How to Write a Systematic Review Manuscript

PRISMA reporting and systematic review publication
MetaSyn Academy guide to write a systematic review manuscript, illustrating the decisions documented by Systematic review manuscript template
By Dr. Esmaeel Saeedy Robat, Founder of MetaSyn Academy · Published meta-analyst (Nature Human Behaviour, 2026) · Evidence checked August 2026

Scientific writing after the review is complete

A systematic review manuscript is an argument built from a completed evidence process

The difficult part of writing a systematic review manuscript is not deciding whether the paper needs an Introduction, Methods, Results and Discussion. Those labels are familiar. The difficult part is making the sections behave like parts of the same scientific object: the Introduction asks the question that the review actually answers; the Methods explain how the evidence was obtained and synthesized; the Results report the evidence those methods produced; and the Discussion interprets those findings without exceeding their uncertainty.

A reporting checklist cannot create that coherence automatically. PRISMA 2020 specifies information that should be reported, but the checklist is not a manuscript generator and it does not establish one compulsory journal structure.12 A paper can contain each checklist item and still be difficult to follow if its Methods and Results do not correspond, its tables repeat the prose, its limitations are generic, or its conclusion uses stronger language than its own findings allow.

This is why systematic-review writing deserves to be treated as a separate methodological communication problem. Conduct, reporting and writing overlap, but they answer different questions. Conduct asks whether the review was performed appropriately. Reporting asks whether readers can see what was done. Writing determines whether the reported information forms a clear, internally consistent scientific argument.

Writing quality cannot repair methodological weakness

Clear prose can make a weak method easier to evaluate; it cannot make the method stronger. Conversely, rigorous methods can lose much of their value when the manuscript does not report or connect them clearly. The purpose of manuscript writing is therefore faithful communication, not cosmetic improvement.

IMRaD is a useful map, not a law of systematic reviews

The Introduction–Methods–Results–and–Discussion structure became dominant in major medical journals during the twentieth century. In a historical survey of four leading medical journals, Sollaci and Pereira found that IMRaD began appearing in the 1940s, reached about 80% of sampled articles in the 1970s, and became the only pattern in their sample during the 1980s.4 That history demonstrates adoption, not scientific proof that IMRaD is the uniquely optimal structure.

Its usefulness is functional. The four sections help readers locate four different kinds of information. The Introduction explains why the investigation was needed. Methods describe how the evidence was generated. Results tell readers what was found. Discussion considers what those findings mean. These questions provide a useful teaching model even when the target journal renames, merges or subdivides the sections.

PRISMA 2020 uses similar section groupings because systematic-review reporting naturally follows this logic, but that should not be misread as an IMRaD mandate. The reporting requirement survives when headings change. A journal may place registration information in Methods, Other Information or a declaration panel. Detailed search strategies may be supplementary. A conclusion may be a separate heading or the final paragraph of Discussion. The manuscript still needs to communicate the same underlying information.

Four intellectual functions of a systematic review manuscript Introduction, Methods, Results and Discussion are shown as four aligned functions: why the synthesis was needed, how the evidence was produced, what the review found and what the findings can mean. Think in functions before thinking in headings Journal headings may change; these four intellectual jobs still need to be performed. 1 Introduction Why was this synthesis needed? Rationale + objective 2 Methods How was the evidence produced? Completed review process 3 Results What did the review find? Evidence + uncertainty 4 Discussion What can the findings mean? Interpretation + scope Preserve the functions even when the journal renames, merges or subdivides the sections. Four intellectual functions of a systematic review manuscript Compact mobile display of Introduction, Methods, Results and Discussion as four distinct intellectual functions. 1 · Introduction Why was this synthesis needed? Rationale + objective 2 · Methods How was the evidence produced? Completed review process 3 · Results What did the review find? Evidence + uncertainty 4 · Discussion What can the findings mean? Interpretation + scope Headings may change; the intellectual functions remain.
Figure 1. IMRaD as a functional map. The four questions are a pedagogical model rather than a PRISMA requirement. They help authors preserve the intellectual separation among rationale, method, evidence and interpretation even when a journal uses different headings.

