Abstract only 62 Background: Standard early endpoints are often poor predictors of overall survival (OS) for immunotherapies. There is a need for new intermediate endpoints to support upcoming combination studies with immune checkpoint inhibitors. We investigated the predictive performance of on-treatment tumor growth rate (TGR) estimates to predict OS in atezolizumab clinical trials in patients (pts) with metastatic urinary carcinoma (mUC). Methods: Atezolizumab showed durable response, prolonged OS and good tolerability supporting its therapeutic use in mUC (Rosenberg, Lancet 2016; Balar, lancet 2017; Powles, EACR-AACR-SIC 2017). A multivariate model linking baseline patient characteristics and TGR (estimated using time profile of sum of longest diameters (SLD) of target lesions, RECIST 1.1) to OS was developed using phase 2 data (IMvigor210) and validated to predict phase 3 trial (IMvigor211). Results: To estimate TGR, patients needed to have at least one post-baseline measurement: n = 365/429 (85%) in IMvigor210 (102 in cohort 1, 1L cisplatin-ineligible and 263 in cohort 2, prior platinum-treated 2L+) and n = 745/902 treated (83%) in IMvigor211 (prior platinum-treated 2/3L). A lognormal model with independent baseline prognostic factors (alkaline phosphatase, ECOG performance status and number of metastatic sites) and estimated TGR was developed. The model predicted OS distributions in the two IMvigor210 cohorts in all comers as well as by PD-L1 expression on tumor-infiltrating immune cells (IC). In IMvigor211, SLD profiles in atezolizumab and chemotherapy arms crossed at about 40 weeks with more initial shrinkage in chemotherapy-treated pts but slower overall TGR in atezolizumab-treated pts. The IMvigor210 OS model predicted OS hazard ratio (HR) in IMvigor211 in all comers with post-baseline tumor scans (model-predicted HR prediction interval: 0.80 0.70-0.92 vs. 0.74 observed) as well as by IC status: 0.66 0.50-0.87 vs. 0.70 observed in IC23. Conclusions: Model-based TGR is a promising exploratory endpoint and is being investigated to support early decisions in signal-seeking atezolizumab combination trials in mUC.