Systematic Review Writing and Rewriting for Researchers and Evidence Teams
Pubrica’s subject matter experts, information specialists and statisticians conduct systematic reviews and meta-analyses for academic researchers, clinicians, pharmaceutical and device teams, HTA agencies and policy makers. We develop the review, or the systematic literature review (SLR) an HTA submission requires, from the protocol onwards, or rewrite a review you already have. A registered protocol, a reproducible search and two independent screeners on every review, so another team could rerun your search and reach the same result. You receive the manuscript with the PROSPERO record, search strategy, screening and extraction files, risk-of-bias tables and analysis files.
- Protocol development and PROSPERO registration, with the question framed in PICO, PIRD, CoCoPop or SPIDER before a single record is screened
- A reproducible search written with MeSH and Emtree terms across MEDLINE, Embase, the Cochrane Library, Scopus and the trial registries
- Dual independent screening in Covidence or Rayyan, with every exclusion reason logged
- Risk of bias and certainty graded with Cochrane RoB 2, ROBINS-I, QUADAS-3 or JBI, and GRADE per outcome
- Rewriting a review you already have that was rejected or returned with major revisions, starting with a written methodological critique before any editing
Protocol registered before the search ranPROSPERO
Databases searched, strategy reported in full6 sources
Records screened by two reviewers independently4,476
Risk of bias assessed study by studyCochrane RoB 2
Certainty of evidence graded per outcomeGRADE
Protocol → search → screening → extraction → synthesis → manuscript
- PRISMA 2020 reported
- Protocol registered first
- Two reviewers, independently
- Search strategy delivered
- Risk of bias assessed
Service Scope
Four Systematic Review Service Areas, Available Individually
Pubrica provides systematic review and meta-analysis services in stages as well as end to end, because many authors come to us part-way through, with a protocol that needs a search or extracted data that needs a synthesis and a manuscript. You can outsource the whole review or only the stage you need.
Types of Systematic Reviews
Nine Types of Systematic Review We Conduct
Pubrica provides end-to-end systematic review writing and evidence synthesis services for different research goals, study designs and evidence needs. Choosing the wrong review type is a costly mistake, because it often becomes apparent only at peer review, after the search has been run. Tell us your question and our subject experts will advise which type of review it requires before you commit.
Systematic Reviews by Purpose
The nine review types above are organised by method. The next section organises the same work by research question, such as treatment effectiveness, diagnostic accuracy, harms or cost-effectiveness, with the reporting standard and appraisal tool each question requires. Our subject experts will confirm which method your question calls for.
A meta-analysis is the statistical synthesis step inside a review rather than a different kind of review. Where your studies are alike enough to pool we do it as part of the work, and it is also available on its own — see meta-analysis, or network meta-analysis where treatments have never been compared head to head and the network geometry and inconsistency have to be reported alongside the estimate. Where the network will not connect, the indirect treatment comparison is done as a matching-adjusted indirect comparison or a simulated treatment comparison, with the population adjustment and its limitations stated.
Reviews by purpose
Systematic Reviews by Purpose and Methodology
Pubrica delivers systematic reviews designed to meet different research objectives, aligned with PRISMA, Cochrane, JBI and PROSPERO, with publication support built in, so the methodological work and the submission are handled as one project. Where the question relates to market access or medical affairs rather than publication, the same team conducts a targeted literature review or evidence map, with any methodological concessions declared. If your question is not listed, tell us what it is and we will design the approach around it.
Each of these is a systematic review; what changes is the question, the evidence it draws on and the reporting standard that governs it. Send Pubrica your question and our subject experts will identify the appropriate review type during the free scoping search, including where a different method is more suitable.
Who we serve
Researchers and Organisations We Support with Systematic Reviews
We support academic researchers and the organisations that use systematic reviews to make decisions, including guideline panels, payers, regulators and research offices.
Other Sectors We Support
Pubrica also builds reviews for regulatory and compliance firms, environmental and climate research organisations, education and social science researchers, financial and economic policy institutes, legal and forensic consultancies, and veterinary and animal health companies. The method does not change with the field; what changes is the appraisal instrument, the reporting guideline and the databases the question needs, and those are settled at protocol stage.
Compliance and Standards
Systematic Review Reporting and Methodological Standards
Pubrica works in compliance with this guidance. Each standard governs a different part of the review, and a Pubrica editor checks at least three of them before your manuscript reaches a reviewer. Every standard is listed below at the version we work to, with the date the list was last checked, because standards are revised over time.
Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA 2020)
Twenty-seven checklist items and the flow diagram, worked through against your manuscript. The extensions go with the design: PRISMA-ScR for scoping reviews, PRISMA-DTA for diagnostic accuracy, PRISMA-NMA for network meta-analysis, PRISMA-S for the search itself.
Cochrane Handbook and MECIR conduct standards (Cochrane)
PRISMA governs how a review is reported; the Cochrane Handbook governs how it is conducted. Dual independent screening, dual extraction, pre-specified subgroups and a search designed by an information specialist for each review.
International Prospective Register of Systematic Reviews (PROSPERO)
The protocol is registered before screening begins, which is what separates a systematic review from a literature search written up afterwards. Many journals now decline reviews that were never registered, and every later deviation has to be declared against the record.
Grading of Recommendations, Assessment, Development and Evaluation (GRADE)
Certainty rated high, moderate, low or very low for each outcome, with the reason for every downgrade written down — risk of bias, inconsistency, indirectness, imprecision, publication bias — and presented as a summary-of-findings table a guideline panel can use.
Quality appraisal and risk-of-bias instruments (AMSTAR 2, ROBIS, RoB 2, ROBINS-I, NOS, JBI)
Two layers. Cochrane RoB 2, ROBINS-I, QUADAS-3, the Newcastle-Ottawa Scale and the JBI critical appraisal tools assess the studies you include. AMSTAR 2 and ROBIS assess your review, and readers increasingly apply them to published reviews.
PRISMA 2020
27 items · BMJ 2021
How the review is reported. The 2020 flow diagram separates records found in databases from those found by other methods, which the 2009 diagram did not, and editors check which version is submitted.
