Steroids for Kawasaki disease: the devil is in the detail
- Correspondence to Professor Michael Levin, Department of Paediatrics, Imperial College London, Norfolk Place, London W2 1PG, UK;
- Received 11 September 2012
- Revised 11 September 2012
- Accepted 12 September 2012
- Published Online First 19 October 2012
In the 4 decades since the first description of Kawasaki disease (KD), the disorder has emerged as an important and common childhood condition with worldwide occurrence.1 KD is now recognised as the commonest cause of acquired heart disease in children in developed countries, and as an important cause of long term cardiac disease in adult life.2 The aetiology remains unknown but current thinking suggests that KD is an unusual inflammatory response to one or more as yet unidentified pathogen(s) occurring in genetically predisposed individuals.
Although the majority of patients with KD recover without long term consequences, the disorder is associated with vasculitis affecting the coronary arteries, and occasionally other muscular arteries, which results in coronary artery aneurysms (CAA) in over 20% of untreated patients.2 ,3 In all, 2%–3% of untreated patients die as a result of coronary artery thrombosis, myocardial infarction or, rarely, aneurysm rupture. Patients with giant (8 mm or more) CAA are at a long term risk of developing aneurysm thrombosis or coronary artery stenosis and myocardial infarction even years after the acute illness.2 In view of the frequency and severity of coronary artery complications, there has been intense interest in treatments to reduce the risk of CAA.
Intravenous immunoglobulin (IVIG) has been shown to reduce the risk of aneurysm formation to approximately 5%.4 ,5 Patients who do not respond to IVIG are at an increased risk of developing CAA and there is thus a need for additional treatments. Steroids are an effective treatment for other forms of vasculitis and their potential use in KD was considered soon after the disease was recognised. Early retrospective analyses of non-randomised treatment revealed that steroids were associated with increased risk of CAA.6 However, as the sickest patients were the most likely to have been given steroids, the adverse outcome noted may have reflected selection bias.
Subsequent reports evaluating steroid use in KD have produced conflicting findings which have been difficult to interpret, as several studies differed in patient inclusion criteria and used differing steroid dosage, duration and route of administration.7–13
Chen et al14 report their meta-analysis of published studies comparing the incidence of CAA in patients treated with IVIG plus corticosteroids or IVIG alone, which they hoped would establish the overall efficacy and safety of IVIG plus corticosteroids combinations in the primary treatment of KD.
They followed current guidelines for the conduct of meta-analyses, and included all published studies up to March 2012. Their methodology used defined criteria for quality assessment, including evaluation of criteria for diagnosis, study design, follow-up and blinding. They identified nine randomised studies meeting their quality criteria (seven conducted in Japan and two in the USA), involving 1011 patients (536 received IVIG plus corticosteroids and 475 IVIG alone). Statistical analysis was undertaken using standard approaches for meta-analysis and showed that significantly fewer patients receiving IVIG plus corticosteroids developed CAA than those receiving IVIG alone (7.6% vs 18.9%; OR: 0.3; p<0.001). The benefit was found in several subgroups including the six prospective randomised controlled studies, three studies that enrolled only patients at high risk of IVIG resistance, and studies using prednisolone or intravenous methyl prednisolone. They found no significant differences in the frequency of severe adverse events between the steroid and non-steroid treatment groups.
The meta-analysis of the published studies of Chen et al provides convincing evidence that steroids, combined with IVIG as initial treatment, reduce the overall risk of CAA in KD. The study raises a number of important questions for clinicians treating children with KD: Is the evidence in favour of steroid use sufficiently strong that steroids should be included with IVIG as a standard initial treatment of KD? Should all patients with KD receive steroids or only a selected group with a high risk of poor outcome and, if so, how should the patients at a higher risk be identified? Which steroid should be used, by what route and for what duration? Is steroid treatment safe?
Answers to these questions are essential to guide clinical decision making, but are not readily apparent from the meta-analysis or the studies on which they were based.
Meta-analysis methodology seeks to identify common truth by combining data from multiple studies, and the analysis of Chen et al should provide some security that the findings are robust. Unfortunately, the idiom ‘The devil is in the detail’ is applicable as there are differences between the studies included in their meta-analysis that make the interpretation of the findings more complicated than might be apparent at first sight.
