Published
The PRISMA-DTA checklist has 27 items, and most review teams find out what several of them mean only when a peer reviewer asks where the answer is. This guide applies every item to one real, open-access diagnostic test accuracy review, names the section where each item is answered, and marks the rows that need a judgement call. You will see what a complete answer looks like and where even a careful team leaves a small gap.
The PRISMA-DTA checklist extension was published by McInnes and colleagues in JAMA in 2018 and builds on the general PRISMA statement. If you need the general version first, read our PRISMA 2020 guide and then come back for the diagnostic items. This page covers the diagnostic test accuracy (DTA) extension only, and the full item wording is in the official PRISMA-DTA checklist, which is also listed on the EQUATOR Network page for PRISMA-DTA.
In Simple Words
PRISMA-DTA is the reporting checklist for systematic reviews that pool how well a test detects a disease. It asks for everything a standard systematic review reports, plus the points that matter for tests, such as the clinical role of the test, the unit you count, how you handled different cut-offs, and the 2 x 2 table of every study. Below, all 27 items are checked against a real published review, so you can see what to write and where to put it.
What the PRISMA-DTA checklist adds to PRISMA
A diagnostic accuracy review asks a different question from a treatment review. Instead of participants, intervention, comparison and outcome, it asks about participants, the index test and the target condition, often shortened to PIT. The data behave differently too. Sensitivity and specificity move together as the positivity threshold changes, studies use different cut-offs, several readers may interpret the same test, and some results come back indeterminate. The PRISMA-DTA group wrote the checklist to make those choices visible.
The extension came from a systematic review of reporting problems, a three-round Delphi process and an in-person consensus meeting, and it produced 27 items plus a separate checklist for abstracts, as McGrath and colleagues describe. The official checklist numbers its rows 1 to 27, labels two DTA-only rows D1 and D2, and has no rows numbered 15 or 22. The explanation and elaboration paper in the BMJ gives the reasoning behind each row.
Reporting the meta-analysis method matters more than it first appears. Hierarchical pooling has been shown to give results that differ from non-hierarchical pooling, so reviewers expect to see exactly which model you fitted. In an assessment of 100 DTA reviews, the clinical setting and the definitions of the target condition, the index test and the reference standard were reported in fewer than one third of them. Those are the rows where this guide spends the most time.
The commentary by the same group is direct about scope. It states, “This extension of the PRISMA statement should be used for all DTA systematic reviews performed moving forward.” (McGrath et al., Systematic Reviews, 2019, CC BY 4.0).
How we chose and checked the example
We searched PubMed for diagnostic accuracy reviews that state adherence to PRISMA-DTA and offer full text without a paywall. We kept one whose licence we could confirm three ways. Crossref lists the version of record under CC BY 4.0, Europe PMC flags the article as CC BY, and the licence statement inside the full-text XML names the Creative Commons Attribution License. The article also ships its own PRISMA-DTA checklist as supplementary file S2, which let us compare our reading with the authors’ page mapping.
Each row of the PRISMA-DTA checklist below carries the section where the answer sits, the evidence in the article, and a call. Reported means the item is answered where the checklist expects it. Judgement call means two careful readers could score the row differently, or the numbers in the article need reconciling before they can be trusted. Every quotation is 25 words or fewer, sits in quotation marks, and was checked against the full text. This is a reporting check, so we did not re-analyse the data or assess the authors’ conclusions.
The review behind this PRISMA-DTA checklist example
The article is “Circulating microRNAs as biomarkers for diabetic retinopathy stage identification, a DTA systematic review and meta-analysis” by Martínez-Santos and colleagues, published in PLOS One in November 2025. It pools 16 studies with 1849 participants and 21 microRNAs across three comparisons, which are healthy controls against diabetic retinopathy, type 2 diabetes without retinopathy against retinopathy, and non-proliferative against proliferative disease. It uses a bivariate random-effects model, QUADAS-2 for risk of bias and GRADE for certainty of evidence, and its Methods state adherence to PRISMA-DTA.