PRISMA-S
2021
How the search itself is reported: databases with platform and date, the full strategy for at least one database, and every supplementary method.
PRISMA-ScR
2018
The extension for scoping reviews, where the question is what exists rather than what works, and no effect is pooled.
PRISMA-NMA
2015 · update due
Network meta-analysis reporting. An update is registered with EQUATOR and was planned to finish in 2026, so this is one to re-check before submission.
Cochrane Handbook
version 6.5 · August 2024
The conduct standard behind almost every methodological decision in a review, from eligibility criteria to how missing data is handled.
MECIR
version August 2023
Cochrane’s methodological expectations, unchanged by the 6.5 Handbook update, which define what a review must do to meet the conduct standard.
PROSPERO
CRD, University of York
Prospective registration before screening begins. The register moved to a new platform on 7 March 2025, with duplicate-topic detection at submission.
RoB 2
22 August 2019 version
Risk of bias in randomised trials, by domain rather than by score. The cluster-randomised and crossover variants were both revised on 18 March 2021.
ROBINS-I V2
20 November 2025
Still labelled a draft by its authors and subject to change. We use it and state this, rather than citing the 2016 original.
ROBINS-E
version 24 March 2024
The exposure counterpart, for studies where the thing under study was not assigned by anyone.
ROBIS 1.2
Bristol · 2016 paper
Risk of bias in the review itself rather than in its included studies; this instrument assesses the review we produce.
AMSTAR 2
BMJ 2017
Critical appraisal of an existing systematic review, and the tool a reviewer will reach for if your review is the one being assessed.
QUADAS-3
Ann Intern Med · 17 Feb 2026
Supersedes QUADAS-2, which is now filed under the tool’s own history page. Diagnostic accuracy reviews written to the 2011 version are now out of date.
Newcastle-Ottawa Scale
never versioned
Widely required for observational studies and never formally revised or dated by its authors. We state this rather than assigning a version.
JBI Manual for Evidence Synthesis
JBI · 2024 edition
Prevalence, qualitative, mixed-methods and umbrella reviews. JBI revises the online manual continuously while the downloadable version lags, and says so itself.
GRADE Handbook
updated October 2013
Certainty of evidence across five domains for downgrading and three for upgrading. GRADE carries no version number of its own — the handbook’s date is the only one there is.
GRADE-CERQual
Implementation Science 2018
Confidence in findings from qualitative evidence syntheses. A draft revision for the Cochrane-Campbell qualitative handbook has been posted since February 2025.
ICMJE Recommendations
January 2026
Authorship, contributorship, disclosure and trial registration. The January 2026 update added a section on artificial intelligence in publishing and one on author access to data.
Eighteen standards, each at the version we actually work to, checked on 21 September 2026. Three entries above carry a flag: QUADAS-2 was superseded in February 2026, ROBINS-I V2 is still a draft, and the PRISMA-NMA update may be published soon. Where a standard is revised, we update this table and confirm the version in force at the start of your project.
ICMJE, MOOSE, SWiM, ENTREQ and Other Standards
ICMJE governs the declarations: authorship on a review means contributing to its conception, conduct or interpretation rather than supplying the topic, and funding, competing interests, data availability and any use of AI in screening or drafting have to be stated, because a review is evidence that informs treatment.
Beyond that, the reporting guideline follows the design: observational syntheses under MOOSE, narrative syntheses under SWiM, qualitative evidence synthesis under ENTREQ, and prognostic reviews under CHARMS and PROBAST. The extraction template, the appraisal instrument and the reporting checklist are chosen together at protocol stage, based on the question, rather than adjusted once writing is underway.
Responsible Use of AI
Responsible Use of AI in Systematic Reviews
We define which decisions an AI tool may support, rather than how much of the screening it can do. Every included study is read by two reviewers, and your manuscript states where a tool was used.
Why We Do Not Publish an AI Accuracy Figure
Accuracy figures for screening tools are usually reported by the companies that sell them.
For the same reason, Pubrica does not quote an accuracy figure of its own. Instead, we tell you which tool was used on your review, at which stage, and show you the records it ranked low that our reviewers included.
Methodology
How Our Systematic Review Process Works
The first two stages determine the quality of the review, which is why a Pubrica subject expert and an information specialist work on the project from the protocol stage rather than from drafting. Once the search has run, the remaining stages follow the methods fixed in the protocol.
Question, scope and protocol
We frame your topic as an answerable question using the framework that fits it: PICO for interventions, PIRD for diagnostic accuracy, CoCoPop for prevalence and SPIDER for qualitative work. We first run a research gap and novelty check against PROSPERO and the published literature to confirm that the same question has not recently been reviewed. The protocol specifies eligibility, outcomes, subgroups and the analysis before screening, and Pubrica’s PROSPERO registration support covers submission and any reviewer queries.
Search strategy and database searching
A line-by-line strategy built from controlled vocabulary and free text — MeSH in MEDLINE, Emtree in Embase, with Boolean operators, proximity operators and truncation adapted for each platform rather than pasted between them. Run across MEDLINE, Embase, the Cochrane Library, CINAHL, Web of Science, Scopus, PsycINFO and the subject databases your question needs, plus trial registries and grey literature. The strategy is handed to you in full, dated and reproducible.
Screening in Covidence or Rayyan
Records are deduplicated in EndNote or Zotero, then screened on title and abstract by two reviewers working independently in Covidence or Rayyan, with a third arbitrating disagreements. Full texts are retrieved and assessed the same way, and every full-text exclusion is recorded with its reason, because this list is a required part of the paper and cannot be reconstructed afterwards.
Data extraction and risk-of-bias assessment
Structured extraction into a piloted template — population, intervention, comparator, outcomes, effect estimates, funding — done in duplicate and reconciled. Risk of bias assessment is run study by study with the instrument the design calls for, and is available from Pubrica on its own where the extraction is already done: Cochrane RoB 2 for randomised trials, ROBINS-I for non-randomised studies, QUADAS-3 for accuracy studies, the Newcastle-Ottawa Scale or JBI tools for observational work.
