ABSTRACT
Two forms of within-study biases can potentially invalidate the assumptions of a meta-analysis: biases to internal validity (internal biases) and biases to external validity (external biases). Internal biases are caused by methodological flaws in studies, while external biases are caused by differences between the designs of available studies and the target research question. In this chapter, methods for assessing the presence of internal biases in meta-analyses are described and meta-epidemiological research studies that have estimated the magnitudes of various internal biases are discussed. Methods for adjusting for internal biases using meta-epidemiological evidence are provided, as well as methods for adjusting for internal and external biases using elicited expert opinion. Methods developed specifically for addressing biases caused by missing data within studies are described separately and adjustment for internal biases in the special case of network meta-analysis is discussed.
