Advances in the meta-regression of heterogeneous studies: lessons learned from EPA’s analysis of inorganic arsenic

Challenges can arise when analyzing a group of published epidemiology studies due to heterogeneity in how effect measures are reported, exposure metrics, and other methodological issues. In this presentation, we will discuss techniques developed in the context of exposure or dose-response modeling of inorganic arsenic (iAs) exposure to facilitate meta-analyses in the presence of heterogeneous study designs and reporting. These methods include (1) effective count derivation, where covariate-adjusted relative risks and odds ratios are used to estimate the counts of cases and controls that would have been observed if covariates in all exposure groups were the same as those in the referent group, i.e., the "effective counts”; (2) dose conversion to convert various study-specific iAs exposure metrics into a uniform measure of intake; and (3) meta-analytic dose response methods, including both Bayesian hierarchical modeling for bladder cancer and categorical regression, where health outcomes are grouped into severity categories to facilitate meta-analysis, for diseases of the circulatory system (DCS). These approaches allow us to maximize the set of studies eligible for inclusion in meta-analysis. However, while these methods successfully address issues arising from heterogeneous reporting, use of such approaches can introduce uncertainty in the analysis. This presentation will discuss how improved reporting in published epidemiology studies can decrease such uncertainty and enhance the utility of meta-analyses across studies. Disclaimer: The views expressed are those of the authors and do not necessarily represent the views or policies of the US EPA.