Shared bone marrow-spleen-lymph node hypermetabolic phenotype on 18F-FDG PET/CT in AOSD and undiagnosed FUO is...

To characterize the bone marrow-spleen-lymph node hypermetabolic phenotype (BSL-HMP) identified by 18F-FDG PET/CT in patients with fever of unknown origin (FUO) and to explore whether imaging-defined inflammatory burden was associated with subsequent treatment intensity.

This single-center retrospective cohort study included FUO patients who attended the Department of Rheumatology and Immunology at the Third Affiliated Hospital of Soochow University between January 2020 and January 2025, underwent 18F-FDG PET/CT, and exhibited BSL-HMP. After systematic exclusion of known diseases, patients were divided into an AOSD group and a FUO group that did not meet the Yamaguchi criteria at baseline, and differences between the two groups were compared. Exploratory unsupervised phenotyping of PET/CT features was performed using weighted Gower distance combined with PAM clustering. A mixed-effects model for repeated measures (MMRM) was further applied to analyze longitudinal changes in glucocorticoid dose across different clusters. An exploratory PET/CT-based model was additionally developed and internally evaluated to characterize cluster membership.

A total of 95 patients were included, comprising 40 patients with AOSD and 55 patients with FUO. Although the FUO group showed PET/CT features broadly similar to those of the AOSD group, they had overall lower inflammatory marker levels and less frequent rash and sore throat. During follow-up, 8 patients in the FUO group eventually received an alternative definite diagnosis, while 47 remained undiagnosed. Unsupervised clustering classified the patients into high- and low-inflammatory-burden groups, with 80% of AOSD patients and 42% of FUO patients assigned to the high-inflammatory-burden group. Glucocorticoid dose trajectories differed significantly between clusters in the unadjusted MMRM (group-by-time interaction, P = 0.007) and after adjustment for age, sex, baseline diagnosis, C-reactive protein, and ferritin levels (P = 0.005). The difference remained significant after additional adjustment for baseline glucocorticoid dose (P = 0.029). DMARD use was also more frequent in the high-burden cluster (P = 0.011).

BSL-HMP is a shared, nonspecific systemic inflammatory imaging phenotype. Among patients exhibiting this phenotype, PET/CT-based phenotyping identified different inflammatory burdens associated with subsequent glucocorticoid dose trajectories and DMARD use. These exploratory and hypothesis-generating findings suggest that PET/CT-derived inflammatory burden may help characterize patient heterogeneity, although further validation in independent prospective cohorts is warranted.