OBJECTIVES

This analysis aimed to explore survival heterogeneity and estimate proportion of potential long-term survivors (LTS) among previously untreated, I/P risk aRCC patients who received immune-checkpoint inhibitors (ICIs) NIVO+IPI in the CheckMate 214 trial by applying mixture cure models (MCMs).

METHODS

MCMs were fitted to PFS data from two successive database locks (DBL) with 48- and 60-months (m) of follow-up. In the model, all patients were subject to risk of non-disease-related mortality but only non-LTS were subject to risk of progression and additional disease-related mortality. Age- and gender-specific general population mortality rates in the model were based on life tables from UK Office of National Statistics. Time-to-event outcomes for non-LTS were modeled with standard parametric distributions whose parameters were derived simultaneously with LTS rates via maximum likelihood estimation. Candidate models were evaluated based on statistical goodness-of-fit criteria, and their visual fits to observed survival and hazard trends in the trial.

RESULTS

Estimated LTS rates ranged between 29.0%-34.9% in the 48-m DBL and between 29.8%-34.7% in the 60-m DBL. In the 60-m DBL, generalized gamma MCM provided the best statistical fit with the widest 95% confidence interval (CI) around the lowest LTS rates (29.8%, 95% CI: 22.8% - 38.0%), whereas exponential MCM generated the best visual fit to both observed survival and hazard trends with the narrowest 95% CI around the second-highest LTS rates (33.5%, 95% CI: 28.2% - 39.2%). When compared to observed PFS rate at 60 m, model predictions from the 48-m DBL had a deviation between -0.3% and 1.9%.

CONCLUSIONS

The emergent PFS plateau for I/P risk NIVO+IPI patients in the trial was adequately captured with MCMs. Estimated LTS rates were consistent across most models and both DBLs. Results support the paradigm shift in the natural history and treatment of aRCC gained by the combination of ICIs.