The review suits a worked example because it has a real flow diagram, 2 x 2 data for every study, a per-study risk-of-bias grid, pooled estimates with intervals, subgroup and sensitivity analyses, and GRADE tables. A review without a meta-analysis would leave item D2 and the pooled-results part of item 21 untested. Its gaps are small and typical, which is what makes them useful.
PRISMA-DTA checklist items 1 to 4 and D1 (title, abstract, introduction)
The opening rows of the PRISMA-DTA checklist are the easiest to satisfy, yet D1 is the one most often thin. It asks for the intended use and clinical role of the test, and for the accuracy that would make the test worth using.
| Item | What the item asks | Where in the article | Evidence | Call |
|---|---|---|---|---|
| 1 | Identify the report as a systematic review, with or without meta-analysis, of diagnostic test accuracy studies. | Title | The published title ends with “A DTA systematic review and meta-analysis”. | Reported |
| 2 | Provide a structured abstract that follows PRISMA-DTA for abstracts. | Abstract and S1 Table | The abstract names the design, the guideline, six databases with a last search date of January 2025, QUADAS-2, GRADE and pooled sensitivity and specificity with intervals. The separate abstract checklist is supplied as S1 Table, and we did not score its items here. One sentence about the bivariate model appears twice in the abstract Methods, a copyediting fix that costs nothing. | Judgement call |
| 3 | Describe the rationale for the review in light of what is already known. | Introduction, final two paragraphs | The authors state that an earlier meta-analysis “did not perform any stratification by disease stage or type of control group”, and build their three comparisons on that gap. | Reported |
| D1 | State the clinical background, the intended use and clinical role of the index test, and the minimally acceptable accuracy where relevant. | Introduction, paragraphs 2 and 3 | The role is clear. Fluorescein angiography is described as invasive and resource-intensive, and the authors call for “non-invasive, accessible biomarkers that could support early detection and improve screening and risk stratification”. No minimum sensitivity or specificity that would make the test worth using is stated. | Judgement call |
| 4 | State the question in terms of participants, index test and target condition. | Introduction, final paragraph, and Methods | The aim is “to evaluate the diagnostic accuracy of circulating miRNAs in DR”, and the Methods define population, index test and target condition under the PIT structure. | Reported |
PRISMA-DTA checklist items 5 to 16 (methods)
The Methods block of the PRISMA-DTA checklist holds twelve rows and most of the DTA-specific content. Rows 13, 14 and D2 are where a DTA review differs most from a standard one, so read them closely. For a refresher on handling cut-offs, see our note on threshold effects, and for pooling between-study variability see the heterogeneity guide.
| Item | What the item asks | Where in the article | Evidence | Call |
|---|---|---|---|---|
| 5 | Indicate whether a review protocol exists, where it can be found, and the registration number if there is one. | Methods, first paragraph, and Discussion | The Methods say “This systematic review was not registered.” and the authors repeat the point in their limitations. Their own supplementary checklist carries a question mark against this row and a footnote explaining that the protocol was not registered. The text does not say whether an unregistered written protocol existed. | Judgement call |
| 6 | Specify participants, setting, index test, reference standard, target condition and study design, plus report characteristics such as years and language, with a rationale. | Methods, eligibility criteria | Participants, index test (RT-qPCR, microarray or sequencing in serum, plasma, aqueous humor or extracellular vesicles), target condition and the reference (fluorescein angiography and fundus examination) are specified. The authors “included studies published from 2014 onward”. Language and publication status are not stated, and the search strings in S1 Text start in 2000, so the reason for the 2014 start is the missing piece. | Judgement call |
