Background
Clinicians need simple, clinic-ready rules to sort children with allergic rhinitis (AR) by near-term symptom burden. We examined whether peripheral blood eosinophil percentage (Eos %) yields useable thresholds for first-pass triage.
Methods
We performed a cross-sectional secondary analysis of a public pediatric dataset, including children with AR and non-missing Eos % and Total Nasal Symptom Score (TNSS). “High burden” was prespecified as TNSS ≥21 (cohort upper quartile). We estimated ROC/AUC, derived the Youden cut-off with 1000-bootstrap CIs, and identified targeted operating points emphasizing high sensitivity (≥0.80) or high specificity (≥0.80). Clinical utility was assessed using decision curve analysis (DCA) for threshold probabilities 0.15–0.30. Sensitivity analyses varied the TNSS cut-point and stratified by sex.
Results
Among 199 children, 26.6 % met the high-burden definition. Eos % discrimination was modest (AUC 0.606; 95 % CI 0.518–0.699). The Youden cut-off was 6.8 % (95 % CI 2.8–9.1), yielding sensitivity 0.53 and specificity 0.68 (accuracy 0.64; PPV 0.37; NPV 0.80). A ∼3.7 % rule achieved sensitivity 0.81/specificity 0.33, whereas an ∼8.5 % rule achieved specificity 0.81/sensitivity 0.32. DCA supported fixed rules in the 6–8 % band compared with treat-all/none for pt = 0.15–0.30. Findings were directionally consistent using TNSS ≥20 or ≥22 and across sex strata (boys: AUC 0.569, optimal 6.8 %; girls: AUC 0.691, optimal 3.1 %).
Conclusions
Eosinophil percentage (Eos %) exhibited modest discriminatory ability (AUC 0.606; 95 % CI 0.518–0.699) for triage in pediatric allergic rhinitis. Low thresholds of ∼3–4 % for screening, 6–8 % for balanced triage (optimal ∼6.8 %), and ∼8–9 % for confirmation offered limited utility when combined with symptom assessments. Due to weak performance, CI proximity to 0.5, and false-positive risk (e.g., PPV 0.37 at optimal cutoff), Eos % is inadequate alone and serves only as a weak complementary signal; prospective validation is essential.
扫码关注我们
求助内容:
应助结果提醒方式:
