Rigor Reviewed

Rigor-Reviewed Manuscript

AI-based triage and decision support in mammography and digital tomosynthesis for breast cancer screening: a paired, noninferiority trial.

Name: Adcurare Rigor Record — AI-based triage and decision support in mammography and digital tomosynthesis for breast cancer screening: a paired, noninferiority trial.
Item: AI-based triage and decision support in mammography and digital tomosynthesis for breast cancer screening: a paired, noninferiority trial.
Rating: 59
Author: Adcurare

Elías-Cabot E, Romero-Martín S, Raya-Povedano JL, Rodríguez-Ruiz A, Álvarez-Benito M

Paper source

Read full text (PMC)Journal

Embed this Rigor badge

Adcurare Rigor Review — 4★ (Strong), 0 critical, 7 high issues for “AI-based triage and decision support in mammography and digital tomosynthesis for breast cancer screening: a paired, noninferiority trial.”

Links back to this Rigor Record.

Adcurare Editorial

This rigor review was examined and confirmed by Adcurare Editorial · July 6, 2026

Adcurare Rigor Review

Rigor Reviewed

4 / 5

Strong — minor issues to address

Interventional Study

Critical

High-priority

How this rating was calculated▸

Started at 5★ — deductions

IntegrityIntegrity concern−0.5★
ReportingStudy design partially met−0.25★
ReportingBiological variables partially met−0.25★

Model-reported confidence: highReviewer agreement: moderateEngine: 7.6.0Source: full text

Not verified in this report

No reported statistical tests were found to recompute.

Problems surfaced3 types

Biological variables underreported (sex, age, strain)
Study-design details incomplete (controls, blinding, power)
Internal contradictions in the reported numbers

This Adcurare Rigor Review uses AI Rigor Reviewers trained on a curated corpus of high-fidelity and retracted papers, with expert supervision and curation. It can still make mistakes; verify each finding against the source before relying on it.

The paper presents a well-conceived prospective trial with generally sound methodology, but several reporting gaps reduce confidence: blinding of radiologists is not described, missing-data handling is not pre-specified, biological variable reporting (weight, health status, race/ethnicity) is incomplete, ethical compliance lacks explicit reference to international standards, and the statistical analysis code is not shared. A minor typo in Table 3 and imprecise p-values should also be corrected.

Three independent reviewers evaluated all eight dimensions, supplemented by a copyedit pass and verification components (citations, statistics, reproducibility, preregistration, integrity, claim audit). The statistics verification had 0 tests recomputed (no test statistics with df or effect estimates with CIs were identified), so statistical correctness is not verified beyond internal consistency. The integrity flag for Table 3 denominator discrepancy is addressed in actions.

12 major claims checked against the paper's own evidence: all adequately supported.

1 integrity concern flagged (0 high).

Reporting

Eight Dimensions6 pass · 2 warn

The introduction cites numerous retrospective and prospective studies on AI in breast cancer screening, acknowledges their strengths and limitations (e.g., retrospective nature, simulation studies), and provides a logical rationale for the current prospective trial. The hypothesis follows directly from the cited evidence, and the study addresses limitations of prior work by using a prospective design.

Sub-criteria

prior work citedADEQUATE

premise rationaleADEQUATE

limitations addressedADEQUATE

Per reviewer (3) — the votes behind the verdict

PASSReviewer 1· 95% conf

The scientific premise is well-established, with extensive prior work cited and a clear rationale linking the hypothesis to the study objectives.

Evidence

direct quote[Introduction, paragraph 1-4]

“Over the last few years, several retrospective studies have shown that radiologists improved their cancer detection accuracy when using an AI system as concurrent reading support.”

direct quote[Introduction, paragraph 5]

“The Artificial Intelligence in Breast Cancer Screening Program in Córdoba (AITIC) clinical trial is a prospective, paired, noninferiority, accuracy study (NCT04849776) that aims to confirm prospectively the results obtained in these studies by evaluating whether an AI system can be used to omit human reading completely in a large proportion of the screening exams classified as low risk for cancer.”

direct quote[Introduction, paragraph 4]

“Several prospective studies have been designed to overcome the limitations of retrospective studies and to address and quantify the real-life impact of AI in breast cancer screening.”

PASSReviewer 2· 95% conf

The paper clearly establishes the scientific foundation, citing prior retrospective and prospective AI screening studies and addressing the need for prospective validation.

