Kawasaki disease (KD) is the most common form of generalized vasculitis in children younger than 5 years. The etiology is unknown, and no specific diagnostic test exists thus far. KD was first described in Japan in the 1960s,1 and the first case in Germany was reported in 1979.2 Because of its effect on coronary arteries, KD results in substantial morbidity and mortality. In the United States and Japan, it is considered the most common cause of acquired heart disease in children.3,4
The diagnostic criteria originally described in Japan evolved because it was increasingly recognized that the “classical” clinical case definition was inadequate for KD diagnosis in young children.5 This led to underdiagnosis of KD and children who could otherwise have benefitted from treatment were missed. Historically, the reported incidence was also underestimated because these cases (known as “incomplete” KD cases) were not included in the estimates.
Annual incidence varies geographically and by ethnic background. Since KD was first described, Japan [which has the highest reported annual rate on a global basis (239 per 100,000 in 2010)]6 has witnessed a continuous increase in incidence rates, well documented through 21 serial nationwide surveys. Similar increases have been observed in neighboring Taiwan.7 In regions of predominantly European Caucasian racial background, the reported incidence is lower although it varies considerably between countries and over time. Rates in children younger than 5 years in the United States (20.8 per 100,000 in 2006)8 and Canada (26.2 per 100,000 in 2006)9 exceed those in Europe, Australia, New Zealand and Chile, where reported incidence rates range from 3.6 to 15.2 per 100,000.10–19 Consistent seasonal fluctuations in temperate latitudes of the Northern Hemisphere (peaking in winter) have been described. This suggests a seasonal exposure, varying over wide geographic areas.20
Ultimately, the factors influencing recent global trends are uncertain, but they may in part be attributable to increased physician awareness, a more sensitive case definition, and improved diagnostics.12,21 Most epidemiologic studies on KD rely on retrospective case reports and disease registries that suffer from well-recognized limitations, including incomplete data entry and underreporting.10 In Germany, we conducted a prospective nationwide KD surveillance and adjusted for case underreporting using capture–recapture (CRC) methodology. We describe the epidemiology, seasonal and geographic variation, and clinical and laboratory characteristics of KD in children younger than 5 years.
MATERIALS AND METHODS
Study Design and Case Definition
Prospective, national surveillance was conducted between January 1, 2011, and December 31, 2012. Experience with national surveillance suggests that there is considerable underreporting,21,22 so to account for this, we cross-validated the national surveillance data with hospital record data in a CRC analysis conducted in 2 German states, as described later. This study included all cases of KD diagnosed in Germany in children younger than 5 years treated with intravenous immune gamma globulin (IVIG) to avoid overreporting of infants with suspected KD. Diagnoses were classified as complete or incomplete according to the American Heart Association guidelines.23 Complete cases included those with persistent fever (≥5 days or fever for <5 days if it did not persist post-IVIG treatment) and with ≥4 of the principal clinical features [ie, (1) changes in extremities, such as erythema and edema of hands and feet, or desquamation of fingertips; (2) polymorphous exanthema; (3) bilateral conjunctivitis without exudate; (4) changes in the lips and oral cavity and (5) cervical lymphadenopathy].
Incomplete cases included those with fever and <4 clinical features with detection of coronary artery aneurysms (CAAs) or dilatation or those with laboratory evidence of systemic inflammation (C-reactive protein ≥30 mg/L) in combination with at least 3 other abnormal supplemental laboratory findings, namely (1) increased alanine transaminase, (2) albumin ≤3.0 g/dL, (3) leukocyturia, (4) anemia for age, (5) leukocytosis (≥15,000/mm³) and (6) thrombocytosis (≥450,000/mm³). CAA was diagnosed according to the Japanese Ministry of Health criteria for aneurysms as a lumen >3 mm in children younger than 5 years or a diameter 1.5 times the size of the surrounding segment or a clearly irregular lumen.24 The Japanese criteria were used in preference to the coronary artery–specific Z score as they were better known and well established in Germany at the time of the study. Cases were excluded if their treatment did not include IVIG or they did not otherwise meet the case definition.
