The Long-term Cardiac and Noncardiac Prognosis of Kawasaki Disease: A Systematic Review

The systematic review protocol was designed in accordance with the Preferred Reporting Items for Systematic reviews and Meta-Analyses.¹⁵  The protocol was registered in the International Prospective Register of Systematic Reviews database (CRD42020186273). The search was performed in June 2020 by a trained librarian (Q.M.). Searched databases included Ovid Medline, Embase, and Cochrane Central Register of Controlled Trials from inception to June 2020 (Supplemental Tables 3–5). Search terms were relevant to KD and prognosis. A combination of literature review and expert consensus (R.S.M.Y. and B.W.M.) was performed to facilitate the search for noncardiac diseases that may be of interest. Examples of noncardiac diseases included arthritis and sensorineural hearing loss (Supplemental Table 6). We evaluated the results and references of 4 related systematic and literature reviews to identify additional records.^1,16–19 

Upon removal of duplicates, 2 investigators (J.J.Y.L. and E.L.) independently reviewed abstracts of each reference to identify articles for a full review. The same investigators obtained the full articles of eligible abstracts and determined each article’s eligibility for inclusion independently. Discrepancies were resolved by consensus, with additional input from J.W. and B.F. as required.

Studies were eligible according to the following criteria: (1) individuals with a diagnosis of KD as children (≤18 years), (2) randomized trials or observational studies (cohort, case-control, cross-sectional studies), (3) studies that report any clinical outcome, (4) an average (mean and median) follow-up of ≥1 year after KD diagnosis, and (5) studies written in English, Japanese, Chinese, or Korean. We excluded CAA as an outcome and, instead, considered it as a risk factor for cardiac events. Exclusions included:

1. case reports and series, literature reviews, systematic reviews, expert opinions, or qualitative studies;

2. abstracts without full results;

3. studies written in a language that was not English or one of the Asian languages specified; and

4. studies where the average follow-up was <1 year after diagnosis.

J.J.Y.L. and E.L. independently extracted data from each English study with a standardized extraction form. Recorded results were reviewed for consistency and accuracy. For studies written in Chinese or Korean, 1 investigator with competency in the language (E.L. or J.J.Y.L.) extracted the data. Machine translation software was used to translate studies written in Japanese.

The following data were extracted: author, publication year, country, study design, number of patients with KD and/or comparators, age at diagnosis, sex, ancestry, percentage with CAA, and follow-up duration. The outcome(s), analysis, and results were recorded. Results were descriptively analyzed.

One author (J.J.Y.L.) assessed the quality of included studies. Cohort and case-control studies were evaluated with the Newcastle-Ottawa scale (NOS).²⁰  Studies were assigned points according to 3 quality domains (selection, comparability, and outcome). Studies scored the maximum comparability score if they controlled for KD exposure in the design or analysis. Follow-up completion needed to be ≥80% to reduce the likelihood of bias. The maximum NOS score was 9; higher scores reflected a higher quality assessment. Cross-sectional studies were evaluated with the Joanna Briggs Institute (JBI) checklist.²¹  The JBI checklist is comprised of 8 items, with each item scored as “Yes,” “No,” “Unclear,” or “Not Applicable.” The number of items that scored Yes were counted to provide a maximum JBI score of 8. Higher scores reflected a higher quality assessment. Randomized trials were evaluated with the Jadad scale.²² 

A total of 4621 records were identified and 731 duplicate references were subsequently removed (Fig 1). From 3890 records, the title and abstract screening process yielded 151 articles for full-text review. Interrater reliability among the 2 investigators after the title and abstract screening process was moderate (Cohen’s κ: 0.44). In total, 74 studies met the criteria and were included. Of the 77 articles excluded, 32 were excluded for irrelevant outcomes (studies only reported CAA as an outcome), 21 for short follow-up duration, 11 for incorrect study design (4 reviews, 3 case series, 3 case reports, and 1 opinion/editorial), 1 for abstract only, 6 for excluded language, and 6 (3 Chinese and 3 Japanese studies) were unavailable for full-text review. From the excluded studies, 8 studies evaluated a noncardiac outcome but did not meet eligibility. These included obstetrical outcomes (n = 1, case series), arthritis (n = 2, both excluded for short follow-up), systemic autoimmune rheumatic diseases (n = 1, case report), renal complications (n = 1, short follow-up), infections (n = 1, short follow-up), and allergic diseases (n = 2, both excluded for short follow-up).

