Drs. Alfredsson and Karlson contributed equally to this work.
Ambient Air Pollution Exposures and Risk of Rheumatoid Arthritis
Article first published online: 1 JUL 2013
Copyright © 2013 by the American College of Rheumatology
Arthritis Care & Research
Volume 65, Issue 7, pages 1190–1196, July 2013
How to Cite
Hart, J. E., Källberg, H., Laden, F., Costenbader, K. H., Yanosky, J. D., Klareskog, L., Alfredsson, L. and Karlson, E. W. (2013), Ambient Air Pollution Exposures and Risk of Rheumatoid Arthritis. Arthritis Care Res, 65: 1190–1196. doi: 10.1002/acr.21975
- Issue published online: 1 JUL 2013
- Article first published online: 1 JUL 2013
- Accepted manuscript online: 11 FEB 2013 03:14PM EST
- Manuscript Accepted: 29 JAN 2013
- Manuscript Received: 2 OCT 2012
- NIH. Grant Numbers: R01-AR49880, CA87969, P60-AR047782, K24-AR0524-01, P01-CA87969, ES017017
Environmental factors may play a role in the development of rheumatoid arthritis (RA). We previously observed increased RA risk among women living closer to major roads (a source of air pollution). Herein, we examined whether long-term exposures to specific air pollutants were associated with RA risk among women in the Nurses' Health Study (NHS).
The NHS is a large US cohort of female nurses followed up prospectively every 2 years since 1976. We studied 111,425 NHS participants with information on air pollution exposures as well as data concerning other lifestyle and behavioral exposures and disease outcomes. Outdoor levels of different size fractions of particulate matter (PM10 and PM2.5) and gaseous pollutants (SO2 and NO2) were predicted for all available residential addresses using monitoring data from the US Environmental Protection Agency. We examined the association of time-varying exposures 6 and 10 years before each questionnaire cycle and cumulative average exposure with the risk of RA, seronegative (rheumatoid factor and anti–citrullinated peptide antibody negative) RA, and seropositive RA.
Over the 3,019,424 person-years of followup, 858 incident RA cases were validated by medical record review by 2 board-certified rheumatologists. Overall, we found no evidence of increased risk of RA, seronegative RA, or seropositive RA with exposure to the different pollutants and little evidence of effect modification by socioeconomic status or smoking status, geographic region, or calendar period.
In this group of socioeconomically advantaged middle-aged and elderly women, adult exposures to air pollution were not associated with an increased RA risk.
Rheumatoid arthritis (RA) is a chronic systemic inflammatory disease affecting approximately 1% of the adult population ([1-4]). Epidemiologic studies have shown that the risk of developing RA is associated with exposure to cigarette smoke, silica, and mineral oil ([5-20]), suggesting that respiratory exposures activating the immune system may lead to RA.
Using data from the Nurses' Health Study (NHS), a large prospective cohort of US women who are registered nurses, we previously examined the association between the incidence of RA and distance of residence to the nearest major road as a marker of traffic exposure (). We observed a 30% increased risk of RA in women living within 50 meters of a major road compared with women living at least 200 meters away, suggesting a possible association between RA and air pollution. In a recent analysis from the Swedish Epidemiological Investigation of Rheumatoid Arthritis (EIRA) case–control study, we examined the association of RA risk with the exposure to several specific air pollutants, including SO2, NO2, and particulate matter <10 microns in aerodynamic diameter (PM10) from local sources (traffic and home heating) within Stockholm County (). While there were no consistent overall associations between air pollution and RA risk, the results suggest that selected pollutants (NO2 and SO2) were associated with increased risk of RA. Our objective in the present study was to conduct similar analyses examining the associations of specific air pollutants with the risk of RA within the NHS.
Box 1. Significance & Innovations
- Respiratory exposures, including cigarette smoke, silica exposure, and mineral oil, have been associated with an increased risk of rheumatoid arthritis (RA).
- We have previously observed an elevated risk of RA in women living close to roadways and an association between exposure to certain air pollutants and risk for RA in a Swedish case-control study. Therefore, we aimed to examine if adult exposures to air pollution were associated with increased risks for RA in the Nurses' Health Study.
- Predicted ambient exposures to air pollution (SO2, NO2, and particulate matter) during adulthood were not associated with an increased risk of incident RA in this group of socially advantaged older women.