A number of details, specific to each of the studies included in the meta-analysis, need to be considered before definitive conclusions can be drawn.
First, as seven of the nine studies in the meta-analysis were undertaken in Japan and only two in the USA, are the results applicable outside of Japan, where awareness of KD results in earlier diagnosis than in other countries (as evident from the later mean date of initial treatment in non-Japanese trials)? There is some concern that KD in Japan may behave differently from the disease in non-Japanese populations (highlighted by the failure of the score used to predict IVIG non-response in Japan (the Kobayashi score))15 to accurately predict non-response to IVIG in the USA.15
Second, do methodological differences between the studies make broad conclusions invalid without qualification? These differences are illustrated by two of the methodologically strongest studies, the American Pediatric Heart Network study10 and the recently reported Japanese RAISE study.11 While the American study evaluated the use of intravenous methyl prednisolone (30 mg/kg) given as a single dose in unselected patients with KD, RAISE evaluated low dose (2 mg/kg) intravenous prednisolone followed by prolonged oral prednisolone in patients selected on the basis of predicted IVIG resistance using the Kobayashi score. These studies have produced conflicting results, with steroids conferring significant benefit in the Japanese RAISE study but there being a lack of overall benefit in the American Pediatric Heart Network study (although benefit was observed on subsequent analysis in the IVIG unresponsive group). Given the predominance of Japanese studies and the large numbers of patients included selected on the basis of IVIG resistance, the meta-analysis may say more about the beneficial effect of steroids in Japanese patients meeting the criteria for IVIG resistance than about an overall benefit in unselected patients.
Third, while the meta-analysis has not revealed any increase in severe adverse events associated with steroid use, the total number of patients in the analysis receiving steroids is only about 500. We have no information as to the potential risks in terms of immunosuppression and severe opportunistic infections, or osteonecrosis of the femoral head that might be seen if much larger numbers of children are exposed to high doses and prolonged courses of steroids. As 80% of patients with KD respond to IVIG alone, it is reasonable to ask whether the additional benefit of including steroids justifies the potential risk of exposing the entire population of children with KD to steroid treatment. For patients who are at a higher risk of IVIG non-response (and thus CAA), the benefits of steroid treatment seem to outweigh the potential risks. However, the Kobayashi score used by the Japanese investigators failed to accurately predict IVIG non-response in studies in the USA.16 Clinicians outside Japan are thus faced with the question of how to identify the subgroup of patients with IVIG resistance who appear to benefit most from additional treatment with steroids.
Fourth, the meta-analysis does not help resolve the question of which steroid, route of administration or duration of treatment should be used. The meta-analysis may be weighted by the largest studies such as the RAISE study which utilised low doses (2 mg/kg) of intravenous and then oral prednisolone. However, as a range of different doses, routes and durations of steroids were included in the meta-analysis, clinicians are left with no clear answer on the best steroid regime to use.
What should clinicians treating KD do in the light of the current data? The meta-analysis and recent RAISE study support administration of steroids to patients at the greatest risk of IVIG non-response and CAA. In Japan, the Kobayashi score appears to identify these patients, but outside Japan, it seems reasonable to offer steroid treatment in addition to IVIG to patients with features of the most severe disease (and therefore the greatest likelihood of developing CAA), including the very young, and those with markers of severe disease, including intense inflammation, liver dysfunction, hypoalbuminaemia, anaemia and organ dysfunction. Steroids should also be given to patients who do not respond promptly to initial IVIG. All patients treated with steroids should be followed to identify adverse effects including osteonecrosis and intercurrent infection.
Finally, the therapeutic uncertainties raised by the recent studies and highlighted above can only be answered by further randomised trials. As KD is relatively rare, the KD research community should follow the lead of childhood cancer researchers who established the concept that all childhood cancer cases should be included in randomised studies to continually evaluate new treatment regimes. A similar approach is needed for KD and a rolling programme of treatment trials is needed to address the issues of patient selection, dose, route and safety of steroid treatment as well as use of other promising anti-inflammatory modalities including anti-TNF and interleukin 1 agents. Studies are also needed to identify better biomarkers to predict severe disease, CAA and the need for additional treatments, and which work in different populations. It is time for a concerted international drive to improve the evidence base for the treatment of this important childhood disease.
Competing interests None.
Provenance and peer review Commissioned; internally peer reviewed.