Breastfeeding and vitamin D supplementation reduce the risk of Kawasaki disease in a German population-based case-control study

Kawasaki disease (KD) is a vasculitis of unknown etiology, which can be complicated by the development of coronary artery aneurysm (CAA). It is a rare disease, primarily affecting children under five years of age. The incidence among Asian children is much higher than among Caucasians. In our 2011–2012 cohort, we estimated an incidence of 7.2/100,000 in Germany (children < 5 years) [1, 2]. By contrast, for the year 2012 Japan reported the highest incidence worldwide, with 265/100,000 children (aged 0–4 years) diagnosed [3]. Cause and pathogenesis of KD are unclear. A combination of different risk factors, for disease development, along with a genetic predisposition, so far has been assumed [2, 4, 5]. To further investigate this assumption, we analysed the impact of different risk factors for KD during the perinatal period and infancy. Week of gestation, birth weight and type of parturition are known to impact health in later life [6]. Caesarean delivery, for example, is known to be a risk factor for multiple chronic immune diseases, including asthma and juvenile idiopathic arthritis [7]. Similarly, vitamin D supplementation and breastfeeding are well-known for their health benefits: In Germany, vitamin D supplementation generally is recommended for the first year of life. Besides playing an important role in the bone and calcium balance [8], vitamin D also regulates other cell functions and supports the immune system overall [9]. Furthermore, epidemiological studies on vitamin D deficiency show potential associations with cardiovascular diseases [10] and vasculitis [11].

For its health and other benefits, breastfeeding is preferable to formula feeding. Among other advantages, breastfeeding seems to protect against a variety of infectious diseases [12] and non-breastfed children have a higher risk of developing allergies and bronchial asthma later in life [13]. A Japanese study already has indicated a potential protective effect in KD [14].

During the perinatal and infancy periods, these and other factors potentially play a role in the pathogenesis of KD. Therefore, in a population-based, retrospective case-control study we investigated the association and impact of birth characteristics, breastfeeding practices and vitamin D supplementation on KD.

Our study was designed as an epidemiological case-control study. KD cases were recruited from the population-based German Pediatric Surveillance Study (ESPED) from January 2012 to December 2014 [15]. For the ESPED report, pseudonymous identifiers were used. A standardized questionnaire was sent to reporting physicians requesting clinical and laboratory details from each KD case. Additionally, we requested that reporting physicians ask parents (via written informed consent) to agree to disclosure of their child’s identity so that our group could conduct follow-up questioning. In cases where such consent was provided, it became possible for us to validate the original ESPED data via discharge letters, recorded laboratory values and echocardiographic findings. For details see Additional file 1.

Due to data protection regulations, we only were able to include those KD cases where parents had provided written informed consent. Data regarding breastfeeding practices, vitamin D supplementation and birth characteristics were collected via a standardized questionnaire, and responses were provided online and in telephone interviews (see Additional file 2).

In order to obtain control cases from the KD patient’s own immediate environment, control cases (n = 326) were recruited from among the personal contacts of the KD case. First, the parents of KD cases were asked whether they might be able to recruit one of their child’s friends to take part in the survey. If no immediate friend of the KD cases was available or interested, then the treating paediatrician was asked to find a child who met the following criteria: (1) same sex, (2) same place of residence (≤ 50 km distance), (3) same age (≤ 6 months age difference) and (4) no medical history of KD. Once KD and control cases had been established, data were collected in an anonymous way through an online questionnaire. Case-control-pairing was conducted by using the ESPED pseudonymous identifier. In some cases, more than one control case was able to be recruited. For these KD cases, the control case was selected manually accordingly to best fit, which was based upon: (1) age, (2) gender and (3) place of residence. The best fitting case-control-pair was used for the statistical analysis.

In accordance with AHA guidelines, in cases where fever persists for over 36 h, German recommendations stipulate the administration of a second IVIG therapy course. Our study defined such cases as refractory to IVIG.

Out of a total of n = 631 reported KD cases, n = 69 cases did not fulfil the AHA case definitions. Follow-up with complete data in response to our questionnaire was available in n = 308 cases. For details see Fig 1.

