Invited commentary: Particulate matter-mortality exposure-response relations and threshold

In this issue of the Journal, Daniels et al. (1) report an analysis of the shape of the exposure-response relation between daily concentrations of par ticulate matter air pollution (PM) and mortality. They provide additional e pidemiologic evidence of the absence of a population-based "no-effects" thr eshold level for PM within relevant ranges of exposure. Their results sugge st that the PM-mortality exposure-response relation is near linear, with mo rtality risk occurring even at concentrations below current regulatory leve ls. This analysis is the latest of several important contributions to the l iterature on particuate matter and mortality by Samet, Zeger, and various c olleagues. In the early 1990s, following the publications of several studies that sugg ested a link between daily mortality and PM pollution at relatively low con centrations (2-7), Samet and others argued that the findings could not be a dequately interpreted and they encouraged new studies (8). Samet and Zeger then led a reanalysis effort (9, 10) that largely replicated the PM-mortali ty associations observed in selected early studies while refining and contr ibuting to applicable statistical methodologies. Most recently they, along with Dominici and others, have been developing approaches to incorporate mu ltiple cities in a comprehensive analysis of daily time-series mortality an d air pollution in the United States (11, 12). They have provided insights relating to the consistency of the observed PM-mortality associations (13), the importance of mortality displacement (or harvesting) (14), issues rega rding measurement error (15), alternative ways to control for weather varia bles (10), and now the shape of the exposure-response relation (1). While this excellent work represents an important contribution, it does not stand alone. This commentary provides some context and perspective with re spect to the development of the epidemiologic literature and the public hea lth relevance of these findings.