Article Text

Predicting IVIG resistance in UK Kawasaki disease
  1. Sarah Davies1,
  2. Natalina Sutton1,
  3. Sarah Blackstock1,
  4. Stuart Gormley1,2,
  5. Clive J Hoggart3,
  6. Michael Levin1,2,
  7. Jethro A Herberg1,2
  1. 1Department of Paediatric Infectious Diseases, St Mary's Hospital, Imperial College Healthcare NHS Trust, London, UK
  2. 2Section of Paediatrics, Imperial College London, London, UK
  3. 3Department of Genomics of Common Disease, Imperial College London, London, UK
  1. Correspondence to Dr Jethro A Herberg, Section of Paediatrics, Imperial College, Norfolk Place, London W2 1PG, UK; j.herberg{at}


The Kobayashi score (KS) predicts intravenous immunoglobulin (IVIG) resistance in Japanese children with Kawasaki disease (KD) and has been used to select patients for early corticosteroid treatment. We tested the ability of the KS to predict IVIG resistance and coronary artery abnormalities (CAA) in 78 children treated for KD in our UK centre. 19/59 children were IVIG non-responsive. This was not predicted by a high KS (11/19 IVIG non-responders, compared with 26/40 responders, had a score ≥4; p=0.77). CAA were not predicted by KS (12/20 children with CAA vs 25/39 with normal echo had a score ≥4; p=0.78). Low albumin and haemoglobin, and high C-reactive protein were significantly associated with CAA. The KS does not predict IVIG resistance or CAA in our population. This highlights the need for biomarkers to identify children at increased risk of CAA, and to select patients for anti-inflammatory treatment in addition to IVIG.

  • Kawasaki Disease
  • Kobayashi Score
  • IVIG-resistant
  • prediction

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

What is already known on this topic?

  • Early switch-off of the inflammatory response in Kawasaki disease, using anti-inflammatory medications, reduces the incidence of coronary artery abnormalities (CAA).

  • The Kobayashi score (derived from clinical and laboratory values) has been validated to predict treatment failure (resistance to intravenous immunoglobulin (IVIG)) in Japan.

  • Japanese children with a high Kobayashi score given early adjunctive steroid treatment alongside IVIG were less likely to develop CAA.

What this study adds?

  • The Kobayashi score did not predict IVIG resistance or coronary artery abnormalities (CAA) in a UK cohort.

  • Previously described biomarkers of severe Kawasaki disease (KD) (haemoglobin, albumin, C-reactive protein) were poor discriminators of IVIG resistance or CAA.

  • This study does not provide evidence to support the use of the Kobayashi score in assessment of UK children with KD; alternative biomarkers are needed.


Kawasaki disease (KD) is an acute inflammatory disorder of unknown origin associated with vasculitis affecting medium-sized vessels. Coronary artery abnormalities (CAA)—aneurysms or dilatations—are the most serious complication of KD, which is the commonest cause of acquired heart disease in children in developed countries. Up to 20% of children with KD fail to respond to the initial standard treatment with intravenous immunoglobulin (IVIG) and aspirin, and have persisting fever and inflammation (see Eleftheriou et al1 for a recent guideline including source references). The vasculitis leading to CAA begins early in the illness, leaving little time for additional therapies in IVIG non-responders before vascular damage occurs. CAA occur most frequently in IVIG non-responders,1 suggesting that there is a critical window for switching off the inflammatory process and preventing long-term coronary artery damage.

Early administration of steroids together with IVIG improved coronary artery outcomes in Japanese children predicted to be at risk of IVIG non-response,2 ,3 and meta-analysis of published trials suggests a reduction in CAA in patients receiving steroids. However, there is on-going debate about the optimal steroid regimen and which patients with KD should receive steroids. The Kobayashi score (KS), based on clinical and laboratory data, has been reported to identify patients with KD at high risk of IVIG resistance. In a Japanese population, a KS ≥4 has been validated as predictive of IVIG resistance and of increased risk of CAA.4 However, the KS failed to predict IVIG resistance in the USA, where it had low sensitivity but moderate specificity for high-risk disease.5 As the KS has not been tested in a UK population, we performed a retrospective analysis to investigate the predictive value of the KS and other proposed prognostic markers (hypoalbuminaemia, deranged liver function, jaundice and anaemia) for IVIG resistance and coronary artery aneurysms in a UK cohort.


