Study Approach
We conducted a two-site, cross-institutional study. We developed a multivariable risk model based on clinical parameters at one institution and externally validated it at another, directly comparing its utility for save biopsy reduction to established risk calculators (including the MRI-ERSPC risk calculator) and simple clinical strategies such as PSA density and normalized ADC.
Key Insights
1. PSA Density: The All-Time Classic
PSA density (PSAd) remains a robust tool for biopsy reduction, especially at lower risk thresholds (≤10%). In our cohort, a PSAd threshold of 0.08 ng/ml/ml allowed for significant biopsy reduction-210 unnecessary biopsies avoided per 1,000 men at a 10% risk threshold-without missing csPCa, confirming its value as a robust, practical decision tool.
2. ADC Ratio-A Surprising Contender
Normalized ADC (cutoff 0.81) performed on par with PSA density for risk stratification, avoiding 223 unnecessary biopsies per 1,000 men at a 10% risk threshold. Our findings suggest normalized ADC could serve as an imaging-based alternative to PSAd, but this requires further validation.
3. Lesion Volume: Limited Clinical Utility
Contrary to some previous studies advocating the use of lesion size/volume metrics for PI-RADS 3 biopsy decisions, decision curve analysis demonstrated no net benefit in our study. Inflammatory changes can mimic tumor appearance and extent in this subgroup, limiting its use for risk stratification.
4. Local Model Fit Outperforms “One-Size-Fits-All” Models
Our locally developed risk model (variables: age, prostate volume, PSA density, prior negative biopsy status) outperformed previously published models-even though it used the same clinical variables. How is that possible? The answer lies in calibration. By tailoring the model to local referral patterns, and patient demographics, we achieved higher net benefit and avoided 547 unnecessary biopsies per 1,000 men at a 10% risk threshold-without missing csPCa. This underscores the importance of adapting risk models to the specific clinical environment in which they are applied.
Clinical Implications
- If local model development and calibration are feasible, this approach provides the highest benefit for biopsy decision-making in PI-RADS 3 lesions.
- Where this is not possible, PSA density (and potentially normalized ADC, pending further validation) are easy-to-implement alternatives.
- Joint Department of Medical Imaging, Princess Margaret Hospital, Sinai Health System, University of Toronto, Toronto, ON, Canada.
- Department of Diagnostic and Interventional Radiology, Cantonal Hospital Frauenfeld, Frauenfeld, Switzerland.
- Department of Diagnostic and Interventional Radiology, Klinikum rechts der Isar, Technical University of Munich, Munich, Germany.
- Division of Urology, Department of Surgical Oncology, Princess Margaret Cancer Centre, University Health Network, Toronto, ON, Canada.
- The University of Ottawa, Ottawa, ON, Canada.
- Institute of Medical Science, University of Toronto, Toronto, ON, Canada.
- Department of Surgery, Division of Urology, Sunnybrook Health Sciences Centre, University of Toronto, Toronto, ON, Canada.
- Department of Surgery, Division of Urology, Mount Sinai Hospital, Toronto, ON, Canada.
- Lunenfeld-Tanenbaum Research Institute, Toronto, ON, Canada.