National Surveillance: German Pediatric Surveillance Unit (ESPED)
KD cases were identified nationally through the hospital-based German Pediatric Surveillance Unit (ESPED).25 Data on rare pediatric diseases are actively collected from all children’s hospitals, pediatric departments and departments of pediatric surgery (n = 423) on a monthly basis. KD was introduced to the register for the purposes of the study, announced in the journal of the German Society of Pediatrics (monthly publication) in advance of the study, and all participating hospitals were informed by letter. The diagnosis was established by the reporting physician who was sent a standardized questionnaire to provide clinical information, laboratory and echocardiographic findings. Reported KD cases were independently reviewed and validated by an adjudicating committee including a pediatric cardiologist (B.S.), a pediatric rheumatologist (M.H.) and a specialist in pediatric infectious disease (R.B.)
Regional Surveillance: Hospital Records From Baden Württemberg and Saxony
To determine the degree of KD underreporting to ESPED, cases were cross-validated against hospital discharge records using the KD International Statistical Classification of Diseases and Related Health Problems 10th revision code (ie, M30.3). Records were assessed retrospectively in all pediatric hospitals in 2 geographically distant states from the north-east and the south-west of Germany: Baden Württemberg (BW; total population: 10,569,111) and Saxony (population: 4,050,204). These states together account for 18% of the total population in Germany and are representative of the national population in terms of age distribution and ethnic makeup (based on the self-reported nationality in the 2011 census).
Population denominator data, population density per square kilometer, proportion of land under farming and proportion of high-risk ethnic groups resident in the federal state were extracted from the Federal Statistics Office database (https://www.destatis.de/).
Differences between proportions were tested using Pearson χ2 or Fischer exact test, and between means using 2-sided independent sample t tests. Logistic regression was used to estimate the relationship between laboratory findings and age group (age < 1 year vs. age ≥ 1 year). The outcome is reported as an odds ratio (OR) with 95% confidence interval (CI). Incidence rates were estimated as the annual number of reports of KD per 100,000 children younger than 5 years in Germany, corrected for case under-ascertainment in ESPED using a 2-source CRC method.26,27
KD cases residing in Saxony or BW were included in the CRC estimate. The principle assumptions of CRC are26 as follows: (1) perfect case linkage between sources; (2) the same catchment probability between individuals; (3) independence between sources; (4) a closed study population. Hospital cases were linked with ESPED reports using 5 variables: hospital number, age at onset of illness, gender, month/year of admission and the first 3 digits of the residential postal code. Where cases were detected through hospital surveillance, but not reported to ESPED, the treating physician was contacted to complete an identical ESPED questionnaire. To account for possible heterogeneities in capture probabilities (ie, that some subgroups of the population are more or less likely to be recorded than others), we present the Chapman CRC estimator stratified by age and gender.26–28 To validate our stratified CRC estimate, we also report Chao lower bound estimator (including CIs),27,29 which relaxes the assumption that sources are independent.
Spatiotemporal Characteristics of Cases
Seasonal variation was examined using ordinary regression to fit a sine curve to the time series of weighted monthly case counts. The outcome is reported as a peak-to-trough ratio, comparing the months of highest and lowest incidence.30,31 The regional distribution of cases was summarized by federal state for ESPED reports only (Fig. 1). Regional incidence was compared with the ESPED national incidence using Poisson regression analysis. The rate ratio is reported as a standardized morbidity ratio (SMR) with a 95% CI.
