Extended definition
A systematic review is a structured synthesis of scientific literature on a specific research question, conducted with explicit, reproducible, and preregistered methodology. It is distinguished from narrative reviews by following formal protocol: specified research question (often in PICO format — Population, Intervention, Comparison, Outcome), documented search strategy, predefined inclusion and exclusion criteria, quality appraisal of included studies (risk of bias), structured data extraction, and synthesis — narrative or meta-analytic when data allow. The Cochrane Collaboration established the contemporary gold standard in health with the Cochrane Handbook for Systematic Reviews of Interventions (Higgins et al., 2019); Khan et al. (2003) offered an accessible introduction in “Five steps to conducting a systematic review”. Protocol preregistration (PROSPERO in health, OSF in other fields) is modern standard practice. Variants include scoping review (maps literature without deep synthesis), umbrella review (review of reviews), and living systematic review (continuously updated).
When it applies
Systematic review applies when the question is specific and there is sufficient literature to synthesize — typically >10 relevant studies. It is standard in evidence-based medicine (informs clinical guidelines, regulatory decisions), in applied social sciences (Campbell Collaboration), in education (EPPI-Centre), in environmental sciences (CEE — Collaboration for Environmental Evidence). It applies as the basis for meta-analysis when primary studies allow quantitative synthesis. It is a growing requirement in funding proposals as justification for the necessity of proposed research. In new fields or with broad but heterogeneous literature, scoping review precedes systematic review.
When it does not apply
It does not apply in fields with sparse literature (few, scattered studies) — narrative or scoping review is an alternative. It does not apply for excessively broad questions — systematic review requires a focused question (PICO). It does not replace narrative review in theoretical essays, conceptual debates, and thesis introductions — different functions. It does not apply as a shortcut for quick “state of the art” — rigorous protocol requires months of work with a team. It does not apply without appropriate methodological training: poorly conducted systematic review produces biased synthesis, often worse than no review (impression of solid evidence where there is systematic bias).
Applications by field
— Health: Cochrane Library, gold standard; Campbell in mental health; PROSPERO required for registration. — Education: EPPI-Centre conducts reviews on educational policy and practice. — Software engineering: systematic review in SE has its own tradition since Kitchenham (2004); specific guidelines. — Environmental sciences: CEE sets standards for evidence in conservation and management.
Common pitfalls
The first pitfall is confusing systematic review with exhaustive narrative review — systematic requires preregistered protocol, not just broad reading. The second is incomplete search strategy: limiting to one database (PubMed only) or to English excludes relevant literature; search in ≥3 databases is minimum, with complementation by reference searching and gray literature. The third is failing to appraise risk of bias of included studies — Cochrane Risk of Bias 2 (RoB 2) and ROBINS-I are standard instruments. The fourth is meta-analysis without testing heterogeneity: high signals that studies should not be combined statistically; narrative synthesis is more appropriate. The fifth is ignoring publication bias: studies with significant results are published faster and in more visible journals; funnel plot and Egger’s test are standard diagnostics.