Vaccination and Risk of Childhood IgA Vasculitis

We prospectively enrolled consecutive children (<18 years old) with incident IgAV consulting from July 2011 to March 2016 in the Pediatric Rheumatology Department of Bicêtre University hospital. We also enrolled children with incident IgAV consulting from 2012 to 2014 in 3 pediatric hospitals located in Val-de-Marne, a suburb of Paris, who participated in an incidence study of IgAV in Val-de-Marne²³ and those hospitalized in the Pediatric Nephrology Department of Trousseau University Hospital, which is specialized in managing IgAV with severe renal involvement. The latter patients were identified through the computerized hospital discharge database. All IgAV diagnoses were verified by European League Against Rheumatism, Paediatric Rheumatology International Trials Organization, and Paediatric Rheumatology European Society criteria for IgAV. The criteria were purpura with lower-limb predominance plus 1 of 4 criteria: acute diffuse abdominal pain, leukocytoclastic vasculitis or proliferative glomerulonephritis with predominant immunoglobulin A deposits, arthritis or arthralgia of acute onset, and renal involvement defined by proteinuria >0.3 g per 24 hours (>30 mmol/mg urine albumin/creatinine ratio) or hematuria ≥2+ on dipstick testing.²⁴ 

All children with available vaccination records were included. In France, parents are given health booklets detailing the vaccinations of each child. The appropriate booklet pages were copied during hospital consultations or hospitalizations; when this information was not immediately available, parents were contacted by phone and asked to forward copies by e-mail, fax, or postal mail. For children hospitalized in the Pediatric Nephrology Department of Trousseau University Hospital, vaccines administered and vaccination dates were obtained from hospital discharge summaries, which routinely included this information.

For each patient, a study-specific case report form was completed by the emergency physician, the treating hospital pediatrician, or the principal investigator (M.P.) with the following information: demographics (date of birth, sex), date of IgAV onset, vaccination history during the year before the disease onset (dates and types), and the main clinical and laboratory findings. The case report form also included information on whether the child had an infection during the 3 weeks before IgAV onset and the type of infection, if applicable.

The risk of developing IgAV postvaccination was analyzed by case-crossover analysis, a variant of a traditional case-control study, in which each case serves as its own control. The method consists of comparing, for a given individual, exposure to a potential risk factor during an “index period” immediately before the disease onset and exposure during 1 or several previous “control” periods.^25,26 As is true for other case-only methods too, case-crossover analysis is particularly well adapted to study the effects of transient exposures (eg, vaccination) on acute disease (eg, IgAV), and it has the advantages of eliminating potential confounding linked to different characteristics of control subjects and facilitating participant recruitment by including only cases.

In our setting, the date of IgAV onset was the date of the first symptom associated with the disease. An index or control period was considered exposed if at least 1 vaccine was administered during that period. On the basis of previously reported cutoffs and possible pathogenic mechanisms, we considered that vaccination could trigger IgAV within a maximum of 3 months.²⁷ By using the history of vaccinations given during the 12 months before IgAV onset, each patient’s index period corresponded to the 3 months before IgAV onset, and control periods were the 3 consecutive 3-month periods before that index period. From the results obtained, the risk of IgAV attributable to vaccination in the 3 months after vaccination was calculated as Pe (OR − 1)/1 + Pe (OR − 1), where Pe is the frequency of vaccine exposure during the 3 months before IgAV onset, and OR is the estimated ratio for the link between vaccination and IgAV onset.

The robustness of the primary analysis was tested by sensitivity analyses in which researchers examined different exposure periods: 2 months (ie, 1 index and 5 control periods), 1.5 months (1 index and 7 control periods), and 1 month (ie, 1 index and 11 control periods). Furthermore, to identify potential high-risk populations or confounding by time-varying factors, we computed ORs stratified by the season of IgAV onset, the year of onset, the presence of an infectious episode during the 3 weeks before the onset, sex, and age (<5, 5–10, and >10 years). To identify a potential cumulative effect of vaccinations on IgAV risk, an additional subgroup analysis involved the number of vaccinations administered during the 12 months before IgAV onset (1 or >1). To study whether IgAV risk was associated with a specific vaccine, we also calculated ORs stratified by vaccines grouped into 4 categories (against tetanus-diphtheria–inactivated polio vaccine [Td-IPV] with or without acellular pertussis, hepatitis viruses, meningococci, or MMR); other vaccines were not considered in this analysis because they were poorly represented.

The sample size required for the study was calculated by using the formula for paired case-control studies because of a lack of an equivalent formula for case-crossover studies.²⁸ With an estimated 5% vaccine exposure rate for a 3-month period (which corresponds to the French Public Health Council recommendations for Td-IPV boosters every 5 years and potential other vaccines²⁹), a 3:1 control/index period ratio, 80% statistical power and 2-tailed α level of .05, a minimum detectable OR of 2.5, and a null-exposure correlation factor between the index and control periods, we estimated that 160 cases had to be enrolled. The correlation coefficient value was a priori estimated to be null by assuming no correlation of probability of vaccination between 3-month periods.

