Integrated survival analysis.

Abstract

e14695 Background: Clinical trial efficacy may be determined by the primary endpoint(s) of median progression free survival (mPFS) and/or median overall survival (mOS) from the Kaplan-Meier estimate, and secondary endpoints of overall response rate (ORR) and duration of response (DOR). In the treatment of metastatic cancers, single agent immunotherapy tends to have lower ORR and longer DOR in comparison to chemotherapy or targeted tyrosine kinase inhibitor therapy, which tends to have higher ORR and shorter DOR. This leads to intersecting or crossing Kaplan-Meier survival curves, which are harder to interpret. In immunotherapy trials, mPFS is a poor surrogate for mOS, and both may underrepresent the clinical benefit over time. Methods: We aimed to develop novel metrics that more effectively represent the Kaplan-Meier estimate. The graphical curve of the Kaplan-Meier estimate may be represented by the equation as a function of time. The integration of f ( x), which mathematically represents the area under the curve and clinically the cumulative or integrated survival benefit with time, is: ∫0t f( x) dx = F(t) − F(0). Kaplan-Meier curve equation with constants C and A may be approximated by: ∫0t C e-A x dx = C e-A t ]0 t = C( e-A t – 1). The average integrated clinical benefit with time is: 1/ t ∫0t f( x) dx = [ F( t) − F(0)]/ t. Results: The Phase 3 KeyNote-061 trial of metastatic gastric cancer treated with single agent Pembrolizumab immunotherapy versus Paclitaxel chemotherapy was use for integrated PFS and iOS calculation. The Kaplan-Meier curves intersected or crossed in KeyNote-061 trial. We show the established ORR, mPFS, mOS, DOR. Integrated PFS (iPFS) and iOS were calculated with greater increase in iOS and much less decrease in iPFS for Pembrolizumab than Paclitaxel treatment. Conclusions: Integrated survival is a novel analysis that integrate the Kaplan Meier estimates and better represent the entire median survival and drug DOR may be combined into a single useful metric for overall clinical benefit. [Table: see text]