| 7 | List all information sources with dates of coverage and the date last searched. | Methods, search strategy | Six databases are named and “The last update of this review was on January 20, 2025.” The six grey-literature records from ARVO abstracts appear in the flow diagram and Results but are not listed as a source in the Methods. | Judgement call |
| 8 | Present the full search strategy for every database so it can be repeated. | Methods and S1 Text | The Methods point to “S1 Text in S1 File”. We opened that file. It gives full strings for PubMed, CENTRAL, Web of Science, Scopus, ClinicalTrials and ScienceDirect, with Boolean operators and date limits. | Reported |
| 9 | State the process for selecting studies. | Methods, study selection | Two reviewers “assessed studies for eligibility based on predefined inclusion and exclusion criteria”, disagreements went to consensus, and a third reviewer arbitrated when needed. | Reported |
| 10 | Describe data extraction, including piloting, duplication and any contact with investigators. | Methods, study selection and data extraction | The extracted fields are listed and the full dataset is in S3 Table. Independent duplicate extraction and contact with study authors are not stated, although duplication is stated for eligibility screening. | Judgement call |
| 11 | Provide definitions for the target condition, index test, reference standard and other characteristics such as clinical setting. | Methods, eligibility, and Table 1 | Retinopathy stages are tied to fluorescein angiography and fundus examination, and Table 1 lists specimen, platform and normalization for each study. The columns do not include clinical setting. | Judgement call |
| 12 | Describe how risk of bias and applicability were assessed. | Methods, quality assessment | QUADAS-2 was “applied independently by two reviewers across the four standard domains”, with disagreements settled by consensus. | Reported |
| 13 | State the principal accuracy measures and the unit of assessment, for example per patient or per lesion. | Methods and Results | Sensitivity, specificity and AUC are named. The unit of assessment is not stated, and some studies contribute more than one marker, because the sensitivity analysis refers to “miRNA-specific entries”. The unit decides how many data points enter the pooled model. | Judgement call |
| 14 | Describe how data were handled, including multiple thresholds, readers, indeterminate results, grouping of tests and different reference standards. | Methods, statistical analysis | Thresholds are addressed directly. The authors write “Since most studies did not report diagnostic thresholds, a hierarchical HSROC model was not applicable.” The three control-group comparisons are analysed separately. Readers, indeterminate results and mixed reference standards are not mentioned, and one line for each would close the row. | Judgement call |
| D2 | Report the statistical methods used for the meta-analysis. | Methods, statistical analysis | Pooling used “a bivariate random-effects model, recommended for diagnostic test accuracy meta-analyses”, with likelihood ratios, diagnostic odds ratio, SROC curves, Cochran Q and I-squared. The software is named as STATA 18 with the MIDAS package for diagnostic test accuracy meta-analysis. | Reported |
| 16 | Describe additional analyses such as subgroup, sensitivity and meta-regression, and say which were pre-specified. | Methods, statistical analysis | The meta-regressions are described as “pre-specified univariable meta-regressions”, and leave-one-out, influence diagnostics and Deeks funnel testing are listed. Pre-specification is stated for the meta-regression only. | Judgement call |
PRISMA-DTA checklist items 17 to 23 (results)
The Results rows of the PRISMA-DTA checklist test whether a reader can rebuild your numbers. Item 20 is the most demanding, because it asks for 2 x 2 data for every analysis, which is also what lets others re-run the meta-analysis. If you want to check pooled sensitivity and specificity from your own 2 x 2 tables, try the free diagnostic meta-analysis calculator or the diagnostic accuracy calculator.
