Improved survival with third-line CAR-T therapy vs alternative treatment in DLBCL: A propensity score–matching analysis

Introduction Chimeric antigen receptor T-cell (CAR-T) therapy has transformed the management of relapsed/refractory diffuse large B-cell lymphoma (r/r DLBCL). Initially approved for use in the third line (L3) setting and beyond, CAR-T has increasingly been adopted earlier in the second line (L2). However, many patients continue to receive CAR-T as L3 therapy due to prior treatment choices or L2 CAR-T ineligibility (e.g., late relapses). The aim of this analysis is to compare survival outcomes between CAR-T and alternative L3 therapies using propensity score matching (PSM) to adjust for key clinical differences. Methods We analyzed 240 consecutive r/r DLBCL patients prospectively enrolled in the NiHiL project (NCT03199066), all initially treated with R-CHOP-like regimens and receiving L3 therapy between 2020–2023. Baseline characteristics, diagnosis(dg)-to-L3 interval, clinical parameters at L3 (age, ECOG performance status [PS], and year of L3), L3 and L4 treatment patterns and outcomes were collected. Patients with ECOG PS 0–2 at L3 (n=163) were included; those with prior L2 CAR-T failure (n=6) were excluded, resulting in 157 eligible patients. PSM (1:1, caliper 0.2, nearest neighbor) was performed between patients with the intention of receiving CAR-T and those treated with alternative L3 therapies. The matching was based on age, ECOG PS, and dg-to-L3 interval. The endpoints were event-free survival (EFS) and overall survival (OS), both calculated from the L3 initiation (for CAR-T patients defined as the start of bridging [holding] therapy, or apheresis if no bridging). Results Of the 157 included patients, 75 were intended-to-CAR-T and 65 received CAR-T (axi-cel n=26, or tisa-cel n=39), while 82 received alternative L3 treatments (novel therapies n=27, salvage chemotherapy n=14, other therapy n=41). PSM yielded 104 matched patients (52 per group). We first compared the outcomes of patients included in the PSM (i.e., matched; n=104) versus those excluded (n=53). Among patients intended for CAR-T (n=75), individuals excluded from PSM (n=23) were younger (P<0.01), and had longer dg-to-L3 interval (P<0.01); these unmatched CAR-T patients had better EFS at L3 (P=0.05; OS P=0.29) compared to matched CAR-T patients (n=52). In alternatively treated patients (n=82), those who were not matched by PSM (n=30) were older (P<0.01), had longer dg-to-L3 interval (P<0.01), but did not show survival differences compared to matched patients (n=52; EFS P=0.48, OS P=0.59). PSM (n=104) achieved good balance across key covariates at L3 initiation between CAR-T and alternatively treated cohorts (both n=52): median age 64 vs 66 years (SMD 0.180), ECOG PS 1–2 in 35 (67%) vs. 37 (71%) cases (SMD 0.083), median dg-to-L3 interval of 19 vs 15 months (SMD 0.148). Disease characteristics at initial diagnosis were also comparable between groups: clinical stage III–IV (81% vs 83%, ns), elevated LDH (69% vs 71%, ns), extranodal involvement >1 site (38% vs 48%, ns), IPI score 3–5 (60% vs 60%, ns), and primary refractory or early relapsing disease (46% vs 57%, ns). Of alternatively treated patients included in the PSM (n=52), 20 (38%) received novel therapies (Pola-BR n=14, bispecific antibodies n=4, lenalidomide plus rituximab n=1, ibrutinib n=1), 12 (23%) salvage, and 20 (38%) received other treatments. After median follow-up of 20.9 months (living patients), EFS was significantly improved in the matched CAR-T group (43.5% vs 24.7% at 18 months, HR=0.62, 95% CI 0.40–0.98, P=0.04) compared to matched alternatively treated patients, with a trend towards prolonged OS (66.9% vs 43.1% at 18 months, HR=0.72, 95% CI 0.43–1.21, P=0.22). Among the matched patients who ultimately received CAR-T (n=48 out of 52), 18-month EFS was 47.2% and 18-month OS was 68.2%. Of these, 21 (44%) required subsequent L4 therapies (novel agents n=11, salvage n=1, other n=9). In the alternatively treated group (n=52), 31 (60%) required L4 therapy (CAR-T n=8, novel agents n=8, salvage n=4, other n=11). Conclusion After adjusting for key clinical covariates, L3 CAR-T therapy was associated with significantly improved EFS compared to alternative L3 treatments in r/r DLBCL. Despite its increasing use in earlier lines, CAR-T remains a valuable and effective treatment option in the L3 and beyond for eligible patients not previously exposed to this treatment modality.