Systematic Review Report Template (PRISMA 2020)

Completed systematic review report: PRISMA 2020 items, flow diagram, RoB 2 table, meta-analysis results, reporting bias, GRADE summary of findings

科研完整PRISMA 2020 (27 items, BMJ 2021;372:n71); RoB 2; GRADEsystematic review report diagram math更新于 2026年10月29日

在编辑器中打开下载 .md

使用方法: Use it when writing up a finished systematic review or meta-analysis against PRISMA 2020. Fill the flow diagram counts first, then the risk-of-bias and synthesis tables, and write the abstract last. The main pitfall is numbers that do not reconcile between the flow diagram, the study table and the meta-analyses.

预览

Systematic Review Report: [Review title]

Report template for a completed systematic review, with or without meta-analysis. The sections follow the topics and order of the PRISMA 2020 statement (27 items, published in BMJ 2021;372:n71), including the flow diagram, risk-of-bias tables and a GRADE summary of findings. PRISMA item numbers appear in brackets, for example [16a], so you can fill the official checklist quickly. Topics are written in original words; use the template with the official PRISMA 2020 checklist, abstract checklist and flow diagram from prisma-statement.org. This is a reporting format, not medical, statistical or methodological advice. Replace every [bracketed] value and every row marked Example.

Document control

Field	Value
Document ID	[SRR-001]
Review title	[Full title as registered]
Registration	[PROSPERO CRD number / OSF Registries DOI]
Protocol version reported against	[Protocol v1.2, YYYY-MM-DD]
Review type	[Systematic review with meta-analysis / without meta-analysis / update of an earlier review]
Reporting guideline	PRISMA 2020 [plus extensions used, for example PRISMA-S for searches, SWiM, PRISMA-NMA]
Date of last search	YYYY-MM-DD
Version	[0.1]
Status	[Draft / In review / Approved]
Owner	[@lead-reviewer]
Approvers	[@second-reviewer], [@statistician], [@information-specialist]
Last updated	YYYY-MM-DD
Classification	[Confidential until publication]

Approvals

Role	Name	Decision	Date
Lead reviewer	[@name]	[Approved]	YYYY-MM-DD
Statistician	[@name]	[Approved]	YYYY-MM-DD
Information specialist	[@name]	[Approved]	YYYY-MM-DD
Content expert or guarantor	[@name]	[Approved]	YYYY-MM-DD

Revision history

Version	Date	Author	Change summary
0.1	YYYY-MM-DD	[@name]	First full draft after synthesis
0.2	YYYY-MM-DD	[@name]	Example: Search rerun before submission; one new trial screened and excluded
1.0	YYYY-MM-DD	[@name]	Approved for journal submission

Title and abstract
Introduction
Registration, support and availability
Methods
Results
Discussion
Summary of findings
PRISMA cross-reference, glossary and references

1. Title and abstract

The title and abstract let readers find the review and judge it quickly.

1.1 Title [1]

[Intervention] for [condition] in [population]: a systematic review [and meta-analysis]

Example: "Supervised exercise therapy for chronic low back pain in adults: a systematic review and meta-analysis of randomised trials."

1.2 Structured abstract [2]

Heading	Content
Objectives	[Main question in one sentence]
Eligibility criteria	[Population, intervention, comparator, outcomes, designs]
Information sources	[Databases and registers, date of last search]
Risk of bias	[Tool used, for example RoB 2]
Synthesis	[Model, effect measure]
Results	Example: 14 randomised trials (2,442 participants). Pain at about 12 weeks: mean difference -1.17 points on a 0 to 10 scale (95% CI -1.55 to -0.78; 12 trials, 2,172 participants; low certainty). Adverse events: risk ratio 1.45 (1.07 to 1.96; 8 trials; low certainty).
Limitations	[Main limitations of the evidence and of the review process]
Interpretation	[One balanced sentence]
Funding and registration	[Funder; registration number]

2. Introduction

The introduction states the problem, the gap and the exact questions.

2.1 Rationale [3]

[Describe the condition, its burden and current practice. Summarise earlier reviews and explain the uncertainty this review resolves.]

Example: The most recent review of exercise for chronic low back pain searched to 2017 and did not assess risk of bias with RoB 2. Several large trials have been published since.

