Abstract Background:Glofitamab is a CD20xCD3 T-cell engaging bispecific antibody with a novel 2:1 (CD20:CD3) format, which redirects T cells to eliminate B cells, including those that cause malignant disease. In the Phase III STARGLO trial (NCT04408638), glofitamab in combination with gemcitabine-oxaliplatin (Glofit-GemOx) demonstrated a significant overall survival benefit compared with rituximab-GemOx in patients with relapsed or refractory (R/R) diffuse large B-cell lymphoma (DLBCL; Abramson et al. Lancet 2024). While cytokine release syndrome (CRS) can be observed with glofitamab, events typically occur with Grade 1/2 severity per American Society for Transplantation and Cellular Therapy grading, and are mostly confined to the first treatment cycle. Komanduri et al. (ASH 2021) reported a predictive model based on 8 baseline factors (CRS-risk score CRS-RS) that allowed accurate classification of patients at risk of Grade ≥2 CRS after the first glofitamab dose; a streamlined version using 5 baseline factors (CRS-RS.5p) was subsequently derived from the original CRS-RS (Gritti et al. Blood Advances 2024). Here, we assess the performance of CRS-RS.5p in glofitamab-exposed patients with R/R DLBCL from the Glofit-GemOx arm of the STARGLO trial, including subgroup analyses in those who had received ≥2 prior lines of therapy (third-line plus 3L+). Methods:The previously described CRS-RS.5p model, with the same weights and cutoffs, was applied. The multivariate CRS-RS.5p model included the first glofitamab dose and a weighted sum of 5 baseline factors: age 64 years, lactate dehydrogenase (LDH) 280 units/L, white blood cells 4.5 x 109 cells/L, Ann Arbor stage III/IV, and sum of the product of the perpendicular diameters ≥3000mm2. Weights reflected the predictive strength of risk factors in a dose-adjusted logistic regression model and the relative stability of risk parameters in a multivariate model resulting from random forest analyses. Patients were classified as low-risk (CRS-RS.5p4) or high-risk (CRS-RS.5p≥4) for CRS following the first dose of glofitamab (2.5mg). To control the rate of false negative predictions, missing data were input at the highest levels for each respective parameter. After patients were identified as low-risk for CRS, negative predictive value (NPV) was used to demonstrate the likelihood that a patient classified as low-risk would not develop Grade ≥2 CRS after the first glofitamab dose. CRS risk predictions were performed at baseline (pre-obinutuzumab) in glofitamab-treated patients from the STARGLO trial. Data have been collected prospectively since February 2021. Results:Imputation of missing values (baseline LDH) was required for 2/172 cases.Of the 172 patients from the Glofit-GemOx treatment arm who received at least one dose of glofitamab (2.5mg), CRS occurred in 35.5% (61/172; Grade 1, 70.5%; Grade 2, 24.6%; Grade 3, 4.9%) within 168 hours of the first glofitamab dose. In the 3L+ setting, CRS events were reported in 32.8% (21/64; Grade 1, 66.7%; Grade 2, 28.6%; Grade 3, 4.8%) of patients.In the overall population, the majority (77.0%) of CRS events occurred within 24 hours of the first glofitamab dose. CRS-RS.5p accurately predicted Grade ≥2 CRS in 12 (CRS-RS.5p≥4) out of 18 (Grade ≥2 CRS) cases with a true positive prediction rate of 0.67 (95% confidence interval CI: 0.41–0.87). In the 3L+ setting, CRS-RS.5p accurately predicted Grade ≥2 CRS in 6 (CRS-RS.5p≥4) out of 7 (Grade ≥2 CRS) cases with a true positive prediction rate of 0.86 (95% CI: 0.42–1.00).In the overall population, the majority (104/172) of those predicted to be low-risk did not develop Grade ≥2 CRS after the first glofitamab dose (98/104; NPV=0.94; 95% CI: 0.88–0.99); similar results were observed in the 3L+ setting (38/39; NPV=0.97; 95% CI: 0.87–1.00). Conclusions: Using data from glofitamab-exposed patients in STARGLO, theCRS-RS.5p 5-parameter model accurately ruled out patients with R/R DLBCL who were at low-risk of Grade ≥2 CRS. Results were consistent between the overall population and patients who had received ≥2 prior lines of therapy, and with previously reporting findings. These data further emphasize the potential application of this model for accurate risk stratification and as a clinical decision tool to guide the need for additional monitoring for patients at high risk of Grade ≥2 CRS.
Abdulhaq et al. (Mon,) studied this question.