Assessing and exploring heterogeneity

Abstract

Meta-analysis is a statistical method for combining the results of studies included in the systematic review. It is justified only when the potentially eligible studies are similar enough. However, some differences in their clinical or methodological characteristics are inevitable as no two studies are expected to be identical in the true sense. Clinical heterogeneity is caused by diversity in important characteristics such as participants, interventions, comparators, or outcomes (in extreme cases, 'apples vs. oranges'). Methodological heterogeneity involves differences in the design (e.g., randomised vs. quasi-randomised) and methodological quality of studies (e.g., masked vs. non-masked allocation) included in a systematic review. A fair amount of clinical judgement is thus necessary to decide whether or not studies are similar enough to be combined in a meta-analysis. Statistical heterogeneity in a meta-analysis means that the between-study variation in the effect of intervention varies beyond the extent expected by chance alone. This chapter is focussed on understanding, assessing and handling heterogeneity from various sources in meta-analysis.