AUTHORS
Jeet Patel, DO, Matthew P Dandan, DO2, Karthik Gangu, MD3
AFFILIATIONS
1 The University of Kansas Medical Center, Olathe, Kansas, United States
2 Medical College of Wisconsin, Milwaukee, Wisconsin, United States
3 The University of Kansas Medical Center, Olathe, Kansas, United States
ABSTRACT
Background: Cytokine release syndrome (CRS) and immune effector cell–associated neurotoxicity syndrome (ICANS) are described by incidence and maximum grade. Whether their kinetics—time to onset and duration—carry prognostic information beyond grade is unknown at registry scale. Earlier onset may signal more intense in vivo CAR-T expansion, whereas prolonged toxicity may reflect sustained inflammation and greater exposure to immunosuppressive rescue therapy. Both could plausibly influence infection, response, and survival, but neither has been studied as a predictor in a large real-world myeloma cohort.
Objectives: To characterize CRS and ICANS kinetics after idecabtagene vicleucel (ide-cel) and to test whether their onset and duration predict clinically significant infection, disease response, and survival beyond maximum toxicity grade.
Methods: We analyzed all 821 commercial ide-cel recipients in the CIBMTR public-use dataset MM23-01a (infused 2021–2023). Primary exposures—CRS onset, CRS duration, ICANS onset, and ICANS duration—were modeled continuously (per interquartile range [IQR]) and dichotomized at the cohort median (early vs. late onset; short vs. prolonged duration). To address the conditional-onset/immortal-time structure, toxicity onset was entered as a time-dependent covariate in Cox models for overall (OS) and progression-free survival (PFS), complemented by a day-14 landmark analysis. Infection and response were modeled using multivariable logistic regression among patients who experienced toxicity. Models adjusted for maximum grade, age, comorbidity, baseline cytopenias, LDH, ISS stage, prior lines, and prior BCMA therapy. Proportional hazards were assessed by scaled Schoenfeld residuals and non-linearity by restricted cubic splines; missing covariates were handled by multiple imputation (M=25) with complete-case analyses in parallel.
Results: CRS occurred in 657 patients (80.0%; median onset 2 days [IQR 1–2; range 1–26], median duration 3 days [IQR 2–5]) and ICANS in 220 (26.8%; median onset 2.5 days [IQR 2–5], median duration 4 days [IQR 2–9]); grade ≥3 events were uncommon (CRS 24 [2.9%], ICANS 39 [4.8%]). The CRS kinetic parameters were interrelated: earlier onset was weakly associated with both longer duration (Spearman ρ=−0.28) and higher grade, and longer duration with higher grade (ρ=0.22). Over a median follow-up of 6.7 months, there were 227 deaths, 407 PFS events, and 367 clinically significant infections; median OS was 15.6 months (67% at 12 months) and median PFS 8.8 months. In time-dependent Cox models that appropriately attributed person-time to the unexposed state, developing CRS was independently associated with superior OS (HR 0.64, 95% CI 0.45–0.92; p=0.015) and PFS (HR 0.46, 0.35–0.60; p<0.001); once resolution was accounted for, an active-CRS state was not associated with mortality (HR 0.85; NS), suggesting that the observed benefit reflects developing CRS itself rather than time spent in the active toxicity state. Among CRS patients, prolonged (vs. short) duration was associated with higher complete-response odds (OR 1.90, 1.24–2.90; p=0.003) and, in the day-14 landmark cohort, with better PFS (HR 0.76, 0.58–0.99; p=0.045); the PFS association was robust to multiple imputation (HR 0.78, 0.62–0.98) and the exclusion of patients with <30 days follow-up, whereas the complete-response association attenuated under imputation (OR 1.37; p=0.09) and is considered exploratory. CRS onset timing was not independently associated with OS or PFS. Neither CRS nor ICANS kinetics were associated with clinically significant infection after adjustment (e.g., early vs. late CRS onset OR 0.96, 0.60–1.53); the higher crude infection rate among CRS patients (48% vs. 33%) showed no kinetic dose-response. Proportional-hazards assumptions held, and splines did not improve on linear terms.
Conclusion: In 821 commercial ide-cel recipients, the occurrence and duration of CRS—rather than its onset timing—carried favorable prognostic information for disease control and survival, with no association between toxicity kinetics and infection. ICANS kinetics were not independently associated with response or survival. These data reframe prolonged CRS as, on balance, a marker of effective CAR-T activation rather than an ominous sign, and suggest that its duration—more than its timing—adds prognostic value to the conventional incidence-and-grade description. Prospective, biomarker-anchored confirmation is warranted.
DISCLOSURES
Jeet Patel, DO: No relevant conflicts of interest to declare.