All studies were observational; 51 (69%) were retrospective cohort, 13 (18%) were cross-sectional, 6 (8%) were case-control, and 4 (5%) were prospective cohort. Published years ranged from 1982 to 2020. The studies included a total of 72 487 patients from 13 countries. One study was written in Japanese and the remainder in English. Fifty-five (74%) studies originated from East Asian countries. Cohort and case-control studies had a median NOS score of 6 out of 9 (range: 5–8). Cross-sectional studies had a median JBI score of 6 out of 8 (range: 3–8).

Thirty-six studies reported mortality, 55 reported a cardiac outcome, and 12 reported a noncardiac outcome. The patient population in 28 of 36 (78%) mortality studies, 42 of 55 (76%) cardiac outcome studies, and 3 of 12 (25%) noncardiac outcome studies was confined to patients with KD and without a non-KD comparator group. The distribution of studies by average follow-up duration was as follows: 22 (31%) reported ≥1 to 5 years, 24 (33%) reported >5 to 10 years, 14 (19%) studies reported >10 to 15 years, and 13 (15%) reported >15 years of follow-up. One study did not report the average follow-up duration but was included because patients were followed for a minimum of 2 years.²³ 

Thirty-six studies evaluated mortality in the KD population (Table 1; Supplemental Tables 7 and 8). Seventeen studies reported all- cause mortality in patients with and without CAA over an average follow-up period of 4 to 24 years. Eight of these studies compared mortality relative to a non-KD comparator group; of these, 7 publications were based on the same patient cohort that was reassessed at different follow-up intervals. Mortality estimates ranged from 0% to 6%. Of the 10 studies that reported a mortality prevalence, 9 studies reported a prevalence of ≤1%.

Nakamura et al published 7 studies that reported standardized mortality ratios (SMRs) observed in 6576 Japanese patients with KD, diagnosed between 1982 and 1992, until 2009 (Supplemental Table 8).^24–30  Approximately 15% (n = 1003) developed cardiac complications after KD but the studies do not provide details. It is unclear if these children had CAA. After the acute phase, patients with KD did not appear to be at increased risk for all-cause mortality compared with the general population.^24–30  Conclusions about the long-term mortality in patients with KD and cardiac sequelae varied by publication. The most recent study concluded that male patients with early cardiac complications were at an increased risk of death after 23 years of follow-up (SMR: 2.3 [95% confidence interval (CI), 1.2–3.9]).³⁰ 

Nineteen studies evaluated outcomes exclusively in patients with CAA (Supplemental Table 9). None of these studies compared mortality outcomes with a comparator group. Mortality estimates ranged from 0% to 25% over an average follow-up from 2 to 20 years. Only 3 studies (16%) reported a mortality prevalence of ≤1%.

Nine studies reported survival probabilities over time (Fig 2; Supplemental Table 10).^31–38  Survival ranged from 92% to 99% at 5 years,^32,37,38  92% to 99% at 10 years,^{31–35,38,39}  85% to 99% at 20 years,^{33–35,38,40}  63% to 95% at 25 years,^33,34  88% to 94% at 30 years,^34,36,38,39  and 69% at 35 years.³⁷  Tsuda et al reported the lowest survival, with 85% (95% CI, 73%–92%) and 79% (95% CI, 66%–88%) at 10 and 20 years, respectively.³⁴  However, this likely relates to the selection criteria because only patients with a history of MI were included. Chih et al also reported a low 35-year survival of 69% in 27 patients with large CAA, but it is unclear how many patients achieved this follow-up since the median follow-up duration was 7 years.³⁷ 

None of the 19 studies compared survival with non-KD comparators. Four studies compared survival among subgroups within the KD population. Kitamura et al compared 54 patients who underwent CABG with a saphenous vein graft to 114 patients with an internal thoracic artery (ITA) graft and observed higher survival with ITA grafts.³¹  Tadokoro et al compared patients who received single grafts (n = 53) to those with multiple grafts (n = 49) during CABG, and found no survival differences. Tsuda et al reported mortality in 245 patients with KD and large CAA (≥8 mm), according to bilateral (n = 141) or unilateral (n = 104) involvement. Bilateral CAA was defined as having aneurysms in right and left coronary arteries. The 30-year survival for patients with bilateral CAA was significantly lower than those with unilateral involvement (87% [95% CI, 78%–93%] vs 96% [95% CI, 85%–99%]).³⁵  Tsuda et al also determined in a separate cohort of 214 patients with KD that there were no significant survival differences in patients with large CAA (≥8 mm) compared with those with medium CAA (≥6 to <8 mm) at 30-years follow-up (large: 94% [95% CI, 84%–98%], medium: 96% [95% CI, 77%–99%]). No deaths were observed in patients with small CAA (<6 mm).³⁶ 