National Surveillance Through ESPED
Nationwide, 338 children younger than 5 years were reported to ESPED during 2011 and 2012. Three hundred and one cases remained for evaluation after exclusion of 14 duplicate records, 11 cases that were not treated with IVIG, 4 children without permanent German residence and 8 children for whom KD diagnosis later was canceled. On further review, an additional 29 cases (9.6%) did not meet the case definition despite treatment with IVIG and were excluded. Of 272 validated cases identified, 122 were hospitalized in 2011 and 150 in 2012. Mean age was 1.9 years (standard deviation: 1.4) and 66.1% (179 of 271) were male. Overall, 79.8% (217 of 272) were classified as complete KD cases and 20.2% (55 of 272) as incomplete cases. Of the incomplete cases, 58.2% (32 of 55) were echocardiographic positive and 41.8% (23 of 55) were positive by laboratory findings. ESPED reports in BW and Saxony accounted for 14.3% of cases nationally (39 of 272). There was no difference in the age distribution nationally or in either of the 2 states surveyed, but in Saxony, there was a trend toward more female cases (7 of 11, 63.6%) compared with BW (12 of 28, 42.9%) or other regions nationally (73 of 232, 31.5%; Pearson χ2 test: 5.95, P = 0.055).
Regional Surveillance: Hospital Records in BW and Saxony
Of the 65 hospitals invited to participate in BW and Saxony, the response rate was 97% (63 of 65). Hospital surveillance identified 89 children with KD during the study period, of whom 4 were nonresident in BW or Saxony, and 10 did not meet the inclusion criteria. Of the remaining 75 records, 36 were also notified in ESPED, whereas 39 cases were notified through hospital records only. The latter were classified as complete KD (n = 31) and incomplete (n = 8, 5 of which were echocardiographic positive and 3 met laboratory criteria only). There was no difference in the age or gender distribution of cases notified through hospital records.
Incidence in Germany
Cases classified according to the American Heart Association algorithm were included in the CRC incidence estimate (n = 39 ESPED and n = 75 hospital records). The true number of cases and corresponding incidence rates according to the Chapman estimate and the Chao lower bound estimate are reported by age and gender in Table 1. These estimates yield an overall incidence of 6.4 per 100,000 (95% CI: 6.2–7.2; ie, Chapman estimate) and 7.2 per 100,000 (95% CI: 6.4–10.1; ie, Chao estimate), respectively, in children younger than 5 years. This translates into point estimates of 435 and 489 cases nationally in children younger than 5 years during the study period, an underestimate of the true incidence of 37%–44%.
Clinical Characteristics of KD Cases
Nationally, 315 cases were identified through ESPED and/or hospital surveillance. Incomplete KD cases accounted for 20.3% of cases (n = 64/315), of which 42.2% (n = 27/64) were diagnosed based on the laboratory criteria only. Younger children were more likely to present with incomplete KD than older children [mean age, 2.0 years (complete) vs. 1.2 years (incomplete, t = −4.06, P < 0.001)]. Of children younger than 1 year (n = 76), 19.7% (15 of 76) were diagnosed on the basis of a positive echocardiogram and 23.7% (18 of 76) on laboratory findings only, compared with 9.2% (22 of 239) and 3.8% (9 of 239) in children older than 1 year, respectively. Children younger than 1 year were more likely to have leukocytosis (OR: 2.4; 95% CI: 1.4–4.0; P = 0.002) and thrombocytosis (OR: 3.1; 95% CI: 1.7–5.9, P < 0.001) than children older than 1 year. There were high rates of anemia in children younger than 6 months (73.0%, 27 of 37), but this proportion did not differ significantly from older children (ie, 63.2%, 151 of 239) or children aged 6 months to 1 year (58.9%, 23 of 39), P = 0.410. Among all cases, risk factors associated with the development of CAAs were (1) young age (<1 year), (2) male gender and (3) delayed treatment with IVIG. Children with CAA were more likely to have received repeat IVIG treatment (Table 2). Of cases that did not meet the inclusion criteria (n = 34 in total), 24 had 2 or 3 clinical signs but no additional laboratory or echocardiographic findings allowing them to be defined as incomplete cases. A higher proportion of children older than 1 year did not meet the case definition (12.1%, 33/272) than children younger than 1 year (1.3%, 1 of 77; Pearson χ2 (1 df) = 8.0, P = 0.005).
