We found no evidence to support the hypothesis that childhood vaccine-exposure triggers HAEs in children with UCDs. Given their vulnerability to catabolic decompensation and HAEs in the setting of fever or acute inflammatory or infectious response, it is particularly important to avoid vaccine-preventable diseases in patients with UCDs. Thus, we conclude that the known preventive benefits of immunization likely outweigh any residual theoretical risk of triggering HAEs in these patients.
The UCD registry is an unprecedented resource for the study of the safety of vaccines in vulnerable patient populations, and our study is, to our knowledge, the largest controlled comparison published on this topic. Given the challenges of assembling sufficient sample sizes of patients with rare diseases, the UCD registry represents a successful model that may potentially be extended to the study of vaccine safety in other groups of patients with IEM. The current practice of assessing vaccine safety within the voluntary Vaccine Adverse Event Reporting System is informative but only if IEM are cited as potential predisposing factors on reporting forms. To compare our findings with information in this voluntary reporting system, we researched the past 5 years of their database and found no citations of UCDs in relation to adverse events after vaccines. Likewise, we are unaware of any deaths in patients with UCD occurring in the 2009–2010 H1N1 flu pandemic, although there is no systematic method for reporting such occurrences in patients with UCDs. The extent to which H1N1 immunization was effective at preventing HAEs is unknown, and should be a topic for future research.
This was a cross-sectional study, and capture was estimated at 27% of all eligible patients, the highest recruitment rate for any disorder in the Rare Diseases Clinical Research Consortium.8
Although the general limitations of nonrandom sampling apply to the present study, recruitment was unrelated to the question of vaccination as a possible trigger for HAE, and we do not perceive a substantial potential for bias. One additional limitation of our study design was that immunization data were not prospectively collected and some vaccination records were obtained from primary care providers, if tertiary care hospital records were lacking. Patients with UCDs are highly reliant on metabolic specialty centers; in the experience of the UCDC, it is rare for any patient to be lost to follow-up. Therefore, our clinical ascertainment of HAE is considered to be essentially complete, although it was not possible to document complete vaccination exposures on all patients. We considered the possibility that missing vaccine data could potentially represent a systematic bias against the hypothesis that HAEs could be triggered by vaccines. However, the UCDC was focused on clinical management of UCDs, and when children experienced an HAE, recent vaccine exposure was specifically elicited by study coordinators. Therefore, we would anticipate that any resulting vaccine-exposure ascertainment bias would likely favor temporal correlation of recent vaccines with HAEs, which we did not observe.
For the primary statistical analysis, we relied on the SCCS method, which is considered to be the optimal method for relating recurrent time-dependent events.16,17
The SCCS method assumes variability in the timing of events and exposures. In the neonatal period, this assumption is almost uniformly violated, in that newborns with UCDs receive first exposure to milk protein at essentially the same time as their first HepB vaccine exposure. Therefore, vaccines administered in the immediate neonatal period are essentially uninterpretable by the SCCS method. Although a temporal association between HepB vaccination and HAE is possible, age-adjustment of the Poisson model adequately corrected for the known age effect, and importantly, restricting the analysis to >30 days of age found no significant association even without further age adjustment. In addition, on the basis of our sensitivity analysis of the Poisson count model to HAE clustering, we found that such clustering did not bias our results. We also considered the possibility that physicians may have been less likely to administer scheduled vaccinations to metabolically unstable children who had recently experienced an acute HAE. We acknowledge that our study was not designed to assess the safety of vaccination of recently ill, metabolically unstable children, and advocate that physicians consider waiting ~21 days before administering vaccines unless the risks of vaccine deferral seem to outweigh theoretical concerns about triggering an HAE recurrence.
Our study is the largest that we know of to have investigated an association between vaccination and HAE, and our posthoc estimates indicated 80% power to detect risks less than twofold. Indeed, the expected elevation in relative incidence of HAEs stemming from the diagnosis in the newborn period coinciding with the time of HepB vaccination was readily detected in our study, demonstrating adequate power for risks of 1.5 and 1.9 (). Such apparent risk increases were statistically significant before age adjustment, although the newborn period represents only a small fraction of total observation time, HAE, and vaccine exposures. In addition, given 371 HAEs and 7% of observation time spent within 21-day postvaccination windows, we estimated that our study had 80% power to detect a 1.8-fold risk increase for HAEs caused by vaccination exposure, even when conservatively incorporating into our model a decreasing age effect (ie, decline in nonvaccine-associated risk of hyperammonemia) from birth to older ages) that is known to reduce power substantially. Therefore, we do not consider a type II (false-negative) error to be a likely explanation for our overall null findings. Theoretical small risks of vaccinations not detectable in this study must be weighed against the known benefits of childhood immunizations.