Validation of the PREMM1,2 model for the prediction of MLH1/MSH2 germline mutation carriers in a population- based cohort of colorectal cancer (CRC) patients
10545 Background: PREMM1,2 is a predictive model for estimating the likelihood of finding a mutation in the MLH1 and MSH2 genes developed in a population at risk of Lynch syndrome (Balmaña et al. JAMA 2006). The EPICOLON cohort is a Spanish population-based series of 1222 patients with CRC (Piñol et al. JAMA 2005). Methods: All 1222 individuals underwent microsatellite instability (MSI) and immunohistochemistry (IHC) analysis for MLH1/MSH2. Patients whose tumours exhibited MSI or altered IHC underwent MLH1/MSH2 analysis (n=91). Sensitivity (Se), specificity, positive predictive value (PPV), and the areas under the receiver operating characteristics curves (AUC) for the PREMM1,2 model were calculated and compared with the Edinburgh model (Barnetson et al. NEJM 2006) and the Revised Bethesda guidelines (RBG). Results: Three-hundred and ninety six individuals (32%) had a PREMM1,2 score =5%, 287 (23%) fulfilled the RBG, and 75 (6%) had a Barnetson score =5%. A PREMM1,2 score =5% and the RBG identified all carriers with deleterious mutations (n=8, Se=100%), while a Barnetson score =5% missed 2 mutation carriers (Se= 75%). For PREMM1,2 and Barnetson scores =5% and fulfilment of the RBG, specificities were 68%, 94%, and 77%, respectively; and PPV were 2%, 8%, and 2.8%, respectively. The AUC was 0.93 (95% CI: 0.86–0.99) for the PREMM1,2 model and 0.86 (95% CI: 0.66–1.04) for the Barnetson model. The predictions of carrying a mutation stratified into five groups based on PREMM1,2 scores (<5%, 5–9%, 10–19%, 20–39%, =40%) correlated reasonably with the presence of mutations (0%, 1%, 1%, 13%, and 22%, respectively). Conclusions: In a population of CRC patients, the PREMM1,2 model identifies all mutation carriers of MSH2 and MLH1 using a cutoff of =5%. In addition, it provides quantification of risk that is useful to decide the strategy used for molecular evaluation and risk counselling of patients. No significant financial relationships to disclose.