MATERIALS AND METHODS
The NHS is a long-term prospective cohort study of US nurses. The NHS was initiated in 1976 when 121,700 US female registered nurses ages 30–55 years completed a mailed questionnaire. At the study inception, the women resided in 11 states in various locations across the US (California, Connecticut, Florida, Massachusetts, Maryland, Michigan, New Jersey, New York, Ohio, Pennsylvania, and Texas). However, by 1986, members of the cohort had moved throughout the US, with at least 10 nurses in all 50 states. Followup questionnaires, with response rates >90%, are mailed every 2 years to update information on risk factors and the occurrence of major illnesses. The followup questionnaires also provide a detailed residential history for each participant, which has been available electronically since 1986. Women were included in the current study if they had at least 1 home address within the continental US from 1986–2006 and had no history of RA or other connective tissue disease at baseline in 1976. A total of 111,425 participants were available for analysis.
RA was confirmed in all nurses reporting doctor-diagnosed RA by a connective tissue disease screening questionnaire () followed by medical record review,as detailed previously (). The subjects who self-reported but later denied RA diagnosis, denied permission to obtain medical records, or had a negative screening questionnaire were excluded. For this analysis, we identified a total of 858 confirmed incident RA cases from 1976–2006. Information on the presence of rheumatoid factor (RF) or anti–citrullinated peptide (anti-CCP) antibodies was extracted from the medical records and used to classify RA phenotypes as seropositive RA (RF and/or anti-CCP antibody positive) or seronegative RA (RF and anti-CCP antibody negative). The protocol was approved by the Human Subjects Committee at Brigham and Women's Hospital and participants provided implied informed consent to participate.
The information on residential addresses is updated in the NHS cohort every 2 years as part of the questionnaire mailing process and was geocoded to obtain latitude and longitude for each address from 1986–2006. The predicted monthly information on ambient PM10 and PM2.5 from 1988–2007 was available for each address from spatiotemporal models incorporating a combination of spatial smoothing of US Environmental Protection Agency monitoring values, meteorologic information, and information on geographic information system–based covariates such as elevation, population density, distance to roadways, and point-source emissions ([25, 26]). The data on annual average exposure to SO2 and NO2 from 1985–2000 were available from different prediction models using a similar approach of spatial smoothing of monitoring values and information from geographic information systems–based covariates including elevation, population density, distance to roadways, and distance to and emissions from the nearest power plant (). We assumed that the levels of pollution were constant from 1976 (cohort study start) to 1986 (first available air pollution data) and then used the updated exposure estimates every 2 years based on the geocoded address data. From 1976–1986, 40.8% of the nurses moved to a different address. In sensitivity analyses, we excluded RA cases diagnosed before 1986, including only cases diagnosed after exposure data were available.
We created 3 exposure metrics to examine the different potentially important windows of exposure. To explore the effects of the timing of exposure prior to disease incidence, we created metrics to examine the time-varying annual exposure on the sixth and tenth year prior to each questionnaire cycle. These windows of exposure were originally chosen in the EIRA study on the basis of previous studies demonstrating that antibodies were elevated 5–10 years prior to the diagnosis of RA ([28-30]). Because it is also plausible that longer-term exposure to air pollution is the exposure of interest, we calculated the time-varying cumulative average exposure during the followup period.
Information on potential confounders from the biennial questionnaires is available in the NHS. Therefore, when appropriate, each woman was assigned updated covariate values every 2 years. We examined possible confounding by age (in months), race, age at menarche, parity, total months of lactation, current menopausal status, menopausal hormone use, oral contraceptive use, physical activity, and body mass index; many of these characteristics have been shown to be risk factors for RA ([24, 31]). To control for smoking, we used lifetime smoking history data to calculate pack-years (number of packs per day multiplied by the number of years of cigarette smoking) and current smoking status (current, former, and never) (). To control for individual-level socioeconomic status (SES), we considered several variables including the nurses' education level, occupation of both parents, marital status, and, if applicable, husband's education level. To control for area-level SES, we included area-level information on census tract–level median income and house value.
