Air pollution risk estimates: determinants of heterogeneity.

Information on risk to health for particular populations may be critical for motivating air pollution control, setting air quality standards, and evaluating control measures. This article considers the application of associated risk estimates from one or more external populations to a specific population. This concept is termed "transportability," and the degree of transportability of a risk estimate may be limited by (1) sources and characteristics of the local air pollution mixture, (2) pattern of exposures to the population, and (3) population characteristics determining susceptibility. For any particular estimate in a particular study or in compilations of study findings, multiple potential determinants affect its value, including biological factors, aspects of study methodology, and bias introduced by study design and analysis and by the publication process. Each limitation is discussed in light of current research, and solutions are proposed for addressing these limitations. These include (1) a full description of key aspects of study design and analysis; (2) systematic capture of published risk estimates using large databases; (3) use of multicity study models such as NMMAPS and APHEA; and (4) development of protocols for empirical analyses at the local level to support the rigorous estimation of risk in a broader, international context.