The Introduction should end where the review question begins

A systematic-review Introduction is easily overgrown because the subject already has a substantial literature. Authors can feel pressure to summarize every important primary study, describe every mechanism, rehearse epidemiology and explain previous reviews in detail. That often produces background without a clear reason for the present synthesis.

A stronger Introduction has a directional structure. First, establish the problem that matters. Second, summarize enough existing evidence for the reader to understand the current state of knowledge. Third, identify the uncertainty that remains. Fourth, explain why synthesis is needed now. Fifth, state the review objective explicitly. PRISMA requires the rationale and an explicit objective or question, but the precise rhetorical sequence is a scientific-writing choice rather than a checklist mandate.1

The key test is whether the objective feels inevitable by the time the Introduction reaches it. If the reader could delete the objective and replace it with a completely different review question without changing the preceding paragraphs, the rationale is probably too generic.

Existing systematic reviews deserve particular attention. A new review should not imply an evidence vacuum when previous syntheses exist. Instead, explain what makes the current review necessary: new studies, a broader or narrower population, improved methodology, a different comparison, updated certainty assessment, unresolved disagreement, a new outcome or a genuinely different review question.

Final Methods are a record of what happened, not a memorial to the protocol

Protocols are prospective. Manuscripts are retrospective reports of completed work. Their relationship should be visible but they cannot be identical because reviews encounter unforeseen evidence structures. Current Cochrane reporting guidance states directly that Methods in a completed review should be written in the past tense and describe what was done to obtain the present results and conclusions.3

That principle prevents two opposite errors. The first is copying prospective language into the final paper even when the planned method was never used. The second is rewriting the Methods so completely that readers can no longer see where the completed review differed from its plan.

Suppose a protocol planned meta-regression to explore heterogeneity, but the review ultimately included only a small number of studies. The manuscript should not report the meta-regression as though it occurred. Nor should the planned analysis simply disappear if the change is material. The completed report can state that the analysis was prespecified but not performed because the evidence base did not support it. Transparency is the objective, not artificial protocol rigidity.

The same discipline applies when an unexpected analytical decision becomes necessary. Post hoc work is not automatically illegitimate. The problem arises when it is written as though it had always been prespecified. A manuscript should allow readers to distinguish planned analysis from analysis prompted by the emerging evidence.

The strongest manuscript has a visible Methods–Results correspondence

Methods and Results are often taught as separate sections, but systematic-review writing becomes much easier when they are treated as paired structures. Selection methods produce study-selection results. Risk-of-bias methods produce risk-of-bias findings. Synthesis methods produce synthesis results. Sensitivity-analysis methods produce sensitivity results. A certainty-assessment method, when used, produces a certainty judgment.

This is not a claim that every sentence in Methods must generate a numerical result. Some methods define eligibility, describe software or explain procedures. The correspondence principle applies to major analytical and evaluative components that logically produce observable outputs. It is also not a claim about why the PRISMA developers created particular sub-items; that developer intent should not be inferred without evidence. It is a manuscript-engineering principle derived from the reporting structure itself and reinforced by current Cochrane guidance.3

Methods and Results as a manuscript mirror Four clean rows pair selection, risk of bias, synthesis and an unperformed planned analysis in Methods with the corresponding result or transparent explanation. Methods and Results should recognize each other METHODS RESULTS / EXPLANATION Selection process screening + eligibility decisions Study-selection result records → reports → included studies Risk-of-bias method tool + assessors + judgment process Risk-of-bias findings study + domain judgments Synthesis method meta-analysis or another named synthesis Synthesis result finding + uncertainty + evidence amount Planned analysis not performed do not leave silent protocol residue Explicit explanation why no corresponding result exists Methods and Results correspondence Compact mobile display of four Methods–Results correspondence pairs. METHOD → RESULT / EXPLANATION Selection process screening + eligibility Selection result records → included studies Risk-of-bias method tool + assessors Risk-of-bias findings study + domain judgments Synthesis method named analysis Synthesis result finding + uncertainty Planned analysis not performed report the change Explicit explanation why no result exists Major analytical methods should lead to a visible result or a visible explanation.
Figure 2. The correspondence principle. The important test is whether the paper allows readers to move from each major reported analytical method to its result, or to a transparent explanation when a planned analysis was not performed.

