Abstract
Objective
Recent studies on air pollution and disease have been based on millions of participants within a region or country, relying entirely on register-based confounder adjustment. We aimed to investigate the effects of increasing adjustment for register- and questionnaire-based covariates on the association between air pollution and cardiometabolic diseases.
Methods
In a population-based cohort of 246,766 eligible participants randomly selected across Denmark in 2010 and 2013 and followed up until December 31, 2017, we identified 3,247 myocardial infarction (MI) cases, 4,166 stroke cases and 6,366 type 2 diabetes cases. Based on historical address-information, we calculated 5-year time-weighted exposure to PM2.5 and NO2 modelled using a validated air pollution model. We used Cox proportional hazards models to calculate hazard ratios (HR) with increasing adjustment for a number of individual- and area-level register-based covariates as well as lifestyle covariates assessed through questionnaires.
Results
We found that a 5 μg/m3 higher PM2.5 was associated with HRs (95% CI) for MI, stroke and diabetes, of respectively, 1.18 (0.91–1.52), 1.11 (0.88–1.40) and 1.24 (1.03–1.50) in the fully adjusted models. For all three diseases, adjustment for either individual-level, area-level or lifestyle covariates, or combinations of these resulted in higher HRs compared to HRs adjusted only for age, sex and calendar-year, most marked for MI and diabetes. Further adjustment for lifestyle in models with full register-based individual- and area-level adjustment resulted in only minor changes in HRs for all three diseases.
Conclusions
Our findings suggest that in studies of air pollution and cardiometabolic disease, which use an adjustment strategy with a broad range of register-based socioeconomic variables, there is no effect on risk estimates from subsequent lifestyle adjustment.