Time-varying Cox proportional hazards models were used to assess the relationship between incident RA from 1976–2006 with each of the air pollutants in separate models. Person-time accrued from January 1, 1976 until either the diagnosis of RA, loss to followup, date of death, or the end of followup (whichever happened first). Person-time (3% of the total) was excluded from followup for any period in which the home address was outside of the continental US or pollution predictions were not available. Hazard ratios (HRs) and 95% confidence intervals were calculated based on an interquartile range change (difference between the 75th and 25th percentiles of the distribution) for each pollutant after determining that the dose-response relationship was not statistically significantly different from linear using splines. All Cox models were stratified by age in months and calendar year. Previous analyses in the Swedish EIRA study suggested possible effect modification by SES, with higher risks observed in those with lower SES (). To examine a potential effect modification by SES in the NHS, we performed stratified analyses by the level of the husband's education (the best available indicator of individual SES in this cohort) to obtain category-specific HRs and created multiplicative interaction terms to test for statistical significance (P < 0.05). Smoking status has also been shown to be an important effect modifier of RA in the NHS (); therefore, we similarly examined effect modification by ever/never smoking status. Last, to determine if there were important regional or temporal differences in the risk of RA with pollution, we examined effect modification by US Census region (Northeast, South, Midwest, and West) and calendar period (1976–1986, 1987–1996, and 1997–2006). In sensitivity analyses, we also limited the study followup to the periods in which air pollution predictions were directly available in order to determine if our assumptions of static levels were important. All statistical analyses were performed using SAS, version 9.1.3 ().
Throughout the entire followup period (1976–2006), NHS participants had a mean ± SD age of 55.9 ± 10.9 years. Of the 111,425 participants included, 43% were never smokers and most (73%) had an RN degree (Table 1). As expected in this nationwide study, all of the pollutants examined had a wide range of exposure, with wider distributions for NO2 and SO2 than for particulate matter (Table 2). The distributions of each pollutant were similar for the various exposure metrics examined.
|Age, mean ± SD yearsa||55.9 ± 10.9|
|Pack-years of smoking, mean ± SDb||12.8 ± 18.6|
|Parous, never breastfed||30|
|Parous, breastfed 1–11 months||36|
|Parous, breastfed ≥12 months||16|
|Postmenopausal hormone usec|
|Oral contraceptive use|
|Physical activity, MET hours/week|
|Missing or not applicable||34|
|Less than high school||4|
|Greater than high school||35|
|Median census tract family income, mean ± SD||$59,602 ± $28,252|
|Median census tract household value, mean ± SD||$159,880 ± $128,329|
|Exposure metric||NO2 (μg/m3)||SO2 (μg/m3)||PM10 (μg/m3)||PM2.5 (μg/m3)|
|Sixth year prior|
|Mean ± SD||33.9 ± 14.2||20.4 ± 10.8||28.3 ± 7.1||16.8 ± 3.7|
|Median (IQR)||32.6 (15.3)||21.8 (14.4)||27.2 (7.6)||16.9 (5.1)|
|Min, max||0, 171.4||0, 128.6||5.1, 83.8||2.5, 32.7|
|Tenth year prior|
|Mean ± SD||34.3 ± 14.1||20.9 ± 10.7||28.7 ± 7.0||17.0 ± 3.7|
|Median (IQR)||32.9 (15.0)||22.3 (14.1)||27.6 (7.5)||17.2 (4.9)|
|Min, max||0, 171.4||0, 123.4||5.1, 81.9||2.5, 31.3|
|Mean ± SD||33.3 ± 13.9||19.7 ± 10.4||27.7 ± 6.8||16.5 ± 3.6|
|Median (IQR)||31.9 (14.9)||20.7 (14.1)||26.6 (7.3)||16.6 (4.9)|
|Min, max||0, 171.4||0, 124.1||5.1, 82.1||2.5, 31.4|
During 3,019,423.5 person-years of followup, there were a total of 858 incident cases of RA (58.4% seropositive) among the 111,425 women. Overall, there was no evidence of an increased risk of total RA with exposures to air pollution (Table 3). The HRs were mostly below 1, and in some cases statistically significant inverse effects were observed. Similar patterns were seen in models restricted to the seropositive or seronegative RA phenotypes (Table 4). The results from models stratified by husband's education level as a measure of individual-level SES are shown in Figure 1. Although not statistically significantly different, overall, the group of women who were not married or whose husbands had a high school education or less tended to have slightly higher HRs compared to the group of women whose husbands had greater than a high school education. The results stratified by smoking status are shown in Figure 2. With the exception of SO2, in general, never smokers tended to have lower HRs associated with air pollution exposures compared to ever smokers and only the interaction terms for NO2 and smoking were statistically significant (P = 0.02 for interaction for the sixth year and P = 0.01 for interaction for the tenth year prior to each questionnaire cycle). No statistically significant differences were observed by census region of residence or calendar period. Additionally, our findings from models restricted to periods in which predictions of air pollution were directly available were similar.
