Protocol
Pre-specify the decisions that should not be driven by observed results, then preserve versions and amendments.
Sources and scope · Guide reviewed September 22, 2026
- Coverage
- Health-intervention systematic reviews: protocol control, search, selection, study linkage, appraisal, synthesis, and reporting
- Source basis
- PRISMA, PROSPERO, Cochrane, and PRESS primary method sources
Prepared by the Manusights editorial team. Publisher and institutional instructions take precedence.
Evidence synthesis desk
Control the review.
Preserve the trail.
A systematic review is not a checklist completed at the end. It is a chain of decisions that should remain traceable from the protocol to every reported claim.
7 primary method sources8 control gates7 evidence handoffsEdition 2026.09.22-review-1
Five-page protocol-to-reporting record
Systematic review control map
Record review identity, protocol changes, search coverage, selection flow, study linkage, appraisal, synthesis decisions, and reporting locations. Licensed CC BY 4.0 for the original Manusights framework.
- Edition
- 2026.09.22-review-1
- Format
- License
- CC BY 4.0
- Evidence reviewed
- September 22, 2026
Method boundary
Reporting guidance does not certify the conduct
PRISMA 2020 provides a reporting checklist, expanded explanations, and flow diagrams. It helps a reader see what was done; it does not choose the review design, search coverage, appraisal tool, or synthesis method. Source A
Conduct
Use current method-specific guidance appropriate to the question, eligible evidence, and exact decision.
Reporting
Map the final record to the relevant PRISMA guidance and expose what changed, remains unavailable, or cannot be concluded.
Registration boundary: PROSPERO is for eligible systematic reviews with health-related outcomes. Check current eligibility before assuming it controls the project. Registration and a full protocol are useful linked records, not interchangeable documents. Source C
Original Manusights framework
Eight gates preserve the chain of evidence
Complete the gates as control records, not prose sections. Each gate has a question, a bounded record, and evidence that another team member can inspect.
- 1
Review identity
What review is this, and what evidence unit will answer it?
Control record
Name the review type, objective, eligibility concepts, eligible designs, outcomes, time points, settings, and the distinction between a study and its multiple reports.
Evidence of completion
A current protocol section and a one-sentence boundary explaining what this review will not answer.
- 2
Protocol and change control
Which decisions were made before results could influence them?
Control record
Record the protocol version, date, public location, eligible registry and identifier when applicable, amendment log, decision owner, and reason for each change.
Evidence of completion
A retrievable protocol or registration record plus dated amendments that distinguish prospective decisions from later adaptations.
- 3
Search coverage
Why should these sources and strategies recover the eligible evidence?
Control record
For every database, register, repository, citation route, or other source, record the rationale, platform, exact strategy, run date, coverage limits, restrictions, and peer-review status.
Evidence of completion
Re-runnable strategies, export counts, dates, and a review record from an information specialist or other qualified search peer when used.
- 4
Selection record
Can every included and excluded study be traced through the flow?
Control record
Define deduplication, report-to-study linkage, pilot rules, reviewer roles, disagreement resolution, automation use, full-text reason vocabulary, and count reconciliation.
Evidence of completion
A decision log in which flow counts reconcile and potentially relevant full texts have a specific primary exclusion reason.
- 5
Extraction and study linkage
Can each synthesized value be traced to a study and source report?
Control record
Predefine fields, outcome and time-point rules, report-to-study mapping, extraction roles, transformations, author queries, corrections, and form changes.
Evidence of completion
A piloted form, provenance for each value, and a query or correction log that preserves the original and resolved record.
- 6
Risk of bias
Does the appraisal tool match the design and result being assessed?
Control record
Name the exact tool and version, eligible design, result or outcome assessed, assessor process, supporting information, judgments, rationale, and unresolved disagreements.
Evidence of completion
Completed domain-level records tied to the appropriate result. RoB 2 is specifically for randomized trials and is not a generic quality score.
- 7
Synthesis decision
What combination is scientifically meaningful, and what assumptions does it require?
Control record
Predefine synthesis groups, effect measures, data preparation, missing-data handling, model assumptions, heterogeneity assessment, sensitivity analyses, and the rule for not pooling.
Evidence of completion
A decision record that explains every synthesis, non-synthesis, subgroup, and departure without choosing a model from an I-squared threshold alone.
- 8
Reporting and open materials
Can a reader reconstruct what was planned, done, changed, found, and shared?
Control record
Map PRISMA items to locations and record the flow diagram, registration, protocol, amendments, support, conflicts, extraction materials, analysis data, code, and limitations.