Evidence synthesis and meta-analysis
Where the studies are alike enough to pool, we conduct a meta-analysis in R, Stata or RevMan with the model stated and justified: pooled effect estimates, forest and funnel plots, quantified heterogeneity, pre-specified subgroup and sensitivity analyses, and publication-bias testing. Where they are not, we provide a structured narrative synthesis reported under SWiM. Certainty of evidence is graded per outcome with GRADE, and the summary-of-findings table is built from it.
Manuscript, checks and submission
The full manuscript is written in IMRaD with a PRISMA-compliant structured abstract, the flow diagram, the completed checklist and references in your journal’s style, formatted to the target journal. It then undergoes plagiarism and AI-content screening, proofreading and a final technical review. Journal selection, submission and response to reviewer comments are available if required.
Rewriting a review you already have
Rewriting and Revising Rejected Systematic Reviews
The six stages above assume a new review. Many authors instead have a rejected systematic review, or one returned with requests that go beyond editing. Pubrica handles these as a rewrite, which begins with a written assessment from a subject expert stating what is wrong, what can be repaired and what cannot, so you can decide whether to resubmit on the evidence.
Limits of a Rewrite
A search that missed a database the journal requires, or that was not documented well enough to reproduce, cannot be fixed by editing. It must be re-run and re-screened, because a rewrite that works around it is likely to meet the same objection at the next journal. The same applies to a review written up after the reading was completed.
The critique states which situation applies before you commit to the rewrite, and any re-run search is quoted separately.
Deliverables
Client Inputs and Pubrica Deliverables
Many authors are unsure what they need to prepare before starting. The lists below set out what you provide and what Pubrica delivers.
What you provide
What Pubrica provides
If you have only a question, that is enough to begin. Send it to us and our subject experts will run a free scoping search within two business days, establishing whether the review has already been done, whether there is enough evidence to synthesise and how many records your question returns.
Methodology
Comprehensive Literature Searching and the Screening Funnel
Each database omitted from a search can remove eligible evidence, and the finished paper does not show which evidence was missed. This example funnel from a Pubrica review shows the numbers our information specialists report at every stage, so reviewers can follow the same path.
Review question
Sources searched
Record counts are what the strategy returns from each source. Turn one off and the diagram below rebuilds.
Limits applied
Each limit must be declared in the protocol before the search runs and justified in the paper.
Identification
Records identified from databases5 databases searched
n =7,118Records removed before screeningDuplicates across databases
n =2,918Records from registers and grey literatureTrial registries, theses, reports
n =212Records from citation searchingReference lists and cited-by
n =64Screening
Records screenedTitle and abstract, two reviewers independently
n =4,476Records excludedAgainst the pre-specified criteria
n =4,389Reports sought for retrieval
n =87Reports not retrievedNo full text obtainable
n =5Eligibility
Reports assessed for eligibilityFull text, two reviewers, disagreements arbitrated
n =82Reports excluded, with reasons
n =66- Wrong population — mixed cardiac cohorts with no heart-failure subgroup21
- Wrong intervention — education or diet only, no exercise component16
- No readmission outcome reported13
- Conference abstract with no full report9
- Secondary report of a trial already included7
Included
Studies included in the review
n =16Reports of included studiesProtocols, secondary papers, registry entries
n =2216 of 16 eligible studies found100% of the eligible evidence for this question
The full search finds every eligible study. Remove one database to see which studies are lost. This is why systematic reviews search several databases, and why peer reviewers question searches limited to fewer.
An illustrative worked example. Study labels are anonymised and record counts are representative. The effect shown is typical: searching a single database misses eligible evidence for most questions, and the resulting review gives no indication that evidence is missing.
Why this matters at peer review
The first thing a methods reviewer checks is whether the numbers in your PRISMA flow diagram add up, and the second is which databases you searched. A review based on PubMed alone will be asked why Embase was not searched, and the only remedy is to re-run the search, re-screen and re-extract.
Pubrica sets the search strategy against the question before anything is screened, runs it across the databases the question actually needs, and hands you the strategy line by line so a reviewer, an editor or a reader can run it again.
Our experts
Your Systematic Review Team
Each Pubrica systematic review is conducted by a named team: an information specialist, a subject expert matched to your topic, a second reviewer who screens independently of the first, and a statistician. The handover records which team member completed each stage.
Designs the search strategy line by line for each platform, because MeSH and Emtree do not share headings, and delivers it dated and reproducible so a reviewer can run it again.
Frames the question in the appropriate framework, sets eligibility criteria before screening and determines what the evidence will and will not support, including advising you when a question is not suited to a systematic review.
Screens and extracts independently of the first reviewer. Two reviewers reading the same records without sight of each other’s decisions is a defining requirement of a systematic review. Every disagreement is logged and adjudicated.
Assesses whether the studies can be pooled before any analysis is run, conducts the meta-analysis in R, Stata or RevMan, delivers the analysis files and grades the certainty of each outcome separately.
Your project coordinator introduces the four by name before the protocol is written, and the same four stay with the review to submission. Where a question needs a fifth — a health economist for a cost-effectiveness review, a clinician in a second specialty where the population straddles two — we say so at scoping and name them too.
Sample work
Sample Systematic Review Deliverables
Every Pubrica systematic review returns the manuscript and the working documents behind it — the registered protocol, the search strategy line by line, the extraction and appraisal tables, and the graded summary of findings. Use the tabs to see each one.