There was a clear seasonal distribution of cases (Fig. 2), with the peak of cases occurring on January 22 and a peak-to-trough ratio of 1.6 (95% CI: 1.1–2.2). Cases were widely distributed nationally (Fig. 1). By using ESPED reports only, the SMR in 2 states significantly exceeded the national mean (4.0 of 100,000, 95% CI: 3.5–4.5, as a reference) in 2011 to 2012: Brandenburg (SMRref: 2.3; 95% CI: 1.4–3.7) and Bavaria (SMRref: 1.4; 95% CI: 1.1–1.9). The SMR in Hessen was lower than the national mean (SMRref: 0.4; 95% CI: 0.2–0.7; Fig. 1). Twelve clusters of ≥3 cases in the same 3-digit postal code area were reported across 5 states. Five of these clusters (18 cases) were reported in Bavaria and 2 clusters (6 cases) in Brandenburg. No correlation was found among incidence rates by state and state population density per square kilometer, proportion of land under farming or proportion of high-risk ethnic groups resident in the region (data not shown).
The diagnosis of KD is complex, and the appearance of its classical clinical signs may not be enough to secure a diagnosis. Many national estimates of KD incidence are based on the retrospective case reports or on registries that do not take account of cases diagnosed with the aid of laboratory and/or echocardiographic test results. Without this supplemental information, incomplete cases will be missed and the true incidence underreported.10 In our study, we account for both: incomplete cases using laboratory and echocardiographic findings and underreporting by applying CRC methodology, which allows us to estimate a more realistic range of KD incidence. Depending on the CRC method employed, the incidence in Germany in children younger than 5 years treated with IVIG ranged from 6.4 of 100,000 to 7.2 of 100,000. Underreporting to our primary data source, the national Pediatric Surveillance Unit (ESPED; ref. Website: www.inopsu.com/countries) is known to be high.24,25 To account for this, we supplemented ESPED data with active hospital surveillance in 2 German states (18% of the German population). By using the well-known Chapman estimator,27,28 we obtained a crude estimate of 6.4 of 100,000 (95% CI: 6.2–7.2). However, 2-source CRC suffers from a range of biases, including heterogeneity of capture probabilities and source dependence.26 On stratification by age and gender (2 factors strongly associated with KD), the sum of the subgroup estimates was similar to the crude estimate. However, there was a substantial overlap between the data sets (>90% of ESPED cases “captured” by hospital surveillance), and therefore, likely dependence between the sources that cannot be statistically tested in a 2-source CRC.32 To account for this, we also report Chao estimate29 (7.2 of 100,000; 95% CI: 6.4–10.1), which is believed to be less biased than Chapman estimate when independence between sources is in doubt.27 To validate our estimates, we reviewed the German national discharge records for discharge diagnosis of KD (ie, International Statistical Classification of Diseases and Related Health Problems 10th revision code M30.3). Multiple discharges may be attributed to a single case, and thus, it is unlikely that the incidence of hospital discharges would be less than the true incidence of cases (underreporting notwithstanding). Based on these records, the estimated annual KD incidence for the years 2011 and 2012 was 9.6 of 100.000, which is more consistent with the Chao estimate, being just within the upper bound confidence limit. While acknowledging the limitations of CRC, we consider Chao estimate to be a more reliable reflection of the “true” incidence. Our findings are broadly consistent with recent estimates from neighboring countries, such as those from the Netherlands, where underreporting was not accounted for.10
Historical estimates of KD incidence may be biased by the recent shift in KD case definitions (first published in 2004)21, heightened physician awareness and improved diagnostics. Reports including incomplete cases may also have contributed to recent increases in estimated incidence rates.12 Of 315 KD cases identified in this study, 20% were diagnosed as incomplete KD cases: 12% of those on the basis of positive echocardiography and an additional 8% on laboratory findings alone. This is consistent with other studies where both echocardiographic and laboratory findings have been reported.10,12 In England, where KD incidence has been estimated repeatedly, incidence rates doubled13 from 4.0 of 100,000 in 1991 to 8.4 of 100,000 in 2003.14 However, the most recent estimate of 9.1 of 100,000 in 2012 (Gillian Hall, personal communication, May 2015) suggests that the incidence rates may have stabilized in recent years. Overall, incomplete cases were younger than complete cases, and two thirds of children diagnosed on the basis of laboratory criteria only were younger than 1 year. One third of incomplete cases were diagnosed using echocardiographic criteria, which supports the role of cardiac imaging in detecting cardiac complications and in securing the diagnosis of KD. Basic hematologic parameters (ie, thrombocytosis and leukocytosis) were also more commonly altered in younger children. In the absence of reliable diagnostics for KD, significant underdiagnosis or delayed diagnoses may occur resulting in an increased risk of CAAs, particularly in younger infants. Additional markers to distinguish KD from other clinical entities are the subject of ongoing research (eg, the analysis of specific urine proteomics33), but associations have not yet been confirmed. For this reason, such tests have not yet been implemented in routine clinical practice. In terms of treatment, our patient cohort clearly benefitted from early IVIG administration, and late administration was shown to be a risk factor for development of CAA. Other risk factors included young age and male gender.