Evidence of completion
A completed location map and working public or controlled-access locators, with unavailable materials named rather than implied.
Evidence handoffs
A review fails where information disappears between stages
Audit the payload that must move forward. The next stage should not have to reconstruct the previous stage from memory.
Question Protocol
- Must carry forward
- Review type, objective, eligibility, outcomes, study designs, and planned synthesis logic.
- Failure signal
- The protocol contains generic methods that do not operationalize the question.
Protocol Search
- Must carry forward
- Every eligible evidence type and justified restriction, translated into source coverage and search concepts.
- Failure signal
- Databases are chosen from a default list rather than the eligibility criteria.
Search Selection
- Must carry forward
- Complete exports, run dates, source labels, deduplication provenance, and report identifiers.
- Failure signal
- Counts cannot be reconciled or reports silently disappear during deduplication.
Selection Extraction
- Must carry forward
- Stable study identities, linked reports, inclusion rationale, and full-text evidence.
- Failure signal
- Multiple reports are treated as multiple studies or extraction begins from an unverified citation list.
Extraction Appraisal
- Must carry forward
- Outcome-specific values, source locations, transformations, and design details needed for the selected tool.
- Failure signal
- A global study score replaces result-level reasoning or necessary design information was never extracted.
Appraisal Synthesis
- Must carry forward
- Domain judgments, rationale, uncertainty, and sensitivity-analysis implications.
- Failure signal
- Risk-of-bias judgments appear in a table but do not affect synthesis or interpretation.
Synthesis Report
- Must carry forward
- Exact analysis set, effect measures, assumptions, heterogeneity, sensitivity results, deviations, and unavailable evidence.
- Failure signal
- The manuscript reports a pooled number without preserving why those studies were combined.
Failure patterns
Four shortcuts that break the evidence trail
Reporting replaces conduct
Failure: A completed PRISMA checklist is treated as proof that the methods were valid.
Repair: Use PRISMA to report what happened, then separately justify conduct against the protocol and method-specific guidance.
Registry overreach
Failure: PROSPERO is presented as the universal registry for every review type.
Repair: Check current eligibility first; record another public protocol route when the review is outside scope.
Report-study conflation
Failure: Several publications from one study are counted or synthesized as separate studies.
Repair: Maintain a report-to-study map and choose the result source deliberately.
Automatic pooling
Failure: A meta-analysis is performed because numerical data exist, or a model is chosen from an I-squared threshold.
Repair: Judge whether the synthesis is meaningful, state model assumptions, and preserve a justified no-pooling route.
Method decisions
Search, appraisal, and synthesis need separate warrants
Search coverage follows eligibility
Cochrane instructs intervention-review teams to plan source coverage from the eligibility criteria, preserve strategies and dates, search beyond bibliographic databases where appropriate, and work with an information specialist. PRESS can structure peer review of an electronic strategy, but it does not select the sources for you. Source D Source E
Open the search strategy worksheetRisk of bias is not a global quality score
RoB 2 uses domains and signalling questions for randomized trials, with judgments of low risk, some concerns, or high risk. Other eligible designs require tools whose scopes match those designs. Record the tool version, result assessed, supporting locations, rationale, and assessor process. Source F
A random-effects model does not repair incompatibility
Cochrane warns that meta-analysis can mislead when study designs, biases, diversity, and missing evidence are not handled. Model choice should not be made from a heterogeneity test alone, and a systematic review may legitimately report no meta-analysis. Source G
Publication audit
Eight tests before calling the record reproducible
Run the tests independently. Passing a reporting check does not erase a prospective-control, search, linkage, appraisal, or synthesis failure.
Guidance fit
Pass: The review type maps to the relevant reporting extension and method-specific guidance.
Repair if: PRISMA 2020 is treated as one universal method for every evidence synthesis.
Prospective control
Pass: Protocol versions and amendments reveal which decisions preceded access to results.
Repair if: Later decisions are presented as if they were specified from the start.
Search reproducibility
Pass: A qualified reader can rerun every strategy on the named platform and date.
Repair if: Only keywords, database names, or total hits are retained.
Flow reconciliation
Pass: Source exports, duplicates, screened reports, full texts, exclusions, reports, and studies reconcile.
Repair if: The flow diagram is assembled after writing and cannot be traced to decisions.
Study linkage
Pass: All reports for one study are linked and each extracted result has a stable source locator.
Repair if: Citation rows are mistaken for independent study units.