Results should organize evidence, not merely unload it

A systematic review can produce a large amount of descriptive and analytical information: thousands of screened records, characteristics of dozens of studies, many outcomes, multiple risk-of-bias judgments, subgroup analyses, sensitivity analyses and certainty assessments. Transferring all of that information into prose does not make the manuscript complete; it makes the manuscript difficult to navigate.

Current Cochrane guidance recommends a brief narrative summary of included studies rather than describing each study individually, with more detailed characteristics placed in structured tables or supplementary material.3 This illustrates a wider writing principle: the Results section should help readers understand the evidence architecture while the displays hold the detailed data.

Study-selection prose, for example, should orient the reader to the flow diagram. It does not need to convert every box in the diagram into a sentence. A characteristics paragraph can summarize the range of settings, participant populations, designs or follow-up periods that matter for interpretation while the table carries study-level details.

Synthesis prose should similarly communicate the amount of evidence, the direction and magnitude of the result, its uncertainty and the context needed to understand it. If meta-analysis is not possible, the paper still needs a named and reproducible synthesis method. The SWiM guideline exists precisely because vague labels such as “narrative synthesis” often fail to explain what authors actually did.5

One of the clearest writing errors is vote-counting by significance in narrative prose: “three studies were significant and two were not.” This loses effect magnitude, direction and uncertainty and can create misleading conclusions. Current Cochrane guidance explicitly warns against inadvertent vote counting and recommends reporting results logically regardless of statistical significance, magnitude or direction.3

Results and Discussion need a boundary, but not a wall

The conventional distinction is useful: Results report the findings; Discussion interprets them. But the rule should not be exaggerated into “Results must contain no explanatory language.” Readers still need orientation. Authors can explain which comparison is being reported, define what the estimate represents and identify why a particular display is being discussed.

The stronger boundary is between evidence description and broader inference. A Results sentence such as “the pooled estimate favored the intervention, but the confidence interval included effects ranging from negligible to clinically important benefit” still reports the result. A sentence such as “the intervention should therefore be adopted in routine practice” moves into interpretation and decision-making.

The boundary between Results and Discussion Results contain magnitude, uncertainty and evidence amount. Discussion contains context, limitations and implications. A central interpretation bridge connects them without adding new evidence. Interpretation starts from Results; it must not outrun them Results What was found? magnitude + direction precision + uncertainty amount of evidence Discussion What can it mean? context + comparison limitations + certainty implications + scope INTERPRET without adding new evidence The bridge is legitimate only when interpretive claims remain supported by the reported findings. The boundary between Results and Discussion Compact mobile display of Results, interpretation and Discussion as three separate stages. RESULTS — What was found? magnitude + direction precision + uncertainty amount of evidence INTERPRET without adding new evidence DISCUSSION — What can it mean? context + comparison limitations + certainty implications + scope Interpretation remains constrained by the evidence already reported.
Figure 3. Results–Discussion boundary. Results can include enough explanation to make findings intelligible. Broader contextualization, implications and judgments about meaning belong in Discussion, where they remain constrained by the evidence already reported.

Do not confuse “no evidence of an effect” with “evidence of no effect”

Language becomes especially important when confidence intervals include the null. Current Cochrane reporting guidance identifies the common error directly: when a confidence interval includes no effect, authors should not automatically conclude that the intervention has no effect or is no different from control. Such a conclusion requires the interval to be sufficiently narrow to exclude a meaningful effect in either direction.3

This is a writing issue because familiar statistical phrases can smuggle an interpretation into the Results. “There was no statistically significant difference” does not mean “the interventions were equivalent.” The better sentence reports the observed direction and magnitude together with the uncertainty and, where appropriate, a meaningful-difference threshold.

