Association between Residences in U.S. Northern Latitudes and Rheumatoid Arthritis: A Spatial Analysis of the Nurses' Health Study

Article excerpt

BACKGROUND: The etiology of rheumatoid arthritis (RA) remains largely unknown, although epidemiologic studies suggest genetic and environmental factors may play a role. Geographic variation in incident RA has been observed at the regional level.

OBJECTIVE: Spatial analyses are a useful tool for confirming existing exposure hypotheses or generating new ones. To further explore the association between location and RA risk, we analyzed individual-level data from U.S. women in the Nurses' Health Study, a nationwide cohort study.

METHODS: Participants included 461 incident RA cases and 9,220 controls with geocoded addresses; participants were followed from 1988 to 2002. We examined spatial variation using addresses at baseline in 1988 and at the time of case diagnosis or the censoring of controls. Generalized additive models (GAMs) were used to predict a continuous risk surface by smoothing on longitude and latitude while adjusting for known risk factors. Permutation tests were conducted to evaluate the overall importance of location and to identify, within the entire study area, those locations of statistically significant risk.

RESULTS: A statistically significant area of increased RA risk was identified in the northeast United States (p-value = 0.034). Risk was generally higher at northern latitudes, and it increased slightly when we used the nurses' 1988 locations compared with those at the time of diagnosis or censoring. Crude and adjusted models produced similar results.

CONCLUSIONS: Spatial analyses suggest women living in higher latitudes may be at greater risk for RA. Further, RA risk may be greater for locations that occur earlier in residential histories. These results illustrate the usefulness of GAM methods in generating hypotheses for future investigation and supporting existing hypotheses.

KEY WORDS: disease mapping, generalized additive models, geographic information systems (GIS), prospective cohort study, rheumatoid arthritis. Environ Health Perspect 118:957-961(2010). doi:10.1289/ehp.0901861 [Online 25 March 2010]

Rheumatoid arthritis (RA) is a chronic autoimmune disease with unknown etiology, although epidemiologic studies suggest genetic and environmental factors may play a role. Research on other chronic autoimmune diseases including lupus erythematosus, dermato-myositis, polymyositis, and vasculitis has shown geographic associations with higher latitudes (Gatenby et al. 2009; Hengstman et al. 2000; Somers et al. 2007; Walsh and Gilchrist 2006). Geographic variation in incident RA has been observed at the regional level according to state of residence (Costenbader et al. 2008a). The findings suggested an increased risk of RA for women who lived in the midwestern and eastern United States compared with the west, and the association was stronger with residency at ages 15 and 30 years than at baseline in 1976. In their review, Alamanos et al. (2006) also showed that RA varies geographically in areas of the world, with southern European countries having lower median incidence rates than northern European countries and North America. Ramos-Remus et al. (2007) observed that the mean age of RA onset was much younger among Mexicans than among Canadians. In another study, Anaya et al. (2001) found that RA is rare in African populations.

To explore further the association between location and RA risk, we analyzed individual-level residential data from U.S. women who participated in the Nurses' Health Study (NHS). This prospective cohort study provides information on personal covariates and participant mobility prior to RA onset. Residential histories are particularly useful when exposures of interest are time dependent. We conducted spatial analyses that considered time measured by calendar year and by year of diagnosis for cases or censoring for controls. Generalized additive models (GAMs), a type of statistical model that combines smoothing with the ability to analyze binary outcome data and adjust for examining point data a(Hastie and Tibshirani 1990; Kelsall and Diggle 1998; Webster et al. …