Impact of “Cell-of-Origin” on Outcome after Axicabtagene-Ciloleucel CAR-T Cell Therapy in Relapsed/Refractory Aggressive Non-Hodgkin B-Cell Lymphomas

Introduction: Despite significant advances in characterizing aggressive B-cell non-Hodgkin lymphomas (aNHL) based on molecular and genetic signatures, the Hans algorithm, determined by immunohistochemistry, remains the most commonly used approach to differentiate aNHLs based on the cell-of-origin (COO) into germinal center B-cell (GCB) and non-GCB types. Although the Hans algorithm has demonstrated reliability as a predictor of better outcomes in newly diagnosed GCB cases, conflicting evidence has emerged, particularly in the relapsed/refractory (R/R) setting, and there are no definitive data about the impact of COO in patients (pts) receiving CAR-T cells. Thus, we conducted a retrospective analysis to evaluate the impact of COO extended to R/R aNHLs who underwent treatment with axicabtagene-ciloleucel (axi-cel).

Methods: We retrospectively collected data from pts with R/R aNHLs treated with axi-cel at Fred Hutchinson Cancer Center between February 2018 and August 2022. The Hans algorithm, using CD10, BCL6 and MUM1 markers, was employed to stratify the pts into two cohorts: GCB and non-GCB. Paired biopsies of diagnosis and relapse before CAR-T eligibility were reviewed at our institution. Response to CAR-T was assessed at one month per Lugano PET/CT criteria.

Univariate cox proportional hazard regression was used to estimate associations between baseline factors and outcomes. Statistically significant factors (p<.05) in univariate analysis were further evaluated by multivariate analysis. Continuous variables were compared by Mann-Whitney U test and categorical variables by Fischer’s exact test. Progression-free survival (PFS) was calculated from the date of CAR-T cell infusion to the date of death from any cause, disease progression or relapse. Overall survival (OS) was calculated from the date of CAR-T cell infusion to the date of death from any cause.

Results: Sixty-four pts were included in this analysis, with an equal distribution of pts in the two cohorts: 32 with GCB subtype and 32 with non-GCB subtype. Baseline pts’ characteristics and histology subtypes are listed in Table 1. Transformed follicular lymphoma and primary mediastinal B-cell lymphoma were not included in this analysis. The presence of bulky disease (defined as >5cm diameter) was higher in the GCB subtype (56% vs 22%, p<.001). Although primary refractory cases were more frequent in the GCB cohort (87.5 vs 66%), statistical significance was not reached (NR)(p=.07). The same proportion of pts received bridging therapy in the two groups (53% vs 53%, p=.99). Pts evaluable for response assessment were 62. Median time to best response was 29 days (range, 21-186) for GCB pts and 31 days (range, 22-182) for non-GCB pts (p=.91). The overall response rate was 70% for GCBs (34.5% complete response [CR], 34.5% partial response [PR]) and 75% for non-GCBs (44% CR, 31% PR). After a median follow-up of 24.5 months (range, 9-121), comparable outcomes were observed for both the COO groups: despite the estimated median OS of 14 months (95% CI, 4.73-NR) for GCB vs 40 months (95% CI, 12.55-67.46) for non-GCB, we did not find any significant differences (p=.73) (Figure 1A). Estimated median PFS was 10 months (95% confidence interval [CI], 3.5–16) for GCB and 11 months (95% CI, 0-24) for non-GCB (p=.57) (Figure 1B). The 2-year OS and PFS were: GCB, 45% (95% CI, 30-69) and 44% (95% CI, 29-67) respectively; non-GCB, 55% (95% CI, 38-79) and 29% (95% CI, 16-55), respectively.

Treatment-related toxicities did not differ between GCB and non-GCB, with cytokine-release syndrome rates of 97% vs 94% (p=.99) and neurotoxicity rates of 53% vs 62.5% (p=.61), respectively. The incidence of infectious complications was 25% vs 19% (p=.76).

On univariate analysis, the impact of COO on OS (HR=0.82, 95% CI 0.39-1.73, p=.60) and PFS (HR=1.18, 95% CI 0.61-2.28, p=.62) was undetermined. After adjusting for bulky disease, elevated LDH, absolute lymphocyte count≤1000/uL, performance status≥2, >2 lines of therapy and response status before axi-cel infusion, the impact of COO on OS (adjusted HR [aHR]=1.73, 95% CI 0.71-4.22, p=.23) and PFS (aHR=2.01, 95% CI 0.93-4.35, p=.08), remained undetermined.

Conclusion: We could not confirm an association between COO, as determined by Hans algorithm, and outcomes in R/R aNHLs who received axi-cel at our center. Prospective analyses on larger datasets and other types of CAR-T products are needed.