A clinical risk scoring system for the prediction of early infection after CAR-T therapy.

Abstract

e23330 Background: CAR-T therapy has transformed the management of hematologic malignancies but is often complicated by severe infections, a leading cause of non-relapse mortality. Fever within the first 30 days post-infusion frequently results from either cytokine release syndrome (CRS) or infection, necessitating prompt differentiation and timely clinical intervention. Methods: In this study, we characterized early fever and infection events (days 0-30 after CAR-T infusion), found independent predictors and formulated a risk scoring system of infection in 535 patients undergoing CAR-T therapy from center No. 1 (internal cohort). Then we validated the risk scoring system in an external validation cohort (31 patients) from other research centers and a prospective validation cohort (31 patients) from center No.1. Results: A total of 503 fever episodes were documented among 443 patients in the internal cohort. Most patients (88%) experienced only a single fever episode, with 20% of these initial episodes attributed to infection and 80% to CRS. However, for subsequent fever episodes, infections became predominant, accounting for 82% to 100% of cases. Fever patterns in the external and prospective cohorts were consistent with findings from the internal cohort. Since first fever episodes often overlap with CRS, we utilized significance analysis and logistic regression to analyze infection characteristics, ultimately identifying six routine clinical indicators-neutropenia grades, lymphocyte counts, CRP levels, IL-6 levels, platelet counts, and changes in ferritin levels-as key predictors for constructing a predictive model. The sensitivity across different cohorts ranged from 79% to 100%, while the specificity ranged from 76% to 83%. Based on different levels of infection risk, we stratified the risk of fever events according to the likelihood of infection in predictive model and provided tailored anti-infection recommendations. The risk score has been incorporated into an online calculator accessible to the public (http://tongji.dcykjmb.com/). Conclusions: In this study, we developed a universal risk score system of infection based on patient characteristics at the onset of fever within 30 days following CAR-T infusion, aiming to rapidly provide the recommends for the clinical to treatment the fever during CAR-T therapy.