Children treated for KD at a London tertiary referral infectious diseases centre during an 8-year period (2005–2013) were identified from hospital clinical coding data. Of 78 children, 70 had case notes available. Data were extracted from case records on a standard proforma including, as part of an approved audit at Imperial College Healthcare NHS Trust: clinical features of KD, laboratory data for each day until IVIG was received (full blood count, C-reactive protein (CRP), AST/ALT, albumin, bilirubin and sodium), details of all treatments, echocardiography results and demographic data. The KS was calculated based on the worst pre-IVIG results. We defined IVIG resistance as lack of defervescence within 24 h of administration, to match the definition of Kobayashi.4 Data were analysed using the Mann–Whitney and Fisher’s exact tests for categorical variables, and logistic regression for multivariate analyses. We aimed to determine if the KS was a useful predictor of IVIG resistance in our cohort.


Sixty-nine of 70 children received IVIG and aspirin during the acute illness (there was one retrospective diagnosis). There were no cases of recrudescent fever among our patient group. Ten cases had inadequate data recorded to calculate the KS (figure 1): these patients were not significantly different in sex, age or referral method; seven were IVIG-resistant and three developed CAA. Of the 59 remaining cases, 41 were male (69.5%), and the mean age was 33 months (range 2–113 months). Black children were over-represented and white children under-represented compared to London population norms (20% and 22% in the study, and 13% and 60% in the 2011 census (p=0.04 and p=<0.001 for black and white groups, respectively).6

Figure 1

Sample selection flowchart. After the exclusion of patients with incomplete information, and one patient who did not receive IVIG, 59 patients were eligible for the study. The Kobayashi score (KS) was calculated as follows: a score of 2 points is given for fever for ≤4 days at diagnosis; ALT ≥100 IU/mL; and ≥80% neutrophils. And 1 point is given for age <1 year; CRP ≥100 mg/L; and platelets ≤300×109/L. A score of 4 or more indicates ‘high-risk’ disease. IVIG, intravenous immunoglobulin; KD, Kawasaki disease.

KS and IVIG resistance

Of the 59 cases with full data, 37 patients (63%) were identified as having a high-risk KS (score ≥4) (figure 1). Nineteen of 59 had IVIG resistance, and required second-line treatment for persisting fever, including a second dose of IVIG (n=13), steroids (n=11), infliximab (n=4) and ciclosporin (n=1). KS ≥4 was not a sensitive predictor of IVIG resistance (11 of 19 resistant cases, sensitivity 58%). The specificity of the KS score was low (14 of 40 IVIG sensitive cases had KS <4, specificity 35%) (table 1). The proportion of children with KS ≥4 in the IVIG-resistant and sensitive groups (11 of 19 vs 26 of 40) was not different (p=0.77).

Table 1

Kobayashi score (KS) in children with and without IVIG resistance and coronary artery abnormalities

Children referred for tertiary management from other hospitals included those referred for refractory disease. Fourteen of 25 referred patients had received IVIG before transfer, and in consequence, a higher proportion of the referred group required second-line treatment than ‘walk-in’ children admitted directly from our emergency department (11/25 vs 8/34; Fisher’s exact, p=0.16). We compared the predictive value of the KS in each group (table 1). A KS ≥4 had poor sensitivity and specificity for referred patients (55% and 29%, respectively) and for walk-in patients (63% and 38%, respectively).

The KS was calculated for each patient, and patients were assigned to a ‘high-risk’ cohort (KS ≥4) or a ‘low-risk’ cohort (KS <4). The table shows the number of patients in the high- and low-risk categories, in relation to the IVIG response and the development of CAA. The top section refers to ‘walk-in’ patients who were admitted via the emergency department. The middle section refers to children referred for specialist management, who had a higher burden of complex KD. The lower section refers to the entire cohort taken together. The positive predictive value and the negative predictive value of the KS are shown for prediction of IVIG resistance and CAA development in the combined cohort.

