Fine particulate matter and nonaccidental and cause-specific mortality: Do associations vary by exposure assessment method?

Background: There is considerable heterogeneity in fine particulate matter (PM_2.5)-mortality associations between studies, potentially due to differences in exposure assessment methods. Our aim was to evaluate associations of PM_2.5 predicted from different models with nonaccidental and cause-specific mortality.

Methods: We followed 107,906 participants of the Nurses' Health Study cohort from 2001 to 2016. PM_2.5 concentrations were estimated from spatiotemporal models developed by researchers at the University of Washington (UW), Pennsylvania State University (PSU), and Harvard TH Chan School of Public Health (HSPH). We calculated 12-month moving average concentrations and we used time-varying Cox proportional hazard ratios (HRs).

Results: There were 30,242 nonaccidental deaths in 1,435,098 person-years. We observed high correlations and similar temporal trends between the PM_2.5 predictions. We found no associations of UW, PSU, or HSPH PM_2.5 with nonaccidental mortality, but suggestive positive associations with cancer, cardiovascular, and respiratory disease mortality. There were small differences in HRs between the PM_2.5 predictions. All three predictions showed the strongest associations with cancer mortality: HRs (95% confidence interval, expressed per 5 µg/m³ increase) were 1.06 (1.01, 1.12) for UW, 1.08 (1.03, 1.13) for PSU, and 1.05 (1.00, 1.10) for HSPH. In a subset restricted to participants who were always exposed to PM_2.5 below 12 µg/m³, we observed positive associations with nonaccidental mortality.

Conclusion: We found that differences between PM_2.5 exposure assessment methods could lead to minor differences in strengths of associations between PM_2.5 and cause-specific mortality in a population of US female nurses.