Confirming the Validity of the New EULAR/ACR Classification Criteria for Pediatric Chronic Nonbacterial Osteomyelitis

Abstract

We aimed to investigate the performance characteristics of the new European Alliance of Associations for Rheumatology (EULAR) /American College of Rheumatology (ACR) classification criteria for pediatric CNO, in a distinct real-world clinical setting and compare it to the performance characteristics of the existing Jansson and Bristol diagnostic criteria. Single-center cross-sectional study including children ≤18 years, diagnosed with CNO, acute infectious osteomyelitis (AOM) and bone malignancy from June 1, 2018, to May 31, 2024. Patients with other mimicking conditions, immunodeficiency, sickle cell disease, or those previously included in the original development cohort were excluded. Patients were divided into 2 groups: a CNO group and a non-CNO group (AOM and bone malignancy), from the latter a representative, random, sample was selected ( www.random.org ). Demographic, clinical, laboratory, imaging and pathology data were collected at disease onset. EULAR/ACR criteria were retrospectively applied to the entire cohort, independently from the initial diagnosis. Patients with an aggregate score ≥ 55 points were classified as having CNO. A secondary analysis was conducted excluding patients with missing data. Classification results using the new criteria were compared with the final clinical diagnosis based on physician assessment (criterion standard). The same analysis was applied to Jansson and Bristol criteria. Sensitivity and specificity, positive likelihood ratio (LR+) and post-test probability (PTP) (pre-test probability: 20%, based on CNO prevalence in our center) were calculated. Of 164 children included, 82 had CNO, 41 were randomly selected from 274 cases of AOM, and 41 from 849 cases of bone malignancy. The median age was 10 years (IQR 3-16), with 62% girls and 37% boys, 33% had a bone biopsy. Overall, 40% scored ≥ 55 and 60% did not, with 19 false positive and 3 false negative. The EULAR/ACR criteria demonstrated 77% sensitivity and 96% specificity, with a LR+ of 19.25 and PTP of 83%. In our secondary analysis excluding patients with incomplete data, results remained consistent (sensitivity 79%, specificity 96%, LR+ 19.75, PTP 83.2%). In comparison, the Jansson criteria showed 78% sensitivity, 67% specificity, LR+ of 2.36, and PTP of 37%. The Bristol criteria yielded 89% sensitivity, 70% specificity, LR+ of 2.97, and PTP of 43%. Based on its favorable sensitivity and specificity, especially in comparison to existing criteria, the new EULAR/ACR criteria appeared to be more effective in distinguishing CNO from AOM and bone malignancy at disease onset, results consistent with findings from the original validation cohort. Best Abstract by a Rheumatology Post- Graduate Research Trainee Award