Objective: There is evidence that adding steroids to the Kawasaki disease (KD) therapy reduce
coronary artery aneurysms (CAA). This benefit is particularly seen when starting steroids
early in the course of disease - at best supplementary to the first dose of i.v. immunoglobulins
(IVIG). Therefore, the recent American Heart Associations´ (AHA) statement on KD from
2017 recommend supplementary steroids for primary treatment, but only in patients
at risk for being refractory to IVIG. Different scores (Kobayashi, Egami and Sano)
are predictive in identifying these “high-risk” patients in the Japanese KD population.
The challenge is however to identify high-risk KD patients in Caucasian children.
Do the Japanese scores or application of up-to-date statistical methods allow to predict
response to standard IVIG therapy and to determine the risk of persistent CAA in patients
with KD in a mainly Caucasian population in Germany?
Methods: Data on 442 children with KD (German population-based survey in 2013 and 2014) were
used to evaluate the Japanese risk scores (Kobayashi, Egami and Sano). The prognostic
validity for being refractory to IVIG treatment and for predicting the risk of CAA
at 4 weeks and one year after the acute disease was assed. Additionally, an up-to-date
statistical approach (Random Forest) was applied to identify a potentially more valid
score.
Results: 301 children were eligible for assessment of their response to IVIG treatment. Among
those, 177 children with known identity, could be followed-up for one year to identify
persistent CAA.
Although all scores were significantly associated with being refractory to IVIG (relative
risk range between 2.32 and 3.73), the prognostic properties were low (likelihood
ratio positive: 1.83 - 4.57; sensitivity in the range of 0.28 to 0.53). Only the Sano
score was significantly associated with having a CAA at four weeks (relative risk
2.5), but with low sensitivity (0.23). None of the scores was a significant predictor
of CAA one year after acute illness. Unfortunately, application of additional up-to-date
statistical analysis such as Random Forest did not yield a more valid score.
Conclusion: None of the available Japanese risk scores, nor our Random Forest approach appears
to be appropriate for identifying high-risk Caucasian children with KD and therefore
is not helpful for clinicians in charge for those children.