Seasonal variation in Germany was confirmed, with a winter peak in KD hospitalizations. ESPED report rates by federal states were variable, and the SMR exceeded the national mean in 2 states, ie, Bavaria and Brandenburg. This may simply reflect regional report bias. No correlation was found between report rates and potential risk factors including population density, proportion of land under farming and proportion of high-risk ethnic groups at state level. However, formal testing for spatial-temporal clustering would be required to investigate these findings further.
There were some limitations to this study (in addition to CRC limitations discussed previously). To prevent overreporting of infants with fever of unknown origin who were also suspected to have KD, we only included children who were treated with IVIG. Therefore, some children with true KD—who were not treated with IVIG—will have been missed.
The authors thank all the participating pediatricians reporting KD cases.
1. Kawasaki T. [Acute febrile mucocutaneous syndrome with lymphoid involvement with specific desquamation of the fingers and toes in children]. Arerugi. 1967;16:178–222
2. Cremer HJ. Acute febrile mucocutaneous lymphadenopathy syndrome in Germany
? Pädiat Praxis. 1979;21:75–82
3. Taubert KA, Shulman ST. Kawasaki disease. Am Fam Physician. 1999;59:3093–102, 3107
4. Sonobe T. A summary of the epidemiologic surveys on Kawasaki disease conducted over 30 years. JMAJ. 2005;48:30–33
5. Burns JC, Glodé MP. Kawasaki syndrome. Lancet. 2004;364:533–544
6. Nakamura Y, Yashiro M, Uehara R, et al. Epidemiologic features of Kawasaki disease in Japan: results of the 2009-2010 nationwide survey. J Epidemiol. 2012;22:216–221
7. Lue HC, Chen LR, Lin MT, et al. Epidemiological features of Kawasaki disease in Taiwan, 1976-2007: results of five nationwide questionnaire hospital surveys. Pediatr Neonatol. 2014;55:92–96
8. Holman RC, Belay ED, Christensen KY, et al. Hospitalizations for Kawasaki syndrome among children in the United States, 1997-2007. Pediatr Infect Dis J. 2010;29:483–488
9. Lin YT, Manlhiot C, Ching JC, et al. Repeated systematic surveillance of Kawasaki disease in Ontario from 1995 to 2006. Pediatr Int. 2010;52:699–706
10. Tacke CE, Breunis WB, Pereira RR, et al. Five years of Kawasaki disease in the Netherlands: a national surveillance study. Pediatr Infect Dis J. 2014;33:793–797
11. Salo E, Griffiths EP, Farstad T, et al. Incidence of Kawasaki disease in northern European countries. Pediatr Int. 2012;54:770–772
12. Heuclin T, Dubos F, Hue V, et al.Hospital Network for Evaluating the Management of Common Childhood Diseases. Increased detection rate of Kawasaki disease using new diagnostic algorithm, including early use of echocardiography. J Pediatr. 2009;155:695–9.e1
13. Harnden A, Alves B, Sheikh A. Rising incidence of Kawasaki disease in England: analysis of hospital admission data. BMJ. 2002;324:1424–1425
14. Harnden A, Mayon-White R, Perera R, et al. Kawasaki disease in England: ethnicity, deprivation, and respiratory pathogens. Pediatr Infect Dis J. 2009;28:21–24