| Item | What the item asks | Where in the article | Evidence | Call |
|---|---|---|---|---|
| 17 | Give the numbers screened, assessed and included, with reasons for exclusion, ideally in a flow diagram. | Results, first paragraph, and Figure 1 | Figure 1 shows identification, screening, eligibility and inclusion, and the exclusion reasons add to 38 (8 + 4 + 6 + 20). The duplicate step does not reconcile. The text says “After the removal of 270 duplicates, 180 records were screened” while the figure labels 270 as the records remaining. With 454 database records and 6 grey-literature records, neither reading arrives at 180. | Judgement call |
| 18 | For each study, give citations and key characteristics including participants, clinical setting, design, target condition, index test, reference standard, sample size and funding. | Results, Table 1 | Table 1 gives country, design, comparison, marker, specimen, platform, normalization and total sample size. Participant presentation, clinical setting and funding of the included studies are not tabulated in the main article. | Judgement call |
| 19 | Present risk of bias and applicability concerns for each study. | Results, Figure 2 | Figure 2 is a grid of 16 studies across four risk-of-bias domains and three applicability domains, with summary bars alongside. | Reported |
| 20 | For each analysis, report TP, FP, FN and TN with accuracy estimates and intervals, ideally with forest or ROC plots. | Results, Table 1 and Figure 3 | Table 1 lists TP, TN, FP and FN with sensitivity and specificity for each entry, and Figure 3 carries forest plots and SROC curves for all three comparisons. | Reported |
| 21 | Describe test accuracy and its variability, with pooled results and intervals if a meta-analysis was done. | Abstract and Results | Pooled sensitivity, specificity and I-squared are given for each comparison. The study counts differ by section. The abstract and the accuracy paragraphs give 7, 9 and 12 studies, while the characteristics paragraph gives “6 studies analyzed CTL vs DR” and 7 and 9 for the other two comparisons. The abstract also reports an AUC interval of 0.82 to 0.89 for the first comparison where the Results report 0.84 to 0.92. | Judgement call |
| 23 | Report the results of additional analyses such as subgroup, sensitivity and meta-regression analyses. | Results, Tables 2 to 4, Figures 5 to 7 | Subgroups by origin, specimen, sample size, expression pattern and normalization are in Tables 2 to 4, meta-regressions in Figure 6, influence and outlier checks in Figure 7, and Deeks funnel plots in Figure 5. | Reported |
PRISMA-DTA checklist items 24 to 27 (discussion and funding)
The closing rows of the PRISMA-DTA checklist are about interpretation and transparency. The certainty of evidence in row 24 is usually handled with GRADE, which our GRADE certainty of evidence guide explains, and publication bias testing for DTA data has its own methods covered in our note on publication bias methods.
| Item | What the item asks | Where in the article | Evidence | Call |
|---|---|---|---|---|
| 24 | Summarize the main findings, including the strength of evidence. | Results (GRADE), Tables 5 to 7, Discussion | GRADE tables for each comparison give pretest probability, pooled estimates and certainty. The Methods name Deeks funnel testing, while the GRADE paragraph refers to a “Contour-Enhanced Funnel Plot” analysis, so the named test should be made consistent. | Judgement call |
| 25 | Discuss limitations of the included studies and of the review process. | Discussion, limitations | The authors list that “the protocol was not registered in Prospero”, the concentration of studies in Chinese populations, case-control and cross-sectional designs, laboratory heterogeneity and the lack of reported cut-offs. | Reported |
| 26 | Interpret the results against other evidence and state implications for research and for the clinical role of the index test. | Discussion and Conclusion | The Discussion compares the work with two earlier reviews, and the Conclusion calls for markers to be “validated in well-designed prospective cohorts” and combined with established clinical assessment. | Reported |
| 27 | Describe funding sources and the role of funders. | Funding Statement | Internal and external funders are named, and “The funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.” | Reported |
What the 27 rows tell you
Across the 27 rows of the PRISMA-DTA checklist we scored 13 as Reported and 14 as Judgement call, and none as missing. That pattern is typical for a careful team. The review reports every section the checklist expects, and the calls cluster in rows that ask for specifics, such as the unit of assessment, the clinical setting and the handling of indeterminate results.
Four PRISMA-DTA checklist rows where reviews lose marks
Clinical role and minimum accuracy (D1)
Name the decision the test informs. Then, where you can, say what sensitivity and specificity would make the test worth using. The example gives the role, which is a non-invasive aid to staging, but it leaves the minimum unstated. One sentence turns a descriptive aim into a testable one.
Unit of assessment (item 13)
Say whether the unit is the patient, the sample, the lesion or the marker. In the example, some studies contribute more than one marker, so the unit decides how many entries enter the pooled model and which standard errors are valid. A reviewer who cannot tell will ask.
