A real-world comparison of tisagenlecleucel and axicabtagene ciloleucel CAR T cells in relapsed or refractory diffuse large B cell lymphoma

Abstract

Axicabtagene ciloleucel (axi-cel) and tisagenlecleucel (tisa-cel) have both demonstrated impressive clinical activity in relapsed/refractory (R/R) diffuse large B cell lymphoma (DLBCL). In this study, we analyzed the outcome of 809 patients with R/R DLBCL after two or more previous lines of treatment who had a commercial chimeric antigen receptor (CAR) T cells order for axi-cel or tisa-cel and were registered in the retrospective French DESCAR-T registry study ( NCT04328298 ). After 1:1 propensity score matching (n = 418), the best overall response rate/complete response rate (ORR/CRR) was 80%/60% versus 66%/42% for patients treated with axi-cel compared to tisa-cel, respectively (P < 0.001 for both ORR and CRR comparisons). After a median follow-up of 11.7 months, the 1-year progression-free survival was 46.6% for axi-cel and 33.2% for tisa-cel (hazard ratio (HR) = 0.61; 95% confidence interval (CI), 0.46-0.79; P = 0.0003). Overall survival (OS) was also significantly improved after axi-cel infusion compared to after tisa-cel infusion (1-year OS 63.5% versus 48.8%; HR = 0.63; 95% CI, 0.45-0.88; P = 0.0072). Similar findings were observed using the inverse probability of treatment weighting statistical approach. Grade 1-2 cytokine release syndrome was significantly more frequent with axi-cel than with tisa-cel, but no significant difference was observed for grade ≥3. Regarding immune effector cell-associated neurotoxicity syndrome (ICANS), both grade 1-2 and grade ≥3 ICANS were significantly more frequent with axi-cel than with tisa-cel. In conclusion, our matched comparison study supports a higher efficacy and also a higher toxicity of axi-cel compared to tisa-cel in the third or more treatment line for R/R DLBCL.

Fig. 1. Patient flow diagram for PSM analysis.

¹Manufacturing failure (n = 3), uncontrolled infection (n = 3), waiting for infusion (n = 3), patient decision (n = 1), leukapheresis failure (n = 1), acute coronary syndrome (n = 1), concomitant malignancy (n = 1) and progression of another malignancy (n = 1). ²Patients with PMBCL histology were excluded because tisa-cel has no approval for this histology. ³Patients with ≥25% of missing data for matching covariates were removed from the matching step.

Fig. 2. Survival of the whole cohort of patients treated with commercial tisa-cel or axi-cel from the French DESCAR-T registry before any matching.

a, OS since CAR T order (n = 809, blue line) or since CAR T infusion (n = 729, red line). For 80 patients, a CAR T product was ordered, but patients did not proceed until infusion due to disease progression or death, physician decision or other reasons (see patient flowchart). b, PFS from CAR T infusion (n = 729). Shaded areas correspond to the 95% confidence bands using the Hall–Wellner method. CL, confidence limit.

Fig. 3. Survival according to CAR T product after PSM.

a, Propensity score reflects the probability of receiving tisa-cel or axi-cel conditional on an exhaustive list of 14 pre-infusion covariates. PSM is based on matching patients with similar propensity score. Comparability according to each covariate (for instance, disease stage depicted here) of the resulting two sub-cohorts of patients receiving one CAR T or the other is checked using SMDs (Extended Data Fig. 3). b, DOR according to CAR T product (axi-cel, n = 168, red line; tisa-cel, n = 138, blue line) (P = 0.11). c, DOR according to CAR T product and response quality (complete response (CR) versus partial response (PR); P = 0.30 and P = 0.90, respectively) (axi-cel and CR, n = 126, red line; tisa-cel and CR, n = 88, blue line; axi-cel and PR, n = 42, brown line; tisa-cel and PR, n = 50, green line). d, PFS according to CAR T product (P = 0.0003). e, OS according to CAR T product (P = 0.0072). P values were calculated using two-sided log-rank tests. No adjustment was made for multiple comparisons. Shaded areas correspond to the 95% confidence bands using the Hall–Wellner method. CL, confidence limit; NA, not assessable.

Extended Data Fig. 1. Overall survial (OS) for patients with a CAR-T product order who did not proceed until infusion according to CAR-T type.

Extended Data Fig. 1 For 80 patients, a CAR-T product was ordered but was never infused (n = 42 for axi-cel, n = 38 for tisa-cel) due to disease progression (n = 60), physician decision (n = 6) and other reasons detailed in the patient flow diagram in Fig. 1 (n = 14). Shaded areas correspond to the 95% confidence bands using Hall-Wellner method. P value was calculated using a two-sided logrank test.

