Nonlinear association between fever-to-IVIG interval and clinical outcomes in Kawasaki disease: a retrospective cohort study of 6,921 children

The American Heart Association recommends intravenous immunoglobulin (IVIG) for Kawasaki disease (KD) within 10 days of fever onset, ideally within 7 days, but the optimal timing remains unclear. Prior studies suggested that very early treatment may increase IVIG resistance, but were limited by small samples. We retrospectively analyzed 6,921 KD patients at a single center (2012–2025), stratified by fever-to-IVIG interval into ≤ 4 days ( n = 998), 5–7 days (reference, n = 4,483), 8–10 days ( n = 1,025), and ≥ 11 days ( n = 415). Outcomes were IVIG resistance and coronary artery dilation (Z-score > 2.0). Nonlinearity was assessed by restricted cubic spline logistic regression (primary) and quadratic models (sensitivity), adjusted for age, sex, incomplete KD, high-sensitivity C-reactive protein (hs-CRP), albumin, platelets, neutrophils, and hemoglobin. IVIG resistance showed a U-shaped association with fever-to-IVIG interval, confirmed in both spline and quadratic models (restricted cubic spline nonlinearity likelihood ratio test χ²=117.51, df = 2, P < 0.001; curve minimum at day 7.4). Compared with the 5–7-day reference (9.2%), resistance was higher in the ≤ 4-day group (18.6%; crude odds ratio [OR] 2.25, 95% confidence interval 1.86–2.72) and the ≥ 11-day group (16.4%; OR 1.93, 1.46–2.55; both P < 0.001). Coronary artery dilation rose progressively with delay (≤ 4d: 13.6%, OR 1.06, P = 0.62; 8–10d: 20.6%, OR 1.74; ≥11d: 28.0%, OR 2.60; both P < 0.001). The composite adverse outcome was lowest at 5–7 days (20.0%), and patterns were consistent across age, hs-CRP quartile, KD subtype, and coronavirus disease 2019 era subgroups. The progressive coronary risk with delay was confirmed using a stricter aneurysm-level Z-score threshold (≥ 2.5) and was stable across two calendar eras (2012–2018 and 2019–2025). Fever-to-IVIG interval showed a U-shaped association with IVIG resistance, whereas coronary dilation risk rose with delay; the composite outcome was lowest at 5–7 days. These findings suggest the timing–outcome relationship is more complex than a linear earlier-is-better assumption, but there is insufficient evidence to recommend intentional IVIG delay once KD is diagnosed.