National Morbidity, Mortality, and Air Pollution Study. Part III: Concentration–Response Curves and Thresholds for the 20 Largest US Cities

In Part III of the National Morbidity, Mortality, and Air Pollution Study (NMMAPS), Dr. Daniels and colleagues at Johns Hopkins University and Harvard University evaluated the shape of the relation between PM₁₀ concentrations measured at fixed monitoring sites and daily mortality among residents from all causes (excluding accidental causes), from all cardiovascular and respiratory causes combined, and from causes other than cardiovascular-respiratory disease. They developed new statistical methods to address the shape of this relation as well as to characterize the uncertainty of estimated threshold values for each city and for all 20 cities combined. The investigators evaluated three concentration–response models: log-linear models, in which the relation between PM₁₀ and the (log) risk of mortality is assumed to be a straight line; spline models, in which the method estimates a concentration–response curve when little can be assumed about its shape; and threshold models, in which the method assumes no effect until a specified threshold level is exceeded (after which the relation is assumed to be log linear).