Introduction:

IEC-EC has emerged as a unique toxicity of BCMA-targeted CAR-T therapy in multiple myeloma (MM) patients (pts), most commonly with ciltacabtagene autoleucel (cilta-cel) (G. Fortuna et al, Blood Cancer J, 2024). While colitis associated with other cancer therapies such as checkpoint inhibitors is well characterized, IEC-EC may follow distinct mechanisms including possible lymphoproliferative processes that potentially could occur with other CAR-T products. We hypothesized that a pharmacovigilance (PV) analysis of the FDA Adverse Event Reporting System (FAERS) would reveal gastrointestinal (GI) toxicity signals across CAR-T therapies. We specifically analyzed serious adverse event (SAE) reports for GI toxicities associated with CAR-T therapies versus with checkpoint inhibitors to better compare their frequency and clinical features.

Methods:

We analyzed all SAE reports from FAERS through July 2025 for 5 approved CAR-T products, including CD19-targeted therapies in lymphoma pts. Events were coded using standardized MedDRA terms: “diarrhea,” “immune-mediated enterocolitis,” and “autoimmune colitis.” SAEs were defined per FDA criteria (death, hospitalization or other serious outcomes). We did not assess causality but evaluated biological plausibility using literature on time-to-onset patterns. Signal detection employed three PV methods: reporting odds ratio (ROR), proportional reporting ratio (PRR), and empirical Bayes geometric mean (EBGM). Multiple testing correction was applied using Benjamini-Hochberg and Bonferroni methods (10 comparisons). We compared these findings against corresponding SAE reports for checkpoint inhibitors along with published data sets including USA MM Immunotherapy Consortium data and registrational trials of each product. We built machine learning models to predict GI toxicities using key FAERS features (drug, demographics, geography), applying SMOTE for class imbalance and evaluated with standard performance metrics.

Results:

Among 13,581 SAE reports, 55% of patients were male with a median age of 66 (57.9% ≥65 years), and 70.9% were from the USA. Reports were distributed as follows: axicabtagene ciloleucel (axi-cel, 57.6%, n=7828), cilta-cel (19.5%, n=2647), brexucabtagene autoleucel (brexu-cel, 11.9%, n=1615), idecabtagene vicleucel (ide-cel, 6.9%, n=933), and lisocabtagene maraleucel (liso-cel, 4.7%, n=641). Diarrhea occurred in 1.69% (n=230) of pts. Highest rates were seen with ide-cel (4.61%), followed by cilta-cel (1.89%), brexu-cel (1.49%), liso-cel (1.40%) and axi-cel (1.39%). These rates mirrored CARTITUDE-4 grade 3-4 diarrhea findings (3.8%). IEC-EC occurred in 0.22% (n=30), mostly with cilta-cel (1.02%, n=27). Axi-cel had only 3 (0.04%) cases, whereas the other CAR-T products had none. PV analysis revealed strong signals for colitis with cilta-cel (ROR: 126.35, 95% CI: 86.31–184.98) and for diarrhea with ide-cel (ROR: 2.91, 95% CI: 2.15–3.96), both significant after correction (p<0.001). For the consortium cohort, median time from CAR-T infusion to IEC-EC was 92.5 days (range, 22-210). Compared with checkpoint inhibitors, cilta-cel-associated IEC-EC rates (1.02%) exceeded those for the PD-1 agents nivolumab (0.23%), pembrolizumab (0.13%) and the CTLA-4 inhibitor ipilimumab (0.46%). Finally, XGBoost models predicting IEC-EC and diarrhea achieved accuracies of 0.91 and 0.89, respectively, with strong precision and recall metrics (>0.86).

Conclusions:

Our PV analysis reveals distinct GI SAE profiles among CAR-T therapies, with cilta-cel demonstrating a particularly high signal for IEC-EC (ROR: 126.35, p<0.001). However, 3 cases occurred with axi-cel as well. The reported colitis rates well exceeded those observed with established checkpoint inhibitors, suggesting unique pathogenic mechanisms and worsened clinical severity of IEC-EC versus checkpoint-associated colitis. These findings, along with emerging evidence of CAR T-cell tissue infiltration and potential lymphoproliferative complications, indicate that novel pathophysiologic processes are involved in IEC-EC that are distinct from classical autoimmune colitis.

This content is only available as a PDF.
2025
Sign in via your Institution