Sample Evidence Synthesis
The screening funnel earlier on this page shows how the search was narrowed. These figures show the synthesis of the included studies from the same worked review, prepared by Pubrica’s statisticians and subject experts. Both are delivered with your review as editable analysis files rather than static images.
| Trial | D1 | D2 | D3 | D4 | D5 | Overall |
|---|---|---|---|---|---|---|
| Multicentre · United States, 2009 | Low risk of bias | Some concerns | Low risk of bias | Low risk of bias | Low risk of bias | Some concerns |
| Multicentre · Italy, 2013 | Low risk of bias | Low risk of bias | Low risk of bias | Low risk of bias | Low risk of bias | Low risk of bias |
| Multicentre · China, 2018 | Some concerns | Low risk of bias | Low risk of bias | Some concerns | Low risk of bias | Some concerns |
| Home-based · Netherlands, 2019 | Low risk of bias | Low risk of bias | Low risk of bias | Low risk of bias | Low risk of bias | Low risk of bias |
| Multicentre · United Kingdom, 2012 | Low risk of bias | Low risk of bias | Low risk of bias | Low risk of bias | Low risk of bias | Low risk of bias |
| Pragmatic · Belgium, 2021 | Low risk of bias | Some concerns | Some concerns | Low risk of bias | Low risk of bias | Some concerns |
| Nurse-led · Australia, 2010 | Some concerns | Low risk of bias | Low risk of bias | Some concerns | Low risk of bias | Some concerns |
| Multicentre · Spain, 2017 | Low risk of bias | Some concerns | Low risk of bias | Low risk of bias | Low risk of bias | Some concerns |
| Registry-posted · Canada, 2023 | Low risk of bias | Low risk of bias | Some concerns | Low risk of bias | High risk of bias | High risk of bias |
| Single-centre · Germany, 2004 | Some concerns | High risk of bias | Low risk of bias | Some concerns | Low risk of bias | High risk of bias |
| Single-centre · Brazil, 2016 | Low risk of bias | High risk of bias | Some concerns | Low risk of bias | Low risk of bias | High risk of bias |
Each square above represents one trial, sized by its weight and coloured by its risk-of-bias rating in the table beneath; the diamond shows the pooled result, which lies clear of the line of no effect.
Systematic review handover
Intervention review · 16 included studies · PRISMA 2020
4,476 records screened · 6 sources · 16 studies included
- Population
- Adults with chronic heart failure, any ejection fraction, recruited after a confirmed diagnosis
- Intervention
- Supervised or home-based exercise training of at least eight weeks
- Comparator
- Usual care, or an intervention with no exercise component
- Primary outcome
- All-cause hospital readmission within twelve months
- Secondary
- Mortality, six-minute walk distance, quality of life, adverse events
- Designs
- Randomised controlled trials, including cluster and cross-over designs
- Pre-specified subgroups
- Supervised against home-based; preserved against reduced ejection fraction
- Registration
- PROSPERO, submitted before screening began; two later deviations declared
MEDLINE (Ovid) · run and dated · translated line by line for Embase, CENTRAL, CINAHL and Web of SciencePRISMA-S reported
16 studies · 22 reports · extraction form supplied as a spreadsheetDual and reconciled
Summary-of-findings table · GRADE applied per outcomeEvery downgrade reasoned in writing
Redacted extract · client and topic details changedPage 1 of 4
The handover includes the manuscript and every file needed to verify it. Use the tabs to see each part.
The search strategy is the document methods reviewers most often request and authors often cannot produce afterwards. You receive it line by line, with the record count each line returned and the date it was run, translated for every database.
The extraction and appraisal tables are delivered as working spreadsheets rather than static images, so you can respond efficiently when a reviewer asks for a sensitivity analysis or for the pooling to be re-run without the high-risk trials.
On request, we will send a sample handover from a review similar to yours, with the same question type, synthesis method and reporting guideline.
Systematic review handover · redacted sample
Intervention review · 16 included studies · PRISMA 2020
4,476 records screened · 6 sources · Cochrane RoB 2 · GRADE per outcome. Client names, journal names and identifying particulars have been removed or changed. On request, we will send a sample from a review similar to yours.
1Protocol summary
Registered on PROSPERO before screening began. Two deviations from the registered protocol arose during the review and are declared in the manuscript: one subgroup was dropped because too few trials reported it, and the follow-up window for the primary outcome was fixed at twelve months after two trials were found to report at both six and twelve.
| Element | As registered |
|---|---|
| Framework | PICO |
| Population | Adults with chronic heart failure, any ejection fraction, diagnosis confirmed |
| Intervention | Supervised or home-based exercise training of at least eight weeks |
| Comparator | Usual care, or an intervention with no exercise component |
| Primary outcome | All-cause hospital readmission within twelve months |
| Secondary outcomes | All-cause mortality, six-minute walk distance, health-related quality of life, serious adverse events |
| Eligible designs | Randomised controlled trials, including cluster and cross-over designs |
| Subgroups | Supervised against home-based delivery; preserved against reduced ejection fraction |
| Language and date | No language restriction; no date restriction |
| Reporting guideline | PRISMA 2020, with PRISMA-S for the search |
| Appraisal instrument | Cochrane RoB 2 |
| Certainty framework | GRADE, applied per outcome |
2Search strategy, as run
Reproduced here for MEDLINE (Ovid). The strategy was translated line by line for each other platform rather than pasted across, because Emtree and MeSH do not share headings and the proximity syntax differs. Every version is supplied with the handover, dated, with the record count each line returned.
| # | Line | Records |
|---|---|---|
| 1 | exp Heart Failure/ | 142,318 |
| 2 | (heart adj2 failure or cardiac failure or HFrEF or HFpEF).ti,ab,kf. | 189,047 |
| 3 | 1 or 2 | 214,806 |
| 4 | exp Exercise Therapy/ or exp Cardiac Rehabilitation/ | 88,412 |
| 5 | (exercis* or train* or rehabilitat* or physical activity).ti,ab,kf. | 1,204,558 |
| 6 | 4 or 5 | 1,229,730 |
| 7 | 3 and 6 | 6,904 |
| 8 | 7 and Cochrane highly sensitive RCT filter (sensitivity-maximising version) | 1,842 |
Databases searched: MEDLINE (Ovid) 1,842 · Embase 2,310 · Cochrane CENTRAL 1,104 · CINAHL 486 · Web of Science 1,376 · ClinicalTrials.gov and WHO ICTRP 212 · citation searching 64. All searches run on the same date and re-run before submission.
