Differentiating Periodic Fever, Aphthous Stomatitis, Pharyngitis, Cervical Adenitis (PFAPA) and Syndrome of Undifferentiated Recurrent Fever (SURF): A Comparative Study of Clinical Features, Treatment Response, and Remission.
To apply proposed classification criteria for periodic fever, aphthous stomatitis, pharyngitis, cervical adenitis (PFAPA) and syndrome of undifferentiated recurrent fever (SURF) to a heterogeneous cohort of children with recurrent fever syndromes and to evaluate the differences in phenotype, treatment response, and disease outcomes. We conducted a retrospective cohort study of 235 children referred to the Autoinflammatory Clinic at The Hospital for Sick Children between 2016 and 2024. Patients were classified as PFAPA or SURF using validated Eurofever/Paediatric Rheumatology International Trials Organisation (PRINTO) classification criteria for PFAPA and empirical indications for SURF. Clinical features; response to corticosteroids, colchicine, and tonsillectomy; and time to remission were compared. Phenotypic clusters were identified using principal component analysis (PCA); log-rank test and Cox proportional hazards regression were used to assess predictors of remission. Of 235 patients, 155 (66%) met PFAPA criteria and 80 (34%) were classified as SURF. Patients with PFAPA more commonly exhibited classical symptoms (aphthous ulcers, pharyngitis, cervical lymphadenopathy), whereas those with SURF were characterized by gastrointestinal and systemic features (abdominal pain, arthralgia, fatigue). Data-driven clustering identified 2 predominant patterns, underscoring heterogeneity: a classic PFAPA cluster and a gastrointestinal-dominant cluster. Neither corticosteroid nor colchicine response was associated with phenotype or remission. Time to remission was shorter in PFAPA than in SURF (median 4.8 vs 5.7 years; = 0.04). Cox regression confirmed SURF classification was an independent predictor of delayed remission (hazard ratio 0.68; = 0.04). Structured classification distinguishes PFAPA and SURF by phenotype and disease trajectory. Data-driven clustering revealed 3 overlapping phenotypic patterns, supporting a continuum of autoinflammatory expression and emphasizing the need for individualized diagnostic and therapeutic approaches.