The same principle applies beyond conventional intervention meta-analysis. The manuscript should distinguish an uncertain estimate from a confidently negligible association. It should distinguish inconsistent evidence from evidence supporting true variation. It should distinguish absence of reported harms from evidence establishing safety.

Spin happens when the wording becomes stronger than the evidence

Yavchitz and colleagues developed a classification of 39 types of spin in systematic reviews and meta-analyses, covering misleading reporting, misleading interpretation and inappropriate extrapolation.6 The value of that taxonomy is not a single prevalence estimate. It is the recognition that distortion can enter at several stages of writing.

Selective emphasis is one route. An author may highlight a favorable secondary outcome while giving less prominence to the prespecified main outcome. Interpretation is another. A manuscript can acknowledge high risk of bias in one section and then write a confident causal conclusion as though the limitation no longer exists. Extrapolation is a third. Findings in one population can be written as if they apply to a broader population that the review never studied.

Harms require particular care. A 2024 framework for spin in harms reporting identified 12 types of harms-related spin across reporting, interpretation and extrapolation and showed why phrases implying safety can be misleading when adverse-event evidence is sparse, incompletely reported or imprecise.7 “No statistically significant increase in adverse events” is not automatically equivalent to “the intervention is safe.”

A useful editing question

For every strong adjective or recommendation in the Discussion and Conclusion, ask: Which result authorizes this wording? Then ask whether the confidence interval, risk of bias, missing evidence, certainty assessment and population scope still authorize it. If the wording cannot survive that audit, revise the wording rather than the evidence.

Certainty should constrain language without turning the manuscript into a GRADE tutorial

When a review formally assesses certainty or confidence, the writing should reflect that assessment. Cochrane recommends language such as “probably” for moderate-certainty evidence and “may” for low-certainty evidence in its GRADE-based reviews.3 Those terms are useful examples, but they should not be transformed into universal vocabulary for every evidence-synthesis methodology.

Some review types use different confidence frameworks; some do not use a formal certainty system at all. A manuscript should therefore report the framework it actually used. The broader principle is transferable: the degree of confidence in the evidence limits the degree of confidence expressed in the conclusion.

Calibrating conclusion language to the evidence Six aligned stages move from result through precision, limitations, confidence, scope and final wording. Each stage can narrow the strength of the defensible conclusion. A conclusion is the end of an evidence filter Each stage can narrow the claim that remains defensible. RESULT direction magnitude PRECISION confidence interval LIMITATIONS bias + missing evidence + variation CONFIDENCE formal certainty if assessed SCOPE population + setting outcomes WORDS claim only what survives the filter Strong prose is not confident prose by default It is prose whose confidence matches the evidence. The point estimate starts the interpretation; the full evidence determines how strongly the manuscript can speak. Calibrating conclusion language to the evidence Compact mobile sequence from result through precision, limitations, confidence and scope to final wording. 1 · RESULT direction + magnitude 2 · PRECISION confidence interval 3 · LIMITATIONS bias + missing evidence + variation 4 · CONFIDENCE formal certainty when assessed 5 · SCOPE population + setting + outcomes 6 · WORDS claim only what survives the filter Confidence in the prose should match confidence in the evidence. Each stage can narrow the defensible conclusion.
Figure 4. Conclusion calibration. The point estimate is only the beginning of interpretation. Precision, limitations, formal certainty when assessed, and the scope of the included evidence all determine how strongly the manuscript can speak.

Write two different kinds of limitations

PRISMA 2020 separates limitations of the evidence from limitations of the review process, and the distinction is reinforced by both current Cochrane and JBI guidance.138

Limitations of the included evidence

These arise from the studies and body of evidence: risk of bias, sparse data, imprecision, inconsistency, indirectness, incomplete measurement, selective reporting, limited populations or other constraints on what the evidence can support.

