Genomic Profiling of Early-Onset Metastatic Chinese Colorectal Cancer (mcrc) Patients: the Development of a PFS Prediction Model for First Line Treatments.

Xicheng Wang,Lin Shen,Jing Zhang,Lu Zhang,Xinru Mao,Ting Hou,Hao Liu,Han,Jian Li,Jing Gao,Yanyan Li

DOI: https://doi.org/10.1200/jco.2018.36.15_suppl.e15657

IF: 45.3

2018-01-01

Journal of Clinical Oncology

Abstract:e15657 Background: The incidence of CRC is rising in early-onset patients despite a decline in overall incidence. In this study we interrogated the mutation spectrum of mCRC patients diagnosed before age 55 and established a model for predicting progression-free survival (PFS) for 1st line treatments, integrating both clinical and molecular features. Methods: Variables with p-values < 0.05 were entered into multivariate analyses using Cox stepwise regression model to select independent predictors as input. A final model was selected using a backward step-down process. Results: We performed targeted sequencing on 187 resected or biopsied tumor tissue samples and matched white blood cells of early onset mCRC patients using a panel consisting of 520 cancer-related genes. We identified 13.9% patients (26/187) with pathogenic germline mutations in cancer susceptibility genes. Comparing to TCGA data set, our cohort revealed a distinct mutation profile, reflected by significantly higher mutation frequencies in TP53, BRAF, RET, EP300 and a lower frequency in APC in patients with left-sided primaries. In patients with right-sided primaries, our cohort showed significantly higher frequencies in TP53, DOTIL and lower frequencies in APC, EP300, CREBBP and FBXW7. To predict the probability of 12-month and 18-month PFS of patients undergoing 1st line chemotherapy± biologic agents, we constructed a nomogram incorporating both clinical and molecular features, including T stage, tumor location, BRAF V600E, ARID1A, PIK3CG, the presence of germline pathogenic mutations and mutations affecting the proofreading domains of POLE /POLD1, achieving a C-index of 0.683. The predictive power drops to 0.62 if only stage and location, the two most commonly used predictors, are used as input. We confirmed the performance of our model by grouping patients into 3 subgroups based on total points obtained, and demonstrated significant difference in PFS (p < 0.01) independent of treatments. Conclusions: We revealed a distinctive genomic landscape for early onset Chinese mCRC patients and developed a PFS prediction model for patients undergoing 1st line treatments.