5073 Background: Active surveillance is increasingly used for insignificant prostate cancer (PCa). In order to identify suitable patients, risk scores have been developed which use pre-operative factors. We evaluated the accuracy of 9 separate tools developed to identify patients harbouring insignificant PCa in 2613 patients who underwent radical prostatectomy for Gleason 3+3 PCa. We have developed and validated a novel risk score to correctly identify insignificant PCa for use in unscreened patient cohorts using non-dichotomised clinical predictors. Methods: 2799 patients who would have been candidates for AS (Gleason score 6 only) patients underwent robotic radical prostatectomy between 2006 and 2016 at a tertiary referral center. The volume and grade of tumour in the resected prostate was analysed. Inignificant PCa was defined as Gleason 3+3 only, index tumour volume <1.3 cm3 , total tumour volume <2.5 cm3 (updated ERSPC definition). 2613 patients were included in the final analysis. We computed the accuracy (specificity, sensitivity and area under the curve (AUC) of the receiver operator characteristic) of 9 predictive tools. Multivariate logistic regression with elastic net regularisation was used to develop a novel tool to predict insignificant prostate cancer using age at diagnosis, baseline PSA, TRUS volume, clinical T-stage, number of positive cores and percentage of positive cores as predictors. This tool was validated in an external cohort of 441 unscreened patients undergoing surgery for Gleason 6 PCa. Results: All of the predefined tools rated poorly as predictors of insignificant disease as none of them reached the required AUC threshold of 0.7. The new tool performed well in training and validation cohorts. Conclusions: Pre-existing predictive tools to identify indolent PCa have a poor predictive value when applied to an unscreened cohort of patients. Our novel tool shows good predictive power for insignificant PCa in this population in training and validation cohorts. The inherent selection bias due to analysis of a surgical cohort is acknowledged. [Table: see text]