ESTRO 2022 - Abstract Book

S360

Abstract book

ESTRO 2022

Figure 2

PD-0416 Derivation and external validation of a RAPID Risk score for predicting significant prostate cancer

M. Peters 1 , D. Eldred-Evans 2 , M. J. Connor 3 , M. Bertoncelli Tanaka 3 , H. Bhola-Stewart 3 , T. T Shah 3 , S. Ahmad 4 , M. Noureldin 2 , K. Wong 4 , H. Tam 3 , D. Hrouda 3 , M. Winkler 2 , P. van Rossum 5 , P. Kurver 5 , S. Gordon 4 , H. Qazi 6 , H. U. Ahmed 2 , U. Giovanni Falagario 7 , I. Jambor 8 , A. Briganti 9 , T. Nordström 10 , G. Carrieri 7 , L. Powell 11 , S. Joshi 12 , E. Pegers 12 1 University Medical Center Utrecht, Department of Radiotherapy, Utrecht, The Netherlands; 2 Imperial College London, Urology, London, United Kingdom; 3 Imperial College London, Urology , London, United Kingdom; 4 Epsom and St Helier University Hospitals NHS Trust, Urology, London, United Kingdom; 5 University Medical Center Utrecht, Department of Radiotherapy , Utrecht, The Netherlands; 6 St George’s Healthcare NHS Trust, Urology, London, United Kingdom; 7 University of Foggia, Department of Urology and Organ Transplantation , Foggia, Italy; 8 University of Turku, Radiology, Turku, Finland; 9 IRCCS Ospedale San Raffaele, Urological Research Institute, Milan, Italy; 10 Karolinska Institute, Department of Medical Epidemiology and Biostatistics, Stockholm, Sweden; 11 St George's University Hospitals NHS Foundation Trust, Urology, London, United Kingdom; 12 RM Partners: West London Cancer Alliance, Urology Pathway Group, London, United Kingdom Purpose or Objective Although multi-parametric MRI (mpMRI) has high sensitivity for significant prostate cancer (sPCa), the lower specificity, particularly with equivocal (PIRADS 3) lesions, can result in numerous unnecessary biopsies. We aimed to develop and externally validate a simple scoring system to predict which men with a visible MRI lesion might consider avoiding an immediate prostate biopsy. Materials and Methods We analyzed data from 1,189 men with a visible MRI lesion from a prospective multi-institutional registry. A total of 14 clinical and radiological MRI variables known to be associated with risk of sPCa were evaluated. Using multivariable logistic regression, we derived a parsimonious risk score for different definitions of sPCa. The performance of the risk score was internally validated and then externally validated in 1825 men across 6 international cohorts. For the final risk scores nomograms were created and analyses of clinical benefit were performed. Results For the primary endpoint (detection of ≥ Gleason 3+4), the simplified RAPID risk score included age, PSA density (logged), prior negative biopsy, prostate volume and MRI Score (PIRADS or LIKERT score) (figure 1, available at: https://rapidriskscore.shinyapps.io/RapidRiskScore1/). The model had a discriminative ability (c-statistic or AUC) of 0.82 after internal validation which was similar in different years of data collection and in the external validation cohorts. This simplified model was compared to alternative models with additional clinical variables (including Afro-Caribbean ethnicity, abnormal DRE and a positive family history) and the clinical differences were negligible With thresholds for ≥ Gleason 3+4 prostate cancer from 1-20% predicted by the model, the model was able to reduce the amount of biopsies up to 10.7%, while missing up to 1.9% of cancers. The full models reduced biopsies up to 11.6%, albeit missing up to 2.4% of cancers (figure 2). Conclusion The basic RAPID risk score is a simple five-item score which provides a standardized tool for prediction of clinically significant prostate cancer in men with a visible MRI lesion. It is available online as an app and its performance has been externally validated across multiple external, geographically distinct and independent datasets. In the context of a risk- stratified MRI pathway, this tool can support patients and clinicians making decisions regarding the need for prostate biopsy.