Limitations of the review process

These arise from choices or constraints in the synthesis itself: search limitations, unavailable reports or data, contestable eligibility decisions, unsuccessful author contact, protocol amendments, data transformations, synthesis choices or other processes that could have influenced the findings.

Mixing these categories weakens interpretation. Saying “the review was limited by small samples” is imprecise if the small samples belonged to the included primary studies. Saying “the evidence was limited by our database restrictions” reverses the same distinction in the other direction.

A limitations paragraph becomes useful when it explains consequence rather than merely names a problem. Instead of “heterogeneity was high,” ask what kind of variation existed and how it changes confidence in a single summary effect. Instead of “only English-language studies were included,” explain whether this restriction could plausibly have changed the evidence base and therefore the interpretation.

A practical writing sequence is: identify the limitation, explain its source, state which part of the interpretation it affects, and avoid claiming a direction of bias when the direction is genuinely unknown. This sequence is a reasoning framework, not an official PRISMA rule.

Implications are not the same as recommendations

Systematic reviews inform decisions, but most do not themselves perform all of the work needed to issue formal clinical or policy recommendations. Decisions can require values, preferences, feasibility, resources, equity, acceptability and implementation considerations beyond the effects synthesized by the review.

The Discussion can therefore state what the evidence implies without pretending that evidence synthesis and guideline development are the same process. A review might conclude that an intervention probably reduces symptoms but that evidence for harms is sparse. That information may be highly relevant to a guideline panel. It is not automatically equivalent to “all clinicians should use the intervention.”

Future-research implications should also be specific. “More research is needed” communicates little. A useful manuscript identifies what remains unresolved: whether evidence is absent in a particular population, whether existing studies use unreliable outcomes, whether follow-up is too short, whether imprecision is the central problem, or whether future research needs a different design rather than simply a larger number of studies.

The Conclusion is not where uncertainty disappears

Conclusions are often short, which makes every word carry disproportionate interpretive weight. A qualifier removed for brevity can turn an uncertain result into a definitive claim. A population omitted from the sentence can make a narrow result sound universal. A statement about lack of statistical significance can be transformed into a claim of equivalence.

The strongest Conclusion answers the review question using the narrowest level of confidence that remains defensible after the evidence has been considered. It does not need to repeat every effect estimate, but it should not contradict the estimates, confidence intervals or certainty judgments presented earlier.

Research on spin illustrates the danger. Yavchitz et al. identified unsupported clinical recommendations and selective emphasis among the more serious forms of spin in systematic reviews.6 In a later diagnostic-accuracy sample, multiple forms of actual and potential overinterpretation remained detectable even in high-impact journals, underscoring that prestige does not remove the need for disciplined conclusion writing.9

Supplements should extend the manuscript, not hide the methodology

Word limits make supplementary reporting necessary. Full search strategies, extended risk-of-bias explanations, excluded-study lists and additional analyses can exceed what belongs comfortably in the main text. PRISMA itself anticipates distributed reporting by allowing essential information to appear in the main report or supplementary material where appropriate.2

The mistake is to treat a supplement as a dumping ground. A reader should know why the supplement exists, what it contains and where the manuscript refers to it. The main Methods cannot say only “additional details are in the supplement” when the omitted details are essential to understanding what was done.

The same principle applies to repositories. A data-availability statement is not strengthened by an impressive-looking URL that leads to an empty project, an inaccessible private folder or a version that does not correspond to the published analyses. Availability statements should describe real availability.

Declarations are part of the scientific record

PRISMA 2020 separates support, competing interests and availability of data/code/materials into Items 25, 26 and 27.1 These statements influence how readers interpret the research process and how easily others can inspect or reuse the work. They should therefore be prepared alongside the manuscript rather than added from memory during submission.

Contributorship is related but distinct. CRediT is a standardized taxonomy of 14 contributor roles and was formalized as ANSI/NISO Z39.104-2022.10 It describes the nature of contributions; it does not determine who qualifies for authorship. Authorship criteria remain governed by the relevant journal and editorial framework.