3PRISMA 2020 flow
| Stage | n |
|---|---|
| Records identified from databases | 7,118 |
| Records identified from registers and grey literature | 212 |
| Records identified by citation searching | 64 |
| Duplicates removed before screening | 2,918 |
| Records screened on title and abstract | 4,476 |
| Records excluded at screening | 4,389 |
| Reports sought for retrieval | 87 |
| Reports not retrieved | 5 |
| Reports assessed for eligibility at full text | 82 |
| Reports excluded with reasons | 66 |
| Studies included in the review | 16 |
| Reports of included studies | 22 |
Full-text exclusions, with reasons
| Reason | n |
|---|---|
| Wrong population — mixed cardiac cohorts with no heart-failure subgroup | 21 |
| Wrong intervention — education or diet only, no exercise component | 16 |
| No readmission outcome reported | 13 |
| Conference abstract with no full report | 9 |
| Secondary report of a trial already included | 7 |
The full-text exclusion list is a required part of the published review. It is compiled as screening proceeds, because it cannot be reconstructed afterwards.
4Data extraction and risk of bias
Every study was extracted independently by two reviewers into a form piloted on four studies, then reconciled. Eleven disagreements arose; all were resolved by discussion, and three required the corresponding author to be contacted. Where a number was obtained from an author rather than from the published report, its provenance is recorded in the extraction table.
| Study | RoB 2 | Note |
|---|---|---|
| Multicentre RCT · United States, 2009 | Some concerns | Largest trial; drives the pooled estimate |
| Single-centre RCT · Germany, 2004 | High | Deviations from intended intervention |
| Multicentre RCT · Italy, 2013 | Low | — |
| Nurse-led RCT · Australia, 2010 | Some concerns | Indexed in CINAHL alongside MEDLINE only |
| Home-based RCT · Netherlands, 2019 | Low | — |
| Pragmatic RCT · Belgium, 2021 | Some concerns | Indexed in Embase only |
| Multicentre RCT · China, 2018 | Some concerns | Published in Chinese; translated for extraction |
| Registry-posted RCT · Canada, 2023 | High | Results posted to registry, never published |
Extract only — eight of sixteen studies shown. Overall: four at low risk, nine with some concerns, three at high risk. The high-risk trials are removed in a pre-specified sensitivity analysis, and the pooled estimate holds.
Two findings worth noting from the search. One included trial is indexed in Embase and in no other database searched, and one exists only as results posted to a trial registry. A review of this question based on MEDLINE alone would have missed both, with no indication in the finished manuscript.
5Summary of findings (GRADE)
| Outcome | Studies | Certainty | Reason for downgrade |
|---|---|---|---|
| All-cause readmission, 12 months | 11 (2,847) | Moderate | Risk of bias — three trials at high risk |
| All-cause mortality, 12 months | 9 (2,208) | Low | Risk of bias; imprecision — interval spans benefit and harm |
| Six-minute walk distance | 12 (3,014) | Moderate | Risk of bias |
| Health-related quality of life | 8 (1,942) | Low | Inconsistency; indirectness — four instruments, two unvalidated in heart failure |
| Serious adverse events | 5 (1,106) | Very low | Not pooled — no shared definition; reported narratively under SWiM |
Heterogeneity for the primary outcome was low and the direction of effect was consistent across every trial that measured it. Publication bias was assessed by funnel plot and Egger test for the primary outcome only, since fewer than ten trials reported each of the others.
6What is handed over
- The manuscript in IMRaD, with a PRISMA 2020-compliant structured abstract, formatted to the target journal’s instructions for authors.
- The completed PRISMA 2020 checklist and the flow diagram as an editable figure.
- The search strategy for every database, line by line, with record counts and the dates run, reported to PRISMA-S.
- The screening and extraction spreadsheets, including the full-text exclusion list with reasons.
- The risk-of-bias assessments per study and per domain, as a traffic-light figure and as data.
- The analysis files — R, Stata or RevMan — so the synthesis can be re-run rather than trusted.
- The PROSPERO record and a written note of every deviation from it.
Pubrica supports the work; authorship remains yours. Where our contribution should be acknowledged under ICMJE, the handover drafts the wording for you to approve.
Prepared by Pubrica. Pubrica is a trading name of Guires®, registered in England, Wales and India.
Download this as a PDFAs it arrived
Search: (diabetes AND exercise AND HbA1c) AND 2015:2025[dp] — run in all fields, in one database, with a ten-year limit.
- Free text only. No MeSH explosion, so every record indexed but not phrased that way is invisible.
- No truncation and no adjacency, so “physical activity”, “aerobic training” and “glycated haemoglobin” are all missed.
- A date limit nobody justified, which has to be defended in the protocol or removed.
- 4,212 records, most of them noise, and no way to reproduce the search from the paper.
As it went out
Two concept blocks, each built as exploded MeSH OR focused free text with truncation and adjacency, no date limit, translated line by line for Embase, CENTRAL, CINAHL, Scopus and Web of Science, and reported to PRISMA-S.
- 2,860 records after de-duplication, and 41 relevant records the first search never returned.
- 1,352 fewer records to screen saved about 34 reviewer hours at our dual-screening rate.
- Peer-reviewed against the PRESS checklist by a second information specialist, an item that editors check.
- Printed in full in an appendix, so that anyone can re-run the search.
The search strategy has more influence on a systematic review than any other stage, yet it is often rushed. A strategy that returns four thousand records to find forty studies spends most of its screening budget on irrelevant records. Every figure above comes from one project and can be reproduced from the appendix delivered with it.
Benefits
Benefits of a Pubrica Systematic Review
A rigorously conducted systematic review informs clinical guidelines, health policy and future research. Each benefit below can be verified on your own review after delivery.
Workload estimator
Estimate Your Systematic Review Screening Workload
The workload of a systematic review depends mainly on how many records the search returns, rather than on the difficulty of the topic. Enter your own figures and the estimator converts them into reviewer hours, elapsed weeks and cost for each stage: title and abstract screening, full-text assessment, data extraction and risk of bias, with dual screening doubling the reading time and adding time to resolve disagreements.
01
Enter the size of your search to see the reviewer hours, elapsed weeks and cost for each stage. Every rate is an editable field, so you can replace our defaults with your own.