Evidence

direct quote[Introduction, paragraph 4]

“Several prospective studies have been designed to overcome the limitations of retrospective studies and to address and quantify the real-life impact of AI in breast cancer screening”

direct quote[Introduction, paragraph 5]

direct quote[Introduction, paragraph 4]

“Several prospective studies have been designed to overcome the limitations of retrospective studies”

PASSReviewer 3· 95% conf

The paper establishes a clear scientific foundation by citing prior work on AI in breast cancer screening and explaining how this prospective trial addresses limitations of retrospective studies.

Evidence

direct quote[Main text, Introduction, paragraphs 1-5]

“Over the last few years, several retrospective studies have shown that radiologists improved their cancer detection accuracy when using an AI system as concurrent reading support... Several prospective studies have been designed to overcome the limitations of retrospective studies...”

direct quote[Main text, Introduction, paragraph 5]

“This has been simulated retrospectively in two large-scale studies with DBT population screening, demonstrating a negative predictive value of AI of 99.91% (ref. 17) and 99.97% (ref. 18) for the screening exams classified by AI as lowest risk.”

direct quote[Main text, Introduction, paragraph 6]

“This paired, noninferiority trial prospectively evaluates the use of an AI system in a real-world population-based breast cancer screening program for the autonomous and automated reading of a large number of examinations without requiring radiologist involvement in terms of workload, CDR and RR.”

The study uses a paired design, so randomization of participants to strategies is not applicable. Power analysis and inclusion/exclusion criteria are adequately reported. However, blinding of radiologists to the AI strategy is not described (only 'standard double-blind reading' is mentioned without elaboration), and no outlier handling or missing data approach is stated. These omissions reduce the assessment of rigor.

Sub-criteria

controlsN/A

power analysisADEQUATE

blinding levelsNOT REPORTED

outlier handlingNOT REPORTED

randomization unitN/A

inclusion exclusionADEQUATE

randomization methodN/A

replicate distinctionN/A

independent replicationN/A

Per reviewer (3) — the votes behind the verdict

WARNReviewer 1· 85% conf

The study design is generally sound, but several aspects of experimental rigor are inadequately reported, particularly blinding of readers and handling of outliers.

Evidence

direct quote[Methods, paragraph 7]

“Screening readings were randomly assigned according to availability of the nine dedicated breast radiologists in the hospital’s radiology department (3 to 21 years of experience in breast cancer screening), regardless of whether they were readings with or without AI assistance and regardless of the reader’s experience.”

direct quote[Methods, Sample size]

“An alpha (type I error estimate) value of 5% and a beta (type II error estimate) of 20% (representing a power for the study of 80%, 1 − beta) was used.”

direct quote[Methods, Clinical trial design and patients]

“After signing informed consent, women were excluded from the clinical trial if they had symptoms or signs of suspected breast cancer; if they had breast prostheses; or if their images could not be processed by the AI system.”

WARNReviewer 2· 85% conf

The design is adequately described with a power analysis, inclusion/exclusion criteria, and a paired noninferiority framework, but lacks explicit blinding details and missing-data handling.

Evidence

direct quote[Methods, Procedures]

“Screening readings were randomly assigned according to availability of the nine dedicated breast radiologists in the hospital's radiology department, regardless of whether they were readings with or without AI assistance”

direct quote[Methods, Sample size]

“An alpha (type I error estimate) value of 5% and a beta (type II error estimate) of 20% (representing a power for the study of 80%, 1 − beta) was used... the sample size needed was defined as 27,000 women”

direct quote[Methods, Clinical trial design and patients and Figure 1]

“women were excluded from the clinical trial if they had symptoms or signs of suspected breast cancer; if they had breast prostheses; or if their images could not be processed by the AI system”

PASSReviewer 3· 85% conf

The study uses a paired noninferiority design with pre-specified sample size, clear inclusion/exclusion criteria, and appropriate control arm. Blinding is not described but the open-label nature is implied. Outlier handling is adequate via the flow chart.

Evidence

direct quote[Methods, Sample size]

“For the sample size calculation, the method by Connor for paired studies was used... sample size needed was defined as 27,000 women in order to demonstrate the noninferiority of the AI strategy over the standard strategy globally.”

direct quote[Methods, Clinical trial design and patients]

“Women participants aged 50 to 71 years old... were eligible... women were excluded from the clinical trial if they had symptoms or signs of suspected breast cancer; if they had breast prostheses; or if their images could not be processed by the AI system.”

absence[Methods, Clinical trial design and patients]

No mention of blinding or open-label justification.