AI disclosure adds another current layer. The January 2026 ICMJE recommendations state that authors should disclose AI-assisted technologies used in submitted work and describe how they were used. Writing assistance belongs in acknowledgments; AI used for data collection, analysis or figure generation belongs in Methods. Human authors remain responsible for accuracy, attribution and integrity, and AI tools cannot qualify as authors.11

This is an editorial recommendation rather than a new PRISMA item, and journal policies vary. A manuscript-writing workflow should therefore record the target journal’s current AI policy rather than assume every publisher implements ICMJE identically.

Adapt a stable scientific core to the journal instead of rewriting the science for every journal

Journal formatting varies: word limits, abstract formats, heading labels, table and figure caps, reference styles, supplementary-file systems, data policies and declaration placement can all differ. Those differences are real, but they do not require authors to rethink the scientific logic of the review each time the journal changes.

The practical distinction is between scientific content and submission packaging. The review still needs a rationale, transparent Methods, reportable Results, interpretation, limitations and declarations. The target journal determines where and how those elements appear.

This is why drafting directly into a highly restrictive journal format can be risky. If the journal’s limit becomes the organizing principle too early, important information may be compressed or omitted before the manuscript’s scientific structure is stable. A better workflow is to establish a coherent core, then adapt its expression to the target venue while using supplements and concise cross-referencing intelligently.

Consistency is a separate writing task

A manuscript can be individually well written at the paragraph level while still containing contradictory numbers or messages. The study count in the Results may differ from the flow diagram. A revised forest plot may contain a corrected estimate while the prose retains the previous value. A certainty rating may be downgraded without the Discussion language being revised.

The problem is especially important because many readers interact with only part of the research report. A systematic review of research on healthcare abstracts found persistent reporting problems and inconsistencies between abstracts and full texts, with Results and Conclusions among the areas showing the greatest inconsistency.12 Detailed abstract-writing solutions belong elsewhere, but the manuscript-level lesson is clear: every public surface should refer to the same review.

Analytic reproducibility is related but different. A 2024 investigation of meta-analyses in clinical psychology evaluated whether published analyses could be independently reproduced and identified data availability and reporting practices as important barriers.13 That evidence should not be confused with proof of within-manuscript numerical inconsistency. One problem concerns recalculating the analysis; the other concerns whether the paper’s own representations agree.

Final editing should move across the manuscript, not only down the manuscript

Line editing moves sentence by sentence. Consistency editing moves sideways: Methods to Results, Results to figure, figure to Discussion, Discussion to Conclusion, manuscript to supplement, manuscript to registry, and final result to abstract. Systematic-review manuscripts need both.

A practical order for writing the final manuscript

There is no single mandatory drafting sequence, but one order reduces avoidable rework. Begin by freezing the evidence displays and major analyses. Draft Methods from the completed review rather than the protocol alone. Draft Results against those Methods and displays. Use the Methods–Results correspondence check before interpretation is written. Only then write the Discussion and Conclusion, because interpretation should depend on the final results rather than an earlier analytical version.

The Introduction can be drafted earlier, but it should be revisited after the Results are stable. Reviews sometimes evolve enough during conduct that the final evidence question becomes more precise than the opening rationale. The Introduction must still lead to the objective the manuscript actually answers.

The detailed abstract should come after the full manuscript has stabilized because it compresses the review rather than defining it. The declaration and supplement package should then be verified against the final manuscript, not against an earlier submission draft.

Writing stage Main question Common failure Final check
Introduction Why was this review needed? Background without a specific evidence gap. Does the rationale lead directly to the stated objective?
Methods What was actually done? Prospective protocol language retained after methods changed. Do Methods describe the completed review and disclose material amendments?
Results What evidence did those methods produce? Selective emphasis, unexplained missing analyses or prose duplication of displays. Can each major result be traced back to a reported method?
Discussion What can the findings mean? Repetition of Results or interpretation stronger than uncertainty allows. Are interpretation and implications constrained by limitations and confidence?
Conclusion What is the narrowest defensible answer? Certainty, scope or harms disappear from the final sentence. Could every strong claim be defended directly from the Results?
Publication package Can readers inspect the complete research record? Broken cross-references, stale supplements or incomplete declarations. Have every referenced file, identifier and availability statement been verified?

