A systematic review and meta-analysis of nonrelapse mortality after CAR T cell therapy

Abstract

Although chimeric antigen receptor (CAR) T cell therapy represents a transformative immunotherapy, it is also associated with distinct toxicities that contribute to morbidity and mortality. In this systematic review and meta-analysis, we searched MEDLINE, Embase and CINAHL (Cochrane) for reports of nonrelapse mortality (NRM) after CAR T cell therapy in lymphoma and multiple myeloma up to March 2024. After extraction of causes and numbers of death, we analyzed NRM point estimates using random-effect models. We identified 7,604 patients across 18 clinical trials and 28 real-world studies. NRM point estimates varied across disease entities and were highest in patients with mantle-cell lymphoma (10.6%), followed by multiple myeloma (8.0%), large B cell lymphoma (6.1%) and indolent lymphoma (5.7%). Entity-specific meta-regression models for large B cell lymphoma and multiple myeloma revealed that axicabtagene ciloleucel and ciltacabtagene autoleucel were independently associated with increased NRM point estimates, respectively. Of 574 reported nonrelapse deaths, over half were attributed to infections (50.9%), followed by other malignancies (7.8%) and cardiovascular/respiratory events (7.3%). Conversely, the CAR T cell-specific side effects, immune effector cell-associated neurotoxicity syndrome/neurotoxicity, cytokine release syndrome and hemophagocytic lymphohistiocytosis, represented only a minority of nonrelapse deaths (cumulatively 11.5%). Our findings underline the critical importance of infectious complications after CAR T cell therapy and support the comprehensive reporting of NRM, including specific causes and long-term outcomes.

Figure 1.. Study retrieval and identification for meta-analysis.

Flow diagram displaying the process for study inclusion and exclusion for the systematic review and meta-analysis of NRM following CAR T-cell therapy following the PRISMA guidelines.

Figure 2.. Forest plot of NRM point estimates across all study cohorts and stratified by treatment setting.

orest plot illustrating NRM point estimates and 95% confidence intervals using a random effects model. Studies are ordered by disease entity from top to bottom: indolent lymphoma (IL, blue), large B-cell lymphoma (LBCL, yellow), mantle cell lymphoma (MCL, grey), and multiple myeloma (MM, red). The total point estimate across all included CAR T-cell cohorts is highlighted in bold black, while the aggregated values for the treatment setting subgroups are highlighted in bold grey. Heterogeneity measures including I² are depicted (I² between 50% to 75% indicates moderate-to-high study heterogeneity). Abbreviation: NRM = non-relapse mortality.

Figure 3.. CAR T-cell products impact NRM point estimates in a disease-specific manner

A Bubble plot demonstrating NRM point estimates in relation to follow-up times for indolent lymphoma (IL, blue), large B-cell lymphoma (LBCL, yellow), multiple myeloma (MM, red), and mantle cell lymphoma (MCL, grey). Each bubble represents one cohort. Bubble size indicates the total number of patients per cohort. The aggregated NRM point estimate for all cohorts within one entity is encircled in black. B Depiction of aggregated NRM point estimates and 95% confidence intervals (CIs) for the different disease entities and the overall study cohort (“All”). C, E Comparison of aggregated NRM point estimates and 95% CIs across the different CAR T-cell products for LBCL (C) and MM (E). The p-values for the comparisons of NRM point estimates by disease entity and CAR T-cell product were calculated using the test for subgroup differences (random effects model). D, F Multivariable meta-regression analysis using random effects models for study features in LBCL (D) and MM (E). The integrated forest plot indicates the model estimates with 95% CI for each study variable (e.g., CAR T-cell product, treatment setting, treatment line, treatment era, follow-up time). Reference levels for the calculation of model estimates and respective p-values are depicted for each study feature. Abbreviations: NRM = non-relapse mortality; CT = clinical trials; RWS = real-world studies.

Figure 4.. Distribution of causes of non-relapse deaths across treatment settings and disease entities.

A Top: donut plot displaying causes of death among the entire study cohort. Defined causes of death are shown in different colors, undefined and unclassifiable causes of death are depicted in grey colors. Additional pie charts subdivide non-relapse deaths from infections (blue tones, right), other malignancies (green tones, middle), and cardiovascular/respiratory events (red tones, left). B-C Comparison of the causes of non-relapse deaths across the treatment setting (B) and between disease entities (C). Chi-square distribution test was used for statistical testing. Abbreviations: AML = acute myeloid leukemia, CRS = cytokine release syndrome, CT = clinical trial, HLH = hemophagocytic lymphohistiocytosis, ICANS = immune effector cell-associated neurotoxicity syndrome, MDS = myelodysplastic syndrome, NOS = not otherwise specified, RW = real-world.