Predictive models for biopsy upgrading in active surveillance for prostate cancer: a scoping review

Purpose: Active surveillance (AS) is an established approach for managing favorable risk prostate cancer. However, it remains underutilized, partly due to concerns about missing higher-risk disease. Predictive models that integrate patient-specific factors may help with personalized risk stratification. This scoping review summarizes published predictive models in active surveillance and provides future directions for inquiry. Methods: Databases (PubMed, Embase, Scopus) were searched for studies that proposed or validated predictive models of biopsy upgrading during active surveillance. Studies were eligible if they used ≥ 2 predictor variables with an objective upgrading outcome. Data extraction was performed in duplicate with a standardized form. Discrepancies were resolved by consensus. Results: The search identified 2,886 unique records, of which 28 studies met inclusion criteria. All were observational (24 retrospective, 3 prospective, 1 case-control) with sample sizes ranging from 71 to 7,813 participants. Predictive models incorporated clinical and pathological features, with a growing number integrating imaging or molecular markers. Model discrimination was moderate to high (AUC 0.60–0.90; c-index 0.61–0.81), though external validation and clinical implementation were limited. Conclusion: Available predictive models show promise for individualized risk stratification in active surveillance for prostate cancer, but their continued validation, accessibility and integration into clinical practice are necessary next steps before widespread adoption.