Abstract
Abstract
In oncology trials, overall survival and progression-free survival (PFS) are typically summarized with Kaplan–Meier curves, but medians cannot be estimated when follow-up is early. We evaluated whether exponential-decay regression applied to the earliest quartile of Kaplan–Meier data could approximate mature medians. A total of 114 phase III immunotherapy trials encompassing 420 treatment arms were analyzed, yielding 618 evaluable survival curves across overall survival and PFS endpoints. For immunotherapy overall survival, predicted and reported medians showed no significant difference (P = .59; R2 = 0.76; concordance correlation coefficient = 0.87). For immunotherapy PFS, predictions modestly overestimated medians (P < .001; R2 = 0.50). Nonimmunotherapy arms demonstrated overestimation for both overall survival (R2 = 0.61) and PFS (R2 = 0.68). Pooled overall survival performance was good (R2 = 0.70). Early exponential modeling may offer a transparent tool for interim overall survival estimation in immunotherapy trials, warranting prospective validation.