Runs in your browser · nothing sent anywhere355h
Total reviewer hours
17.7wk
At 10 hours a week each
38
Studies likely to be included
This is reviewer time, not elapsed time. Retrieving full texts your library does not hold, waiting on authors for missing data, and the fact that nobody screens for eight hours a day all push the calendar out well beyond 17.7 weeks. The single biggest lever is not speed — it is the search. A search that returns 4,200 records to find 38 studies is spending 79 hours to discard 92% of what it found, and an hour spent with an information specialist on the strategy usually saves more than an hour of screening.
Method. Straight arithmetic on the rates in the fields, which are our own defaults from completed reviews and are meant to be overwritten with yours. Dual screening doubles the screening, assessment and extraction time and adds an allowance for resolving conflicts; it also halves the number of studies wrongly excluded, which is why Cochrane requires it. Automation-assisted prioritisation can cut title and abstract time substantially, but it does not remove the requirement for two humans on the included set.
What the estimate is for
It is reviewer time, not elapsed time, and it is built to be adjusted to your own figures: the rates are fields rather than constants, because a review of trials in one database behaves nothing like a review of qualitative studies across six.
Use it to see early whether the search is the main source of workload. A strategy that returns four thousand records to find forty studies spends most of its screening budget on irrelevant records, and an hour with an information specialist on the search usually saves more time than faster screening. This is one of twelve calculators and checkers on our free research tools page, all of which run in your browser and ask for nothing.
Systematic Review Pricing Based on Protocol Scope
What a systematic review costs depends on how much evidence there is to handle — how many databases the question needs, how many records come back, whether the studies can be pooled, and whether you are starting from a topic or from a draft that came back from review. Send what you have and we return a scope, a price and a date within two business days.
The scoping search is free and answers three questions first: whether this review has already been done, whether there is enough evidence to review, and whether the question is narrow enough to complete. We would rather tell you this at the start of a project than months into it.
Packages
Systematic Review Service Packages
Every package is run by a Pubrica subject expert with a dedicated project coordinator, and every one returns the working documents, not only the manuscript. What changes is how much of the review we build and how much you bring. Each figure is a starting point, not a quotation: what moves it is the volume of evidence your question returns, which is what the screening estimator above is for. We confirm the figure in writing after the free scoping search, before anything begins. Reviews commissioned by pharmaceutical, device, HEOR and market access teams are scoped and quoted separately, because a submission-grade SLR is priced on the dossier it has to survive rather than on the journal it is aimed at.
Focused
10–15 business days
A narrower question and up to three databases. Quoted on capacity rather than on urgency, so a short timeline is never a surcharge on your deadline.
Complete
30–40 business days
The standard timeline for a full review, with time for a revision round before your own deadline.
Extended
60–120 business days
Larger scopes, meta-analysis with GRADE, and submission support through to a decision.
Structured evidence summary
Early-career researchers and postgraduates working from studies they have already collected. A structured summary rather than a full systematic review — registration, dual screening and risk-of-bias assessment are the three things that make it one, and they are added at Complete.
From $580up to 3 databases · 300 records after de-duplication
Delivery 10–15 business days
- ✓Initial protocol development and an answerable question
- ✓Search strategy across up to three databases
- ✓Study selection and screening reported under PRISMA
- ✓Narrative synthesis of the findings
- ✓Language editing, plagiarism and AI-content screening
Add later: PROSPERO registration; more databases; meta-analysis.
View a sample Select Evidence summaryComplete systematic review
Academics, clinicians and healthcare professionals starting from a question
From $850up to 4 databases and grey literature · 500 records · 12 studies
Delivery 30–40 business days
- ✓Full protocol development and PROSPERO registration
- ✓Multi-database search across up to four databases, plus grey literature
- ✓Dual independent screening in Covidence or Rayyan
- ✓Risk-of-bias assessment — Cochrane RoB 2, ROBINS-I, NOS or JBI
- ✓Data extraction and qualitative or quantitative synthesis
- ✓Manuscript formatted to your target journal's guidelines
Add later: meta-analysis; GRADE; submission support.
View a sample Select StandardPublication-ready with meta-analysis
Researchers and groups targeting a high-impact journal or an evidence submission
From $1,450as Standard, with up to 20 studies pooled
Delivery 60–120 business days
- ✓Everything in Standard
- ✓Meta-analysis with forest plots and the statistical model justified
- ✓Subgroup, sensitivity and publication-bias analyses
- ✓GRADE certainty rating and a summary-of-findings table
- ✓Journal selection and submission handled for you
- ✓Response to reviewer comments and revision through to a decision
Add later: network meta-analysis; cost-effectiveness; living update schedule.
View a sample Select AdvancedRewriting and enhancement
Authors whose review was rejected, or came back with major revisions
From $690one manuscript · one round of reviewer comments
Delivery 15–25 business days
- ✓Methodological critique of the existing manuscript, written first
- ✓Structural and methodological repairs, with what cannot be repaired named
- ✓Rewriting against the reviewer and editor comments received
- ✓Language editing and reference reformatting
- ✓Plagiarism recheck and AI-content screening before resubmission
Add later: a re-run search where the original one will not survive review.
View a sample Select RewritingTurnaround is measured in working time and depends on the volume of evidence rather than on effort: a question returning nine hundred records is a different job from one returning nine thousand. Where a funder deadline, thesis submission or conference date is fixed, tell us the date and we will confirm whether it can be met and what would need to be declared if the search were narrowed to meet it.
Add-On Services
Additional Services Available with Any Package
These are extras rather than kinds of review, and most authors ask for them once the synthesis is written and the journal is chosen. Each can be added to a package or ordered on its own, and each is quoted separately so you can see what it costs.
Quality Assurance
Our Service Commitments and Limitations
A systematic review can run for several months, so our service terms are set out in full below for you to review before you commit.
The Pubrica guarantee
Revision Support Until Agreed Requirements Are Met
We continue revising the review until it meets the requirements agreed within the scope. Rather than offering a refund, which would end the engagement with the work incomplete, we keep working until the agreed requirements are met.
01
Unlimited revisions within the agreed scope
There is no limit on the number of revision rounds within the agreed scope for 365 days from delivery. If the work does not meet the agreed requirements, we revise it. Additional work, such as another database, a second outcome or a re-run search, is quoted separately, and we confirm which applies before any work starts.
