Microarray-based tumor molecular profiling to direct choice of cisplatin plus S-1 or oxaliplatin plus S-1 for advanced gastric cancer: A multicentre, prospective, proof-of-concept phase 2 trial.

48 Background: Theoxaliplatin/S-1 (SOX) and cisplatin/S-1 (SP) regimens are interchangeably used in the management of advanced gastric cancer (AGC). We previously developed a classification tool using gene expression signatures which successfully stratified gastric cancer cell lines and primary tumour samples according to their sensitivity to SOX or SP ( Tan et al, Gastroenterology 2011). We now report the first prospective study to evaluate the feasibility and efficacy of using a genomic classifier to tailor treatment in this setting. Methods: Pts with histologically-confirmed locally-advanced, metastatic and recurrent GC were recruited from 3 centres in Singapore and South Korea. Tumours were analysed using the Affymetrix HG-U133 Plus 2.0 array and results were used to classify pts as G1 (oxaliplatin-sensitive), G2 (cisplatin-sensitive) and G3 (status unclear or gene expression not available). G1 and G2 pts were matched to SOX and SP regimens respectively, while G3 pts were assigned SOX. The primary endpoint was best overall response; secondary endpoints were turnaround time and biomarker analyses. Results: Between July 2, 2010, and Apr 2, 2015, we screened 85 AGC pts. 74 pts received at least 1 cycle of treatment and were evaluable for analysis. Median turnaround time was 7 working days (IQR, 5–9). The misclassification rate was 6%. After an initial 30 pts in the G1 subgroup were treated with SOX, subsequent pts ( N = 13) classified as G1 received the SP regimen. The ORR were 44.8%, 8.3%, 26.7% and 55.6% for G1 SOX, G1 SP, G2 SP ( N = 19), G3 SOX ( N = 12) respectively; and was higher in G1 pts treated with SOX compared with SP ( P = 0.033). Post hoc genomic reclassification based on Lei et al (Gastroenterology 2013) confirmed the utility of the metabolic subtype as a predictive marker of benefit from chemotherapy (log rank P value for PFS = 0.004). Conclusions: This bench-to-bedside effort establishes that molecular profiling to direct choice of conventional chemotherapy for AGC is possible within a reasonable timeframe. The clinical utility of our genomic classifiers in question are promising but warrant further investigation. Clinical trial information: NCT01100801.