Development and internal validation of prediction models for biochemical failure and composite failure after focal salvage high intensity focused ultrasound for local radiorecurrent prostate cancer: Presentation of risk scores for individual patient prognoses.

ABSTRACT

PURPOSE: Patient selection for focal salvage remains difficult. Therefore, we developed and internally validated prediction models for biochemical failure (BF) and a composite endpoint (CE) following focal salvage high intensity focused ultrasound (HIFU) for radiorecurrent prostate cancer. MATERIALS AND METHODS: A prospective HIFU registry identified 150 cases (November 2006-August 2015). Recurrence was assessed with multiparametric magnetic resonance imaging (MRI) combined with template prostate mapping biopsies, targeted biopsies, or systematic transrectal ultrasound-guided biopsies. Metastatic disease was ruled out with a positron emission tomography-computed tomography and a bone scan. Focal salvage HIFU consisted of quadrant-ablation, hemi-ablation, or index-lesion ablation. Cox-regression was used for BF (Phoenix-definition) and CE (BF/MRI+/biopsies+/local or systemic treatment/metastases+/prostate cancer specific mortality+). Internal validation was performed using bootstrap resampling (500 datasets) after which C-statistic and hazard ratios were adjusted. Models were calibrated and risk scores created. RESULTS: Median follow-up was 35 months (interquartile range 22-52). Median biochemical disease-free survival (DFS) was 33 months (95% CI 23-45). Median CE-free survival was 24 months (95% CI 21-35). After multivariable analysis, DFS interval after primary radiotherapy, presalvage prostate-specific antigen (PSA), PSA-doubling time, prostatic volume, and T-stage (both MRI based) predicted BF. For the CE, PSA-doubling time was not predictive but additionally, primary Gleason score was. The adjusted C-statistics were 0.68 and 0.64 for BF and CE, respectively. Calibration was accurate until 48 months. The risk scores showed 3 groups, with biochemical DFS of 60%, 35%, and 7% and CE-free survival of 40%, 24%, and 0% at 4 years. CONCLUSION: Our model, once externally validated, could allow for better selection of patients for focal salvage HIFU.