Background: Many patients (pts) with RR-DLBCL treated in the real-world are not offered ASCT despite its curative potential. Understanding the impact of ASCT on OS in the real world is critical to better position novel therapies as an alternative to ASCT. Methods: We retrospectively analyzed real-world pts with RR-DLBCL treated between 1/1/10-12/31/16 in 126 community-based hematology/oncology practices across the US. Pts treated with first line (1L) R-CHOP or DA-EPOCH-R were grouped into 2 categories: early relapse (progressive disease [PD] on treatment within 365 days from completing last cycle) and late relapse (any 2L treatment 366-730 days from last cycle of 1L). Pts were stratified whether ASCT was performed after 2L. Pts receiving rituximab monotherapy in 2L or later were excluded from analysis. A period of ≥6 months of clinical inactivity at end of follow-up served as a proxy for death. Median (95% CI) OS were estimated by Kaplan-Meier method, and adjusted hazard ratios (HR) were estimated by Cox proportional hazard models to adjust for differences in pt characteristics (sex, age at DLBCL diagnosis, time to relapse (2L), 1L R-CHOP, and comorbidities). Results: For the 430 pts, median age was 65 (range 18-84) years and 61% were male. Following 1L, over 86% (n = 371) had early relapse, and 14% (n = 59) had late relapse. Across all lines, 33% (n = 144) pts received a salvage regimen intended for transplant, with 38% (n = 55) of these pts proceeding to ASCT. For all pts at first relapse, median OS was 13.8 mo. from initiation of 2L. Median OS was significantly longer among those who underwent ASCT compared to those who did not (21.4 mo. vs 10.5 mo.; adjusted P< 0.01); HR = 0.51 (95% CI 0.32-0.81). Median OS shortened with each subsequent line of therapy—more so for non-ASCT group with a median OS at 2L of 10.5 mo., 3L of 9.4 mo., and 4L+ of 3.9 mo. for non-ASCT. Conclusions: Although 33% of real-world pts with RR-DLBCL received salvage regimens intended for ASCT, only 13% of all evaluable pts eventually underwent ASCT. The low utility of ASCT and the poor OS rates in non-ASCT pts in this real-world population demonstrate an unmet need for novel therapies in this setting.