The methodological quality of meta-analyses in behavioral and social sciences: leads for enhancements according to AMSTAR 2

Objectives: To investigate the methodological quality in meta-analyses (MAs) published in behavioral and social sciences.
Methods: We evaluated a random sample of 207 MAs indexed in PsycINFO database in 2016. Two authors independently extracted methodologic characteristics of all MAs with the 16 items of AMSTAR2 (A MeaSurement Tool to Assess systematic Reviews 2). AMSTAR2 allows to assess the methodological quality of MAs that classifies them in 4 categories: critically low, low, moderate and high.
Results: From the 207 MAs, according to AMSTAR2 criteria, 95% of MAs were rated as critically low. Statistical methods were appropriate and publication bias well evaluated in respectively 87% and 70% of the MAs. However, much improvement is needed in data collection and analysis: Publication of the research protocol (in 11% of MAs), comprehensive literature search strategy (44%), assessment (37%) and interpretation (29%) of risk of bias in individual included studies and presentation of excluded studies (11%).
Conclusions: The methodological quality of MAs of our sample was critically low, according to AMSTAR2 criteria. Some efforts could improve tremendously the methodological quality of behavioral and social sciences MAs and thus gain in robustness and reliability.