Cerebral ¹⁸F-FDG PET/CT Metabolism as Diagnostic Signature for Central Nervous System Toxicity After Immune...

Our aim was to investigate probable biomarkers specific to immune-related central nervous system toxicity (CNST) in cancer patients treated with immune checkpoint inhibitors (ICI) by analysis of ¹⁸F-FDG PET/CT images. Methods: Cancer patients receiving ICI treatment were enrolled in a multicenter observational study that analyzed regional metabolic changes before and during CNST onset from January 2020 to February 2022. In 1:1 propensity score-matched pairs, the regional SUV_mean of each bilateral brain lobe of CNST patients (CNST+) was compared with that of patients who had central nervous system infections (CNSIs) and patients without CNST or CNSI (CNST-). In a validation cohort, patients were recruited from February 2022 to July 2023 and followed up for 24 wk after the start of ICI. Early changes in regional SUV_mean at 5-6 wk after therapy initiation were evaluated for ability to predict later CNST onset. Results: Of 6,395 ICI-treated patients, 2,387 underwent prognostic ¹⁸F-FDG PET/CT and 125 of the scanned patients had CNST (median time from ICI treatment to onset, 9 wk; quartile range, 2-23 wk). Regional ¹⁸F-FDG PET/CT SUV_mean changes were higher in CNST+ than in CNST- patients (117 patient pairs) but were lower than in CNSI patients (50 pairs). Differentiating analysis reached an area under the curve (AUC) of 0.83 (95% CI, 0.78-0.88) for CNST+ versus CNST- and of 0.80 (95% CI, 0.72-0.89) for CNST+ versus CNSI. Changes in SUV_mean were also higher before CNST onset than for CNST- (60 pairs; AUC, 0.74; 95% CI, 0.66-0.83). In a validation cohort of 2,878 patients, preonset changes in SUV_mean reached an AUC of 0.86 (95% CI, 0.79-0.94) in predicting later CNST incidence. Conclusion: Brain regional hypermetabolism could be detected during and before CNST clinical onset. CNST may be a distinct pathologic entity versus brain infections defined by ¹⁸F-FDG PET/CT brain scans. Regional SUV differences may be translated into early diagnostic tools based on moderate differentiating accuracy in our study.