Appraisal fit
Pass: The exact risk-of-bias tool, version, target result, and judgment rationale are preserved.
Repair if: A convenience checklist or global quality score is applied across incompatible designs.
Synthesis warrant
Pass: Every pooling or non-pooling decision names scientific comparability, assumptions, and uncertainty.
Repair if: Software defaults or a heterogeneity threshold make the decision.
Report traceability
Pass: Checklist locations, protocol, amendments, data, code, forms, and limitations have working locators or explicit availability statements.
Repair if: The report claims reproducibility while key materials are unnamed or inaccessible.
Evidence and maintenance
Source record and limits
The web guide, PDF, and source record share one edition. The framework links reporting, conduct, and project control without pretending that one source governs every review type.
- APRISMA 2020
PRISMA Executive / Statement, checklist, expanded checklist, and flow diagrams. Supports: Reporting-system boundary, checklist and flow-diagram routes, explanation and elaboration, original-versus-updated review templates, and the need to select relevant extensions.
- BPRISMA for Protocols
PRISMA Executive / PRISMA-P statement and checklist. Supports: Development and reporting of systematic review protocols and the distinction between protocol guidance and final-review reporting.
- CPROSPERO
Centre for Reviews and Dissemination, University of York / Registry purpose and current registration route. Supports: Prospective registration for eligible systematic reviews with health-related outcomes, transparency, reporting-bias reduction, and duplication awareness.
- DCochrane Handbook Chapter 4: Searching for and selecting studies
Cochrane / Version 6.5.1; sections 4.2-4.6. Supports: Question-led source selection, information-specialist collaboration, comprehensive searching, documentation, report-to-study linkage, selection records, and limits of automation.
- EPRESS Peer Review of Electronic Search Strategies: 2015 Guideline
Canada's Drug Agency / Guideline statement and explanation and elaboration. Supports: Structured peer review of electronic search strategies and a traceable review record; it is not a substitute for selecting appropriate sources.
- FRoB 2: revised Cochrane risk-of-bias tool for randomized trials
Cochrane Bias Methods Group / Tool scope, domains, signalling questions, and judgments. Supports: RoB 2's randomized-trial scope, domain structure, signalling questions, judgment categories, and the need to use the exact current tool and guidance.
- GCochrane Handbook Chapter 10: Analysing data and undertaking meta-analyses
Cochrane / Version 6.5.1; analysis and heterogeneity sections. Supports: Effect-measure and model assumptions, uncertainty, heterogeneity, sensitivity analysis, the option not to pool, and the warning not to choose a model from a heterogeneity test.
Boundary: A planning, change-control, and reporting record for health-intervention systematic reviews. Methods must still be selected for the exact question, eligible study designs, field, protocol, team, and controlling guidance. Recheck all seven sources after a material guidance change and no later than 90 days after this review. Recheck registry eligibility, journal instructions, and every method-specific tool at the time of use.
Common questions
Separate the control records
Is PRISMA a method for conducting a systematic review?
PRISMA is a reporting guideline. Its checklist and flow diagrams help authors report why the review was done, what methods were used, and what was found. They do not replace a protocol or method-specific conduct guidance. The control map therefore records reporting and conduct separately.
Must every systematic review be registered in PROSPERO?
No. PROSPERO is a prospective registry for eligible systematic reviews with health-related outcomes. Confirm the current eligibility criteria before relying on it. If a review is outside scope, preserve the protocol through another suitable public route and report the registration or protocol status accurately.
Does every systematic review need a meta-analysis?
No. A systematic review may conclude that quantitative pooling would not provide a meaningful summary. The decision should follow the question, eligible designs, outcomes, clinical and methodological diversity, and protocol, not the availability of software or a single heterogeneity statistic.
Can one risk-of-bias tool be used for every included study?
Not safely. Tool scope must match the eligible design and the result being assessed. For example, Cochrane RoB 2 is specifically for randomized trials. Mixed-design reviews may need more than one current instrument and an explicit plan for using the resulting judgments.
Cite this resource
Citation formats
Free to use and cite, with attribution. See the reuse policy for what you may republish.
- APA
- Manusights. (2026). Systematic review control map. Edition 2026.09.22-review-1. https://manusights.com/resources/systematic-reviews
- MLA
- Manusights. "Systematic Review Control Map." Manusights, 22 Sept. 2026, https://manusights.com/resources/systematic-reviews.
- Vancouver
- Manusights. Systematic review control map [Internet]. 2026 Sep 22. Available from: https://manusights.com/resources/systematic-reviews