Background: Immune checkpoint inhibitors enhance immunologic antitumor activity and ongoing studies are exploring the potential of such inhibitors to provide long term (OS) benefits across several cancer types. We describe an exploratory analysis of the association between tumor shrinkage (TS) and OS in patients (pts) with melanoma or NSCLC receiving ipilimumab or nivolumab, immune checkpoint inhibitors that augment T-cell activity by blocking CTLA-4 and programmed death-1 receptors, respectively. Methods: TS and OS associations for ipilimumab in pts with advanced melanoma were assessed using data from 4 phase II studies (CA184-004/007/008/022, n=351). TS and OS associations for nivolumab were assessed in pts with advanced melanoma (n=97) or NSCLC (n=112) using data from cohorts of a phase I study (CA209-003). A nonlinear mixed-effects model was used to describe longitudinal tumor burden data; TS was determined from the model as % decrease from baseline. The relationship between TS and OS was determined by a multivariate Cox proportional-hazards model. Results: Across the pt populations tested, the risk of death decreased with increasing % TS (hazard ratio [HR] coefficients in all 3 models and associated 95% confidence intervals [CIs] <1.0). At 30% TS (corresponding to magnitude of RECIST partial response), the association between TS and OS for pts with melanoma was comparable for ipilimumab or nivolumab treatment (HR [95% CI] of 0.602 [0.543–0.668] and 0.502 [0.381–0.662], respectively, relative to no change in % TS). At 30% TS, the association between TS and OS for NSCLC pts treated with nivolumab was comparable with that seen for melanoma pts (HR [95% CI] of 0.361 [0.252–0.517]). Conclusions: An association was found between the extent of TS and OS across melanoma pts receiving ipilimumab or nivolumab and NSCLC pts receiving nivolumab in this exploratory retrospective analysis. Measuring the extent and timing of TS may have use in predicting potential OS benefits of immune checkpoint inhibitors; however, this observation would need to be prospectively evaluated in data from a well-controlled study. Additional exploratory analyses are ongoing to assess correlates to OS benefit.