2.2 Objectives [4]

ID	Objective
OBJ-01 (Example)	Estimate the effect of supervised exercise therapy compared with usual care or no treatment on pain intensity at about 12 weeks in adults with chronic low back pain
OBJ-02 (Example)	Estimate the effect on the risk of adverse events
OBJ-03	[Further objective, for example function or return to work]

3. Registration, support and availability

These items let readers compare the review with its plan and find the underlying data and code.

3.1 Registration and protocol [24a, 24b, 24c]

Item	Detail
Registry and number [24a]	[PROSPERO CRD number / OSF DOI], registered YYYY-MM-DD
Where the protocol can be found [24b]	[Registry record, journal reference or repository link]
Amendments to registration or protocol [24c]	See section 4.11

3.2 Support and competing interests [25, 26]

Item	Detail
Funding and other support [25]	[Funder, grant number]
Role of funders [25]	Example: The funder had no role in design, conduct, analysis, interpretation or the decision to publish
Competing interests [26]	[Declarations per author. Name any reviewer who authored an included study and how that study was handled.]

3.3 Availability of data, code and other materials [27]

Material	Available?	Where
Data extraction forms	[Yes / No]	[Repository link]
Extracted data	Example: Yes	[Repository DOI]
Analysis code	Example: Yes	[Code repository, tagged release]
Full search strategies	Yes	Appendix A
Risk-of-bias judgements with support	[Yes]	Appendix C

4. Methods

The methods follow the order of the PRISMA 2020 methods items.

4.1 Eligibility criteria [5]

Element	Criterion
Population	Example: Adults aged 18 or over with low back pain lasting 12 weeks or more
Intervention	Example: Exercise therapy supervised by a health professional
Comparator	Example: Usual care, waiting list or no treatment
Outcomes	Example: Pain intensity (primary), function, adverse events
Study design	Example: Randomised controlled trials
Report characteristics	[Years, languages, publication status, registry results included]
Grouping for synthesis	[For example by comparator type: usual care versus no treatment]

4.2 Information sources [6]

Source	Platform or URL	Date last searched
MEDLINE	[Ovid]	YYYY-MM-DD
Embase	[Ovid / Elsevier]	YYYY-MM-DD
CENTRAL	[Cochrane Library]	YYYY-MM-DD
Trial registries	[ClinicalTrials.gov, WHO ICTRP]	YYYY-MM-DD
Other methods	[Reference lists, citation searching, contact with authors]	YYYY-MM-DD

4.3 Search strategy [7]

[State that full strategies for every database, register and website, with filters and limits, appear in appendix A. Name who designed and who peer reviewed the search, for example with the PRESS checklist.]

4.4 Selection process [8]

[Number of reviewers who screened each record and each report, whether they worked independently, how disagreements were resolved, and any automation tool used, with how it was used and validated.]

4.5 Data collection process [9]

[Number of reviewers who extracted data, independence, how missing data were obtained from investigators, and any automation tool.]

4.6 Data items [10a, 10b]

ID	Outcome	Measures accepted	Time point and selection rule
OUT-01 (Example)	Pain intensity	Any validated 0 to 10 or 0 to 100 scale, converted to 0 to 10	Closest to 12 weeks within 6 to 26 weeks
OUT-02	[Function]	[Instruments]	[Time point]
OUT-03 (Example)	Adverse events	Participants with any adverse event	End of intervention

Other variables [10b]: [age, sex, symptom duration, intervention dose and provider, comparator detail, setting, funding source]. State any assumptions made about missing or unclear information.

4.7 Study risk of bias assessment [11]

[Tool and version, for example RoB 2 (version of 22 August 2019) for randomised trials, assessed per outcome; number of assessors and independence; how overall judgements were reached.]

4.8 Effect measures [12]

Outcome type	Effect measure
Continuous, same scale	Mean difference (MD)
Continuous, different scales	[Standardised mean difference (SMD), Hedges' g]
Binary	[Risk ratio (RR), with absolute risk difference for the summary of findings]

4.9 Synthesis methods [13a to 13f]

Sub-item	Method
Eligibility for each synthesis [13a]	Example: All trials with usable pain data at 6 to 26 weeks
Data preparation [13b]	[Conversion of scales, imputation of missing SDs, handling of multi-arm trials]
Tabulation and display [13c]	[Forest plots, characteristics tables]
Model and software [13d]	Example: Random-effects, REML estimate of $\tau^2$ , Hartung-Knapp-Sidik-Jonkman confidence interval; [software and version]
Heterogeneity [13e]	[ $I^2$ , $\tau^2$ , prediction interval; prespecified subgroups]
Sensitivity analyses [13f]	[Excluding high risk of bias; alternative model]

Heterogeneity is summarised with Cochran's $Q$ and

I^2 = \max\left(0,\ \frac{Q - df}{Q}\right) \times 100\%

with $df = k - 1$ for $k$ studies.

