AUTHORS
Utkarsh Goel, MD, Catalin Dragomirescu2, Mohamed Hashem3, Diana Basali, MD4, Roy Chedid, MD5, Louis Samuel Williams, MD6, Sandra Mazzoni, DO7, Shahzad Raza, MD8, Jack Khouri, MD9
AFFILIATIONS
1 Mayo Clinic, Rochester, Minnesota, United States
2 Cleveland Clinic, Cleveland, Ohio, United States
3 Cleveland Clinic, Cleveland, Ohio, United States
4 Cleveland Clinic, Cleveland, Ohio, United States
5 University of Cincinnati, Cincinnati, Ohio, United States
6 Cleveland Clinic, Cleveland, Ohio, United States
7 Cleveland Clinic, Cleveland, Ohio, United States
8 Cleveland Clinic, Cleveland, Ohio, United States
9 Cleveland Clinic, Cleveland, Ohio, United States
ABSTRACT
Background
Cytokine release syndrome (CRS) and Immune effector cell-associated neurotoxicity syndrome (ICANS) can be predicted using Endothelial Activation and Stress Index (EASIX) scores, which include routine laboratory parameters and inflammatory markers. Existing studies predicting CRS/ICANS in multiple myeloma (MM) have typically considered assessments at a single time point (e.g., pre-apheresis or pre-infusion). However, the treatment path from apheresis to CAR-T infusion is complicated and often involves bridging therapy (BT), which can alter disease burden and inflammatory markers, highlighting the need for a dynamic risk stratification approach.
Methods
We included patients with MM treated with standard-of-care idecabtagene vicleucel (ide-cel) or ciltacabtagene autoleucel (cilta-cel) at our institution from Aug 1, 2021, and Jan 31, 2024. We calculated EASIX [lactate dehydrogenase (LDH) (U/L) x creatinine (Cr) (mg/dL) / platelet count (109/L)], modified EASIX [mEASIX: LDH x CRP / platelet count], simplified EASIX [sEASIX: LDH / platelet count], and CAR-HEMATOTOX (CAR-HT, ≥2 considered high) serially at six peri-infusion timepoints: apheresis, LD, pre-infusion (day 0), and days +1, +2, +3 post-infusion.
We used logistic regression to assess relations between respective scores at each timepoint and occurrence of CRS/ICANS. To reduce skewness, we log2 transformed continuous EASIX scores (ORs thus represent changes associated with doubling of the score). Next, we evaluated serial log2 EASIX/mEASIX/sEASIX across the six peri-infusion timepoints and performed exploratory trajectory clustering to identify distinct peri-infusion patterns.
Results
A total of 103 patients (58 cilta-cel, 45 ide-cel) with a median of 5 prior lines (range 4-14) were included. Ninety percent had triple-class refractory disease; 22% had extramedullary disease prior to infusion; 62% received BT (ORR 17% with BT). Overall, 76% patients developed any grade CRS (27% ≥ grade 2, 7% ≥ grade 3) and 25% developed any grade ICANS (9% ≥ grade 2, 5% ≥ grade 3).
All scores showed limited discrimination for any grade CRS at all peri-infusion timepoints. In contrast, several scores demonstrated a strong association with ≥ grade 2 CRS and ICANS. Pre-apheresis EASIX (OR 1.4 per doubling of score; 95% CI 1.1-1.9; AUC 0.68) and sEASIX (OR 1.5; 95% CI 1.1-2.1; AUC 0.69) were associated with ≥ grade 2 CRS, with consistent associations through later timepoints (EASIX AUCs 0.71-0.73 from pre-LD through day +3; sEASIX AUCs 0.71-0.73). Pre-apheresis EASIX (OR 1.3; 95% CI 1.04-1.7; AUC 0.68) and sEASIX (OR 1.4; 95% CI 1.04-1.9; AUC 0.64) also predicted any-grade ICANS. CAR-HT showed association with ICANS at later timepoints (e.g., day +3, high CAR-HT OR 4.4; 95% CI 1.4-13.8; AUC 0.64).
We then assessed serial EASIX trajectories among patients with complete data at all timepoints (n=96). Median log2(EASIX) increased modestly from 0.12 at apheresis to 0.69 pre-infusion. Values then remained relatively stable, ranging from 0.60 to 0.66 across days +1 to +3. Stratification by response to BT did not reveal major differences in EASIX trajectories.
We identified three trajectory groups that were already separated at apheresis and remained distinct across the peri-infusion period: persistent low EASIX (n=56; median log2[EASIX] -0.45; IQR -0.76 to -0.08), intermediate EASIX (n=30; median log2[EASIX] 1.01; IQR 0.44-1.60), and persistent high EASIX (n=10; median log2[EASIX] 4.11; IQR 1.18-4.67). Parallel exploratory analyses using mEASIX and sEASIX showed similar patterns. Any-grade CRS rates were similar across persistent low, intermediate, and persistent high EASIX groups (73%, 73%, and 80%). Clinically meaningful toxicity increased stepwise across groups, including grade ≥2 CRS (16%, 30%, 60%), any-grade ICANS (9%, 37%, 60%), and grade ≥2 ICANS (0%, 10%, 40%). The clusters also predicted PFS, OS. Median PFS (95% CI): NR/9.8 (8.2-NR)/1.0 (0.5-NR); OS: 26.7 (26.3-NR)/NR (13.8-NR)/1.8 (0.5-NR) months in low/intermediate/high clusters.
Conclusion
EASIX-based risk profiles were apparent as early as apheresis and were maintained through infusion and the early post-infusion period. These findings support early toxicity risk stratification beginning at apheresis, with serial assessments helping confirm the stability of these risk profiles across the peri-CAR-T treatment course. These results may help inform toxicity prophylaxis approaches and feasibility of outpatient administration.
DISCLOSURES
Utkarsh Goel, MD: No relevant conflicts of interest to declare.