , ¹⁷ and RA flare. ⁴² Future mechanistic studies could help distinguish whether fire smoke triggers initial RA‐related autoimmunity itself or drives later downstream inflammatory processes such as epitope spreading.

The other pollutants studied including CO, overall PM_2.5, the third quartile of fire smoke PM_2.5, and SO₂ were not associated, or even inversely associated, with RA and RA‐ILD. One reason for the negative PM_2.5 results could be the elevated point estimates starting in the second decile. Another reason could be that prior studies showing an association between overall PM_2.5 and RA ¹⁶ , ¹⁷ , ¹⁸ had higher levels than ours and others showing no association. ²⁰ , ²¹ Although ours and some prior studies have not observed an association between SO₂ and RA, ¹⁹ , ²³ this study and the others did not include exposure data >10 years before RA onset, when it was previously found to be associated. ²² However, given the lack of reported protective associations in prior literature and known inflammatory nature of these compounds, ⁴³ , ⁴⁴ , ⁴⁵ these “protective” associations are likely spurious. We speculate they may have resulted from selection bias due to the nonrandom missingness of pollutant data, especially for the nonfire smoke pollutants, which had higher missingness and wider grids. These inverse associations for other pollutants raise the question of whether the observed positive associations for fire smoke PM_2.5 and NOx are underestimated due to a similar underlying bias. Regardless, these inconsistencies highlight the need for more robust monitoring systems in the United States for a more accurate assessment of the connection between pollutants and human health.

Strengths of this study include using the largest available integrated health data system in the United States, robust data linkages, and longer duration of pollutant exposure history compared to previous studies. There are also limitations. Perhaps most importantly, selection bias, for example due to increased missing data in controls and individuals from rural areas, may have biased results, creating potentially spurious protective associations we observed for some pollutants. Results may not generalize well to other populations because the VA population has an overrepresentation of male sex and smoking history, which likely impact inhalant exposures and RA onset. ⁸ , ⁹ , ²² The use of administrative algorithms for RA and RA‐ILD likely resulted in misclassification of outcomes, though this would also be anticipated to bias results toward the null. Misclassification of pollutant exposures may have biased results, especially due to sparse monitoring in rural areas, using monitors up to 50 km away for nonfire smoke pollutants and relying on quarterly postal address data that do not account for travel. The greater frequency of missing covariate data in controls may have resulted in selection bias, though imputed results showed minimal differences. We were not able to adjust for potential confounders such as socioeconomic status, ²² season, ⁴⁶ occupational exposures ⁹ including burn pits, ⁸ or genetics. ⁹ We chose not to adjust for certain respiratory diseases such as asthma or chronic obstructive pulmonary disease because such diseases may mediate the association between pollutants and RA. ¹¹ Results may have been due to chance due to multiple testing. Sensitivity analyses by geographic region and smoking status could be biased due to the unmatched, unconditional model design. Finally, this study did not capture recent high fire smoke years (2020–2023) or pollutant exposure data >10 years before RA diagnosis, in which certain pollutants have been most strongly associated with RA. ²²

In summary, greater fire smoke PM_2.5 exposure was associated with RA‐ILD and RA in the subsequent one to five years. Fossil fuel–related NOx was also associated with incident RA, though several other pollutants were not, or negatively, associated with RA. Given the potential sources of bias with current monitoring systems, these findings spark a call to action to develop more comprehensive monitoring systems in the United States. Confirmation of these findings would suggest that addressing wildfire smoke and fossil fuel emissions may have large‐scale population‐level health benefits by reducing the risk of RA and RA‐ILD.

ACKNOLWEDGMENTS

We thank Alexander Maas for the inspiration and Emily Fischer and Bonne Ford for their important contributions to the fire smoke algorithm necessary to conduct this project.

AUTHOR CONTRIBUTIONS

All authors contributed to at least one of the following manuscript preparation roles: conceptualization AND/OR methodology, software, investigation, formal analysis, data curation, visualization, and validation AND drafting or reviewing/editing the final draft. As corresponding author, Dr Kronzer confirms that all authors have provided the final approval of the version to be published, and takes responsibility for the affirmations regarding article submission (eg, not under consideration by another journal), the integrity of the data presented, and the statements regarding compliance with institutional review board/Declaration of Helsinki requirements.

Supporting information

Disclosure form

Appendix S1: Supplementary Information