|Timing of time-varying pollution||NO2 (15 μg/m3)||SO2 (14 μg/m3)||PM10 (7 μg/m3)||PM2.5 (5 μg/m3)|
|Sixth year prior|
|Model 1a||0.92 (0.85–0.99)||1.00 (0.92–1.09)||0.91 (0.85–0.98)||0.94 (0.86–1.03)|
|Model 2b||0.92 (0.85–0.99)||1.01 (0.92–1.10)||0.92 (0.86–0.99)||0.95 (0.87–1.05)|
|Model 3c||0.94 (0.87–1.01)||1.00 (0.91–1.10)||0.93 (0.86–1.00)||0.95 (0.87–1.04)|
|Tenth year prior|
|Model 1a||0.91 (0.85–0.98)||0.99 (0.91–1.08)||0.91 (0.85–0.98)||0.93 (0.85–1.02)|
|Model 2b||0.91 (0.84–0.98)||1.00 (0.91–1.09)||0.92 (0.87–1.05)||0.95 (0.86–1.04)|
|Model 3c||0.92 (0.85–1.00)||0.99 (0.90–1.08)||0.93 (0.86–0.99)||0.95 (0.85–1.03)|
|Model 1a||0.91 (0.84–0.98)||0.99 (0.91–1.09)||0.90 (0.84–0.97)||0.93 (0.85–1.02)|
|Model 2b||0.91 (0.84–0.98)||1.00 (0.91–1.10)||0.91 (0.85–0.98)||0.95 (0.86–1.04)|
|Model 3c||0.92 (0.85–1.00)||0.99 (0.90–1.09)||0.92 (0.85–0.99)||0.94 (0.86–1.04)|
|Timing of time-varying pollution||Seronegative RA (n = 357 cases)||Seropositive RA (n = 501 cases)|
|NO2 (15 μg/m3)||SO2 (14 μg/m3)||PM10 (7 μg/m3)||PM2.5 (5 μg/m3)||NO2 (15 μg/m3)||SO2 (14 μg/m3)||PM10 (7 μg/m3)||PM2.5 (5 μg/m3)|
|Sixth year prior|
|Model 1a||0.93 (0.83–1.04)||0.98 (0.86–1.12)||0.99 (0.89–1.10)||0.99 (0.86–1.14)||0.90 (0.82–0.99)||1.02 (0.91–1.14)||0.85 (0.77–0.94)||0.90 (0.80–1.02)|
|Model 2b||0.93 (0.83–1.04)||0.99 (0.86–1.13)||1.00 (0.91–1.11)||1.01 (0.88–1.16)||0.90 (0.82–1.00)||1.03 (0.92–1.15)||0.86 (0.78–0.95)||0.92 (0.81–1.03)|
|Model 3c||0.97 (0.86–1.10)||0.96 (0.83–1.11)||1.02 (0.92–1.14)||1.01 (0.87–1.16)||0.92 (0.83–1.01)||1.03 (0.91–1.16)||0.87 (0.79–0.96)||0.93 (0.82–1.05)|
|Tenth year prior|
|Model 1a||0.93 (0.83–1.04)||0.97 (0.85–1.11)||0.99 (0.89–1.09)||0.98 (0.86–1.13)||0.89 (0.81–0.98)||1.02 (0.91–1.14)||0.85 (0.77–0.94)||0.89 (0.80–1.00)|
|Model 2b||0.93 (0.83–1.04)||0.97 (0.85–1.12)||1.00 (0.91–1.11)||1.00 (0.87–1.15)||0.89 (0.81–0.99)||1.03 (0.91–1.15)||0.86 (0.78–0.95)||0.91 (0.81–1.02)|
|Model 3c||0.97 (0.86–1.09)||0.95 (0.82–1.09)||1.02 (0.92–1.13)||0.99 (0.86–1.14)||0.91 (0.82–1.00)||1.03 (0.91–1.16)||0.86 (0.78–0.95)||0.92 (0.81–1.03)|
|Model 1a||0.92 (0.82–1.04)||0.97 (0.84–1.11)||0.98 (0.88–1.10)||0.99 (0.86–1.15)||1.02 (0.90–1.14)||0.89 (0.80–0.98)||0.84 (0.76–0.93)||0.89 (0.78–1.00)|
|Model 2b||0.93 (0.83–1.04)||1.03 (0.68–1.55)||1.00 (0.91–1.11)||0.99 (0.94–1.04)||1.03 (0.91–1.16)||0.89 (0.80–0.98)||0.85 (0.77–0.94)||0.90 (0.80–1.02)|
|Model 3c||0.97 (0.86–1.10)||0.95 (0.82–1.10)||1.02 (0.91–1.14)||1.01 (0.87–1.17)||1.03 (0.91–1.16)||0.90 (0.81–1.00)||0.85 (0.77–0.95)||0.91 (0.80–1.03)|
In this cohort of middle-aged and elderly women, we found little evidence of adverse effects of total ambient exposures to air pollution on the incidence of RA. Our results are similar to our previous findings in the Swedish EIRA case–control study, where no consistent associations were observed with local source–specific levels of the same pollutants. However, unlike in the EIRA analyses, we did not observe increased risks in the analyses of any RA phenotypes with exposures to NO2 and SO2.