4.10 Reporting bias and certainty assessment [14, 15]

[Methods for assessing missing results (for example comparison with registry records, funnel plots and Egger's test when 10 or more studies contribute) and for rating certainty of evidence per outcome (for example GRADE, using the Core GRADE guidance).]

4.11 Differences from the protocol

ID	Change	Date	Reason	Before results known?
DEV-01 (Example)	Added a regional database	YYYY-MM-DD	Search peer review found trials indexed only there	Yes
DEV-02 (Example)	Used REML instead of DerSimonian-Laird	YYYY-MM-DD	Methods guidance updated; prespecified in the analysis plan	Yes

5. Results

This section reports what was found and how much confidence it deserves.

5.1 Study selection [16a, 16b]

The flow diagram traces records from identification to inclusion.

Stage	Databases and registers	Other methods
Records identified	Example: 2,056 (1,842 + 214)	23
Duplicates removed	618	Not applicable
Records screened	1,438	23
Records excluded at title and abstract	1,312	0
Reports sought / not retrieved	126 / 4	23 / 1
Reports assessed	122	22
Reports excluded	109	20
Reports included	13 (12 studies)	2 (2 studies)

Example: reports excluded at full text, with reasons: wrong population 38, wrong intervention 29, wrong comparator 17, not randomised 20, no usable outcome data 5 (total 109). Reports excluded from other methods: not randomised 12, wrong intervention 8 (total 20). The list of excluded reports with reasons is in appendix B.

5.2 Study characteristics [17]

Study	Country	n (int / ctl)	Intervention	Comparator	Pain measure	Follow-up
ST-01 (Example)	[Country]	60 / 58	Group exercise, 8 weeks	Usual care	NRS 0 to 10	12 weeks
ST-04 (Example)	[Country]	200 / 196	Individual physiotherapy exercise, 12 weeks	Usual care	VAS 0 to 100	12 weeks
ST-13 (Example)	[Country]	80 / 78	Pilates, 10 weeks	Waiting list	Not usable (median only, no spread)	10 weeks
[ST-nn]	[Country]	[n / n]	[Intervention]	[Comparator]	[Measure]	[Time]

Example: 14 trials randomised 2,442 participants. Twelve trials (1,090 intervention, 1,082 comparator, 2,172 in total) reported usable pain data; ST-13 and ST-14 (270 participants) did not.

5.3 Risk of bias in studies [18]

RoB 2 domains: D1 randomisation process, D2 deviations from intended interventions, D3 missing outcome data, D4 measurement of the outcome, D5 selection of the reported result. Judgements: Low, Some concerns, High. Outcome assessed: pain intensity (OUT-01).

Study	D1	D2	D3	D4	D5	Overall
ST-01 (Example)	Low	Low	Low	Some concerns	Low	Some concerns
ST-02	Low	Low	Low	Low	Low	Low
ST-03	Some concerns	Low	High	Some concerns	Low	High
ST-04	Low	Low	Low	Low	Low	Low
ST-05	High	Low	Some concerns	Some concerns	Low	High
ST-06	Low	Low	Low	Some concerns	Low	Some concerns
ST-07	Low	Low	Low	Low	Low	Low
ST-08	Low	Some concerns	Low	Some concerns	Low	Some concerns
ST-09	Low	Low	Low	Some concerns	Low	Some concerns
ST-10	Low	Low	Low	Some concerns	Low	Some concerns
ST-11	Some concerns	Low	Low	Some concerns	High	High
ST-12	Low	Low	Low	Low	Low	Low

Example: 4 trials at low risk, 5 with some concerns and 3 at high risk of bias for pain. The common concern (D4) was self-reported pain by participants who knew their allocation.