Risk factors that may reduce survival in the KD population with CAA include the presence of medium or large CAA,³⁶  multiple CAA,³⁵  previous history of cardiac events,³⁴  and poor cardiac function.³⁴ 

Fifty-five studies evaluated cardiac outcomes in the KD population. The most commonly reported outcomes were MI (38 studies, 69%) and cardiac interventions (29 studies, 53%) (Supplemental Table 11). Twenty-one of 55 studies reported MACE as a composite outcome, of which 17 reported MACE-free survival over time. Of the studies that reported on chronic cardiac diseases, 10 (19%) reported on hypertension, 11 (20%) on hyperlipidemia, and 10 (19%) on vascular markers, suggesting early atherosclerosis.

Twenty-four (63%) of 38 studies evaluated outcomes exclusively in patients with CAA or a history of CABG; MI prevalence ranged from 0% to 35%. This is in contrast to studies that included patients regardless of CAA, where MI prevalence ranged from 0% to 7%.

Three population-based studies, which included patients with KD, regardless of CAA, reported lower MI events compared with the remaining studies.^41–43  Chiang et al reported an MI prevalence of 0.2% in 8148 Taiwanese patients.⁴¹  Holve et al compared cardiac outcomes in 546 KD patients with 2218 non-KD comparators. After an average follow-up of 15 years, the age-adjusted incidence rate of MI was 0.3 per 1000 patient-years in the KD group, compared with 0 in matched comparators.⁴²  In a US population-based study, Coon et al reported 113 patients with MI from 17 809 patients with KD (0.6% prevalence) over a median follow-up time of 8 years.⁴³ 

Seventeen studies reported on the proportion of the population that remained free from MACE over time (Supplemental Table 12). Composite outcomes differed, but typically included MI, cardiac interventions, and cardiac-specific death. With the exception of 1 study, all studies were in patients who developed early cardiac complications after KD, mostly CAA.⁴⁴  MACE-free survival estimates ranged from 72% to 100% at 5 years,^{33,39,45–48}  66% to 91% at 10 years,^{32–34,37,44–46,49–51}  57% to 84% at 15 years,^38,39  29% to 74% at 20 years,^{33,34,37,44,52}  25% to 68.5% at 25 years,^33,34,38,39  and 36% to 96% at 30 years^35,36  (Fig 3).

Several studies evaluated MACE according to risk factors. Friedman et al reported that the size of CAA at diagnosis (for each unit increase in z-score) was associated with an increased odds ratio of 1.1 for MACE.⁵³  Similar findings were reported in 3 other studies which stratified patients according to CAA size.^36,49,54  Chih et al suggested that the progressive increase in CAA size is a risk factor for reduced MACE-free survival.³⁷  Tsuda et al reported that the MACE-free survival was worse with bilateral large aneurysms compared with those with unilateral large aneurysms (21% vs 59%).³⁵ 

Of 10 studies that reported on hypertension in the KD population, 1 study reported an increased risk and the remainder demonstrated no difference (Fig 4; Supplemental Table 13). The majority of these studies (67%) were cross-sectional. Novelli et al reported a hypertension prevalence of 44% in a cohort of patients, but only 2 of 8 patients developed persistent hypertension after the acute phase.⁵⁵  Holve et al compared the prevalence of hypertension in a KD cohort to age- and sex-matched comparators and found a significant difference in hypertension prevalence at baseline (3% vs 2%), although the incidence rate of hypertension was not evaluated.⁴² 

Studies on hyperlipidemia showed mixed results (Fig 4; Supplemental Table 13). Two studies concluded an increased risk compared with healthy comparators,^56,57  5 reported no difference,^58–61  and 4 were inconclusive.^17,62–64  All studies reported on biochemical variations rather than assessing for clinically- relevant hyperlipidemia.