All participants are women (justified by breast cancer screening context). Median age and interquartile range are provided. Breast density and screening round are given. However, weight is not reported, health status is only indirectly inferred, and race/ethnicity was not individually recorded (only noted as majority Caucasian). Comorbidities are not mentioned.

Sub-criteria

demographicsINADEQUATE

sex reportedADEQUATE

sex justifiedADEQUATE

age weight healthINADEQUATE

housing conditionsN/A

species strain sourceN/A

Per reviewer (3) — the votes behind the verdict

PASSReviewer 1· 95% conf

Biological variables and subject characteristics are well-reported, including sex, age, breast density, and screening history.

Evidence

direct quote[Table 1]

“Women | 31,301 (100.0%)”

direct quote[Table 1]

“Age at screening, years, median (interquartile range) | 59 (54–64)”

direct quote[Table 1]

“Breast density (BI-RADS 5th edition) | A (almost entirely fatty) | 6,445/31,301 (20.6%) | B (scattered fibroglandular tissue) | 14,569/31,301 (46.5%) | C (heterogeneously dense) | 8,625/31,301 (27.6%) | D (extremely dense) | 1,662/31,301 (5.3%)”

WARNReviewer 2· 85% conf

Sex, age, and breast density are reported, but weight and health status are missing, and demographics are incomplete (race/ethnicity not individually recorded, no comorbidities).

Evidence

direct quote[Methods, Clinical trial design and patients]

“All women aged 50 to 71 years, regardless of socioeconomic status, were considered.”

paraphrase[Results and Table 1]

“Table 1 reports age, screening round, breast density, but no race/ethnicity or comorbidity data.”

direct quote[Methods, Procedures]

“Race and ethnicity were not individually recorded, but the majority of the target population was Caucasian.”

WARNReviewer 3· 80% conf

Age and sex are reported, but weight and health status are not reported. Race/ethnicity is mentioned only descriptively as 'majority Caucasian'. No justification is given for studying only women (single-sex), though this is inherent in breast cancer screening. Overall, reporting of biological variables is incomplete.

Evidence

direct quote[Table 1]

“Women | 31,301 (100.0%)”

direct quote[Table 1]

“Age at screening, years, median (interquartile range) | 59 (54–64)”

direct quote[Methods, Clinical trial design and patients]

“Race and ethnicity were not individually recorded, but the majority of the target population was Caucasian.”

The study received approval from the IRB at Reina Sofía University Hospital of Córdoba Research Ethics Committee (IRB No 4932), written informed consent was obtained from all participants, and compliance with HIPAA and data protection regulations is stated.

Sub-criteria

iacuc statementN/A

informed consentADEQUATE

irb ethics statementADEQUATE

regulatory complianceADEQUATE

Per reviewer (3) — the votes behind the verdict

PASSReviewer 1· 95% conf

Ethical approvals and consent are adequately reported, with a named IRB, protocol number, informed consent, and regulatory compliance.

Evidence

direct quote[Methods, Ethics approval]

“In March 2021, the study protocol and the informed consent received a favorable ruling from the Institutional Review Board (IRB) at Reina Sofía University Hospital of Córdoba Research Ethics Committee (IRB No 4932).”

direct quote[Abstract, Methods, Ethics approval]

“All patients provided written informed consent before enrollment.”

direct quote[Methods, Ethics approval]

“This prospective clinical trial was compliant with the Health Insurance Portability and Accountability Act and the design was preregistered at https://classic.clinicaltrials.gov/ct2/show/NCT04949776 .”

PASSReviewer 2· 95% conf

Ethics approval from a named institutional review board, written informed consent, and regulatory compliance are reported adequately.

Evidence

direct quote[Methods, Ethics approval]

“the study protocol and the informed consent received a favorable ruling from the Institutional Review Board (IRB) at Reina Sofía University Hospital of Córdoba Research Ethics Committee (IRB No 4932).”

direct quote[Methods, Ethics approval]

“All patients provided written informed consent before enrollment.”

direct quote[Methods, Ethics approval]

“This prospective clinical trial was compliant with the Health Insurance Portability and Accountability Act”

WARNReviewer 3· 90% conf

Ethics approval is documented with a named IRB and protocol number, and informed consent is described. However, regulatory compliance is not explicitly stated with reference to a recognized framework such as the Declaration of Helsinki or ICH-GCP, only HIPAA compliance is mentioned.

