A randomized phase II study of coexpression extrapolation (COXEN) with neoadjuvant chemotherapy for bladder cancer (SWOG S1314; NCT02177695)

Purpose: Dose-dense methotrexate-vinblastine-adriamycin-cisplatin (ddMVAC) and gemcitabine-cisplatin (GC) are accepted neoadjuvant regimens for muscle-invasive bladder cancer. The aim of this study was to validate the score from a coexpression extrapolation (COXEN) algorithm–generated gene expression model (GEM) as a biomarker in patients undergoing radical cystectomy. Patients and Methods: Eligibility included cT2-T4a N0 M0, urothelial bladder cancer, ≥ 5 mm of viable tumor, cisplatin eligible, with plan for cystectomy; 237 patients were randomized between ddMVAC, given every 14 days for four cycles, and GC, given every 21 days for four cycles. The primary objective assessed prespecified dichotomous treatment-specific COXEN score as predictive of pT0 rate or ≤ pT1 (downstaging) at surgery. Results: Among 167 evaluable patients, the OR for pT0 with the GC GEM score in GC-treated patients was 2.63 [P ¼ 0.10; 95% confidence interval (CI), 0.82–8.36]; for the ddMVAC COXEN GEM score with ddMVAC treatment, the OR was 1.12 (P ¼ 0.82, 95% CI, 0.42–2.95). The GC GEM score was applied to pooled arms (GC and ddMVAC) for downstaging with an OR of 2.33 (P ¼ 0.02; 95% CI, 1.11–4.89). In an intention-to-treat analysis of eligible patients (n ¼ 227), pT0 rates for ddMVAC and GC were 28% and 30% (P ¼ 0.75); downstaging was 47% and 40% (P ¼ 0.27), respectively. Conclusions: Treatment-specific COXEN scores were not significantly predictive for response to individual chemotherapy treatment. The COXEN GEM GC score was significantly associated with downstaging in the pooled arms. Additional biomarker development is planned.