Identification of patient population with longer survival when treated with S-1 plus cisplatin via predictive enrichment strategy analysis of the FLAGS phase III trial.

Abstract

e15612 Background: FLAGS, a randomized phase III trial, compared S-1, an oral fluoropyrimidine, plus cisplatin (SP) with 5-fluorouracil plus cisplatin (FP) in the first-line treatment for advanced gastric cancer (AGC). The purpose of this analysis was to discover a clinically significant covariate model using the Predictive Enrichment Strategy Analysis (PESA) to identify a patient population most likely to benefit from the treatment. Methods: PESA is a newly introduced concept proposed by the United States Food and Drug Administration and provides robust results. Consensus-based 15 clinical covariates were selected for enrichment analysis and a large cohort with no missing data (889 patients) was analyzed for model development. The models generated were cross-validated 200 times and ROC-AUC for the Cox PH model with the selected variables by the enrichment analysis were computed. Results: ECOGPerformance status (PS = 1) was the strongest covariate and most significant in identifying the enrichment group for SP. In the population with PS = 1, the OS in the SP group was significantly longer than the FP group (Hazard Ratio [HR] = 0.798, 95%CI = (0.66-0.96) p = 0.0166) Other covariates with high potential to be associated with benefit included: diffuse-type histology, presence of peritoneal metastases, and the lack of liver metastases. Various models using different combinations of the covariates were compared using ROC-AUC analysis and found that the model including all four covariates had the best ROC-AUC (0.566, 95%CI = (0.546-0.587)). Conclusions: PESA identified a cross-validated, robust model (PS, diffuse histology, peritoneal mets, and no liver mets). This patient population is associated with improved OS in Advanced Gastric Cancer patients who receive SP. An independent validation of this model will be carried out with data derived from the DIGEST trial. NCT00400179