CAR-T therapy has transformed the treatment of relapsed/refractory B-cell lymphoma, but many patients still relapse or experience potentially severe toxicities. Identifying high-risk patients before treatment could enable more personalized management.

This multicentre real-world study investigated whether quantitative [¹⁸F]FDG PET/CT imaging features combined with clinical data could predict treatment response, survival and immune effector cell-associated neurotoxicity syndrome (ICANS) in 65 patients receiving CAR-T therapy.

Combined models predicted response at 3 and 6 months with AUCs of 0.754 and 0.818, respectively, with the 6-month model reaching 95.5% sensitivity.

Importantly, whole-body metabolic tumor volume (MTVtotal) emerged as a powerful prognostic marker. A 236-mL cut-off stratified patients into markedly different risk groups: median overall survival was 4.73 vs 37.55 months, while progression-free survival was 2.73 vs 11.83 months for high versus low MTVtotal.

The study also reports, to the authors’ knowledge, the first model combining imaging and clinical variables to predict CAR-T-associated neurotoxicity, achieving an AUC of 0.830.

These findings position quantitative PET imaging as a promising non-invasive tool for CAR-T risk stratification and personalized treatment planning.