Evidence

direct quote[Methods, Ethics approval]

“All patients provided written informed consent before enrollment.”

direct quote[Methods, Ethics approval]

“This prospective clinical trial was compliant with the Health Insurance Portability and Accountability Act”

The investigational AI system (Transpara version 1.7, ScreenPoint Medical) is named with manufacturer and version. The statistical software (R version 4.4.2) and libraries are identified. No other biological/chemical resources are used, so the remaining sub-criteria are not applicable.

Sub-criteria

mycoplasma testingN/A

reagents identifiedADEQUATE

organisms identifiedN/A

antibodies identifiedN/A

cell line authenticationN/A

software tools identifiedADEQUATE

Per reviewer (3) — the votes behind the verdict

PASSReviewer 1· 95% conf

Key resources, including the AI system and statistical software, are adequately identified.

Evidence

direct quote[Methods, Procedures]

“This clinical trial used a commercially available AI system for breast cancer detection, Transpara (version 1.7 ScreenPoint Medical).”

direct quote[Methods, Statistical analyses / Code availability]

“The software used to perform the statistical analyses in this study was R, version 4.4.2.”

PASSReviewer 2· 90% conf

The AI system (Transpara v1.7, ScreenPoint Medical) is identified with vendor and version; R software with version is reported for statistical analyses.

Evidence

direct quote[Methods, Procedures]

“This clinical trial used a commercially available AI system for breast cancer detection, Transpara (version 1.7 ScreenPoint Medical)”

direct quote[Methods, Code availability and Methods, Statistical analyses?]

“The software used to perform the statistical analyses in this study was R, version 4.4.2.”

PASSReviewer 3· 95% conf

The AI system (Transpara version 1.7, ScreenPoint Medical) is identified as the investigational product, and mammography devices are specified. Statistical software (R) is identified. No biological reagents or cell lines are used, making other sub-criteria not applicable.

Evidence

direct quote[Methods, Procedures]

“This clinical trial used a commercially available AI system for breast cancer detection, Transpara (version 1.7 ScreenPoint Medical)”

direct quote[Methods, Procedures]

“Images were acquired using four devices: three DM devices (Lorad Selenia, Hologic) and one DBT device (3Dimensions, Hologic).”

direct quote[Code availability]

“The software used to perform the statistical analyses in this study was R, version 4.4.2.”

Statistical tests (McNemar, Wald, z-test) are named. Effect sizes and 95% CIs are reported for primary and secondary outcomes. R version is given. Data presentation includes per-group n, proportions, and CIs. Exact p-values are sometimes reported as inequalities (e.g., P < 0.001) rather than exact numbers. The mathematical plausibility is mostly fine, but Table 3 lists 13,986 DBT instead of 13,968 (a typo).

Sub-criteria

tests namedADEQUATE

exact p valuesINADEQUATE

effect sizes ciADEQUATE

data presentationADEQUATE

software identifiedADEQUATE

assumptions verifiedADEQUATE

mathematical plausibilityWARN

Per reviewer (3) — the votes behind the verdict

PASSReviewer 1· 85% conf

Statistical analysis is largely adequate, with most tests named, effect sizes and CIs reported, and data presented appropriately. However, exact p-values are not always provided.

Evidence

direct quote[Methods, Statistical analyses]

“For the primary outcomes (CDR and RR), we conducted noninferiority testing using the following approach... One-sided 97.5% CIs were established for the difference in proportions... If noninferiority was demonstrated for a given outcome, we proceeded to test for superiority using a paired exact McNemar test.”

direct quote[Results, Table 2]

“CDR | 228/31,301 | 7.3 | 198/31,301 | 6.3 | 1 | 15.2 | P < 0.001 | ... (95% CI) columns”

direct quote[Results, Table 2]

“P < 0.001”

PASSReviewer 2· 90% conf

Most statistical analysis criteria are adequately reported: tests named, effect sizes with CIs, software identified, and data presentation is clear. Exact p-values are partially reported, and a minor denominator typo exists in Table 3.

Evidence

direct quote[Methods, Statistical analyses]

“one-tailed McNemar paired tests were applied for the noninferiority analyses... Wald statistics with or without the Bonett–Laplace adjustment”

direct quote[Results, Table 2]

“P < 0.001”

direct quote[Results, Table 3]

“4,820/13,986”

PASSReviewer 3· 90% conf

Statistical tests are named, effect sizes and CIs are reported, software is identified, and data presentation is adequate. However, test assumptions are not explicitly verified, and p-values are sometimes reported as inequalities (P < 0.001) rather than exact values. The numerical results are internally consistent.

