Meta-Analysis and Systematic Review

Meta-Analysis and Systematic Review

1. Definition, scope & place in the evidence hierarchy

• A systematic review is a review that uses explicit, pre-specified, reproducible methods to identify, select, critically appraise and synthesise all studies addressing a clearly formulated question; meta-analysis is the optional statistical step within it — the quantitative combination of results from two or more separate studies (Cochrane v5.0.1 Ch.9.1.2, p.244).
• A systematic review need not contain any meta-analysis; where studies are too clinically diverse, at risk of bias, or affected by reporting bias, a narrative (qualitative) synthesis is preferable and a pooled estimate may be misleading (Cochrane Ch.9.1.4, p.246; Ch.9.5.3, p.279).
• Distinction from a narrative/traditional review: the systematic review pre-specifies its question, eligibility criteria, search, appraisal and analysis in a protocol, minimising the reviewer's selective, post-hoc emphasis on favoured studies (Cochrane Ch.9.1.7, p.250).
• A general framework for any synthesis (narrative or quantitative) answers four questions: (1) What is the direction of effect? (2) What is the size of effect? (3) Is the effect consistent across studies? (4) What is the strength of evidence for the effect? Meta-analysis addresses 1–3; question 4 additionally requires judgements on study design, risk of bias and certainty (Cochrane Ch.9.1.2, p.244).
• Position atop the evidence pyramid — a rigorously conducted systematic review/meta-analysis of randomised controlled trials sits at the apex of the hierarchy of evidence for therapeutic questions because it integrates the totality of unbiased primary evidence; an individual patient data (IPD) meta-analysis is regarded as the "gold standard" form of systematic review (Cochrane Ch.18.1.1, p.548). [Note: this topic is the synthesis layer — NOT primary trial design.]
• ⚠ Edition caveat: the foundational text here is the Cochrane Handbook v5.0.1 (2008). The modern reporting/appraisal standards that now define best practice — RoB 2, ROBINS-I, GRADE certainty grading, PRISMA 2020 — postdate this edition and are grounded from their respective primary methods documents (see §6, §16, §17).