In our previous study using NHS data, we observed a 30% elevated risk of RA in patients with a residence within 50 meters of a major roadway, suggesting a possible role of air pollution in the increased risk of RA (). However, in the same population, we did not observe an elevated risk of RA with predicted measures of ambient pollution. This may be partially explained by the weak association between the individual pollutants and the specific measurement of the distance to the road used in our previous work. When we restricted the present analysis to women with a street-level geocoded address in 2000 (the same population as our previous analysis), the correlations between the distance to the road and the individual pollutants were in the range of 0.05–0.11 and there was little difference in the distribution of levels by the distance to road categories. Therefore, it is possible that the distance to roadway is a proxy for some other exposures (e.g., noise, neighborhood), which may have confounded the association with an increased risk of RA in our previous analyses, or, alternatively, that the measurements of air pollution exposures in the current study were too imprecise or not strongly associated with the etiologically relevant agent.
The present analysis had several limitations. SES was an important confounder and a significant effect modifier in our previous EIRA analyses (), with higher risk of RA observed among individuals with a lower education level. It is possible that, within the present analysis, the lack of an association between increased RA was due to all of the NHS participants having a high education level and SES because they were employed as nurses at the start of the study; this also likely reduced the generalizability of our findings to women with similar SES in the US. Air pollution exposure data were not available for the full period of followup, meaning that for cases diagnosed from 1976–1985, the air pollution exposure data were from after the onset of RA, with the assumption that air pollution levels were ranked the same in earlier time periods. Our assumption of static levels of air pollution prior to the availability of air pollution predictions likely introduced nondifferential misclassification, which may partially explain our null findings. However, in sensitivity analyses limited to incident RA after air pollution exposure, our results were similar. Our pollution predictions were intended to represent the total levels of each pollutant experienced at each address and had a similar distribution to general levels measured in the continental US. In our previous analysis of the EIRA cohort, we assessed pollution levels only from local sources of traffic and home heating. The modeling approach in the present study included all sources of these pollutants (local and regional), all of which may not have contributed to disease risk. If only local sources of traffic and home heating are associated with RA risk, then our inclusion of additional sources of pollution would lead to increased measurement error, limiting our ability to detect associations. Regardless, our measures of ambient air pollution were based on exposure models that are imperfect predictors of personal exposures, and we do not have information concerning the amount of time that each participant spent at each home address. Last, we lacked information on exposure to these pollutants at locations other than the participants' residence. This prevented us from examining the effect of each nurse's total air pollution exposure on the incidence of RA.
Because cigarette smoking, crystalline silica, and organic solvent exposures, all via respiratory routes, are among the strongest environmental exposures related to increased risk of RA, our hypothesis that ambient air pollution exposure also may be related to RA risk was well founded. Because our study used NHS data, it is the largest population-based cohort followed over many years to examine air pollution exposures prior to the development of RA. However, in this cohort, we were unable to detect any associations between residential air pollution and the risk of RA using several metrics. As discussed previously, it is unclear if this lack of an association was due to limitations in the ability to accurately assess air pollution exposure, or whether air pollution does not represent a source of increased RA risk among middle-aged to older, mainly white women living in the US.
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. Hart 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. Hart, Laden, Costenbader, Klareskog, Alfredsson, Karlson.
Acquisition of data. Hart, Laden, Costenbader, Yanosky, Karlson.
Analysis and interpretation of data. Hart, Källberg, Laden, Costenbader, Karlson.
- 22Ambient air pollution exposures and risk of rheumatoid arthritis: results from the Swedish EIRA case-control study.Ann Rheum Dis2012. E-pub ahead of print., , , , , , et al.
- 33SAS software: version 9. Release 6.12.Cary (NC):SAS Institute;2006.