Kawasaki Disease Following the 13-valent Pneumococcal Conjugate Vaccine and Rotavirus Vaccines

The VSD is a collaboration between the CDC and 9 integrated healthcare organizations that is used to conduct surveillance and epidemiologic research on vaccine safety.¹⁴  The VSD population encompasses approximately 12 million persons per year. VSD electronic data includes patient demographics, inpatient and outpatient encounters, and immunizations and can be supplemented with medical record review. Large cohorts of children can be followed while enrolled in a VSD site, and these data can be used to examine potential associations between vaccines and rare adverse events. For the current study, data were collected from 8 sites: Kaiser Permanente Northern California (CA), Kaiser Permanente Southern California (CA), Kaiser Permanente Washington (WA), Kaiser Permanente Colorado (CO), Denver Health (CO), Marshfield Clinic (WI), Kaiser Permanente Northwest (OR), and Health Partners Institute (MN). The study population consisted of children <24 months of age who were born between January 1, 2006 and December 31, 2017 (RV5, RV1, and PCV13 were introduced in the United States in 2006, 2008, and 2010, respectively). For inclusion in the study, children were required to have continuous enrollment from ≤42 days of life through either the index KD event or 23 months of age.

We conducted a SCRI analysis to compare the occurrence of KD between a risk interval 1 to 28 days postvaccination and an equal length 28-day comparison (control) interval following the risk interval (Fig 1).^15–17  SCRI analysis was chosen for the analytic method since it can inherently account for several important potential confounders including sex, VSD site, and season.¹⁸  The day of vaccination was day 0. The control intervals were days 29 through 56 for doses 1 and 2 (RV1, RV5, and PCV13) and days 43 to 70 for dose 3 (RV5 and PCV13) and dose 4 (PCV13) based on prior literature.^13,19  The latter control interval (43–70 days postvaccination) was preferable to avoid misclassification of the outcome timing caused by uncertainty about the true period of the risk associated with vaccines.¹⁹  However, it is recommended by the CDC that PCV13 be given at 2, 4, 6, and 12 to 15 months of age, RV5 be given at 2, 4, and 6 months of age, and RV1 be given at 2 and 4 months of age.²⁰  As such, the control interval for doses 1 and 2 was needed to fit into a 2-month period to avoid overlapped intervals with a subsequent dose. Vaccines were identified using CVX codes for PCV13, RV1, and RV5.²¹  We examined KD cases following any PCV13 vaccination, any rotavirus vaccinations, simultaneous PCV13 and rotavirus vaccinations, PCV13 vaccination without simultaneous rotavirus vaccination, and rotavirus vaccination without simultaneous PCV13 vaccination. Chart-confirmed KD cases were assigned to risk or control intervals based on the timing of their adjudicated symptom onset. Cases were excluded if the risk or control interval of 1 dose overlapped the risk or control interval of a subsequent dose.

Cases of KD were identified by the International Classification of Diseases, Ninth Revision (ICD-9) code 446.1 (Kawasaki disease) and ICD-10 code M30.3 (acute febrile mucocutaneous lymph node syndrome) assigned in the inpatient setting. The diagnosis was limited to the inpatient setting for specificity as true KD cases are severe and require close monitoring and treatment with high-dose intravenous immunoglobulin (IVIG), which cannot be administered in outpatient settings. For children with multiple KD medical encounters, only the first inpatient visit was included. We conducted manual chart review of cases identified by ICD codes for KD that fell into the risk and control windows to confirm the diagnosis. Because KD symptom onset may precede the ICD code-based diagnosis date, it is possible that some cases with onset during the risk or control windows may be missed. To account for this, KD cases identified by ICD codes that occurred during a risk or control interval as well as cases identified during expanded intervals (ie, 70 days following dose 1 or 2 and 84 days following dose 3 or 4) were also selected for manual chart review (Supplemental Fig 4). We calculated the confirmation rates of this code-based algorithm for identification of incident KD.

The aim of medical record review was to confirm whether subjects had “complete KD” or “incomplete KD,” and to determine the onset date of symptoms. Verified complete KD cases met the CDC epidemiologic case definition for KD (Supplemental Table 3).^8,22  Although the CDC epidemiologic case definition establishes the diagnosis of complete KD, these criteria do not identify all children with the illness and will leave out incomplete KD. As long term cardiac outcomes in children with incomplete KD are comparable to those of children with complete KD, we also used the American Heart Association KD scientific statement to identify incomplete KD.^1,23  Verified incomplete KD was defined as a case has prolonged unexplained fever for ≥5 days (or fever until the date of administration of IVIG if it was given before the fifth day of fever), and 2 or 3 of the 5 principal clinical findings as described in Supplemental Table 3, and compatible laboratory or positive echocardiographic findings.¹  Only chart-confirmed complete KD and incomplete KD were included in the final analyses.

