Performance of PREMM1,2,6, MMRpredict, and MMRpro in detecting Lynch syndrome among endometrial cancer cases.

Abstract

5024 Background: Lynch syndrome (LS), a heritable condition due to a mismatch repair (MMR) gene mutation, accounts for nearly 2% of all endometrial cancer (EC) cases. Prediction models for LS have been developed but have not been evaluated among EC cases. Methods: PREMM1,2,6, MMRpredict and MMRpro risk scores were calculated for 563 population-based, unselected EC cases and 129 clinic-based, high-risk EC cases who underwent genetic evaluation for LS. Discriminative ability of each model was assessed using the area under the receiver operating curve (AUC). Sensitivity and specificity were calculated at 5% cut-off for PREMM1,2,6 and MMRpro, and at 0.5% cut- off for MMRpredict (EC replaced a proximal colorectal cancer (CRC) diagnosis). Results: Pathogenic MMR gene mutations were detected in 14 population-based and 80 clinic-based subjects. PREMM1,2,6, MMRpredict and MMRpro were able to distinguish mutation carriers from non-carriers with AUC of 0.77, 0.79 and 0.75, respectively, for the unselected cohort and AUC of 0.67, 0.61 and 0.66, respectively, for the high-risk cohort. PREMM1,2,6 demonstrated high sensitivity but low specificity (sens and spec = 93% & 5% for unselected cohort and 99% and 2% for high-risk cohort). MMRpredict had 100% sensitivity for both cohorts, and a specificity of 39% and 6.2% for unselected and high-risk EC cases, respectively. MMRpro had 71% sensitivity and 65% specificity among unselected EC cases, and 98% sensitivity and 6% specificity among high-risk cases. More than 95% of high-risk EC cases would have been referred for genetic testing if the aforementioned models’ cut-off points were used as a threshold for referral. Among unselected EC cases, using PREMM1,2,6 would require that 95% of cases be referred for testing. MMRpro missed 29% of mutation carriers in the unselected cohort, precluding its use as a screening tool. MMRpredict, although 100% sensitive, does not provide a useful quantitative estimate of mutation risk given its very low cut-off of 0.5%. Conclusions: In contrast to CRC, prediction models for LS have limited clinical utility in determining which patients with EC should undergo clinical genetic testing for LS, irrespective of clinic or population-based ascertainment. Author Disclosure Employment or Leadership Position Consultant or Advisory Role Stock Ownership Honoraria Research Funding Expert Testimony Other Remuneration Archimedes, Interquest Inc., Myriad Genetics Laboratories Inc. Falco Biosystems