Triangulating meta-analyses: the example of the serotonin transporter gene, stressful life events and major depression

Amy E. Taylor, Marcus R. Munafò

Meta-analysis is intended as a tool for the objective synthesis of evidence across a literature, in order to obtain the best evidence as to whether or not an association or effect is robust. However, as the use of meta-analysis has proliferated it has become increasingly clear that the results of a meta-analysis can be critically sensitive to methodological and analytical choices, so that different meta-analyses on the same topic can arrive at quite different conclusions. We demonstrate the variability in results of different meta-analyses on the same topic, using the example of the literature on the putative moderating effect of 5-HTTLPR genotype on the association between stressful life events and major depression. We also extend on previous work by including a P-curve analysis of studies from this literature, drawn from a previous meta-analysis, in an attempt to resolve the discrepant conclusions arrived at by previous meta-analyses. We highlight the divergent conclusions that can be reached when different methodological and analytical choices are taken, and argue that triangulating evidence using multiple evidence synthesis methods is preferable where possible, and that every effort should be made for meta-analyses to be as unbiased as possible (e.g., conducted by methodologists or as part of an adversarial collaboration between authors from opposing camps).