Value of 18F-FDG PET/CT in the differential diagnosis of AOSD and DLBCL

Introduction: Adult onset still’s disease (AOSD) and diffuse large B-cell lymphoma (DLBCL) are respectively the most common benign and malignant causes of fever of unknown origin (FUO), respectively. Overlapping clinical symptoms and medical imaging findings make their differential diagnosis difficult. 18F-FDG PET/CT has been applied to the systematic evaluation of AOSD and DLBCL, but its ability to distinguish between them is unclear. Our study aimed to screen out 18F-FDG PET/CT image features that effectively discriminated these two diseases, and to establish a feasible scoring model based on PET/CT for the differential diagnosis.

Methods: A development cohort including 70 AOSD patients and 101 DLBCL patients was used to establish a scoring model based on PET/CT images for the differential diagnosis. The scoring model was then validated in a validation cohort (14 AOSD and 28 DLBCL patients). AOSD patients were diagnosed according to Yamaguchi’s criteria, while DLBCL patients were pathologically diagnosed. The Image features of involved bone marrow, spleen, liver, lymph nodes and other organs or tissues were compared between AOSD and DLBCL patients. The diagnosis performance of scoring model was finally evaluated using receiver operating characteristic (ROC) analysis.

Results: Five PET/CT image features that were significant indicators for discriminating these two diseases were selected to establish a 16-point scoring model, including (1) SUVmax of bone marrow≥4.1 (3 points); (2) SUVmax of spleen≥3.4 (1 points); (3) SUVmax of liver≤3.4 (2 points); (4) short axis diameter of lymph nodes≤10.8mm (7 points); (5) splenomegaly (3 points). The model showed a significant ability to discriminate between AOSD and DLBCL with area under curve (AUC) equal to 0.963 (95%CI: 0.922-0.986) in the development cohort, and 0.936 (95%CI:0.816-0.988) in the validation cohort. With the cut-off value of 8 points, the sensitivity, specificity, positive predict value, and negative predict value in the development cohort were respectively 88.6%, 80.2%, 79.5%, 87.8% (Figure 1A), and 78.6%, 75.0%, 61.1%, 87.5% in the validation cohort (Figure 1B).

Conclusions: The PET/CT-based scoring model showed good performance for distinguishing AOSD and DLBCL, and could contribute to the etiological diagnosis of FUO.