Abstract A27: Impact of a validated composite comorbidity score on outcomes in patients treated with CAR T-cell therapy for diffuse large B cell lymphoma (DLBCL): A multicenter real-world evidence (RWE) study

• Published in: Blood cancer discovery 2022-09, Vol.3 (5_Supplement), p.A27-A27
• Main Authors: Shouse, Geoffrey, Kaempf, Andy, Gordon, Max, Yashar, David, Sigmund, Audrey M, Smilnak, Gordon, Bair, Steven M, Mian, Agrima, Fitzgerald, Lindsey, Bajwa, Amneet, Jaglowski, Samantha, Bailey, Neil, Shadman, Mayzar, Patel, Krish, Stephens, Deborah M, Kamdar, Manali, Hill, Brian T, Gauthier, Jordon, Karmali, Reem, Nastoupil, Loretta, Kittai, Adam S, Danilov, Alexey V
• Format: Article
• Language: English
• ISSN: 2643-3249; 2643-3249
• DOI: 10.1158/2643-3249.LYMPHOMA22-A27

Introduction: CAR T-cell therapy (CART) has dramatically improved outcomes for patients (pts) with relapsed/refractory (r/r) DLBCL, but the majority of pts still have poor outcomes due to progressive disease and toxicities. Here we used a machine learning algorithm to rank the prognostic impact of specific comorbidities measured by the Cumulative Illness Rating Scale (CIRS) in DLBCL pts indicated for CART in a multicenter learning cohort (LC) and a separate validation cohort (VC). Methods: pts with r/r DLBCL included in RWE analysis underwent leukapheresis for CART at 10 academic centers. CIRS was assessed at the time of T-cell collection (Salvi et al, 2008). Progression-free survival (PFS) and overall survival (OS) were measured from T-cell collection. Random survival forest (RSF) modeling of PFS and OS was repeatedly applied to random subsets of the LC to determine the most important CIRS categories and comorbidity levels in the presence of known prognostic factors. Cox proportional hazards models of PFS and OS were fit to both cohorts. Associations between comorbidities and CART adverse events were evaluated with Fisher’s exact test. Results: The LC comprised 577 pts, median age of 63 (range, 19-90); 90% had ECOG 0-1. Median number of prior therapies was 3 (range, 1-11). GCB subtype was found in 54% of pts, with 38% non-GCB and 8% unknown. Twenty-seven pts (4.7%) died before CART infusion. Of the 550 pts who received CART, 71% got axicabtagene ciloleucel, 22% tisagenlecleucel, and 7% lisocabtagene maraleucel. The median CIRS score was 7 (range, 0-25) with 54% having CIRS ≥7. The median PFS was 11 months (95% CI: 8 – 15) and OS 30 months (95% CI: 23 – NA), with a median follow-up time of 20 months. Although CIRS ≥7 was significantly associated with PFS (HR=1.26) and OS (HR=1.35) in univariable analysis, it did not remain significant in multivariable models. According to a RSF tree depth-weighted nodal split score, severe comorbidities in the following CIRS categories had the greatest impact on PFS: respiratory, upper GI, renal, and hepatic (denoted Severe4, 9% of pts). When accounting for other significant variables, Severe4 was independently associated with inferior PFS (HR=2.45, p