Proposal of a new diagnostic algorithm for adult-onset Still's disease.

OBJECTIVE

This study was performed to develop a new diagnostic algorithm for adult-onset Still's disease (AOSD).

METHODS

We conducted a multicenter prospective nationwide case-control study in tertiary Internal Medicine, Rheumatology, and Infectious Diseases departments, to include successively patients with suspected AOSD based on the presence of two or more major criteria of Yamaguchi and/or Fautrel classifications. Patients were classified as AOSD or controls according to a predefined procedure. A receiving operating characteristic curve was used to determine the best cutoff value of the points-based score for disease classification. A diagnostic algorithm was developed to help the physician in the diagnostic approach.

RESULTS

A total of 160 patients were included, 80 patients with AOSD and 60 controls with different diagnoses. Twenty patients with incomplete data were excluded. In the multivariate analysis, 6 items remained independently associated with AOSD diagnosis: typical rash (OR: 24.01, 3 points), fever ≥ 39 °C (OR: 17.34, 3 points), pharyngitis (OR: 10.23, 2 points), arthritis (OR: 9.01, 2 points), NLR ≥ 4 (OR: 11.10, 2 points), and glycosylated ferritin ≤ 20% (OR: 1.59, 1 point). AOSD should be considered if the patient satisfies 7 points with a sensitivity of 92.5%, specificity of 93.3%, and accuracy of 92.8% (area under the curve (AUC): 0.97 [95% CI: 0.94-0.99]). The present points-based score was more accurate and sensitive than the Yamaguchi classification (78.8%, 92.5%, p = 0.01) and Fautrel classification (76.3%, 92.5%, p = 0.004). A typical rash associated with a points-based score ≥ 7 points leads to a very likely disease.

CONCLUSION

The proposed new algorithm could be a good diagnostic tool for adult-onset Still's disease in clinical practice and research. Key Points • A diagnostic algorithm was performed to help the physician in the diagnostic approach of AOSD. • The points-based score included in this algorithm had a high sensitivity and accuracy. • This diagnostic algorithm can be useful in the clinical research.