e22077 Background: Different predictive models for Lynch syndrome have recently been developed and their comparative performance in a clinic-based cohort has not been assessed. We aimed to analyze the accuracy of the MMRpro, Barnetson, and PREMM1,2 models to predict MLH1/MSH2 mutation carrier status in 564 unrelated probands with clinical suspicion of hereditary colorectal cancer and compare it with Wijnen model and clinical criteria. Methods: Overall, 538 individuals (95%) underwent mismatch repair (MMR) deficiency screening before germline genetic testing (sequencing with or without large rearrangement analysis) and 26 (5%) performed direct genetic testing. Prediction scores for all individuals were calculated by each model. Sensitivity, specificity, positive predictive value (PPV), and the areas under the receiver operating characteristics curves (AUC) for all models were calculated and compared with the Revised Bethesda Guidelines (RBG). Results: 114 individuals (20%) were mutation carriers (63 MLH1, 51 MSH2). The AUC was 0.95 (95% CI: 0.93–0.97) for MMRpro, 0.87 (95% CI 0.83–0.91) for the Barnetson model, 0.87 (95% CI 0.83–0.91) for PREMM1,2, and 0.75 (95% CI 0.69–0.80) for the Wijnen model (p<0.001). Testing thresholds and specificity at 100% and 90% sensitivity for each model were: 0.001/17% and 0.33/89% for MMRpro, 0.01/9% and 0.07/59% for Barnetson, 0.05/5% and 0.11/58% for PREMM1,2. Sensitivity and specificity of RBG were 86% and 14%, respectively. Calibration was 0.92, 1.05, 0.50, and 1.25 for PREMM1,2,Barnetson, Wijnen, and MMRpro, respectively. Conclusions: In a population of individuals at risk of Lynch syndrome, the MMRpro model has the largest AUC, although the Barnetson and PREMM1,2 model also show adequate discrimination. Any of the models perform better than the RBG and provide quantitative risk estimation of finding a MLH1/MSH2 mutation useful in genetic counselling. No significant financial relationships to disclose.