From response to rupture: Evaluating GI perforation after CD19 CAR‑T in LBCL Free

Introduction:

CD19-directed chimeric antigen receptor T (CAR-T) cell therapy has dramatically improved outcomes in large B-cell lymphoma (LBCL). The gastrointestinal (GI) tract is the most common extranodal site, involved in nearly one-third of cases (Zhao et al. Hematology, 2022). Yet, post–CAR-T outcomes and complications among LBCL patients with GI involvement remain poorly characterized. In this study, we report outcomes for GI involvement—a common scenario encountered when consulting on CAR-T therapy—in a large cohort of LBCL patients treated with CAR-T.

Methods: This retrospective, single-center study included relapsed/refractory LBCL patients who received standard of care (SOC) CD19 CAR-T therapy between January 2018–May 2025. Patient records were reviewed to identify GI involvement through imaging, endoscopy, and pathology. Clinical characteristics and outcomes were compared between patients with and without GI involvement using standard statistical analyses. Kaplan-Meier and Cox proportional hazards models assessed associations between GI involvement and PFS and OS.

Results: Of 398 treated patients, 41 (11%) had radiologic or pathologic evidence of GI involvement at referral. The GI cohort received axicabtagene ciloleucel (axi-cel) (n=36) or lisocabtagene maraleucel (liso-cel) (n=5). Median age of the cohort was 68 years with 56% males. Majority of cases had DLBCL (63%), IPI ≥3 (51%), and stage IV disease (73%). Sites of GI involvement included stomach in 14 (34%), small bowel in 19 (46%), and large bowel in 13 (32%); none had esophageal disease. Involvement was serosal in 6 (15%), luminal in 12 (29%), mucosal in 7 (17%), and unknown in 16 (39%). Seven (17%) had multifocal GI disease.

Bridging therapy was administered to 26 (63%): chemotherapy in 14 (34%), radiation to GI site in 8 (20%), and steroids in 6 (15%). Tweleve (29%) received no bridging. Eight out of 14 evaluable patients had response to site post-bridge, with 7 achieving a PR and one achieving a CR. Residual disease at infusion was documented in 10 (24%), absent in 1 (2%), and unknown in 30 (73%).

In our cohort, 18 (44%) progressed post-CAR-T. Overall, 22 (53%) patients died, with 13 (32%) passing due to progression, 5 (12%) due to infection, one due to secondary malignancy, and one due to obstructive shock. Two passed due to GI perforation-related complications. Grade ≥3 CRS and ICANS occurred in 4 (10%) and 16 (39%) respectively. HLH occurred in one patient. Outcomes were comparable between patients with and without GI tract involvement (median PFS: 15.2 vs. 18.7 months, p = 0.7; median OS: 26.9 vs. 63.3 months, p = 0.06) with a median follow-up of 34.4 months.

Four patients (9.7%) developed GI perforation after receiving SOC axi-cel and bridging therapy (chemo (n=2), radiation to GI site (n=1), steroids and radiation to site (n=1)). Disease types included DLBCL (n=1), HGBCL (n=1), and TFL (n=2), with MYC rearrangements in two. All had IPI scores ranging 2-3 and advanced stage disease (stage III, n=1; stage IV, n=3). Sites of involvement included the small bowel (n=3) and colon (n=1), with one case being multifocal (stomach). Endoscopic findings were mucosal (n=3) and luminal (n=1). The small number of CAR-T–related perforations precluded any meaningful statistical analysis. One patient had site response to bridging (PR), with the other two having progressive disease, and one patient with unknown response status.

All cases had poor outcomes: one developed a small bowel-to colon fistula one month post infusion and died of progression within a year; another had a contained small bowel perforation, relapsed, and died within four months; a third had duodenal perforation following intensive care for Grade 3 CRS/ICANS and died; and the fourth deteriorated rapidly with septic shock and was transitioned to comfort care—autopsy confirmed a necrotic cecal luminal mass with perforation. This group exhibited significantly shorter median PFS and OS: 2.28 (95% CI: 0.95–NR) and 2.38 months (95% CI: 1.35–NR), respectively.

Discussion: In our comprehensive, single-center cohort, approximately 10% of LBCL patients treated with CD19 CAR T therapy exhibited GI tract involvement. Outcomes were similar between patients with and without GI involvement. Perforation remains rare, occurring in <10% of those with GI disease. However, CAR-T–related GI perforation had uniformly OS <1 year, underscoring the importance of optimizing bridging therapy in high-risk cases.