Move from writing logic to manuscript construction

Build the manuscript Systematic review manuscript template

Use the working environment for section drafting, Methods–Results correspondence, display planning, supplements, declarations, journal adaptation and consistency checks.

Open the manuscript template → Place writing in context PRISMA 2020 reporting and systematic review publication

Connect manuscript writing to the broader reporting record from protocol and review conduct to publication, supplements and interpretation.

Explore the reporting framework → Continue with reporting guidance PRISMA reporting and publication guides

Browse methodological articles focused on transparent publication of completed evidence syntheses.

See reporting guides → Use the wider toolkit Free systematic review and meta-analysis templates

Access practical resources for the wider review workflow without turning manuscript writing into a substitute for review conduct.

Browse practical resources →

Frequently asked questions

What is the difference between a PRISMA checklist and a manuscript template?

A PRISMA checklist identifies reporting information that should be present. A manuscript template provides a working architecture for organizing and drafting that information as a coherent scientific report. Checklist completeness and manuscript coherence overlap, but they are not the same problem.

Does a systematic review have to follow Introduction–Methods–Results–Discussion?

No universal reporting rule requires those exact headings. IMRaD is a dominant scientific-writing convention and a useful functional model, while target journals may rename, merge or subdivide sections. PRISMA concerns the information reported rather than one mandatory heading taxonomy.

Should the Methods section describe the protocol or the completed review?

The final manuscript should describe what was actually done. The protocol remains the prospective reference point, so material amendments and planned methods that were not implemented should be reported transparently rather than silently removed or retained as though they occurred.3

Must every Method have a corresponding Result?

Not every procedural sentence generates a result. The correspondence principle applies to major analytical and evaluative components that logically produce outputs. When a planned analysis could not be conducted, the manuscript should explain why rather than fabricate or imply a result.

How much interpretation belongs in Results?

Results need enough orienting language to make estimates and evidence displays understandable. Broader interpretation, explanation in relation to other evidence, implications and recommendations belong mainly in Discussion. The important boundary is between describing evidence and extending its meaning.

What is spin in a systematic review?

Spin is reporting, interpretation or extrapolation that makes findings appear more favorable or certain than the evidence supports. It can include selective emphasis, unsupported practice recommendations, causal overstatement or downplaying uncertainty and harms.67

Should a systematic review make recommendations for clinical practice or policy?

A review can discuss implications, but formal recommendations often require additional considerations such as values, resources, feasibility, equity and implementation. Authors should distinguish an evidence-synthesis conclusion from a guideline or policy recommendation.

What is the difference between limitations of the evidence and limitations of the review?

Limitations of the evidence arise from the included studies and body of evidence. Limitations of the review arise from the review process itself, such as search constraints, unavailable information or contestable analytical decisions. PRISMA and current Cochrane/JBI guidance treat these as distinct reporting problems.

Does CRediT determine who qualifies as an author?

No. CRediT is a standardized taxonomy for describing 14 types of contribution. It improves contributorship transparency but does not determine authorship eligibility, which depends on the journal’s applicable authorship criteria.10

Can AI be used to help write a systematic review manuscript?

Journal policies differ. Current ICMJE guidance allows AI-assisted technologies to be used with human responsibility and disclosure, but AI cannot be listed as an author. Authors remain responsible for accuracy, originality, attribution and verification of generated content.11

Coherence is the final manuscript method

Systematic-review writing is sometimes treated as the easy part that follows the methodological work. It is not. The manuscript determines whether another reader can see the logic of the review without reconstructing it from scattered files and assumptions.