Evidence

direct quote[Methods, Statistical analyses]

“One-sided 97.5% CIs were established for the difference in proportions (AI standard) for the CDR and RR... Noninferiority was declared if the lower limit of the one-sided CI exceeded the predefined noninferiority margin.”

direct quote[Table 2]

“P < 0.001 bc”

absence[Methods, Statistical analyses]

No explicit verification of test assumptions reported.

The data availability statement explicitly names a repository (Zenodo) with a persistent identifier and deposit. The AI system code is proprietary, but the analysis code is not considered bespoke. The statistical software is identified, and no custom code was shared, which is acceptable for a clinical trial using standard methods.

Sub-criteria

code sharingN/A

accession numbersADEQUATE

repository depositADEQUATE

data availability statementADEQUATE

Per reviewer (3) — the votes behind the verdict

PASSReviewer 1· 95% conf

Data availability is adequate, with the deidentified participant dataset deposited in Zenodo with a DOI. Code sharing is not applicable as the analysis used standard software.

Evidence

direct quote[Data availability]

“Individual deidentified participant dataset, data dictionary defining each field, protocol and informed consent form-patients information sheet are publicly accessible via Zenodo at 10.5281/zenodo.17625633 (ref. ).”

direct quote[Data availability]

“Zenodo at 10.5281/zenodo.17625633”

direct quote[Code availability]

“The code for training and developing the evaluated AI algorithm (Transpara version 1.7, ScreenPoint Medical) is part of a proprietary system.”

PASSReviewer 2· 95% conf

The individual deidentified participant dataset is deposited in Zenodo with a DOI, and the data availability statement is concrete. Code sharing is not applicable because no bespoke analysis code was written beyond standard R libraries.

Evidence

direct quote[Data availability section]

direct quote[Code availability section]

“The code for training and developing the evaluated AI algorithm (Transpara version 1.7, ScreenPoint Medical) is part of a proprietary system.”

WARNReviewer 3· 90% conf

The deidentified participant dataset is publicly available via Zenodo with a DOI, which is commendable. However, the statistical analysis code (R scripts) is not shared; only the software version and libraries are mentioned. The AI algorithm code is proprietary and not shared.

Evidence

direct quote[Data availability]

paraphrase[Code availability]

“Paraphrase: The analysis software is R version 4.4.2 with libraries, but no custom script is provided.”

direct quote[Code availability]

“The code for training and developing the evaluated AI algorithm (Transpara version 1.7, ScreenPoint Medical) is part of a proprietary system.”

The study is registered at ClinicalTrials.gov (NCT04849776), methods are detailed enough for replication, a reporting guideline is referenced (Nature Portfolio Reporting Summary), all outcomes appear to be reported, limitations are discussed, conclusions are proportional to the evidence, and funding and competing interests are disclosed.

Sub-criteria

funding coiADEQUATE

trial registrationADEQUATE

reporting guidelineADEQUATE

methods completenessADEQUATE

all outcomes reportedADEQUATE

limitations discussedADEQUATE

conclusions proportionalADEQUATE

Per reviewer (3) — the votes behind the verdict

PASSReviewer 1· 95% conf

Reporting is transparent and complete, with trial registration, detailed methods, discussion of limitations, and proportional conclusions.

Evidence

direct quote[Abstract, Methods]

“ClinicalTrials.gov: NCT04849776 (http://clinicaltrials.gov/ct2/show/NCT04849776) .”

direct quote[Discussion, paragraph 9]

“Our study has the limitation of being a single-site investigation with a screening workflow of double reading without arbitration by expert radiologists in breast imaging and breast screening who have several years of experience in the use of AI.”

direct quote[Acknowledgements, Competing interests]

“In March 2023, EEC received for this study a grant (20,000 euros) from the SEDIM Foundation. The funder had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript. ... A.R.R. is an employee of ScreenPoint Medical. The other authors declare no competing interests.”

PASSReviewer 2· 95% conf

The manuscript is transparent: methods are detailed, trial is registered, primary and secondary outcomes are reported, limitations are discussed, conclusions are proportional, and funding/COI are declared. The only gap is not explicitly naming a reporting guideline (e.g., CONSORT), though the journal's reporting summary is used.