Of the studies that support increased risk, Newburger et al measured lipid profiles in 105 patients after KD and determined a significantly lower high-density lipoprotein (HDL) cholesterol concentration throughout follow-up compared with reference values.⁵⁶  Patients with KD also had significantly lower HDL cholesterol concentrations compared with their nonexposed parents. Similarly, Cheung et al reported significantly lower HDL values in KD children with CAA compared with healthy age-matched children (1.3 vs 1.5 mmol/L).⁵⁷ 

Early atherosclerosis was evaluated via surrogate markers, including endothelial dysfunction and arterial wall thickness or stiffness (Supplemental Table 13). The most commonly used noninvasive sonographic techniques included flow- mediated dilatation (change in the brachial artery during reactive hyperemia),⁶⁵  pulse wave velocity (speed of arterial pressure waves),⁶⁶  and intima-media thickness (thickness of the carotid artery).⁶⁷  Of 10 studies, 7 studies concluded an increased risk, 2 reported no difference, and 1 was inconclusive (Fig 4). All studies compared patients with KD to healthy nonexposed patients. Mean duration from KD diagnosis ranged from 2 to 24 years. In addition, 3 of 10 studies stratified patients with KD according to CAA status; 1 study reported a statistically significant difference in endothelial dysfunction and 1 study reported a statistically significant difference in arterial stiffness in patients with CAA compared with those without.^57,59,64 

Twelve studies were identified; 6 (50%) focused on allergic diseases (asthma, atopic dermatitis, allergic rhinitis, urticaria, allergic conjunctivitis), 2 on cognition and behavior, 1 on malignancy, 1 on health-related quality of life, 1 on exercise performance, 1 on obstetrical outcomes, and 1 on respiratory infections (Table 2 ;Supplemental Table 14).

All studies of allergic diseases reported an increased risk in the KD population, although there were variations on the type of allergic disease that reached statistical significance. Three studies^68–70  reported an increased risk of allergic rhinitis, while 1 study⁷¹  reported no difference. Of the 3 studies that reported an increased risk of allergic rhinitis, the odds ratio ranged from 1.2 to 2.4 (69–71). Two studies^71,72  reported an increased risk of atopic dermatitis, whereas 2 studies^68,70  did not. Kuo et al⁶⁹  reported a significantly increased risk of asthma with an adjusted hazard ratio of 1.5 (95% CI,1.2–1.9), whereas 3 studies did not.^68,70,71 

Yu et al demonstrated in a Taiwanese population-based retrospective cohort study of 3469 patients with KD that the incidence of cancer was increased compared with population comparators, with a standardized incidence rate of 2.9 (95% CI, 1.6–5.2). A total of 11 malignancies were observed. When stratified by follow-up time, the statistically significant increase in cancer risk was evident only beyond 6 years after diagnosis. This was the only study that evaluated malignancy risk in the KD population.

Studies examining the remaining noncardiovascular outcomes, including behavioral, obstetric, and infection outcomes, reported no increased risk.

This systematic review included 74 studies. Overall, survival appears largely unaffected after KD, particularly in those without CAA. Patients who developed CAA appear to be at higher risk for major cardiac events. Few studies evaluated chronic cardiac and noncardiac comorbidities in the population, with mixed or limited results.

The majority of studies demonstrated >90% survival in patients, up to 30 years from KD diagnosis. Given that most studies selected patients with KD and cardiac complications, it is reasonable to assume that reported mortality values are heavily skewed to patients with more severe presentations. Thus, we conclude that the overall survival of the entire KD population, including those without CAA, is likely higher than reported. Indeed, the most notable series of publications based on 1 large Japanese cohort determined that mortality in all patients with a previous diagnosis of KD (with and without CAA) after 23 years of follow-up remained similar to the general population.^24–30 

Numerous studies evaluated MACE-free survival after a KD diagnosis. There was significant heterogeneity in the definitions of MACE among studies, rendering comparisons challenging. As expected, there was more variability in the results, with survival ranging from 36% to 96% at 30 years follow-up. Again, the majority of studies included only children with CAA. Considering that treatment has reduced the incidence of CAA from 25% to 4%, focusing on cardiac outcomes in the larger KD population without CAA is an area of much-needed research that is of high relevance.¹ 

The evaluation of risk factors that impact cardiac prognosis remains limited. Although some studies assessed the characteristics of CAA as risk factors for future MACE (such as size and number of aneurysms), no study to date has evaluated the impact of noncardiac risk factors on future cardiac prognosis. Potential risk factors that could be explored in future studies include those that are associated with higher rates of treatment resistance, including demographic features, type of KD, and biochemical abnormalities.⁷³ 

It is unclear whether our conclusions about the long-term cardiac prognosis are generalizable to all patient populations. The majority of the studies included were from East Asia. Given that studies have previously demonstrated significant variations in both treatment responses and risk factors for treatment resistance according to different ethnic populations, it is possible that ethnicity could also influence cardiac prognosis.^73,74  Further evaluation of the long-term cardiac prognosis in non-East Asian patient populations is needed.

