Second Primary Malignancies after CAR T-Cell Therapy: A Systematic Review and Meta-analysis of 5,517 Lymphoma and Myeloma Patients

Abstract

Purpose: Chimeric antigen receptor (CAR) T-cell therapy is a potent immunotherapy for hematologic malignancies, but patients can develop long-term adverse events, including second primary malignancies (SPM) that impact morbidity and mortality. To delineate the frequency and subtypes of SPMs following CAR-T in lymphoma and myeloma, we performed a systematic review and meta-analysis.

Experimental design: A literature search was conducted in the MEDLINE, Embase, and Cochrane CENTRAL databases. Following the extraction of SPM cases and assignment of malignant origin, we analyzed SPM point estimates using random effects models.

Results: We identified 326 SPMs across 5,517 patients from 18 clinical trials and 7 real-world studies. With a median follow-up of 21.7 months, the overall SPM point estimate was 6.0% (95% confidence interval, 4.8%-7.4%). SPM estimates were associated with treatment setting (clinical trials > real-world studies), duration of follow-up, and number of prior treatment lines, which were each confirmed as independent study-level risk factors of SPM in a meta-regression model. A subgroup meta-analysis of the four trials that randomized CAR-T versus standard-of-care revealed a similar risk of SPM with either treatment strategy (P = 0.92). In a distribution analysis of SPM subtypes, hematologic malignancies were the most common entity (37%), followed by solid tumors (27%) and non-melanoma skin cancers (16%). T-cell malignancies represented a small minority of events (1.5%). We noted disease- and product-specific variations in SPM distribution.

Conclusions: These data raise awareness of SPM as a clinically relevant long-term adverse event in patients receiving CAR T-cell therapy. However, our findings do not indicate that SPM frequency is higher with CAR-T versus previous standard-of-care strategies.

Figure 1.. Forest plot of SPM point estimates across all study cohorts and stratified by disease entity.

Forest plot illustrating SPM point estimates and 95% confidence intervals (95% CI) stratified by disease entity. The results of the random effects model and measures of heterogeneity are shown for each disease entity. Data is presented as point estimate proportion with error bars showing 95% CI. The p-value for the comparison of subgroups was derived from a two-sided test for subgroup differences (random effects model). Heterogeneity measures including I² are depicted (I² between 50% to 75% indicates moderate-to-high study heterogeneity). Abbreviations: IL = indolent lymphoma, LBCL = large B-cell lymphoma, MCL = mantle cell lymphoma, MM = multiple myeloma, SPM = second primary malignancy.

Figure 2.. Subgroup analyses of SPM point estimates

Comparison of aggregated SPM point estimates with 95% CIs for A follow-up time (above median n = 17, below median n = 17), B inclusion of patients with prior HCT (> 35% n = 12, < 35% n = 20), C number of prior lines (> 3 n = 10, < 3 n = 19), D age (< 61 years n = 14, > 61 years n = 11), and E treatment setting (CT n = 21, RW n = 13). To compare follow-up times, prior HCT and therapy line exposure as well as for age, we classified studies using the respective median values. P-values for the comparison of SPM point estimates were derived from the test for subgroup differences (random effects model). Size of included patients per study is illustrated by point size. Abbreviations: CT = clinical trial, HCT = hematopoietic cell therapy, SPM = secondary primary malignancy, RW = real-world.

Figure 3.. Forest plot of SPM point estimates in CAR-T versus standard-of-care arms in a subgroup of randomized clinical trials.

Forest plot illustrating Odds ratios and 95% confidence intervals (95% CI) for the occurrence of SPM in comparison of CAR-T therapy versus standard-of-care (SOC) derived from randomized clinical trials. Pooled Odds ratio was calculated based on a meta-analysis of binary outcomes using a random-effects model. Data is presented as Odds ratio with error bars showing 95% CI. Heterogeneity measures including I² are depicted (I² between 25% to 50% indicates low-to-moderate study heterogeneity). Abbreviations: OR = Odds ratio, SOC = standard-of-care, SPM = second primary malignancy.

Figure 4.. Characterization and distribution of SPM subtypes across disease entities and CAR-T products

A Donut plot displaying major SPM subtypes among the entire study cohort: hematologic (red), solid tumors (blue), non-melanoma skin cancers (dark grey) and not further specified SPM (light grey). B-C Pie charts subdivide hematologic SPMs (red tones, B) and solid tumors (blue tones, C). D-E Comparison of distribution of hematologic and solid SPMs (excluding non-melanoma skin cancers) across disease entities (D) and between individual CAR-T products (E). Studies reporting on multiple entities or CAR-T products without further separation were excluded from the analysis. Chi-square distribution test was used for statistical testing. Abbreviations: MDS = myelodysplastic syndrome, AML = acute myeloid leukemia, MPN = myeloproliferative neoplasms, T/B-NHL = T/B cell non-Hodgkin lymphoma, HPB = hepatopancreaticobiliary, GIST = gastrointestinal stromal tumor, NOS = not otherwise specified, SPM = second primary malignancy, MCL = mantle cell lymphoma, IL = indolent lymphoma, MM = multiple myeloma, LBCL = large B cell lymphoma.