15. Fischer TK, Holman RC, Yorita KL, et al. Kawasaki syndrome in Denmark. Pediatr Infect Dis J. 2007;26:411–415
16. Lynch M, Holman RC, Mulligan A, et al. Kawasaki syndrome hospitalizations in Ireland, 1996 through 2000. Pediatr Infect Dis J. 2003;22:959–963
17. Heaton P, Wilson N, Nicholson R, et al. Kawasaki disease in New Zealand. J Paediatr Child Health. 2006;42:184–190
18. Saundankar J, Yim D, Itotoh B, et al. The epidemiology and clinical features of Kawasaki disease in Australia. Pediatrics. 2014;133:e1009–e1014
19. Borzutzky A, Hoyos-Bachiloglu R, Cerda J, et al. Rising hospitalization rates of Kawasaki Disease in Chile between 2001 and 2007. Rheumatol Int. 2012;32:2491–2495
20. Burns JC, Herzog L, Fabri O, et al.Kawasaki Disease Global Climate Consortium. Seasonality of Kawasaki disease: a global perspective. PLoS One. 2013;8:e74529
21. Reinhardt K, Weiss S, Rosenbauer J, et al. Multiple sclerosis in children and adolescents: incidence and clinical picture—new insights from the nationwide German surveillance (2009-2011). Eur J Neurol. 2014;21:654–659
22. Weiss S, Streng A, Kries R, Liese J, Wirth S, Jenke AC. Incidence of intussusception in early infancy: a capture-recapture estimate for Germany
. Klin Padiatr. 2011;223:419–423
23. Newburger JW, Takahashi M, Gerber MA, et al.Committee on Rheumatic Fever, Endocarditis, and Kawasaki Disease, Council on Cardiovascular Disease in the Young, American Heart Association. Diagnosis, treatment, and long-term management of Kawasaki disease: a statement for health professionals from the Committee on Rheumatic Fever, Endocarditis, and Kawasaki Disease, Council on Cardiovascular Disease in the Young, American Heart Association. Pediatrics. 2004;114:1708–1733
24. Research Committee on Kawasaki Disease. Report of Subcommittee on Standardization of Diagnostic Criteria and Reporting of the Coronary Artery Lesions in Kawasaki Disease. 1984 Tokyo, Japan Ministry of Health and Welfare
25. Göbel U, Heinrich B, Krauth KA, et al. [Process and outcome quality of the German Paediatric Surveillance Unit (ESPED)]. Klin Padiatr. 2010;222:92–97
26. Hook EB, Regal RR. Capture-recapture methods in epidemiology: methods and limitations. Epidemiol Rev. 1995;17:243–264
27. Brittain S, Böhning D. Estimators in capture-recapture studies with two sources. AStA Adv Stat Anal. 2009;93:23–47
28. Chapman C. Some properties of the hypergeometric distribution with applications to zoological censuses. U California Public Stat. 1951;1:131–160
29. Chao A. Estimating population size for sparse data in capture-recaptue experiments. Biometrics. 1989;45:427–438
30. Edwards JH. The recognition and estimation of cyclic trends. Ann Hum Genet. 1961;25:83–87
31. Brookhart MA, Rothman KJ. Simple estimators of the intensity of seasonal occurrence. BMC Med Res Methodol. 2008;8:67
32. Chao A, Tsay PK, Lin SH, et al. The applications of capture-recapture models to epidemiological data. Stat Med. 2001;20:3123–3157
33. Kentsis A, Shulman A, Ahmed S, et al. Urine proteomics for discovery of improved diagnostic markers of Kawasaki disease. EMBO Mol Med. 2013;5:210–220