Diagnostic accuracy of a 'stage-gated' approach for reporting prostate screening MRI: "Is less more?".

Abstract

OBJECTIVES: To evaluate whether a two-step, 'stage-gated' reporting approach could improve the positive predictive value (PPV) of biparametric (bp)MRI for prostate cancer (PCa) screening compared to conventional Likert/PI-RADS scoring. MATERIALS AND METHODS: This retrospective secondary analysis utilised data from IP1-PROSTAGRAM-a prospective, population-based study of men aged 50-69 years who underwent PCa screening with bpMRI, ultrasound and prostate-specific antigen (PSA) testing between October 2018 and May 2019 at two centres (NCT03702439). MRI scans from IP1-PROSTAGRAM were retrospectively evaluated using the 'stage-gated' approach: three radiologists independently reviewed limited MRI sequences (axial T2-weighted and b1500 diffusion-weighted images) and classified scans as positive or negative; if positive, the remaining bpMRI images were reviewed and a hypothetical "decision-to-biopsy" made. The PPV of 'stage-gated' reading was compared to PI-RADS and Likert scores ≥ 4 from the original IP1-PROSTAGRAM bpMRI reports. The reference standard was IP1-PROSTAGRAM biopsy results with grade group (GG) ≥ 2 cancer considered significant. RESULTS: Of 408 participants (median age 57 years [IQR 53, 61]), 405 had MRI scans available for secondary analysis. The prevalence of GG ≥ 2 cancer was 4% (17/405). The 'stage-gated' reporting approach achieved a PPV of 53% (95% CI: 30, 75; 8/15), compared to 29% (95% CI: 15, 47; 8/28) and 30% (95% CI: 17, 46; 11/37) for Likert and PI-RADS ≥ 4 pathways, respectively. The 'stage-gated' approach halved the number of recommended biopsies while maintaining similar cancer detection rates. CONCLUSION: The 'stage-gated' reporting approach, using limited sequences for the initial read, may improve the PPV and benefit-to-harm ratio of MRI-based screening. KEY POINTS: Question The PPV of MRI in PCa screening is low, likely because conventional assessment systems are not optimised for low disease prevalence populations. Findings A two-step, 'stage-gated' reading approach achieved a PPV of 53% (8/15), compared to 29% (8/28) for Likert and 30% (11/37) for PI-RADS scores ≥ 4. Clinical relevance The 'stage-gated' reporting approach, which uses limited sequences for the initial read, may improve the PPV and benefit-to-harm ratio of MRI-based screening by reducing unnecessary biopsies. Prospective evaluation is needed to confirm these findings in real-world screening settings.