Systematic review

A systematic review is a study of studies — it locates every published (and sometimes unpublished) piece of research meeting a pre-specified question, screens each against fixed inclusion criteria, and synthesises what they collectively show. Because the search strategy and decision rules are set before any results are examined, the conclusions follow from the evidence rather than from the reviewers' prior expectations.

Within the hierarchy of evidence a reader will encounter on these pages, systematic reviews sit near the top. A randomised trial answers does this cause that in a controlled setting?; a systematic review asks what do all qualifying trials on this question collectively show? The distinction matters: a systematic review cannot improve on poor underlying data, only aggregate it.

What this design can establish

A well-conducted systematic review can establish the direction and consistency of an effect across multiple independent studies — whether a finding replicates, whether results cluster tightly or vary widely across populations and settings, and whether apparent discrepancies between individual studies dissolve once differences in method or design are accounted for.

When the underlying studies permit it, a systematic review may be paired with a meta-analysis, which pools numerical estimates to produce a single quantitative summary. In that form it can characterise the strength and precision of an association more reliably than any single trial could, and it can detect modest effects that individual studies lacked the statistical power to see.

What it cannot

A systematic review is bounded by the studies that exist. If the underlying trials were conducted on narrow populations — particular ages, health statuses, or patterns of use that do not describe the reader — the review's conclusions do not straightforwardly extend to those groups, even when the published text gives no warning.

A systematic review of poor studies produces a clean summary of poor evidence. The process adds rigour to the synthesis, not to the original data. This is the inference a reader is most likely to get wrong: a systematic review looks authoritative, and encountering a systematic review found... may suggest the finding is settled. What that phrase tells you is that the question was approached carefully; it does not say the answer is strong.

In pharmacology, the evidence base for many substances — particularly those studied primarily outside controlled trial conditions — is thin enough that systematic reviews may conclude only that the record is insufficient. That conclusion is itself informative: an absent study is a stated gap, not evidence of safety or harm.