Validation and Extension of the PREMM 1,2 Model in a Population-Based Cohort of Colorectal Cancer Patients

Abstract

Early recognition of patients at risk for Lynch syndrome is critical but often difficult. Recently, a predictive algorithm-the PREMM(1,2) model-has been developed to quantify the risk of carrying a germline mutation in the mismatch repair (MMR) genes MLH1 and MSH2. However, the model's performance in an unselected, population-based colorectal cancer population as well as its performance in combination with tumor MMR testing are unknown. We included all colorectal cancer cases from the EPICOLON study, a prospective, multicenter, population-based cohort (n = 1222). All patients underwent tumor microsatellite instability analysis and immunostaining for MLH1 and MSH2, and those with MMR deficiency (n = 91) underwent tumor BRAF V600E mutation analysis and MLH1/MSH2 germline testing. The PREMM(1,2) model with a >/=5% cut-off had a sensitivity, specificity, and positive predictive value (PPV) of 100%, 68%, and 2%, respectively. The use of a higher PREMM(1,2) cut-off provided a higher specificity and PPV, at expense of a lower sensitivity. The combination of a >/=5% cut-off with tumor MMR testing maintained 100% sensitivity with an increased specificity (97%) and PPV (21%). The PPV of a PREMM(1,2) score >/=20% alone (16%) approached the PPV obtained with PREMM(1,2) score >/=5% combined with tumor MMR testing. In addition, a PREMM(1,2) score of <5% was associated with a high likelihood of a BRAF V600E mutation. The PREMM(1,2) model is useful to identify MLH1/MSH2 mutation carriers among unselected colorectal cancer patients. Quantitative assessment of the genetic risk might be useful to decide on subsequent tumor MMR and germline testing.