Prognostic variables associated with the development of cytokine release syndrome after bispecific antibodies in non-Hodgkin lymphoma patients – the geltamo (Spanish Lymphoma Group) experience

Abstract

Abstract Introduction Cytokine release syndrome (CRS) is a common toxicity observed during the step-up dosing of bispecific antibody (BsAb) treatment in patients (pts) diagnosed with B-cell non-Hodgkin lymphoma (B-NHL). There is heterogeneous practice regarding hospitalization during the first treatment cycle across centers. Predictive models for CRS based on readily-available parameters could significantly aid decision-making in standard clinical care. The aim of our study was to analyze patient- and disease-related factors in a real-world cohort of B-NHL pts receiving BsAbs and identify those associated with the development of CRS, building a score to optimize hospital resource allocation. Methods We conducted a retrospective, multicenter analysis of pts diagnosed with relapsed/refractory B-NHL and treated with standard-of-care BsAbs in the GELTAMO (Spanish Lymphoma Group) network. Grading of CRS followed ASTCT criteria and management of this adverse event was carried out according to label recommendations and local guidelines. The two main endpoints assessed were the occurrence of CRS (yes vs. no) and the severity of CRS (grade 1 vs. grade ≥2). LASSO logistic regression with minimum lambda was used for variable selection, and the selected factors were then included in a multivariable logistic regression. Model performance was assessed based on discrimination, using the area under the receiver operating characteristic curve (AUC), and calibration, using calibration plots. Internal validation was performed using bootstrap resampling to correct for optimism. Results The study included 164 pts, 93 (57%) with diffuse large B-cell lymphoma (DLBCL) and 71 (43%) with follicular lymphoma (FL). Regarding BsAbs, epcoritamab was administered in 84 pts (51%), mosunetuzumab in 58 (35%) and glofitamab in 22 (13%). In the overall cohort, median age was 64 years (range 52-77), most pts were male (59%), had an ECOG 0-1 (78%) and an advanced stage of disease (80%). 75 pts (46%) developed CRS after BsAbs treatment; 39 pts (52%) had grade 1, 28 (38%) grade 2, 1 (1%) had grade 3 and 1 (1%) grade 4. The CRS grade was not available in 6 pts (8%). Focusing on variables associated with an increased risk of CRS vs. not, we identified a higher rate of pts with IPI/FLIPI score of 4-5 (44% vs 21%, p=0.002). In the same direction, pts with an IPI/FLIPI score of 4-5 had a significantly higher risk of developing grade ≥2 CRS compared to those with a score of 0-3 (63% vs. 26%, p<0.001). The number of extranodal sites involved prior to treatment was significantly associated with the risk of developing a grade ≥2 CRS vs grade 0-1 (57% vs 34%, p=0.024). Prior CAR T-cell therapy was not associated with an increased risk of developing any-grade CRS after BsAbs. Neither lymphoma histology nor the type of BsAbs administered were associated with the development of any-grade of CRS. To predict any grade CRS, variables with the greatest weight in the model were IPI/FLIPI score, absolute lymphocyte count (ALC), lactate dehydrogenase (LDH) and C-reactive protein (CRP) levels, all assessed at the time of BsAb treatment, with an AUC of 0.679 and a corrected AUC of 0.661. Variables included in the score with the greatest weight to identify pts who developed grade ≥2 CRS were >1 extranodal site prior to BsAb treatment and an IPI/FLIPI score of 4-5, with an AUC of 0.7132 and a corrected AUC of 0.702 with an accuracy of 0.80, a precision of 0.88 and a recall of 0.88; also, the calibration plot showed strong correlation between predicted and observed probabilities. These findings indicate that the proposed score performs better at identifying pts at risk of grade ≥2 CRS rather than any-grade CRS. Conclusions Prognostic variables including extranodal involvement and the IPI/FLIPI score could be informative of the risk of developing grade ≥2 CRS. This model was able to identify pts with a high risk of developing ≥2 CRS; however, it was more limited for any-grade CRS. By effectively distinguishing between grade 1 and ≥2 CRS, this prognostic score may serve in the future as a tool to guide decisions regarding which pts would benefit most from inpatient monitoring during the step-up dosing of BsAb therapy.