ABSTRACT
Meta-analysis of individual participant data from multiple prospective epidemiological studies enables detailed investigations of quantitative summaries of evidence but involves a number of analytical challenges. In this chapter, we describe a two-stage individual participant data meta-analysis approach, principally based on Cox proportional hazards regression models, which enables (1) the harmonization of different study designs (e.g., cohort studies, nested case-control studies, and randomized trials); (2) characterization of the shape of exposure-risk relationships; (3) consistent approaches to adjustment for confounders and effect modification; (4) correction for regression dilution bias; and (5) handling of missing data through multiple imputation. The methods are exemplified by analysis of data from the Fibrinogen Studies Collaboration and the Emerging Risk Factors Collaboration.
