Finally the way the crossover patients were analyzed in both control and cdx011 is a head scratcher. There is the inherent bias only patients who is fit or disease stable get the chance to cross over. The company must have felt it compelling to include such patients in both treatment arm as otherwise the PFS difference would be less pronounced. The correct way would be to censor on last SD visit before crossover and include in only the control arm.
I'd agree on PFS analysis alone. I think they did this due to really small sample size for high G and TN patients mainly due to low entry criteria (5% GPNMB). However, OS analysis is done traditionally which is great. Thus the 3 super OS responders (OS 11.3m, 12.3m, 21m) in high G and non TN IC arm (5 of them with 11.3m mOS you mentioned in your post) could be due to crossover to 011, which would make 011 current OS trend more impressive.
From poster:
Patients who initially received Investigator’s Choice and subsequently crossed over to receive CDX-011 (n=15) are included in the PFS analysis for each treatment . These patients, with a median OS of 12.5 months, are assigned to the Investigator’s Choice arm only for OS analysis. Median OS for the remaining IC patients who did not cross over is 5.4 months.