PRISMA 2020

What it is

PRISMA 2020 (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) is the flagship reporting guideline for systematic reviews, published by Page and colleagues in BMJ in 2021 as a major update to the original 2009 statement. It is maintained under the EQUATOR Network and curated by the PRISMA steering group at prisma-statement.org. The statement is a 27-item checklist (organised across Title, Abstract, Introduction, Methods, Results, Discussion, and "Other information" — funding, registration, data/code availability) plus an updated four-phase flow diagram (identification, screening, eligibility, inclusion) that separately tracks records from databases/registers and from other sources, and documents records removed before screening. It defines what a review report must contain — search strategy, eligibility criteria, study selection, data collection, risk-of-bias assessment of included studies, synthesis methods (including any meta-analysis and certainty-of-evidence rating such as GRADE), and results — so that a reader can judge the review's conduct and reproduce its decisions. PRISMA 2020 is endorsed by hundreds of journals and is the expected reporting standard for evidence synthesis submitted to peer review, HTA bodies, and regulators.

When to use

Use PRISMA 2020 whenever you report a systematic review — including a meta-analysis of randomized trials or of observational/non-interventional studies — for a peer-reviewed journal, an HTA/payer dossier (NICE, ICER, CADTH, G-BA), an FDA or EMA submission that relies on a structured evidence synthesis, or any deliverable where the review's conclusions must withstand external scrutiny. Decision rules for choosing the right instrument: (1) For the protocol of the review (registered on PROSPERO or filed as a SAP), use the sibling PRISMA-P, not PRISMA 2020 — PRISMA 2020 governs the final report, PRISMA-P governs the plan. (2) For a scoping review, use the PRISMA-ScR extension; for searching, use PRISMA-S; for abstracts, PRISMA for Abstracts; for living, network meta-analysis, individual-participant-data, or harms-focused reviews, use the corresponding PRISMA extension. (3) PRISMA governs the synthesis; it does not govern the primary studies feeding it — a review of database (claims/EHR/registry) studies is reported with PRISMA, while each included real-world study should itself have followed STROBE/RECORD/RECORD-PE/HARPER.

What it requires

The checklist enforces transparency across the review lifecycle, much of which carries directly into real-world-evidence synthesis: pre-specified, reproducible eligibility criteria (Item 5) framed in PICO/PICOTS terms; a full, reproducible search strategy for every database with dates (Items 6–7); explicit selection and data-collection processes with number of reviewers and automation tools (Items 8–9); pre-specified outcomes and effect measures (Items 10, 12); risk-of-bias assessment of each included study using a named tool (Item 11) — e.g., RoB 2 for trials, ROBINS-I for non-randomized intervention studies — and a certainty-of-evidence rating (Item 15); a transparent synthesis description, including how studies were grouped, the meta-analytic model, heterogeneity handling, and sensitivity analyses (Items 13a–13f); the flow diagram documenting every record's fate (Item 16); and "Other information" items demanding protocol registration, funding/conflicts, and data, code, and analytic-material availability (Items 24–27). For an RWE meta-analysis these map onto the same fitness-for-use, phenotype-transparency, and estimand-alignment concerns the included studies should have addressed — PRISMA makes the synthesis author disclose them rather than letting heterogeneity in design, data source, or time-zero definition disappear into a pooled estimate.

When NOT to use — limitations and common misapplications

PRISMA 2020 is a reporting checklist, not a risk-of-bias instrument and not a quality score. Completing all 27 items certifies that the report is complete and transparent; it says nothing about whether the review was well conducted or whether the included evidence is valid — for that you need a critical-appraisal tool such as AMSTAR 2 or ROBIS applied to the review, and RoB 2 / ROBINS-I applied to the primary studies. Do not compute a "PRISMA score" by counting ticked boxes; the items are not weighted and adherence is not a quality metric. PRISMA does not make a synthesis of observational data causal — pooling confounded estimates yields a more precise confounded estimate, and PRISMA only asks you to report that you assessed bias, not to remove it. Using PRISMA 2020 for a review protocol (PRISMA-P territory), a scoping review (PRISMA-ScR), or a single primary study (STROBE/RECORD-PE/ HARPER) is the wrong-instrument error. Finally, beware checklist-as-theater: citing PRISMA and supplying a flow diagram while omitting the search strategy, the per-study risk-of-bias judgments, or the deviations from protocol defeats the purpose — reviewers increasingly request the completed item-level checklist with page references, not a blanket statement of adherence.

How it maps to this catalog

When the included evidence is real-world data, the PRISMA-reported synthesis inherits the design discipline of its component studies, which this catalog implements concept-by-concept. Eligibility, comparator, and time-zero alignment of the underlying studies are governed by active-comparator-new-user and target-trial-emulation — a meta-analysis should not pool prevalent-user and new-user designs without flagging it. Fitness-for-use and source-data adequacy of the included databases sit with claims-analysis. Outcome and exposure definitions feeding the synthesis are implemented by diagnosis-phenotype-algorithm-1ip-2op-time-window-rwe (and its PPV-validation logic, which PRISMA Item 11/13 forces you to surface). The causal-contrast and intercurrent-event framing that determines whether pooled effects are even commensurable is estimands-ate-att-intercurrent-events-rwe. Confounding control in the included studies — whose adequacy the risk-of-bias item asks you to judge — is implemented by high-dimensional-propensity-score-hdps-rwe. Attrition and missing-data reporting (the flow diagram's RWE analogue) is attrition-and-loss-to-follow-up-rwe. Applied note (claims/EHR/registry): a meta-analysis of database studies of the same drug comparison often mixes fee-for-service and Medicare-Advantage populations, divergent phenotype algorithms, and different washout/time-zero rules; PRISMA 2020 does not adjudicate these but it forces them into the open via the eligibility, risk-of-bias, and synthesis items, so heterogeneity is interrogated rather than averaged away.