2548 Background: In the conduct of randomized trials Kaplan and Meier envisioned rates of censoring as similar between arms, providing accurate assessment of clinical trial results. Censoring is used when patients withdraw consent, leave study due to toxicity, or reach data cut-off without disease progression or death. Censoring can lead to erroneous conclusions as it can be either beneficial or detrimental to the arm under study. Such censoring can also explain how a statistically valid difference in PFS “disappears” when overall survival (OS) is examined. We hypothesized that censoring, especially that due to toxicity, would lead to a discrepancy between DOT and PFS since two different patient populations would be scored. Methods: We reviewed all phase III randomized studies of drugs approved by FDA since 2005 for pts with metastatic solid tumors, looking for DOT and PFS. We used standard statistical analyses using SAS. Results: We identified 55 Phase III studies conducted with abiraterone, axitinib, bevacizumab, cabazitaxel, cetuximab, eribulin, erlotinib, everolimus, ipilimumab, ixabepilone, lapatinib, panitumumab, pazopanib, sorafenib, sunitinib, temsirolimus and vandetinib. DOT was not provided in 27%. Forty-four comparisons (88 arms) were included in the analysis. The median PFS, DOT, delta PFS (difference in PFS between experimental and control arms) and delta DOT were: 161, 126, 51 and 36 days, respectively. The slopes of PFS vs DOT and delta PFS vs delta DOT were 1.16 and 1.03, respectively close to the ideal of 1.0. Five trials fell above the 90% CI boundary with delta PFS/delta DOT of 3 to 36, including two everolimus studies (PNET and breast cancer) two sunitinb studies (RCC and PNET) and one bevacizumab study (E2100, breast cancer). Conclusions: PFS and DOT as well as delta PFS and delta DOT should be concordant. The most likely explanation for a discordance between these values is toxicity-driven censoring and its occurrence raises concerns regarding the degree of efficacy. A greater utilization of “Time to Treatment Failure”, an endpoint that includes toxicity in its definition would be valuable in oncology trials, particularly those with high levels of toxicities.