Matching

A case-control matching was made based on sex, age and place of residence (Figs. 2, 3 and Table 1). The measures of association, i.e. contingency coefficient and Cramer’s V were > 0.9 for age and place of residence. This attests to the high quality of the matching and furthermore shows that the results would have been distorted, had the matching been ignored. With regard to sex, the measures of association were between 0.577 and 0.706, indicating a satisfactory case-control matching. Here, the OR for the matching factor sex was 36,346, meaning that the likelihood that a KD and control case would have the same sex was 36 times higher than that of their having different sexes.

	Matching factors
	Sex	Age	Place of residence
Measures of association
Phi coefficient	0.706
Contingency coefficient	0.577	0.945	0.945
Cramer’s V	0.706	0.961	0.961
Pearson correlation coefficient		0.901

Our study is the first of a nationwide, population-based cohort in Germany. The study investigates various perinatal and infancy factors in association with Kawasaki disease and its disease course. Our data reinforce the findings of a large study from Japan which has indicated that breastfeeding may be protective for KD [14]. Vitamin D supplementation appears to have a minor protective effect for KD as well. However, an influence of these factors on the clinical course (development of CAA, raised systemic inflammatory markers, being refractory to IVIG treatment) was not observed. Since the risk factors age, sex and place of residence were used as matching factors, they could not be included in our analysis of potential risk factors.

Our data regarding vitamin D supplementation showed that KD cases in our cohort received significantly less vitamin D than their healthy controls. In a recently published study that measured vitamin D levels during the acute phase of KD, levels proved to be significantly lower compared to controls [18]. Stagi et al. also have described a reduced serum vitamin D concentration in KD cases. These authors suggested that vitamin D may be reduced by the inflammatory process [19]. Our findings indicate that lower vitamin D levels may already play a role before disease onset, since in our cohort, children with KD had a shorter duration of vitamin D supplementation. However, due to our study design, we were not able to determine the actual vitamin D levels prior to the development of KD.

Because the skin of infants is not able to produce sufficient vitamin D, without supplementation, infants will develop vitamin D deficiency [20]. For this reason, in countries such as the USA, infant food is supplemented with vitamin D. In countries such as Germany, a daily vitamin D supplementation is recommended during the first year of life [8]. The Nutrition Commission of the German Society for Pediatric and Adolescent Medicine recommends 400–500 IU / day during the first year of life [21].

The anti-inflammatory and immunomodulatory effects of vitamin D are well-described in several studies [22]. Vitamin D appears to block factors (e.g., tumor necrosis factor α) that are essential for the activation of proinflammatory cytokines [23]. This raises the possibility of vitamin D being used as complementary therapy in immune-mediated diseases such as KD. Some studies already indicate that reduced vitamin D levels may negatively effect the clinical course. First, reduced vitamin D levels seem to be associated with the prevalence of cardiovascular diseases and the development of CAA [19, 24]. Low vitamin D levels inducing endothelial dysfunction may partially explain this association [25]. Second, Zhang et al. have described significantly reduced vitamin D levels in KD patients who were refractory to IVIG treatment [18]. We investigated both of these factors — CAA development and being refractory to IVIG treatment — but were not able to find a significant association with vitamin D supplementation. Furthermore, due to the anti-inflammatory activity of vitamin D, we also investigated other inflammatory serum markers, including CRP, thrombocytes and leukocytes — ones that might indicate higher disease activity in KD cases. However, the markers did not differ between KD cases with longer vs. shorter vitamin D supplementation. Stagi et al. have reported on a negative association regarding low vitamin D serum levels with CRP titers [19]. Additionally, Jun et al. retrospectively reviewed whether low vitamin D levels were associated with resistance to IVIG therapy in KD cases. They found vitamin D deficiency to be associated with IVIG resistance, but not associated with inflammatory markers [26]. In summary, we conclude that vitamin D could potentially play a role in the inflammatory process of KD. However, for a better understanding of its mechanism, additional prospective studies would be necessary.

