Predicting Active Surveillance Failure for Prostate Cancer Patients in the MRI Era: a Multicentre Transatlantic Cohort Study

Background and Objective. MRI-driven active surveillance (AS) is increasingly used for prostate cancer (PCa) management. To determine the oncological safety of contemporary, MRI-driven AS and identify patients at higher risk of AS failure. Methods. This retrospective cohort study included AS patients with MRI-localised PCa from three US and UK centres. The primary outcome was AS failure, a composite of PCa-specific mortality, metastasis, progression to >GG4, or post-treatment biochemical recurrence. The secondary outcome was disease progression, defined as histological progression to GG3 or progression to locally advanced disease. Hazard ratios (HRs) were estimated using multivariable Cox models, with multiplicity-adjusted log-rank tests comparing event-free survival across subgroups. Key Findings and Limitations. 719 patients (median follow-up 5.2 years) were included. Of those, 629 (87%) had stable disease; 36 (5%) experienced AS failure, including 8 (1%) cases of metastasis and no PCa-related deaths; 54 (8%) had disease progression. Cribriform GG2 histology was the strongest predictor of AS failure (HR 12.7 95% CI, 4.8-33.6; P < 0.001), followed by tumor MRI-visibility (HR 5.0; 95% CI, 1.5-16.5; P = 0.009) and non-cribriform GG2 histology (HR 3.4; 95% CI, 1.6-7.0; P = 0.001). MRI-invisible, non-cribriform GG2 and all GG1 tumors had comparable event-free survival (Padj > 0.05 for both outcomes). The study is limited by retrospective design. Conclusions and clinical implications. Contemporary MRI-driven AS is safe, including for patients with non-cribriform GG2 tumours, particularly those that are MRI-invisible. Conversely, patients with cribriform GG2 disease are at increased risk of AS failure and therefore warrant upfront treatment.