Evidence

direct quote[Abstract and Methods]

“ClinicalTrials.gov: NCT04849776”

direct quote[Discussion, paragraph 8]

“Our study has the limitation of being a single-site investigation with a screening workflow of double reading without arbitration”

direct quote[Methods, Reporting summary]

“Further information on research design is available in the linked to this article.”

PASSReviewer 3· 95% conf

The manuscript provides a thorough description of methods, is registered, reports all outcomes, discusses limitations, and includes funding and COI statements. It does not explicitly reference a reporting guideline (e.g., CONSORT) beyond a Nature Portfolio reporting summary.

Evidence

direct quote[Abstract / Methods]

“ClinicalTrials.gov: NCT04849776”

direct quote[Discussion]

“Our study has the limitation of being a single-site investigation with a screening workflow of double reading without arbitration by expert radiologists... Further studies are needed to validate these findings in other settings.”

paraphrase[Methods, Reporting summary]

“Paraphrase: The paper states 'Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article', but does not name a specific guideline.”

5 copyedit issues flagged: mostly consistency, clarity.

Checked 33 references: 28 verified — 5 not checked.

Reproducibility

4 data/code links checked; 4 live.

Registered (2 IDs: ClinicalTrials.gov). No reporting guideline cited.

Adcurare Rigor Action ItemsFinal pass before journal submission

Ready after minor to moderate edits. The most consequential rigor gaps (blinding description, missing-data handling, code sharing) should be addressed before submission. The copyedit issues (NTC typo, Table 3 denominator) must be fixed. With these changes, the manuscript is suitable for journal submission.

1.
HIGHrigor
Add a statement describing blinding of radiologists to the AI strategy and AI output, or explicitly state that blinding was not feasible and discuss potential bias (Methods, Study design/Procedures).
Lack of blinding description is a major rigor gap that reviewers will flag; transparency about the open-label nature is essential.
2.
HIGHrigor
Pre-specify the handling of missing data and define the analysis population (ITT vs per-protocol) in the Methods (Statistical analyses section).
Without this, the analysis approach is ambiguous and the study's conclusions are less robust.
3.
HIGHdata code
Deposit the statistical analysis R scripts (or equivalent) in a public repository (e.g., Zenodo or GitHub) alongside the dataset, with a reference in the Code availability statement.
Full reproducibility requires the analysis code; its absence limits the utility of the open dataset.
4.
HIGHcopyedit
Correct the registration number typo: change 'NTC04949776' to 'NCT04849776' wherever it appears (Methods, Clinical trial design).
An incorrect registration number undermines traceability and could lead to rejection.
5.
HIGHcopyedit
Correct the denominator in Table 3 for DBT screening readings: change 13,986 to 13,968 to match the text and Table 1.
This internal inconsistency (also flagged by integrity verification) must be resolved before submission.
6.
HIGHreporting
Provide exact p-values (e.g., P=0.0007) instead of inequalities (P<0.001) for all primary outcomes in Table 2, or add a footnote explaining the convention.
Threshold p-values reduce informativeness and are considered inadequate by most journals.
7.
HIGHethics
Add a statement of compliance with the Declaration of Helsinki (or ICH-GCP) in the Ethics approval section (Methods).
Current HIPAA compliance alone is insufficient for an international clinical trial; this omission may concern reviewers.
8.
MEDIUMrigor
Report participants' weight and health status (or explicitly state that these were not collected and acknowledge as a limitation).
Biological variable reporting is incomplete; weight and health status are standard demographic variables.
9.
MEDIUMrigor
Provide individual-level race/ethnicity data if collected, or justify their omission and discuss generalizability limitations (Methods/Table 1).
The current descriptive statement ('majority Caucasian') is insufficient for transparency of the study population.
10.
MEDIUMreporting
Explicitly cite the CONSORT noninferiority trial checklist (or STROBE) in the Methods, and provide the completed checklist as a supplementary file.
While a Nature Portfolio Reporting Summary is linked, naming the guideline improves transparency and meets journal expectations.
11.
MEDIUMrigor
Discuss the open-label nature of the AI strategy and its potential impact on radiologist behavior in the Discussion (Limitations section).
Acknowledging this bias strengthens the paper's credibility and preempts reviewer criticism.
12.
LOWcopyedit
Rephrase 'not noninferior' in the Abstract to 'did not meet noninferiority criteria' for clarity.
Minor clarity improvement for a broader audience.

Full text (56,170 chars)Analyzed: Jul 6, 2026

Read alongside