The perinatal factor of breastfeeding showed a significant protective effect. In general, breastfeeding is recommended for at least six months due to its various health benefits [27]. Breast milk contains lactoferrin and lysozyme, which non-specifically inhibit all types of pathogenic microorganism [28]. In addition, secretory immunoglobulin A confers specific protection against mucosal pathogens [29, 30]. Breastfeeding rarely has been considered in association with KD. Data from a recent study in Japan showed that children who were breastfed were less likely to develop KD [14]. The Japanese authors described a protective effect for both exclusive (OR 0.26; 95% CI 0.12–0.55) and partial breastfeeding (OR 0.27; 95% CI 0.13–0.55) as compared to formula milk [14]. Even partially breastfed children showed a general benefit. Therefore, the protective effect of breast milk probably is not linked to the duration of breastfeeding but rather to substances in either breast milk or colostrum [14], the latter of which contains high levels of secretory antibodies [31]. Additional investigations will be needed to identify these protective substances. In the future, infants who cannot be breastfed might be able to receive these protective substances for protective purposes.

In general, breastfeeding has a positive effect on the outcome of infections in infants. For example, Nishimura et al. have described shorter hospital stays and better outcomes in breastfed children infected with respiratory syncytial virus [32]. However, the effect of breastfeeding on the outcome of KD remains unknown. In our study, we were unable to find an association between the duration of breastfeeding and the appearance of CAA, being refractory to IVIG treatment or the amount of inflammatory serum markers. Therefore, prolonged breastfeeding seems not to be associated with a milder clinical course in KD. Protection against infection by breast milk, the so-called maternal passive immunity, relates to the first six months of feeding [33]. We investigated the potential impact of breastfeeding and vitamin D supplementation on the age of disease onset. However, in shorter-breastfed and/or shorter-supplemented cases, KD did not occur sooner than in other KD cases. It is therefore questionable whether maternal passive immunity has an effect on the age of KD onset.

In conclusion, breastfeeding seems to have a protective effect on the development of KD, supporting the health benefit of general breastfeeding recommendations.

The strength of our case-control study is based upon the large number of KD cases in a Germany-wide cohort, along with the excellent comparability between our KD and control groups. KD cases were reported from all regions of Germany, as were the control cases who were recruited in parallel. A possible bias by the confounder age, sex and place of residence was able to be minimized through a thorough matching process [34]. In addition, all KD cases were evaluated using a standardized questionnaire and they fulfilled internationally-accepted KD criteria [16]. These strict criteria minimized the risk of misdiagnosed KD cases being included in our study. However, the validity of our results should be interpreted in light of our retrospective study design and considered in context of its known limitations. Although KD cases have been reported as part of a prospective surveillance study, data analysis for this study was based upon a retrospective survey conducted over a period of three years. The time period between the acute phase of KD and that of this survey was at least one year. Data quality strongly depends upon the memory ability of the parents, the so-called recall bias. Therefore, we cannot rule out the possibility that families of sick children may have been more likely to remember potential risk factors and/or living conditions than families in the control group did. Another limitation relates to the fact that KD cases were reported from a large number of hospitals, most of whom were without a standardized treatment protocol. KD therapy varies considerably. Some of our KD cases received corticosteroids early in the course of disease that might have impacted the patients’ clinical course, potentially outweighing the influence of vitamin D supplementation and breastfeeding. Finally, a potential confounding between different variables cannot be ruled out. For example, a higher social status often determines a healthier life style. By recruiting control cases via friends and relatives of KD cases, an attempt was made to achieve the best possible comparison regarding socio-economic status. Due to data protection regulations, we were not able to collect additional data regarding the socio-economic status of both KD and control cases.

Another limitation relates to the simultaneous screening for several risk factors. The so-called multiple testing problem (look-elsewhere effect) describes randomly significant findings due to the large number of factors studied [35]. For this reason, our findings should be verified in future prospective studies of larger and more diverse study populations.

In association with KD in our study population, our results indicate protective effects of vitamin D supplementation and further reinforce breastfeeding to have a protective effect. However, these factors seem not to influence the natural course of the disease. Although the effects were small, they nevertheless underscore the overall benefit of both interventions.