|
Abstract:
|
The EU submission of Kymriah in r/r DLBCL relied on a single arm study of 165 enrolled patients, of whom only 111 were infused with Kymriah and included in the main efficacy analysis. The EMA rapporteur cited the high drop out prior to Kymriah infusion as a Major Objection, arguing that the infused set was enriched and efficacy results over-estimated, calling instead for analysis on all 165 patients. Novartis refuted the enrichment argument by noting the similarity in baseline characteristics between enrolled and infused sets, and claimed the better results were due to Kymriah. This argument was formalized statistically for overall survival (OS) using a Cox model with Kymriah infusion as a time-dependent covariate and prognostic factors as fixed covariates. Results demonstrated a strong Kymriah treatment effect and a weaker possible influence of the prognostic factors. The OS benefit of Kymriah was displayed using Extended Kaplan-Meier curves (Snapinn et al., 2005). Also model-based predicted OS curves provided a useful graphical representation showing that the majority of the treatment effect was explained by Kymriah.
|
