The co-expression extrapolation (COXEN) model is used to select genes for predicting tumor cell response to a specific drug treatment. Researchers in a phase 2 trial of neoadjuvant gemcitabine-cisplatin (GC) or dose-dense methotrexate, vinblastine, adriamycin, and cisplatin (ddMVAC) in muscle-invasive bladder cancer (MIBC) used the COXEN model to determine its efficacy at predicting patient response to neoadjuvant chemotherapy.
A total of 167 patients with MIBC were included in the trial and randomized to receive either ddMVAC every 14 days or GC every 21 days, each for 4 cycles. Patient COXEN scores were analyzed for their association with event-free survival (EFS) and overall survival (OS) and by treatment arm. EFS was defined as progression or death before scheduled surgery, patient decision not to undergo surgery, recurrence, or death due to any cause after surgery. Cox regression analysis was used to evaluate the COXEN score or treatment arm association with EFS and OS.
The GC COXEN scores had a hazard ratio (HR) of 0.45 (95% CI, 0.20-0.99; P=.047) in both arms pooled together, but the scores were not significantly prognostic for OS or EFS in their respective arms. In the intent-to-treat analysis (n=226), no significant difference was found between ddMVAC and GC for OS (HR, 0.87; 95% CI, 0.54-1.40; P=.57) or EFS (HR, 0.86; 95% CI, 0.59-1.26; P=.45).
In patients who underwent surgery, pathologic response (pT0 vs downstaging vs no response) was strongly related to superior postsurgical survival, with a 5-year OS rate of 90%, 89%, and 52%, respectively.
The COXEN GC score is a useful tool for predicting response in patients receiving cisplatin-based neoadjuvant treatment, providing estimates of OS and EFS for GC and ddMVAC in the MIBC population. Pathologic response (<pT2) performed well as an intermediate end point and should be used in future phase 2 trials.