5.4 Results of individual studies [19]

Study	MD (95% CI), 0 to 10 scale	Weight (%)	Overall risk of bias
ST-01 (Example)	-1.60 (-2.42 to -0.78)	7.5	Some concerns
ST-02	-0.90 (-1.45 to -0.35)	10.0	Low
ST-03	-2.10 (-3.18 to -1.02)	5.7	High
ST-04	-0.60 (-0.99 to -0.21)	11.5	Low
ST-05	-1.90 (-3.08 to -0.72)	5.1	High
ST-06	-1.20 (-1.89 to -0.51)	8.7	Some concerns
ST-07	-0.40 (-0.89 to 0.09)	10.6	Low
ST-08	-1.50 (-2.44 to -0.56)	6.6	Some concerns
ST-09	-0.80 (-1.54 to -0.06)	8.2	Some concerns
ST-10	-1.40 (-1.99 to -0.81)	9.6	Some concerns
ST-11	-2.40 (-3.38 to -1.42)	6.3	High
ST-12	-0.70 (-1.23 to -0.17)	10.2	Low
Pooled (random effects)	-1.17 (-1.55 to -0.78)	100.0

5.5 Results of syntheses [20a to 20d]

ID	Outcome	Studies (participants)	Effect (95% CI)	Heterogeneity	Item
SYN-01 (Example)	Pain at about 12 weeks	12 (2,172)	MD -1.17 (-1.55 to -0.78)	$I^2$ = 64%, $\tau^2$ = 0.22, prediction interval -2.28 to -0.05	20b
SYN-02 (Example)	Adverse events	8 (1,773)	RR 1.45 (1.07 to 1.96), Mantel-Haenszel	[Report $I^2$ ]	20b
SEN-01 (Example)	Pain, excluding high risk of bias	9 (1,875)	MD -0.92 (-1.24 to -0.61)	$I^2$ = 45%	20d
SUB-01	[Subgroup by comparator type]	[k (n)]	[Effect]	[Test for subgroup differences]	20c

Example: for SYN-01, $Q = 30.64$ with $df = 11$ , so $I^2 = (30.64 - 11) / 30.64 = 64.1\%$ . All 12 point estimates favoured exercise and the prediction interval excludes zero, so the direction of effect is consistent even though its size varies. Removing the three high-risk trials shrank the effect to -0.92 points, which suggests that part of the pooled effect comes from trials with weaker methods.

5.6 Reporting biases [21]

Example: The funnel plot for pain (12 trials) was asymmetric, with smaller trials showing larger effects. Egger's regression intercept was -4.32 (standard error 0.82, $t = -5.28$ , 10 degrees of freedom, $P < 0.001$ ). Two registered trials (registered YYYY and YYYY) have no published results; their investigators did not reply to two requests.

5.7 Certainty of evidence [22]

[Report the certainty per outcome and the reasons for each rating. The full summary of findings is in section 7.]

6. Discussion

The discussion interprets the findings honestly and says what they mean for practice, policy and future research.

6.1 General interpretation [23a]

[Compare the results with other reviews and guidelines. Relate the effect size to the minimal important difference, for example 1 point on a 0 to 10 pain scale, and state whether the confidence interval includes it.]

6.2 Limitations of the evidence [23b]

[Risk of bias: for example self-reported outcomes in unblinded trials]
[Inconsistency and its possible explanations]
[Indirectness of populations, interventions or settings]
[Imprecision and reporting biases]

6.3 Limitations of the review process [23c]

[Language, date or database restrictions]
[Data obtained by imputation or estimation]
[Decisions taken after seeing the data, from section 4.11]

6.4 Implications [23d]

[Implications for practice, policy and research. Phrase them to match the certainty: low-certainty evidence supports "may", not "does".]

7. Summary of findings

Certainty levels: high, moderate, low, very low. Evidence from randomised trials starts at high and may be rated down for risk of bias, inconsistency, indirectness, imprecision and publication bias.

Outcome	Comparator	With exercise	Relative effect (95% CI)	Participants (studies)	Certainty	Reasons
Pain, 0 to 10, about 12 weeks (Example)	[Mean in comparator groups]	MD 1.17 points lower (1.55 lower to 0.78 lower)	Not applicable	2,172 (12 RCTs)	Low	Risk of bias (-1); publication bias (-1)
Adverse events (Example)	74 per 1,000	107 per 1,000 (33 more, 5 more to 71 more)	RR 1.45 (1.07 to 1.96)	1,773 (8 RCTs)	Low	Risk of bias, harms not collected systematically (-1); imprecision, few events (-1)
[Function]	[Value]	[Value]	[Value]	[n (k)]	[Level]	[Reasons]

Example: the absolute effect for adverse events applies the pooled RR to the comparator risk: $65 / 882 = 74$ per 1,000, and $74 \times 1.446 = 107$ per 1,000.

