Kawasaki Disease Treatment Timing: Study of 6,921 Children Reveals a Surprising U-Shaped Risk

Kawasaki disease is the leading cause of acquired heart disease in children in many parts of the world, and for decades clinicians have operated on a simple assumption: the sooner the standard treatment is given after fever begins, the better the outcome. A new retrospective study of 6,921 children, published in BMC Medicine, challenges that linear logic in a striking way. The analysis, conducted by Changjian Li and Yong Zhang of the Department of Cardiology at Wuhan Children’s Hospital, Tongji Medical College, Huazhong University of Science and Technology, found that the relationship between the timing of intravenous immunoglobulin therapy and one key outcome, resistance to that very therapy, is not a straight line at all. Instead, it follows a U-shaped curve, with risk rising both when treatment is given very early and when it is given late. The finding, drawn from one of the largest single-center cohorts ever assembled for this question, is likely to reignite debate about the optimal window for a treatment that has remained essentially unchanged since the 1980s.

Kawasaki disease is an acute, systemic vasculitis that primarily strikes children under five years of age. Its cause remains unknown, although epidemiological patterns, including seasonal clustering and age distribution, have long suggested an infectious or environmental trigger in genetically susceptible children. The disease’s clinical signature is a persistent fever accompanied by features such as rash, conjunctival injection, redness and swelling of the hands and feet, mucous membrane changes, and cervical lymphadenopathy. The danger lies not in the acute symptoms themselves but in the damage the inflamed vasculature can sustain, particularly the coronary arteries. Without treatment, a substantial fraction of affected children develop coronary artery aneurysms, which can lead to thrombosis, stenosis, ischemia, and long-term cardiac complications. Since the introduction of intravenous immunoglobulin, a pooled antibody preparation derived from thousands of plasma donors, the rate of coronary abnormalities has fallen dramatically, which is why prompt treatment has become the cornerstone of management worldwide.

Current guidance from the American Heart Association recommends administering intravenous immunoglobulin within ten days of fever onset, and ideally within seven days. That recommendation rests on decades of observational evidence showing that treatment given after the tenth day of fever is associated with higher rates of coronary artery abnormalities. But within that ten-day window, the guidance has been less precise, and a nagging question has persisted in the literature: does treating at the very earliest signs of fever, before the full inflammatory syndrome declares itself, actually help? Some smaller studies had suggested that children treated within the first four days of fever were paradoxically more likely to resist the first dose of immunoglobulin, requiring retreatment with additional doses or second-line agents such as corticosteroids. Those studies, however, were limited by small sample sizes that made it difficult to separate a genuine biological effect from statistical noise and confounding.

The Wuhan team addressed the question with unusual statistical rigor. They assembled a cohort of 6,921 children with Kawasaki disease treated at their center between 2012 and 2025, extracting data from an integrated electronic health record system encompassing medical records, diagnoses, physician orders, procedures, laboratory tests, imaging, and nursing documentation. Patients were stratified by the interval between fever onset and the first immunoglobulin infusion into four groups: four days or fewer (998 children), five to seven days (4,483 children, serving as the reference), eight to ten days (1,025 children), and eleven days or more (415 children). Two primary outcomes were examined: resistance to intravenous immunoglobulin, defined as persistent or recrudescent fever requiring additional therapy, and coronary artery dilation, defined by a Z-score greater than 2.0 on echocardiographic measurement, a statistical expression of how far an artery’s diameter deviates from normal values adjusted for body size.

The central analytical innovation was the use of restricted cubic spline logistic regression, a technique that allows the relationship between a continuous exposure, here the fever-to-treatment interval, and an outcome to take any shape rather than being forced into a straight line. The spline model was complemented by a quadratic model as a sensitivity analysis, and both were adjusted for a battery of potential confounders including age, sex, incomplete Kawasaki disease presentation, high-sensitivity C-reactive protein, albumin, platelet count, neutrophil count, and hemoglobin. The results were unambiguous. The likelihood ratio test for nonlinearity in the spline model yielded a chi-square statistic of 117.51 with two degrees of freedom and a P value below 0.001, and the fitted curve reached its minimum at day 7.4 of fever, almost exactly the boundary of the guideline’s ideal window.

The pattern across the strata told a consistent story. In the five-to-seven-day reference group, 9.2 percent of children were resistant to immunoglobulin. In the group treated at four days or earlier, resistance jumped to 18.6 percent, with a crude odds ratio of 2.25 and a 95 percent confidence interval of 1.86 to 2.72. The late-treated group fared nearly as badly: among children who received immunoglobulin at eleven days or beyond, 16.4 percent were resistant, corresponding to an odds ratio of 1.93 with a confidence interval of 1.46 to 2.55. Both comparisons were highly statistically significant. In other words, children treated very early and children treated late faced roughly double the odds of treatment failure compared with those treated in the middle window, a symmetric pattern that is the statistical hallmark of a U-shaped dose-response relationship.