Extended Data Fig. 2. Univariate prognostic analysis.

a, univariate analysis for progression-free survival (PFS). b, univariate analysis for overall survival (OS). Blue point represents the value of the hazard ratio (HR) and red segment the value of the 95% confidence interval (CI). The first category in parentheses is taken as the reference category for comparison and HR computation. For instance, for PFS analysis HR is 2.95 for patients with a LDH level twice above the upper limit of the normal (ULN) compared to patients with a normal value. A HR < 1 represents a prognosis factor associated with a prolonged survival while a HR > 1 represents a prognosis factor associated with a shorter survival. A prognostic factor is statistically significant if the 95% CI does not contain 1. Cox univariate model was used for calculating HR and associated two-sided P value. No adjustment was performed for multiple comparisons. All centers were anonymized. Time from last treatment, age and time to first order of the center were dichotomized according to the median value of data distribution. Number of prior treatment, LDH level and bridging were divided into 3 categories (2 vs 3-4 vs >4 prior lines; normal LDH, LDH between 1 and 2 times the ULN and LDH above 2 times the ULN; no bridging vs response to bridging vs no response to bridging, respectively). Time from last treatment represents the time from the start of the last treatment to the time of CAR T infusion. Time from first order of the center represents the time from the order of the first CAR-T in the center to the time of infusion of the CAR T for the patient (as a surrogate of the “center experience” for CAR-T therapy for a given patient). DLBCL, diffuse large B-cell lymphoma; HGBL, high grade B-cell lymphoma; trFL/MZL, transformed follicular lymphoma or marginal zone lymphoma; Resp; complete or partial response to bridging; No Resp, no response (that is stable or progressive disease) to bridging; ECOG, Eastern Collaborative Oncology Group Performance Status; CRP, C reactive protein; LDH, lactate dehydrogenase.

Extended Data Fig. 3. Balance assessment before and after matching.

a, Propensity score (PS) distribution before and after PSM. b, Standardized Mean Differences (SMD) of covariates categories before and after PSM. c, Absolute SMD before and after PSM. d, CAR T products distribution across centers from the DESCAR-T registry before and after PSM (light grey= axi-cel; dark grey=tisa-cel). e, CAR T products distribution according to categorical covariates before and after PSM (light grey= axi-cel; dark grey=tisa-cel). f, Balance assessment according to CAR T product for continuous covariates before and after PSM. Box plot represents 1^st quartile and 3^rd quartile. Line in the middle of the box represents the median. Round symbol represents the mean. Left segment represents distribution from the minimal value. Right segment represents distribution to the maximal value. g, SMD of covariates categories before and after IPTW. b, absolute SMD before and after IPTW.

Extended Data Fig. 4. Response rates and survival according to CAR-T type after inverse probability of treatment weighting.

a, In IPTW, weight of each individual patient is calculated as the inverse of their probability of receiving tisa-cel or axi-cel, assessed by their propensity score. Compared to PSM, it creates a pseudo-population of patients in which patients with a lower likelihood of receiving one CAR-T is over-weighted in the final population. As for PSM, comparability according to each covariate of the resulting 2 pseudo-cohorts of patients receiving one CAR-T or the other is checked using standardized mean differences (SMD, Extended Data Fig. 3). b, DOR according to CAR-T. c, DOR according to CAR-T and response quality. d, PFS according to CAR-T. e, OS according to CAR-T. Shaded areas correspond to the 95% confidence bands using Hall-Wellner method. P values were calculated using two-sided logrank tests. No adjustment was made for multiple comparisons.

Extended Data Fig. 5. Planned subgroup analyses according to age and tumor bulk.

a, PFS according to CAR T product in patients ≤ 70 years. b, PFS according to CAR T product in patients > 70 years. c, OS according to CAR T product in patients ≤ 70 years. d, OS according to CAR T product in patients > 70 years. e, PFS according to CAR T product in patients with ≤ 5 cm tumor bulk. f, PFS according to CAR T product in patients with > 5 cm tumor bulk. g, OS according to CAR T product in patients with ≤ 5 cm tumor bulk. h, OS according to CAR-T in patients with > 5 cm tumor bulk. Shaded areas correspond to the 95% confidence bands using Hall-Wellner method. P values were calculated using two-sided logrank tests. No adjustment was made for multiple comparisons.

Extended Data Fig. 6. Survival according to CART product after PSM or IPTW in the complete case analysis (that is, where all cases with at least one missing value in matching covariates have been removed).

a, DOR according to CART product after PSM. b, PFS according to CART product after PSM. c, OS according to CART product after PSM. d, DOR according to CART product after IPTW. e, PFS according to CART product after IPTW. f, OS according to CART product after IPTW. Shaded areas correspond to the 95% confidence bands using Hall-Wellner method. P values were calculated using two-sided logrank tests. No adjustment was made for multiple comparisons.

Extended Data Fig. 7. Overall survival from CAR-T order (instead of from infusion) according to CAR-T type after PSM or IPTW.

a, OS according to CAR-T after PSM. b, OS according to CAR-T after IPTW. Shaded areas correspond to the 95% confidence bands using Hall-Wellner method. P values were calculated using two-sided logrank tests. No adjustment was made for multiple comparisons.