7.1 GRADE judgements per domain

Outcome	Risk of bias	Inconsistency	Indirectness	Imprecision	Publication bias
Pain (Example)	Serious: 3 of 12 trials at high risk; effect smaller without them	Not serious: consistent direction, prediction interval excludes zero	Not serious	Not serious	Strongly suspected: funnel asymmetry, unpublished registered trials
Adverse events (Example)	Serious: non-systematic collection	Not serious	Not serious	Serious: 160 events in total	Undetected

8. PRISMA cross-reference, glossary and references

This section maps the PRISMA 2020 items to this report and lists terms and sources.

8.1 PRISMA 2020 cross-reference

Item	Topic (paraphrased)	Section
1, 2	Title, abstract	1.1, 1.2
3, 4	Rationale, objectives	2.1, 2.2
5	Eligibility criteria	4.1
6, 7	Information sources, search strategy	4.2, 4.3
8, 9	Selection and data collection processes	4.4, 4.5
10a, 10b	Data items	4.6
11, 12	Risk of bias assessment, effect measures	4.7, 4.8
13a to 13f	Synthesis methods	4.9
14, 15	Reporting bias and certainty assessment	4.10
16a, 16b	Study selection	5.1
17, 18, 19	Study characteristics, risk of bias, individual results	5.2, 5.3, 5.4
20a to 20d	Results of syntheses	5.5
21, 22	Reporting biases, certainty of evidence	5.6, 5.7, 7
23a to 23d	Discussion	6
24a to 24c	Registration and protocol	3.1, 4.11
25, 26	Support, competing interests	3.2
27	Availability of data, code and other materials	3.3

8.2 Glossary

Term	Meaning
GRADE	Grading of Recommendations Assessment, Development and Evaluation: a system for rating certainty of evidence
HKSJ	Hartung-Knapp-Sidik-Jonkman method for the confidence interval of a random-effects mean
MD, SMD, RR	Mean difference, standardised mean difference, risk ratio
Prediction interval	Range in which the true effect of a new similar study is expected to lie
REML	Restricted maximum likelihood estimator of the between-study variance
RoB 2	Cochrane risk-of-bias tool for randomised trials, version 2
ROBINS-I	Risk Of Bias In Non-randomised Studies of Interventions

8.3 References and related templates

Reference	Version or date	Use
PRISMA 2020 statement, Page MJ et al., BMJ 2021;372:n71, https://doi.org/10.1136/bmj.n71	2021	Reporting items and flow diagram
PRISMA 2020 explanation and elaboration, BMJ 2021;372:n160, https://doi.org/10.1136/bmj.n160	2021	Detailed guidance per item
Official checklists and flow diagrams, https://www.prisma-statement.org/prisma-2020	Checked 2026-10-30	Submission files
RoB 2, Sterne JAC et al., BMJ 2019;366:l4898, https://doi.org/10.1136/bmj.l4898	Tool version of 22 August 2019	Risk of bias in randomised trials
ROBINS-I, Sterne JAC et al., BMJ 2016;355:i4919; ROBINS-I V2 at https://www.riskofbias.info	V2 draft posted 20 November 2025	Risk of bias in non-randomised studies
Core GRADE series, BMJ 2025, starting with https://doi.org/10.1136/bmj-2024-081903	2025	Certainty of evidence
Cochrane Handbook for Systematic Reviews of Interventions, https://training.cochrane.org/handbook	Version 6.5, 2024	Methods
science/systematic-review	This collection	Protocol written before the review
science/scoping-review-protocol	This collection	When the question is about the extent of evidence rather than effects
academic/stats-effect-sizes	This collection	Effect measures and conversions
academic/research-paper-imrad, academic/peer-review	This collection	Manuscript and peer review

Standard and licence note

Based on the PRISMA 2020 statement: 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), published open access under the Creative Commons Attribution licence (CC BY 4.0). This template follows the order of its 27 items and sub-items, paraphrased in original text, and reproduces no checklist wording; use the official checklist, abstract checklist and flow diagram from https://www.prisma-statement.org for submission. Risk-of-bias domain names are those of RoB 2 (riskofbias.info), and certainty levels follow GRADE. Check the current versions of these tools before you start, as they are updated. This is a reporting format, not medical, statistical or methodological advice. Template licence: MIT.