Systematic Reviews and Meta-Analysis

Systematic reviews sit at the apex of the evidence hierarchy. They use transparent, replicable, and predefined methods to identify, appraise, and synthesise all relevant studies addressing a specific research question. When done well, they reduce the biases inherent in narrative reviews, where an individual expert may selectively cite evidence supporting a pre-formed opinion.

The steps of a systematic review include: formulating a clear, focused research question (usually in PICO format); developing a comprehensive search strategy across multiple databases; applying explicit, pre-specified inclusion and exclusion criteria; critically appraising each included study for internal validity; and synthesising the results.

A meta-analysis is an optional statistical component of a systematic review that combines quantitative results from two or more studies into a single pooled estimate of effect. Meta-analyses assign greater weight to studies that provide more information — typically larger studies with more outcome events and narrower confidence intervals. Results are displayed in forest plots, where each study is represented by a point estimate and confidence interval, and the pooled result appears as a diamond at the bottom.

Meta-analysis is not always appropriate. It should not be performed when the study designs are too different, when outcomes are not sufficiently similar across studies, or when the quality (internal validity) of the studies is poor. Heterogeneity — statistical variability across study results beyond what would be expected by chance — is a key indicator that pooling may be misleading. I² is a commonly used statistic to quantify heterogeneity.

Publication bias is a major challenge: studies with positive or statistically significant results are more likely to be published than those with null results, skewing the evidence base. Funnel plots can be used to detect publication bias, though they require a sufficient number of studies.

Systematic reviews can be used to develop evidence-based clinical practice guidelines, inform health policy, and identify gaps in knowledge. They are reproducible and transparent — limitations such as inclusion and exclusion criteria are stated explicitly. However, when reviews on the same topic give conflicting results, or when the underlying studies are of poor quality, conclusions may be inconclusive.