Meta-analysis has become a routine component of clinical research, but the software used to conduct it strongly influences how evidence is modelled, interpreted, and translated into practice. This article provides a practical field guide to modern meta-analysis software, explaining how to choose, justify, and implement current methodological standards in day-to-day work. Rather than critiquing outdated tools, the tutorial shows what applied researchers should do to produce analyses that are robust, transparent, and reproducible. The guide distinguishes legacy platforms (RevMan 5.4, MetaDiSc 1.4, Comprehensive Meta-Analysis) from modern implementations (RevMan Web, MetaDiSc 2.0, and script-based solutions such as R and Stata), clarifying which defaults matter, why certain estimators and models have been superseded, and how hierarchical and likelihood-based approaches improve real-world inference. Short, software-agnostic explanations outline when REML or Paule–Mandel estimators are preferable to DerSimonian–Laird, when Hartung–Knapp adjustments are warranted, and why diagnostic accuracy meta-analysis requires bivariate or HSROC models rather than separate pooling of sensitivity and specificity. By positioning software choices within the logic of the modelling workflow, this tutorial enables readers to anticipate limitations, justify methodological decisions, and avoid silent defaults that weaken inference. The objective is not merely technical update but conceptual alignment: treating meta-analysis as a modelling exercise requiring transparency and judgment rather than a mechanical point-and-click task. This field guide is intended as a practical resource for authors, reviewers, and journals aiming to raise methodological standards in applied evidence synthesis.
Javier Arredondo Montero (Tue,) studied this question.