02
Corrections at Our Cost
If an error in our work needs correcting, such as a search line that missed a term, an extraction error or an unmet checklist item, we correct it free of charge.
03
Delivered on the date agreed
Package turnarounds follow the bands above. Where a review falls outside a package, we quote a date we can meet and work to it. If a milestone is at risk, we inform you in advance.
04
Mutual NDA on request
We sign a mutual NDA before you share any material, at no cost. Your question, your data and our engagement remain confidential in either case, and we can sign your organisation’s NDA instead of ours.
05
You Retain Final Decisions
We advise on the question, the limits and the synthesis, and tell you when we disagree with a proposed approach. You make the final decision on everything published in your name.
Average client rating
From 3,000 responses to Pubrica’s post-delivery survey, sent to every client at handover between January 2024 and June 2026. This figure comes from our own records rather than from an external rating service.
Limitations of Our Service We cannot guarantee acceptance, which is the decision of the journal editor. We also cannot guarantee a positive finding: if the evidence is sparse, inconsistent or at high risk of bias, the review reports this. Unlimited revision, within the agreed scope, covers the review we were engaged to build; a changed question, a re-run search after a protocol amendment you decide on, or a second review is new work.
Free Scoping Search
Before you commit, we check whether the project is viable: whether this review has already been registered or published, whether there is enough evidence to synthesise and whether the question is narrow enough to complete. The check is free, and we would rather tell you at the start than months into the project.
Full Search Strategy Provided
We deliver the strategy line by line for every database, with the count each line returned and the date it was run, together with the extraction spreadsheet and the analysis code, so you can reproduce and defend the review independently.
GDPR and data privacy
Pubrica works with researchers in India, across the United Kingdom and Europe, in North America, throughout Asia-Pacific and the Middle East, and in Africa and Latin America, and every project is handled the same way wherever it comes from. Your protocol and your unpublished synthesis are held under GDPR, kept only as long as the project needs them, never used to train anything, and never shared outside your project team.
No ghost authorship
We conduct and write the review but do not act as undeclared authors. Where Pubrica’s involvement should be acknowledged under ICMJE or GPP, as it usually should be when we ran the search and the synthesis, we say so and draft the wording for you to approve.
Client Stories
Client Feedback on Systematic Review Projects
Rephrased from client feedback.
“What arrived was a review that held together as a method, not just as a piece of writing. Every PRISMA item was met, and the structure made the argument far easier to follow than my own draft had.”
“The methodology is what carried it. Reviewers questioned almost nothing about the search or the appraisal, which left the discussion to be about the finding rather than about how we had reached it.”
“They worked to the target journal’s own requirements from the first draft rather than reformatting at the end, and the structured approach to screening and extraction is what got our team over the line.”
FAQs
Systematic Review FAQs
If you are unsure whether your question is answerable as a systematic review, or whether it has already been done, please ask an expert. No quote is required.
There are nine steps, carried out in a fixed order.
- Frame an answerable question in PICO, PIRD, CoCoPop or SPIDER.
- Write a protocol that fixes the eligibility criteria, outcomes and analysis in advance.
- Register it, usually on PROSPERO.
- Build and run a search strategy across the databases the question needs.
- Deduplicate and screen every record on title and abstract, with two reviewers working independently.
- Retrieve and assess the full texts, recording a reason for each exclusion.
- Extract data into a piloted form, in duplicate.
- Assess risk of bias study by study.
- Synthesise, grade the certainty of the evidence and report the whole thing under PRISMA 2020.
None of these steps can be done retrospectively without weakening the review. A review written up after the reading is complete does not meet the definition of a systematic review, however thorough the reading was, and we will tell you at the scoping stage which of the two you have.
A systematic review is the method: a pre-specified question, an exhaustive search, transparent selection criteria, critical appraisal and a synthesis of what the studies found. A meta-analysis is one possible synthesis within it — the statistical pooling of results into a single effect estimate with a confidence interval. Every meta-analysis should sit inside a systematic review; not every systematic review should contain a meta-analysis.
Where studies differ too much in population, intervention or outcome measure, pooling them produces a misleadingly precise estimate, and the appropriate approach is a structured narrative synthesis reported under SWiM. This decision is made at protocol stage, not after the results are available.
By limiting the points at which an author’s judgement can influence the answer. Pre-specifying the criteria stops studies being included because they agree with the expected result. Searching several databases stops the answer depending on what PubMed happens to index. Dual independent screening prevents one reader’s fatigue from causing exclusions. Appraising risk of bias stops a poorly conducted trial carrying the same weight as a well conducted one. Registering the protocol stops outcomes being chosen after the results are visible.
Each of these steps prevents a specific way in which a review could reach a conclusion its evidence does not support.
The question decides. A clinical intervention review is expected to search MEDLINE, Embase and the Cochrane Library at minimum, and reviewers will ask about all three. Nursing and allied health questions need CINAHL; psychology and behavioural questions need APA PsycINFO; agriculture and veterinary questions need CAB Abstracts; anything with trials needs ClinicalTrials.gov and the WHO ICTRP for unpublished and terminated studies. Web of Science and Scopus add citation searching, which finds what subject indexing misses.
On the tooling side, references are managed in EndNote, Zotero or Mendeley, screening is done in Covidence or Rayyan so that two reviewers can work independently, and analysis is run in R, Stata or RevMan. The tools matter far less than the search strategy itself.
Extraction uses a form piloted on three or four studies before it is applied to the rest, because missing fields usually become apparent only after several studies have been extracted. Two reviewers extract every study independently and reconcile their results, which catches most transcription errors. Where a paper reports an outcome only in a figure, or reports a median where a mean is needed, we contact the corresponding author and record the source of the number rather than imputing it.
Analysis follows the pre-specified plan: pooled effect estimates with the model stated, heterogeneity quantified, pre-planned subgroup and sensitivity analyses, and publication-bias assessment where there are enough studies to make a funnel plot meaningful.
There is no legal obligation, but journals increasingly expect it. A growing number of journals decline unregistered reviews outright, AMSTAR 2 treats the absence of a registered protocol as a critical weakness, and without a registration you cannot demonstrate that the outcomes you report are the ones you set out to measure.
