In modern medicine, the results of a comprehensive and methodologically sound meta-analysis bring the most robust, high-quality information to support evidence-based decision-making. With recent developments in newer meta-analytic approaches, iteration of statistical paradigms and software implementations, network and patient-level meta-analyses have recently gained popularity alongside conventional pairwise study-level meta-analyses. However, pitfalls are common in this challenging and rapidly evolving field of statistics. In this regard, guidelines have been introduced to standardize, strengthen and homogenize different aspects of conducting and reporting the results of a meta-analysis. Current recommendations advise a careful selection of the individual studies to be pooled, mainly based on the methodological quality and homogeneity in study designs. Indeed, even if a reasonable degree of variability across study results (namely, heterogeneity) can be accounted for with proper statistics (i.e. random-effect models), no adjustment can be performed in meta-analyses violating the issue of clinical validity and similarity across the included studies. In this context, this statistical primer aims at providing a conceptual framework, complemented by a practical example, for conducting, interpreting and critically evaluating meta-analyses.