Coronary artery outcomes, however, told a different and more sobering story. Unlike immunoglobulin resistance, the risk of coronary artery dilation rose progressively and monotonically with treatment delay. In the earliest treatment group, 13.6 percent of children showed coronary dilation, an odds ratio of 1.06 compared with the reference group and a P value of 0.62, meaning no meaningful difference. In the eight-to-ten-day group, the rate climbed to 20.6 percent, with an odds ratio of 1.74, and in the latest group it reached 28.0 percent, with an odds ratio of 2.60, both highly significant. The investigators confirmed this progressive pattern using a stricter, aneurysm-level Z-score threshold of 2.5 or greater, and demonstrated that it was stable across two calendar eras, 2012 to 2018 and 2019 to 2025, as well as across subgroups defined by age, C-reactive protein quartile, complete versus incomplete disease, and whether the illness occurred before or during the coronavirus disease 2019 pandemic.

When the two outcomes were combined into a composite adverse endpoint, the five-to-seven-day window emerged as the clear optimum, with the lowest composite rate at 20.0 percent. The divergence between the two outcome curves is scientifically intriguing. It suggests that very early treatment, while apparently associated with a higher chance that the first immunoglobulin dose will fail, does not translate into more coronary damage, presumably because the arteries have not yet been exposed to many days of unopposed inflammation and rescue therapy remains available. Late treatment, by contrast, is doubly penalized: the coronary arteries endure a longer inflammatory assault, and the treatment itself is more likely to fail. The biological explanation for early resistance remains speculative. One hypothesis holds that immunoglobulin works most effectively against a specific phase of the inflammatory cascade, and that intervening before that phase peaks leaves the dominant pathological drivers untouched. Another suggests that children who present and are treated within the first few days may represent a phenotypically distinct subgroup with inherently more intense inflammation, although the adjustment for C-reactive protein and other markers was intended to account for precisely this kind of confounding.

The authors are careful about what their findings do and do not justify. Their conclusion states explicitly that while the timing-outcome relationship is more complex than a linear earlier-is-better assumption, there is insufficient evidence to recommend intentional delay of immunoglobulin once Kawasaki disease is diagnosed. That caveat matters enormously in clinical practice. Coronary risk rises with every day of delay, so the U-shaped resistance curve cannot be read as license to wait. What the study does support is a reframing of the ideal target: rather than treating at the first possible moment, clinicians should aim for the five-to-seven-day window, which the data identify as the sweet spot balancing treatment efficacy against coronary protection. For children presenting on day two or three of fever, the findings raise, without resolving, the question of whether a brief period of observation while the diagnosis crystallizes might be preferable to immediate empiric treatment, a decision that must weigh diagnostic uncertainty against the monotonic coronary risk of delay.

The study’s limitations are those inherent to its design. It was retrospective and conducted at a single center, which means treatment protocols, echocardiographic practices, and referral patterns were internally consistent but may not generalize to other populations. The observational nature of the data means that the timing of treatment was not randomized; physicians may have treated sicker-appearing children earlier or later for reasons correlated with outcomes in ways the statistical adjustments could not fully capture. The definition of fever onset also depends on parental recall, introducing potential measurement error. Nevertheless, the sheer size of the cohort, the consistency of the nonlinear signal across two independent statistical models, and the stability of the findings across calendar eras and clinical subgroups lend considerable weight to the central conclusion. For a disease in which the fundamental treatment strategy has been static for four decades, the demonstration that timing effects are shaped like a valley rather than a slope is a genuinely consequential contribution, and it sets the stage for the prospective, ideally randomized, studies that will be needed before any change to the ten-day rule can be contemplated.

Subject of Research: Optimal timing of intravenous immunoglobulin treatment in Kawasaki disease

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

Article References: Nonlinear association between fever-to-IVIG interval and clinical outcomes in Kawasaki disease: a retrospective cohort study of 6,921 children. (n.d.). https://doi.org/10.1186/s12916-026-05222-y

Image Credits: AI Generated

DOI: 10.1186/s12916-026-05222-y

Keywords: Kawasaki disease, intravenous immunoglobulin, IVIG resistance, coronary artery dilation, treatment timing, fever-to-IVIG interval, nonlinear association, restricted cubic spline, retrospective cohort, pediatric cardiology, vasculitis, BMC Medicine