Thresholds, readers and indeterminate results (item 14)
Address every element of the row, even when the answer is that it does not apply. The example handles thresholds well and explains why a hierarchical model was not used. One further line each on readers, indeterminate results and mixed reference standards would close the row completely.
Study characteristics and setting (items 11 and 18)
Clinical setting and how participants presented influence accuracy, and the row asks for them by name. A column for each costs little in a characteristics table and removes a common reviewer comment.
Reconciliation, the pass that clears most judgement calls
Several of the judgement calls above are not missing content at all. They are numbers or names that differ between the abstract, the text, the tables and the figure. In this example the duplicate count, the number of studies per comparison, one AUC interval and the name of the funnel plot test each differ by section. These slips are cheap to fix before submission and costly to leave, because a reviewer who finds one mismatch starts checking everything else.
The fix is a single reconciliation pass in which every count is traced from the flow diagram through the abstract, the Results text and the tables, and every named method is matched between the Methods and the Results. Our PRISMA flow diagram service checks numbers for internal consistency at every stage, and you can run the same pass yourself with the workflow below.
A PRISMA-DTA checklist workflow you can copy
- Open the PRISMA-DTA checklist at the protocol stage. The commentary by McGrath and colleagues cites a study in which authors found reporting guidelines most useful early in the research process, and a protocol that already answers items 4 to 14 saves weeks later.
- Write the target condition, index test and reference standard definitions into a table before screening starts, so items 6 and 11 are answered by design.
- Decide the unit of assessment and the threshold rule in advance, and record both. Date-stamp or register the protocol, and use our PROSPERO registration steps if you choose PROSPERO.
- Extract TP, FP, FN and TN for every analysis into one master sheet, with one row per unique combination of index test, reference standard and threshold. This sheet feeds item 20 and the meta-analysis.
- Assess risk of bias and applicability with QUADAS-2 or its successor, and show the result for each study. Our QUADAS-2 guide walks through the four domains.
- Build the flow diagram from the master counts rather than from memory. Our PRISMA flow chart guide explains each box, and the Flowchart Pro tool below draws it for you.
- Complete the page numbers in the checklist last, then run the reconciliation pass on counts, intervals and method names.
If you are reporting a primary diagnostic accuracy study instead of a review, the right guideline is STARD, and our STARD vs TRIPOD guide explains how to choose between it and TRIPOD.
Frequently asked questions about the PRISMA-DTA checklist
How many items does the PRISMA-DTA checklist have?
It has 27 items, including two DTA-only rows labelled D1 and D2, and it comes with a separate checklist for abstracts.
Should I use PRISMA-DTA or the general PRISMA statement for a diagnostic review?
Use PRISMA-DTA when your review pools diagnostic test accuracy studies. The authors of the extension state that it should be used for all DTA systematic reviews, and the general PRISMA guidance remains useful for background on shared items such as search reporting.
Do I have to register a protocol to follow PRISMA-DTA?
Item 5 asks you to indicate whether a protocol exists and to give registration details if available. The example review answered by stating that it was not registered, which answers the first part of the item, though a registered protocol gives readers more to check.
Which risk-of-bias tool does PRISMA-DTA require?
Item 12 asks you to describe the method you used, and it does not name a tool. The example uses QUADAS-2. A successor, QUADAS-3, was published in Annals of Internal Medicine in 2026, and the same rule applies, which is to describe whichever tool you use and show per-study results.
Free Research Tools from Editverse
- GraphPro: build publication-ready charts, 25 types, with built-in t-test, ANOVA, correlation, and chi-square p-values. Free, no signup.
- StatPro: free statistical calculators and a test-selector tool covering group comparisons, categorical outcomes, time-to-event analysis, and diagnostic accuracy.
- Flowchart Pro: generate PRISMA 2020, PRISMA-ScR, CONSORT, STROBE, STARD, MOOSE, and ARRIVE flow diagrams directly in your browser.
Professional Support for a PRISMA-DTA Compliant Review
Writing to the PRISMA-DTA checklist is easier when someone has already traced every number and every method name through the manuscript. Two Editverse services map directly onto the rows covered above.
