Abstract Introduction: This is a single-arm phase 2 trial testing the efficacy of the investigational dual TORC 1 and 2 inhibitor, sapanisertib (TAK-228), and hormonal therapy (HT) in estrogen receptor-positive patients relapsed or refractory to hormonal therapy and everolimus (NCT02049957). We sought to estimate the potential progression-free survival (PFS) benefit of adding sapanisertib to continued HT treatment by comparing the tumor growth rate, and therefore extrapolated time-to-tumor size progression (TTTSP) in the absence of sapanisertib, to the observed TTTSP in sapanisertib-treated patients. Experimental Procedure: Using the most recent historical prebaseline scans from the subset of response evaluable patients enrolled in the study with scans available (N=40), we estimated the exponential rate of tumor size increase on continued HT alone and extrapolated that rate to a virtual TTTSP, defined per RECIST 1.1 as the time until the sum of diameters (SoD) of target lesions was predicted to exceed 120% of the baseline SoD. We applied Kaplan-Meier (KM) analysis to estimate the potential benefit of sapanisertib added to HT by comparing 3 different PFS estimators to the TTTSP distribution for HT alone: (1) the model-estimated TTTSP obtained by fitting a tumor dynamic model to the on-treatment tumor size data (N = 40), (2) the observed PFS in the same patient population for which prebaseline scans were available (N = 40), and (3) the observed PFS in the entire response evaluable study population (N = 94). Results: Because TTTSP does not account for new lesions or death, the longest times to event were observed using the model-estimated TTTSP of HT with sapanisertib. Accordingly, the strongest separation between the various time-to-event metrics for the HT + sapanisertib arm and the model-based HT-only TTTSP ‘arm’ was observed using the model-estimated TTTSP of HT + sapanisertib (HR = 0.16, 95% CI: [0.07,0.36]). Smaller but statistically significant separation between observed HT + sapanisertib PFS and model-based HT-only TTTSP KM curves was observed in the subset of patients for whom prebaseline scans were available (HR = 0.40, 95% CI: [0.25,0.7]). A similar separation was seen using the full set of 94 evaluable patients in the active arm, irrespective of whether they provided a prebaseline scan (HR = 0.46, 95% CI: [0.3,0.75]), confirming that PFS of the analysis subset was reflective of the larger population. Conclusion: Under the assumption that HT-only patients would have PFS events at least as quickly as the model-based TTTSP, given that progression by new lesions and/or death are not captured by TTTSP, the observed HT + sapanisertib PFS to model-based HT-only TTTSP hazard ratio in the response-evaluable patient subset with prebaseline scans may provide a conservative estimate of what would be observed in a head-to-head HT +/- sapanisertib study in this patient population. Our analysis suggests that using such a modeling approach may aid program decision making by estimating 2 “arms” of time-to-event data from a single-arm study that captures each patient’s most recent prebaseline scan. It predicts that sapanisertib adds benefit to HT in terms of longer times to RECIST progression/TTTSP (median: 5.5 months vs 1.6 months). Citation Format: Dean Bottino, Yaming Hang, Rachel Neuwirth, Chirag Patel, Farhad Sedarati, Douglas V Faller, Annette Schmid. Using prebaseline and on-treatment tumor assessments to compare time to tumor size progression of hormonal therapy + sapanisertib vs an extrapolated hormonal therapy alone from a single-arm phase 2 study [abstract]. In: Proceedings of the AACR-NCI-EORTC International Conference on Molecular Targets and Cancer Therapeutics; 2019 Oct 26-30; Boston, MA. Philadelphia (PA): AACR; Mol Cancer Ther 2019;18(12 Suppl):Abstract nr A073. doi:10.1158/1535-7163.TARG-19-A073