Introduction: Our institution is part of a provincial program providing annual breast MRI screenings to high-risk women. We assessed how MRI experience, background parenchymal enhancement (BPE), and the amount of fibroglandular tissue (FGT) affect the biopsy-proven predictive value (PPV3) and accuracy for detecting suspicious MRI findings. Methods: From all high-risk screening breast MRIs conducted between 1 July 2011 and 30 June 2020, we reviewed all BI-RADS 4/5 observations with pathological tissue diagnoses. Overall and annual PPV3s were computed. Radiologists with fewer than ten observations were excluded from performance analyses. PPV3s were computed for each radiologist. We assessed how MRI experience, BPE, and FGT impacted diagnostic accuracy using logistic regression analyses, defining positive cases as malignancies alone (definition A) or malignant or high-risk lesions (definition B). Findings: There were 536 BI-RADS 4/5 observations with tissue diagnoses, including 77 malignant and 51 high-risk lesions. A total of 516 observations were included in the radiologist performance analyses. The average radiologist's PPV3 was 16 ± 6% (definition A) and 25 ± 8% (definition B). MRI experience in years correlated significantly with positive cases (definition B, OR = 1.05, p = 0.03), independent of BPE or FGT. Diagnostic accuracy improved exponentially with increased MRI experience (definition B, OR of 1.27 and 1.61 for 5 and 10 years, respectively, p = 0.03 for both). Lower levels of BPE significantly correlated with increased odds of findings being malignant, independent of FGT and MRI experience. Summary: More extensive MRI reading experience improves radiologists' diagnostic accuracy for high-risk or malignant lesions, even in MRI studies with increased BPE.
Diagnostic accuracy improved exponentially with increased MRI experience (definition B, OR of 1.27 and 1.61 for 5 and 10 years, respectively, p = 0.03 for both).
selected-observation regression without adequate confounding or dependence control cannot establish causal improvement or a distinct exponential learning curve
Lower levels of BPE significantly correlated with increased odds of findings being malignant, independent of FGT and MRI experience.
selected suspicious observations, unmodelled dependence and qualitative BPE measurement materially limit the adjusted association
There were 536 BI-RADS 4/5 observations with tissue diagnoses, including 77 malignant and 51 high-risk lesions.
directly reported pathology counts are internally coherent
The negative impact of increased BPE on the diagnostic accuracy of BI-RADS 4/5 lesions is evident, leading to unnecessary biopsies, heightened anxiety, and increased follow-up costs.
observational biopsy-yield associations do not identify downstream effects on unnecessary biopsies, anxiety or costs
Derived from the full evaluation — not a separate score.
Strengths
The study extends the cited reader-experience literature into a provincial high-risk screening population, rather than repeating the same diagnostic setting.
↳ Introduction; Discussion, reader-experience comparisons
The analysis distinguishes malignancy alone from malignant-or-high-risk pathology. This makes clear that the reported experience association depends on including high-risk lesions.
↳ Methods §2.3; Results §3.4
Results reports the null malignancy-only experience association, and Limitations explains that multiplicative odds follow from a positive logistic coefficient rather than independently demonstrating an exponential learning curve.
↳ Results §3.4; Limitations, first paragraph
Limitations
Adjustment for BPE and FGT does not separate experience from calendar time, case mix or reader identity. The 516 observations from six eligible readers are analysed without reader-level dependence modelling.
↳ Methods §2.3; Results §§3.3–3.4
The paper describes improved diagnostic accuracy, but measures pathology yield among already suspicious BI-RADS 4/5 observations. It does not estimate sensitivity, specificity or the effect of an education intervention.
↳ Abstract, Findings and Summary; Methods §2.3; Conclusions
Methods ends the cohort in June 2020, while annual PPV3 reporting and Discussion extend through June 2022. The text does not establish which window supports all reported analyses.
↳ Methods §2; Results §3.1; Discussion, first paragraph
The pathology-linked cohort and explicit outcome definitions support a descriptive contribution to high-risk breast-MRI screening research. The experience association is significant only when high-risk lesions join malignancies as positive outcomes, and its multiplicative form follows from the logistic model. Adjustment for BPE and FGT does not resolve reader identity, calendar time, case mix or dependence across observations. The Discussion both omits these decisive limitations and proceeds beyond its acknowledged training and exposure-proxy limitations to an unqualified improvement claim. The conflicting 2020 and 2022 study windows remain a reporting concern, but the text does not permit calculation of their effect on the estimates.
Nabu’s assessment, alongside the field’s view.
Are you an author of this paper?
Limited2.3
Confidence highThe study extends reader-experience evidence into high-risk screening and separates malignancy-only from malignant-or-high-risk outcomes. The contribution remains incremental, and the experience association occurs only under the broader outcome definition.
Pathology-linked outcomes support descriptive biopsy-yield estimates, but the experience analysis does not address central confounding or dependence across observations from the same readers and patients. The causal conclusions therefore exceed the selected-observation design, producing the stacked 2.0 ceiling.
The manuscript explains the two outcome definitions and follows a readable methods-to-results sequence. Its central use of accuracy and exponential improvement materially overstates what a conditional biopsy-yield association establishes.
Discussion engages relevant reader-experience and BPE literature, but omits decisive temporal, case-mix and dependence limitations. Separately, its conclusion contradicts acknowledged training and exposure-proxy limitations, requiring the stacked context ceiling.
Lower confidence on Contribution, Positioning — domain match limited.
Caveats4 of 4 checks
2 reported values do not reconcile, but the key claims do not rest on them.
Research Ethics Board approval and waiver of consent are declared. The supplied text establishes no research-conduct concern; absent availability or disclosure statements are not treated as evidence of misconduct.
Ethics approval: declared
24 of 24 checkable references verified
27 references in manuscript 3 are books, websites or datasets — counted, but not index-checkable
References not verified automatically are resolved manually within 24 hours.
No retraction or expression of concern found in Crossref (includes Retraction Watch).
Sources: Crossref (incl. Retraction Watch) ✓
Where this paper’s evidence sits on the path from initial observation to real-world use.
The findings can motivate an education or workflow study but do not test either intervention. Recommendations for continuing education remain unvalidated applications rather than implementation-ready guidance.
AI-generated, human-governed. Something look off? Contact us to request a review.
Keep reading
Other papers Nabu has evaluated blind, starting with the same field.
Buchanan, Moodie, Batziou, et al. · 2026 · Basic medicine
g is supported by a widely distributed white matter network whose macroscopic wiring capacity, microstructural organisation, and age sensitivity contribute in distinct ways.
TentativeMarkiewicz, Thompson, Wardlaw, et al. · 2026 · Clinical medicine
CL values across 74 scans showed excellent test-retest agreement (ICC = 0.968), improving to 0.987 after exclusion of one attenuation correction related outlier.
SupportedAndrew J. Coristine, Ruud B. van Heeswijk, Matthias Stuber · 2013 · Clinical medicine
The proposed WSA-T2-Prep method was shown to improve fat suppression and vessel sharpness as compared to a CA-T2-Prep technique
Supported