Evidence synthesis
$3,500
Protocol through publication support for a systematic review and meta-analysis, covering search strategy, data extraction, statistical analysis and manuscript writing, with a dedicated methodologist. 8 to 16 weeks depending on scope.
Flow diagram
$100
A PRISMA 2020 flow diagram built from your screening counts, with the numbers checked for internal consistency at every stage. Vector files and high-resolution PNG.
For a complete review from protocol to submission, see our systematic review writing service, which supports PRISMA-compliant evidence synthesis. Our work aligns with COPE, ICMJE and GPP 2022, and follows institutional and country-specific ethics requirements.
Using the PRISMA-DTA checklist well
The PRISMA-DTA checklist rewards specifics. The example review answers every section and still leaves its clearest gains in the rows that ask for the unit of assessment, the clinical setting, the handling of indeterminate results and the minimum useful accuracy. It also shows that most judgement calls are reconciliation problems, which a single careful pass can clear. Fill the PRISMA-DTA checklist early, define the test and the reference standard before screening, trace every count from the flow diagram to the abstract, and your reviewers will have little left to ask. For the shared items of any systematic review, return to the PRISMA 2020 guide.
References
- McInnes MDF, Moher D, Thombs BD, McGrath TA, Bossuyt PM, the PRISMA-DTA Group, et al. Preferred reporting items for a systematic review and meta-analysis of diagnostic test accuracy studies. The PRISMA-DTA statement. JAMA. 2018;319(4):388-396. doi:10.1001/jama.2017.19163. PMID 29362800.
- Salameh JP, Bossuyt PM, McGrath TA, Thombs BD, Hyde CJ, Macaskill P, et al. Preferred reporting items for systematic review and meta-analysis of diagnostic test accuracy studies (PRISMA-DTA). Explanation, elaboration, and checklist. BMJ. 2020;370:m2632. doi:10.1136/bmj.m2632. PMID 32816740.
- McGrath TA, Moher D, McInnes MDF. Steps toward more complete reporting of systematic reviews of diagnostic test accuracy. Preferred Reporting Items for Systematic Reviews and Meta-Analyses of Diagnostic Test Accuracy (PRISMA-DTA). Syst Rev. 2019;8(1):166. doi:10.1186/s13643-019-1090-9. PMID 31296260. PMCID PMC6624884.
- Martínez-Santos M, Ybarra M, Pires ME, Ceresoni C, Martínez-López E, Sancho-Pelluz J, et al. Circulating microRNAs as biomarkers for diabetic retinopathy stage identification. A DTA systematic review and meta-analysis. PLoS One. 2025;20(11):e0335434. doi:10.1371/journal.pone.0335434. PMID 41270025. PMCID PMC12637958.
- Whiting PF, Rutjes AWS, Westwood ME, Mallett S, Deeks JJ, Reitsma JB, et al. QUADAS-2. A revised tool for the quality assessment of diagnostic accuracy studies. Ann Intern Med. 2011;155(8):529-536. doi:10.7326/0003-4819-155-8-201110180-00009. PMID 22007046.
- Reitsma JB, Glas AS, Rutjes AWS, Scholten RJPM, Bossuyt PM, Zwinderman AH. Bivariate analysis of sensitivity and specificity produces informative summary measures in diagnostic reviews. J Clin Epidemiol. 2005;58(10):982-990. doi:10.1016/j.jclinepi.2005.02.022. PMID 16168343.
- Deeks JJ, Macaskill P, Irwig L. The performance of tests of publication bias and other sample size effects in systematic reviews of diagnostic test accuracy was assessed. J Clin Epidemiol. 2005;58(9):882-893. doi:10.1016/j.jclinepi.2005.01.016. PMID 16085191.
- Whiting PF, Tomlinson E, Rutjes AWS, Davenport CF, Yang B, Westwood ME, et al. QUADAS-3. A revised tool for the quality assessment of diagnostic test accuracy studies. Ann Intern Med. 2026;179(4):548-555. doi:10.7326/annals-25-02104. PMID 41698208.