Only cases of KD identified by ICD codes that were confirmed by manual chart review were analyzed. Descriptive analyses were conducted on the study cohort, including by age, sex, race and ethnicity, and season of year when the case occurred (winter = January–March, spring = April–June, summer = July–September, fall = October–December). The rates of chart-confirmed KD with onset in the risk interval were compared with the rates of chart-confirmed KD in the control interval using conditional Poisson regression. As age is a time-varying confounder, we adjusted for the differential risk of KD according to age in each interval using offset terms calculated based on data on incidence of KD by age within the VSD cohort. Additionally, analyses specific to dose number partially accounted for the confounding effect of age.²⁴  As the risk and control intervals are narrow (28 days for each interval) and the incidence rates of KD in the United States were stable during the study period, calendar year of vaccination was not included in the models.²⁵  Finally, a temporal scan analysis was conducted to detect any temporal clustering of chart-confirmed KD cases within 70 days after PCV13 or rotavirus vaccines.^26,27  All statistical tests were 2-sided, and a P value of <.05 considered statistically significant. All data were analyzed using SAS version 9.4 (2019 SAS Institute, Cary, NC). The human subjects review board at each participating site and the CDC approved the study and determined informed consent was not required.

A total of 2 508 207 doses of PCV13 and 2 048 186 doses of rotavirus vaccines were administered to children <24 months of age who were born between January 1, 2006 and December 31, 2017. A total of 655 cases of KD were ascertained by ICD codes. Of these, 511 cases were excluded because they did not occur within the expanded intervals for manual chart review. Of 144 cases that underwent manual chart review, 122 cases were confirmed by manual medical record review, and the symptom onset of 97 cases were in the prespecified risk or control intervals (Fig 2). The chart confirmation rate was 85% overall and comprised 94 cases of complete KD (65%) and 28 incomplete KD (19%). An additional 13 cases (9%) did not meet the criteria for complete and incomplete KD and 9 (6%) were ruled out or lacked the necessary information to make a case determination.

Demographic characteristics of chart-confirmed cases are shown in Table 1. The mean age of children with chart-confirmed KD was 7.1 months. In addition, 63% were male and 40% were Asian. Overall, there was no evidence of an elevated risk of KD following PCV13 or rotavirus vaccines by dose in SCRI analyses using chart-confirmed KD cases (Table 2). In SCRI analyses of chart-confirmed KD cases receiving any PCV13 vaccine with or without simultaneous rotavirus vaccines, there were 29 cases in the risk interval and 41 cases in the control interval (the age-adjusted RR, 0.75; 95% confidence interval [CI], 0.47–1.21). Similarly, in SCRI analyses using chart-confirmed KD cases who received any rotavirus vaccine with or without simultaneous PCV13, there were 24 cases in the risk interval and 41 cases in the control interval (the age-adjusted RR, 0.66; 95% CI, 0.40–1.09). In the temporal scan analysis using the symptom onset date of 97 chart-confirmed KD cases, no statistically significant clusters of KD cases were identified during days 1 to 70 postvaccination (P = .18) (Fig 3).

In this study using US post-licensure vaccine safety monitoring surveillance data supplemented with standardized medical record review, we found no evidence of an increased KD risk following PCV13 or rotavirus vaccination in children under 2 years of age. As consistent with prior studies, males and Asian were more common in chart-confirmed KD cases in this study.^1,24  Although the rate of incomplete KD (23%) is higher than prior reports, this is likely because our KD cases are limited to children less than 2 years of age (the median age, 7.1 months), and incomplete KD is more common in infants.^1,28,29 

Although a prior study showed a nonstatistically significant increased point estimate risk of KD following PCV13, we did not identify a similar trend.¹³  Whereas the prior study used the VSD data between 2010 and 2012 (the number of vaccine doses of PCV13 included was 599 229), we expanded the study period through 2017, which increased the population size and resulted in over 2.5 million doses of PCV13 being included in the current study. As such, we were able to identify more KD cases following PCV13 vaccination than the prior study (70 cases vs 12 cases). Furthermore, we conducted a SCRI analysis with age-adjustment to account for important potential time-invariant and time-varying confounders, which reduces the risk of bias in our results. In fact, our results are consistent with the recent large studies on KD following PCV13 conducted in the Postlicensure Rapid Immunization Safety Monitoring program in the United States and in the Hospital Episodes Statistics system in the United Kingdom.^19,30  Compared with these prior studies, we used manual chart review with stringent criteria for all KD cases identified by ICD codes.