Registration is free, takes a few hours once the protocol is written, and is accepted for reviews that have not yet begun screening. Where PROSPERO does not accept a review type — scoping reviews, for instance — the equivalent is registering on the Open Science Framework or publishing the protocol. Deviations after registration are not a problem in themselves; undeclared deviations are.
An academic team working around other commitments typically takes twelve to eighteen months. With Pubrica running the work, a complete review is usually thirty to forty business days, and sixty to one hundred and twenty where a meta-analysis, GRADE and submission support are included.
What drives the time is the volume of evidence rather than the difficulty of the topic: screening nine thousand records is six times the work of screening fifteen hundred, whatever the question. A rapid review can be delivered far faster using systematic methods with declared shortcuts — fewer databases, single screening with verification, a date limit — and the shortcuts are then reported as limitations rather than hidden.
Six things, in roughly this order of frequency. A search that covered one database, or that cannot be reproduced from what is written in the methods. A PRISMA flow diagram whose numbers do not reconcile. No full-text exclusion list, or one without reasons. Studies pooled that should not have been, usually visible as heterogeneity above seventy-five per cent with no explanation offered. No risk-of-bias assessment, or one performed but never used to weigh the conclusion. And a question that turns out to have been answered by a review published while yours was being written.
Five of these six are determined before any writing begins, which is why the protocol and the search deserve the most investment.
Yes, and most of our work arrives that way. Come with a protocol and we run the search; come with a screened set and we extract, appraise and synthesise; come with a full draft and we rewrite it.
For a rejected manuscript, the first deliverable is a written critique rather than a revision, because it must state whether the problem can be repaired. Structure, reporting, appraisal and synthesis can all be corrected after the fact. A search that missed a required database cannot; it must be re-run and re-screened, and a rewrite that ignores this is likely to meet the same objection at the next journal. We will tell you which situation applies before you commit.
You are. Authorship under ICMJE requires substantial contribution to conception or design, or to acquisition, analysis or interpretation; drafting or critically revising; approving the final version; and agreeing to be accountable for it. Pubrica supports all four but takes none of them in your place, and we do not accept authorship on client work. Where our contribution should appear in the acknowledgements — and on a review where we built the search and ran the synthesis, it should — we say so and draft the wording.
On artificial intelligence: any AI-assisted screening or drafting is declared in the methods, with what was used and for which step, because ICMJE and COPE both now require it and an undeclared tool found later is treated as a research integrity issue rather than a style preference.
By how much evidence there is to handle, not by word count. Four things move the figure: how many databases the question needs, how many records come back after de-duplication, whether the included studies can be pooled or only described, and whether you are starting from a topic, a protocol or a draft that came back from review.
The number of records has the greatest effect, because a search that returns four thousand records to find forty studies spends most of the budget on irrelevant records. You can estimate this before contacting us with the free screening estimator turns your record count into reviewer hours, elapsed weeks and a cost, stage by stage, with every rate an editable field.
Yes, and many should be. Pooling requires the studies to answer a sufficiently similar question with sufficiently similar methods that one estimate describes them all, and often they do not. A review that reports a structured narrative synthesis under SWiM, with effect direction, magnitude and certainty set out study by study and an explicit statement of why pooling was not appropriate, is a stronger paper than one that produces an unreliable pooled estimate.
The decision belongs in the protocol, with the criterion stated in advance, because reviewers object to pooling decisions made after seeing the forest plot.
Five roles, though not necessarily five people. Two independent screeners, because dual screening roughly halves the number of eligible studies wrongly excluded and is what Cochrane requires. An information specialist for the search, which has more influence on the review than any other stage. A content expert who can distinguish a clinically relevant comparison from an irrelevant one. A statistician if any results will be pooled. And a person responsible for the protocol and the PROSPERO record. On our projects these roles are named in the handover, so you can see who completed each stage.
Automation-assisted prioritisation — ranking records so the likely includes surface first — is well established and cuts title and abstract time substantially. Using a large language model to make include or exclude decisions unsupervised is a different thing and is not yet defensible at the screening stage of a review that will be published.
What matters either way is disclosure: journals and PRISMA both now expect you to state which tool was used, at which stage, with what human oversight, and ICMJE added a section on artificial intelligence in publishing in its January 2026 update. We use prioritisation, we say so in the methods, and two humans still read the included set.
Disagreement is the point of screening twice, so it is recorded rather than quietly resolved. Conflicts at title and abstract go to a third reviewer or to discussion, with the rule agreed in the protocol before screening starts.
At full text, every exclusion is recorded with its reason, because those reasons appear in the PRISMA flow diagram and an editor will check that they add up. We report the agreement statistic where the journal expects one, while noting that a high kappa on an easy set of records means less than it appears; what a reviewer actually wants to see is that the disagreements existed and were adjudicated on a stated criterion.
Related services
Related Research Services
Services that support each stage before, during and after a systematic review.
1BEFORE THE REVIEWconfirming the review is needed
Literature review and gap analysisWhether the question has been answered already, and where the evidence actually stops. Experimental designWhen the review finds the evidence missing, the next thing to design is the study that fills it.2INSIDE THE REVIEWspecialist analysis
Meta-analysisPooled estimates, heterogeneity, subgroup and sensitivity analyses, and the model justified rather than defaulted to. Biostatistics and statistical programmingNetwork meta-analysis, dose-response modelling and reproducible code a reviewer can re-run.3AFTER THE REVIEWgetting it published
Journal selectionWhich journals still publish reviews in your field, and which of them your evidence actually fits. Original research articleWhere the review becomes a paper in its own right rather than a chapter of one. Pre-submission peer reviewMethodological issues identified before submission, while the search can still be re-run.Insights
Further Reading on Systematic Review Methodology and Reporting
Three pieces from the Pubrica library on how a systematic review is built and reported.
Request a Free Scoping Search for Your Review Question
Whether your review has already been done, whether there is enough evidence to synthesise, and how many records the question returns — answered in two business days, at no cost and with no obligation.