Our study summarizes the conflicting results when evaluating outcomes such as hypertension, hyperlipidemia, and vascular markers of early atherosclerosis. Most studies concluded no increased risk in hypertension. Hyperlipidemia studies are particularly unclear. A limitation to the majority of hypertension and hyperlipidemia studies is that they often compared blood pressure or biochemical lipid panels among patients with and without KD at one point in time. Furthermore, most studies focused on numerical derangements as opposed to an emphasis on clinical outcomes, such as the number of patients requiring treatment of clinically relevant hypertension or hyperlipidemia.

Our study findings suggest a possible increased risk in the vascular markers associated with early atherosclerosis. Most studies confirmed increased endothelial dysfunction and arterial wall stiffness in patients with KD. However, the actual clinical relevance of these changes, particularly in asymptomatic youth, is not entirely clear. Studies focusing on the prognostic significance of these sonographic derangements report an increase in cardiac events over follow-up time, but primarily in older adults where the baseline incidence rates of cardiac events are higher.^75–77  A Rotterdam study demonstrated that an increase in aortic pulse wave velocity predicted cardiac events in 2835 asymptomatic adults over a 4-year observation period, but the mean age of their cohort was 72.⁷⁸  It is also possible that our conclusions may be influenced by publication bias, where studies with negative results are not published.

Finally, our study highlights the small body of literature that evaluated noncardiac outcomes in the KD population. The evidence supports that allergic diseases, including atopic dermatitis, asthma, and chronic urticaria, are more likely increased after KD exposure. This appears to be consistent with small observational studies which report inflammatory skin eruptions, including urticarial and psoriasis-like eruptions, occurring during or shortly after KD.^79,80  Although all 6 studies reported an increased risk of overall allergic diseases, the studies differed in terms of the specific type of allergic disease that patients were at risk for. We also found one study that concluded an increased risk of malignancy. An increased association with malignancy would be of clinical importance; additional studies are needed to evaluate this relationship. Our search strategy did not discover any studies that evaluated the association between chronic autoimmune diseases and KD exposure. Although there have been case reports of patients with KD who were eventually diagnosed with rheumatic conditions, further evaluation is required.^81,82 

Our conclusions must be considered in light of several possible limitations. Statistical pooling was not applied because of the heterogeneity of the data. Examples of heterogeneity included varying definitions of CAA and MACE. We employed a best-evidence synthesis and methodological quality was assessed. However, studies were not removed according to their methodological rigor. The quality of the studies was modest (median NOS: 6, median JBI: 6). Most case-control and cohort studies lost quality points because the KD cohorts were highly selective, the comparator groups were not properly defined, and the completeness of follow-up was not evaluated. Cross-sectional studies lost points because of insufficient details of how the study population was derived and whether confounder adjustments were performed. The majority of studies did not compare outcomes to a non-KD comparison group. Thus, very few studies generated a clear magnitude of the excess risk after KD.

While we attempted to be comprehensive when identifying noncardiac outcomes associated with KD, we may have missed outcomes for several reasons. First, our study may have missed noncardiac outcomes given that we intentionally selected studies with an average follow-up of ≥1 year. This selection criterion was decided upon a priori to ensure that this study’s objective was focused on the long-term prognosis. As a result, we recognize that some noncardiac outcomes may have not been captured, such as sensorineural hearing loss, a known chronic complication associated with KD.^16,83  Second, because we were interested in all noncardiac outcomes associated with KD, the breadth of our objective may have been difficult to capture within the search strategy. We attempted to identify as many studies as possible by completing a literature review and consulting KD experts of potential noncardiac outcomes that are of interest or importance. Finally, publication bias may have limited our ability to identify studies that explored novel noncardiac outcomes that did not demonstrate increased risk.

Long-term survival does not appear to be significantly impacted in most patients with a previous history of KD. Although studies suggest ongoing cardiac morbidity in KD patients with CAA, it remains unclear whether patients without CAA are at increased risk for cardiac disease later in life. Our knowledge of the predictors for future cardiac disease in patients with KD remains poor. Patients with KD may be at increased risk for noncardiac conditions such as allergic diseases, but more confirmatory studies are needed.

CAA

coronary artery aneurysms

CABG

coronary artery bypass graft

CI

confidence interval

HDL

high-density lipoprotein

ITA

internal thoracic artery

JBI

Joanna Briggs Institute

KD

Kawasaki disease

MACE

major adverse cardiac events

MI

myocardial infarction

NOS

Newcastle-Ottawa scale

SMR

standardized mortality ratio