In contrast to a previous ICD code-based study in Taiwan showing a small increased risk of KD during the third week post dose 2 of RV5 (incidence rate ratio 2.33, 95% CI 1.35–4.00), and fourth week post dose 1 of RV1 (incidence rate ratio 1.98, 95% CI 1.16–3.40), we did not detect any risks for KD among rotavirus vaccine recipients.³¹  Although this may be, in part, caused by the different risk intervals analyzed, it is more likely because of the standardized chart review process we used, which made misclassification of case status less likely. As KD symptom onset could precede physician diagnosis, the ICD code-based results of the study from Taiwan could have been influenced by misclassification of the symptom onset date in relation to the risk or control intervals. In addition, our findings on the safety of rotavirus vaccines are consistent with the results of other small or ICD code-based studies.^32–35 

Our study has several important strengths. First, we conducted manual chart reviews for all KD cases identified by ICD codes. This is particularly critical as diagnosing KD can be challenging and can only be made based on clinical criteria, which may pose a risk of misclassification of the outcome without chart-confirmation. Second, we used the standardized case definitions based on the CDC and American Heart Association criteria, which included supplemental laboratory and echocardiographic findings. Therefore, our criteria for incomplete KD were more stringent than prior studies and provide more accurate classification of the outcome. Third, we used a SCRI analysis with age-adjustment using a large VSD cohort with comprehensive vaccination and follow up information.³⁶ 

This study is subject to limitations. First, we identified cases of KD only in the inpatient setting and we did not include KD cases which did not require hospitalization. However, omission of KD cases would have been rare given that patients with a KD diagnosis generally require inpatient care for high-dose IVIG treatment, which is infused for a much longer time (eg, 10 hours or more) than for other indications and requires close monitoring during and after the infusion. Second, another potential limitation of this study is bias from misclassification of the exposure. It is possible that some children in the study received vaccines outside their VSD health care system, which were not captured in the study databases. We expect this to be rare as we required continuous enrollment in the system up until the time of the diagnosis of KD. Third, although our sample size was relatively large, this study was not powered to conduct subanalyses for risk assessment by outcome type (complete versus incomplete KD) or vaccine brand type (RV1 versus RV5). In addition, the interpretation for the secondary findings by vaccine dose may be limited because of the relatively small number of cases for each dose. Finally, we did not adjust for simultaneous vaccines, such as diphtheria-tetanus-acellular pertussis. However, the majority of vaccines in this age group are given together with other vaccines in accordance with the Advisory Committee on Immunization Practices schedule, and these other vaccines have been previously shown to confer no increased risk of KD.⁸ 

Vaccines including PCV13 and rotavirus vaccines have prevented substantial morbidity and mortality in children and also have other indirect effects, including herd protection, decreasing the use of broad-spectrum antimicrobial agents, and decreasing antimicrobial drug resistance by reducing the circulation of multidrug-resistant bacterial pathogens.^37,38  However, public concerns about the safety of the immunization schedule threaten to undercut these substantial benefits. In particular, there is parental concern that the increasing prevalence of allergic diseases or autoimmune diseases may be associated with the increasing number of vaccines administered during early childhood.³⁹  As KD is thought to be an immunologic phenomenon, our findings of the safety of these vaccines at a young age provide additional reassurance to concerned families, healthcare providers, and the public.

Our extensive chart-confirmed SCRI study showed no evidence of an increased risk of KD following PCV13 or rotavirus vaccinations. These findings provide additional safety evidence of PCV13 and rotavirus vaccines to support the current US childhood immunization schedule.

We thank Vaccine Safety Datalink project managers for their contributions to overall project management, data managers for their contributions to overall data management, and medical record abstractors for their contributions to medical record review; and Joseph Y. Abrams, PhD (Centers for Disease Control and Prevention) for providing critical feedback for this study.

CDC

Centers for Disease Control and Prevention

CI

confidence interval

ICD

International Classification of Diseases

IVIG

intravenous immune globulin

KD

Kawasaki disease

PCV

pneumococcal conjugate vaccine

RR

relative risk

SCRI

self-controlled risk interval

VSD

Vaccine Safety Datalink