KS and coronary abnormalities

We considered whether the KS was able to predict the development of CAA. Twenty of 59 patients developed CAA (including aneurysm or dilatation but not prominent or bright coronaries). Of these, 9/20 patients with CAA were IVIG non-responsive. Ten of 39 without CAA were also IVIG non-responsive (p=0.15). CAA were not predicted by KS ≥4 (12/20, sensitivity 60%). Fourteen of 39 cases without CAA had KS <4 (specificity 36%). The proportion of KS <4 or ≥4 did not vary significantly between groups with and without CAA (p=0.78).

There was a higher incidence of CAA in referred than in walk-in cases (11/25 vs 9/34, Fisher’s exact p=0.17), but the predictive value of the KS for aneurysms was poor for both groups (sensitivity 57% and 64% for walk-ins and referred patients respectively; specificity 36% for both groups.

Incomplete KD

Eighteen patients were classed as having incomplete KD (5 days of fever with two or three typical features). When they were excluded, the KS failed to predict IVIG resistance (KS ≥4 in 22/33 IVIG sensitive and in 7/10 IVIG non-responsive cases; p=1.0) or CAA (KS ≥4 in 10/16 with CAA and 19/27 without CAA; p=0.74).

Laboratory data

We correlated laboratory data for nine parameters previously associated with severe disease with the presence of IVIG resistance and the development of CAA (see online supplementary table S1). No laboratory markers were associated with IVIG resistance. Three laboratory markers correlated with the development of CAA: haemoglobin (p=0.025), albumin (p=0.035) and CRP (p=0.005). However, there was considerable overlap between the results in both groups. CRP showed the greatest difference (median 195 mg/L (CAA) vs 135 mg/L (no CAA)), but remained a poor discriminator (see online supplementary figure S1). We used logistic regression to compare prediction of CAA status by CRP and by the three significant laboratory parameters combined in a multivariate model. The combination of CRP, albumin and haemoglobin offered little advantage over CRP alone (see online supplementary figure S2).


We have demonstrated that in our patient group with KD, the KS is a poor predictor of IVIG resistance and an unhelpful test for the identification of children who may benefit from early adjunctive treatment with steroids. Our findings are consistent with a US study in which KS failed to predict IVIG resistance.5 Although a high proportion of children had a high KS (63%), this correlated poorly with IVIG resistance. Our study was limited by a small sample size. However, the cohort was enriched with referred, severe KD cases, facilitating the assessment of relatively more children with CAA. The KS was not found to be predictive of IVIG resistance or CAA in either the ‘walk-in’ patients or the referred patients. The findings were unchanged when incomplete KD cases were excluded from the overall analysis.

In Japan, the KS has been validated as a predictor of IVIG resistance and CAA development. In our study, the KS failed to predict either. Furthermore, although we identified laboratory markers that were associated with CAA, there was considerable overlap and none of the laboratory markers reliably identified patients in any group.

Our results indicate that neither the KS nor proposed laboratory markers including CRP, albumin and haemoglobin, can reliably identify UK patients at risk of IVIG non-response and CAA, who require additional anti-inflammatory treatment.1 In view of this, we suggest that the findings of the RAISE trial, in which steroids are advocated for the treatment of children with a high KS,3 should not be assumed to be directly applicable to children in the UK without substantiation in further trials. Development of biomarkers that predict CAA rather than IVIG resistance might result in administration of additional anti-inflammatory agents to some children whose disease would respond to IVIG alone, but potentially would reduce the numbers developing CAA.


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:


  • SD and NS contributed equally.

  • Contributors All authors contributed to writing, and approved, the final manuscript. SD, NS, SB and SG compiled and analysed data; SD and NS wrote the first draft; CJH analysed data; ML conceived and supervised the project; JAH supervised the project and developed the manuscript.

  • Funding The research was supported by a National Institute for Health Research (NIHR) Imperial Biomedical Research Centre grant, WMNP_P47845. The views expressed are those of the authors and not necessarily those of the NHS, the NIHR or the Department of Health.

  • Competing interests None.

  • Ethics approval The Governance Team, Imperial College Healthcare NHS Trust, approved this study.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement Investigators who would like to access the raw data used in the article are welcome to contact the corresponding author. There are no additional unpublished data which can be shared.