PRISMA is the checklist for systematic reviews and meta-analyses. This walks the items that make a review reproducible — the search, the selection flow, the risk-of-bias assessment — with a self-check for each.
PRISMA — Preferred Reporting Items for Systematic reviews and Meta-Analyses — is the 27-item checklist and flow diagram ↗ (in its 2020 revision) for systematic reviews, maintained through the EQUATOR Network ↗. A systematic review claims to be a reproducible census of the evidence, so PRISMA exists to make that claim checkable: a reader should be able to see exactly what you searched, what you excluded and why, and how you weighed what remained. Many journals require the completed checklist and the flow diagram at submission.
Register the protocol and report the registration (item 24a). A review pre-registered in PROSPERO is one a reader can trust was not reshaped around the results. State the eligibility criteria (item 5) precisely, and report the full search (items 6–7) — databases, dates, and the complete search string for at least one database, verbatim.
Self-check. Could another team re-run your search from what is printed and arrive at the same set of records? If the search string is paraphrased rather than reproduced, the central promise of a systematic review — reproducibility — is unmet.
The PRISMA flow diagram (item 16a) is the review's backbone: records identified, deduplicated, screened, excluded with reasons, and included. The numbers must reconcile down the diagram and match the text. Describe the selection and data-collection process (items 8–9) — how many reviewers, working independently, and how disagreements were resolved.
Self-check. Do the counts in your flow diagram add up at every stage, and does the final included count match the number of studies in your evidence tables? A mismatch here is a deterministic catch, the kind recomputed automatically in how journals catch statistical errors.
Assess risk of bias in each included study (item 11) with a named tool, and carry that assessment into the conclusions — a pooled estimate built mostly on high-risk studies is not the same evidence as one built on low-risk trials. Describe the synthesis methods (item 13): how effects were pooled, the model used, and — critically — how heterogeneity was quantified and handled. A meta-analysis that pools highly heterogeneous studies into a single number can manufacture false precision.
Self-check. If between-study heterogeneity is high, does your write-up temper the pooled estimate accordingly, or does it report a tidy summary effect as if the studies agreed? Overstating precision from a heterogeneous pool is the meta-analytic cousin of reading an imprecise result as a firm one — see our colectomy cohort case study.
Assess reporting and publication bias (item 14) — small-study effects, funnel-plot asymmetry where enough studies allow it — and rate the certainty of evidence (item 15), commonly with GRADE. Keep the conclusion (item 23) inside that certainty: a review of low-certainty evidence should not close with a high-certainty recommendation.
Self-check. Does your discussion state the certainty of the body of evidence before it states what to do about it? A confident recommendation resting on low-certainty evidence is the systematic review's signature over-claim.
Complete the official 2020 checklist and draw the flow diagram before submission — and use the matching extension (PRISMA-P for protocols, PRISMA-ScR for scoping reviews) where it applies. A structured pre-submission review checks them alongside the statistics: RigorMD maps a review to PRISMA item by item as one of six scored domains, recomputes the reported pooled estimates and flow-diagram counts deterministically, and has two independent engines assess whether the conclusion is calibrated to the certainty of the evidence. It flags; it does not certify, and it does not replace peer review or a statistician's input on design.
See a full sample report →, read how the engine works, or review pricing — the pre-submission review is $25. For other designs, see the STROBE, CONSORT, and TRIPOD walkthroughs.