Presentation (CTI)

OBJECTIVES: In oncology, adjustments of long-term survival extrapolations and analyses of survival heterogeneity borne by long-term survivors often require NCSS. This study compared NCSS trends that are derived from local lifetables and aggregate-level data for the CheckMate-067 study population.
METHODS: Publicly available Kaplan-Meier curves for overall-survival (OS) and melanoma-specific survival (MSS) from the Phase III CheckMate-067 study were digitized to reconstruct pseudo individual-patient level data (IPD) for each arm. Minimum follow-up was 10 years. Pseudo-IPD for OS and MSS were pooled separately across arms to generate a non-parametric NCSS curve which was smoothed to ensure monotonicity over time using isotonic regression. As a benchmark, age- and sex-adjusted lifetables published by World Health Organization were used to generate NCSS curves for each participating country in the trial, which were weighted uniformly across all countries to derive an aggregate NCSS curve for the trial cohort. Weekly NCSS rates and restricted mean survival times (RMSTs) estimated from the trial- and lifetable-based NCSS distributions were compared. The NCSS curves derived from each source were also modeled with standard parametric distributions and splines for comparison of their visual trends.
RESULTS: Lifetable-based NCSS exhibited a more favorable trend than trial-based NCSS with an average overestimation margin of 4% across 10-years after randomization. Estimated 10-year RMSTs from the trial- and lifetable-based NCSS were 9.09 and 9.53 years, respectively. Best-fitting models to lifetable-based NCSS (Gompertz) and trial-based NCSS (lognormal) displayed considerably different visual trends and crossed each other at year 12.75. Based on a log-rank test, the two NCSS curves were statistically indistinguishable from each other with a p-value of 0.958.
CONCLUSIONS: Local lifetables may slightly overestimate medium-term NCSS compared to trial-derived estimates but tend to be more conservative when projecting long-term survival in treatment-naïve advanced melanoma. Further analyses across different tumor types are necessary to assess broader applicability of these results.