The strongest manuscript does not sound more certain than the evidence. It does not use structure to conceal deviations or word limits to justify missing information. It makes the completed review legible: the question leads to the methods, the methods lead to the findings, the findings lead to a bounded interpretation, and the declarations and supplementary objects remain connected to the same scientific record.

Methodological guidance and empirical evidence

  1. Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ. 2021;372:n71. https://doi.org/10.1136/bmj.n71
  2. Page MJ, Moher D, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, et al. PRISMA 2020 explanation and elaboration: updated guidance and exemplars for reporting systematic reviews. BMJ. 2021;372:n160. https://doi.org/10.1136/bmj.n160
  3. Cumpston M, Lasserson T, Flemyng E, Page MJ. Chapter III: Reporting the review [updated July 2026]. In: Higgins JPT, Thomas J, Chandler J, Cumpston M, Li T, Page MJ, Welch VA, editors. Cochrane Handbook for Systematic Reviews of Interventions. Version 6.5. Cochrane; 2026. https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-iii
  4. Sollaci LB, Pereira MG. The introduction, methods, results, and discussion (IMRAD) structure: a fifty-year survey. J Med Libr Assoc. 2004;92(3):364-371. https://pmc.ncbi.nlm.nih.gov/articles/PMC442179/
  5. Campbell M, McKenzie JE, Sowden A, Katikireddi SV, Brennan SE, Ellis S, et al. Synthesis without meta-analysis (SWiM) in systematic reviews: reporting guideline. BMJ. 2020;368:l6890. https://doi.org/10.1136/bmj.l6890
  6. Yavchitz A, Ravaud P, Altman DG, Moher D, Hróbjartsson A, Lasserson T, Boutron I. A new classification of spin in systematic reviews and meta-analyses was developed and ranked according to the severity. J Clin Epidemiol. 2016;75:56-65. https://doi.org/10.1016/j.jclinepi.2016.01.020
  7. Qureshi R, Naaman K, Quan NG, Mayo-Wilson E, Page MJ, Cornelius V, et al. Development and evaluation of a framework for identifying and addressing spin for harms in systematic reviews of interventions. Ann Intern Med. 2024;177(8):1089-1098. https://doi.org/10.7326/M24-0771
  8. Joanna Briggs Institute. JBI Manual for Evidence Synthesis: Discussion. Updated May 2025. https://jbi-global.atlassian.net/wiki/spaces/MANUAL/pages/355828977/4.4.9+Discussion
  9. McGrath TA, Bowdridge JC, Prager R, Frank RA, Treanor L, Dehmoobad Sharifabadi A, et al. Overinterpretation of research findings: evaluation of “spin” in systematic reviews of diagnostic accuracy studies in high-impact factor journals. Clin Chem. 2020;66(7):915-924. https://doi.org/10.1093/clinchem/hvaa093
  10. National Information Standards Organization. ANSI/NISO Z39.104-2022: CRediT, Contributor Roles Taxonomy. Baltimore: NISO; 2022. https://doi.org/10.3789/ansi.niso.z39.104-2022
  11. International Committee of Medical Journal Editors. Recommendations for the Conduct, Reporting, Editing, and Publication of Scholarly Work in Medical Journals. Updated January 2026. https://www.icmje.org/recommendations/
  12. Nascimento DP, Ostelo RWJG, van Tulder MW, Gonzalez GZ, Araujo AC, Vanin AA, Costa LOP. Do not make clinical decisions based on abstracts of healthcare research: a systematic review. J Clin Epidemiol. 2021;135:136-157. https://doi.org/10.1016/j.jclinepi.2021.03.030
  13. López-Nicolás R, Lakens D, López-López JA, Rubio-Aparicio M, Sandoval-Lentisco A, López-Ibáñez C, et al. Reproducibility of published meta-analyses on clinical-psychological interventions. Adv Methods Pract Psychol Sci. 2024;7(1). https://doi.org/10.1177/25152459231202929

Similar Posts

Leave a Reply

Your email address will not be published. Required fields are marked *