The magnitude, shape, and degree of certainty in the association between long‐term population exposure to ambient fine particulate matter (PM2.5) and the risk of premature death is one of the most intensely studied issues in environmental health. For regulatory risk analysis, this relationship is described quantitatively by a concentration‐response (C‐R) function that relates exposure to ambient concentrations with the risk of premature mortality. Four data synthesis techniques develop the basis for, and derive, this function: systematic review, expert judgment elicitation, quantitative meta‐analysis, and integrated exposure‐response (IER) assessment. As part of an academic workshop aiming to guide the use of research synthesis approaches, we developed criteria with which to evaluate and select among the approaches for their ability to inform policy choices. These criteria include the quality and extent of scientific support for the method, its transparency and verifiability, its suitability to the policy problem, and the time and resources required for its application. We find that these research methods are both complementary and interdependent. A systematic review of the multidisciplinary evidence is a starting point for all methods, providing the broad conceptual basis for the nature, plausibility, and strength of the associations between PM exposure and adverse health effects. Further, for a data‐rich application like PM2.5 and premature mortality, all three quantitative approaches can produce estimates that are suitable for regulatory and benefit analysis. However, when fewer data are available, more resource‐intensive approaches such as expert elicitation may be more important for understanding what scientists know, where they agree or disagree, and what they believe to be the most important areas of uncertainty. Whether implicitly or explicitly, all require considerable judgment by scientists. Finding ways for all these methods to acknowledge, appropriately elicit, and examine the implications of that judgment would be an important step forward for research synthesis.