All analyses involved use of Stata 12 (Stata Corp, College Station, TX) and R 3.2.3 (R Project for Statistical Computing, Vienna, Austria). Categorical variables were compared by χ² or Fisher’s exact test or, with ordered variables, by χ² test for trend; for comparisons of observed counts with expected counts (eg, analysis of seasonal distribution of IgAV cases), χ² goodness-of-fit test was used. Quantitative variables were compared by Student’s t test. The results of case-crossover analyses were computed by conditional logistic regression, estimating ORs and 95% CIs. Two-tailed P < .05 was considered significant.

The study protocol was approved by the Advisory Committee on Information Processing in Research in the field of Health (number 12-146bis) and the National Committee for Informatics and Freedom (number 914148).

Among the 193 patients identified (180 [93%] prospectively and 13 [7%] retrospectively), 167 (87%) had complete information concerning the dates and types of vaccines administered during the year before IgAV onset. In total, 42 (25%) patients had received at least 1 vaccine during the year preceding IgAV onset (Table 1): 32 (76%) patients received 1 vaccine, 8 received 2 (19%), and 2 received 3 (5%). Among the 10 patients given several vaccines, 6 were vaccinated on the same day and 4 on different days, with the interval between 2 vaccinations ranging from 5 days to 4 months. Therefore, the 42 vaccinated children received a total of 54 vaccines on 47 distinct dates.

As compared with nonvaccinated children, vaccinated children were older at IgAV onset and differed in distributions of seasons and calendar years of IgAV onset (Table 1). Both subgroups were similar when considering other clinical features and had a similar proportion of infectious episodes during the 3 weeks before IgAV onset (Table 1).

In Fig 1, we show the monthly distribution of the 47 days of vaccinations among the 42 vaccinated children and the monthly distribution of IgAV onset among all 167 children. The monthly and seasonal distributions of the 47 vaccination dates (spring: 13%; summer: 32%; fall: 28%; winter: 28%) did not differ significantly from theoretically expected frequencies (P = .88 and .26, respectively); in contrast, the monthly and the seasonal distribution of IgAV onset (spring: 26%; summer: 13%; fall: 25%; winter: 35%) significantly deviated from expected frequencies (P = .006 and < .001, respectively).

Among all children, 15 (9%) had been vaccinated during the 3-month index period as compared with 4% to 7% during the control periods. In Fig 2, we show the distribution of vaccinated children with IgAV in terms of the interval between the 47 vaccination dates and IgAV onset. It illustrates a progressive increase in the proportion of vaccinated children when considering the 3 consecutive 3-month control periods before disease onset, which was significant when assessed on a quarterly (P_trend = .049) but not a monthly basis (P_trend = .377). In Table 2, we show the distribution of the 54 vaccines in the index and control periods.

In Table 3, we summarize case-crossover analysis results for risk of IgAV onset within 3 months of vaccine exposure. The OR associated with any vaccine exposure within 3 months before IgAV onset was 1.6 (95% CI: 0.8–3.0). On the basis of that estimate and given that 9% of study participants were exposed to a vaccine during the 3 months preceding IgAV onset, the vaccine-attributable risk of IgAV within the following 3 months was estimated at 5% for our population. In terms of vaccine-specific analyses, for combined vaccines against Td-IPV with or without pertussis, hepatitis viruses, meningococci, and MMR, the ORs (95% CIs) were 1.7 (0.8–3.9), 1.7 (0.3–8.3), 1.1 (0.3–4.2), and 2.0 (0.3–12.0), respectively. Sensitivity analyses for 1-, 1.5-, and 2-month index and control periods revealed approximately similar associations for exposure to any vaccine, with ORs (95% CIs) of 1.4 (0.5–3.5), 1.4 (0.6–3.2), and 1.3 (0.6–2.6), respectively.

The results of case-crossover analyses stratified by children or IgAV characteristics revealed no significant association between vaccination and IgAV onset within 3 months (Table 4). However, in these analyses, stronger associations, although not statistically significant, were indicated for children <5 years old (OR: 1.8; 95% CI: 0.4–7.5), boys (OR: 2.1; 95% CI: 1.0–4.8), and winter onset (OR: 3.0; 95% CI: 0.6–14.9). For children <5 years old, the OR associated with vaccination and IgAV onset within 3 months remained nonsignificant, whether an infectious episode occurred during the 3 weeks preceding IgAV onset or not (OR: 2.0; 95% CI: 0.33–12.0 and OR: 1.5; 95% CI: 0.1–16.5, respectively).

In the results of our case-crossover study, it is indicated that the vaccines most frequently administered to children (especially Td-IPV; diphtheria, tetanus, acellular pertussis–inactivated polio vaccine; and meningococcal and hepatitis vaccines) did not significantly increase the risk of IgAV onset during the 3 months postvaccination. Moreover, findings were similar in sensitivity analyses in which we investigated shorter risk periods.

Vaccination is a major preventive measure against infectious diseases, with an estimated avoidance of 2 to 3 million deaths annually.³⁰ However, vaccination programs are regularly associated with the fear they may induce immune-mediated diseases, and such polemics negatively affect public opinion concerning the safety of vaccines.^31,–35 The mechanisms suspected in the triggering of immunologically mediated illnesses by vaccines essentially involve molecular mimicry with self-antigens or bystander effects of adjuvants resulting in activation of the innate immune system.³⁶ Therefore, postmarketing evaluation of the tolerability of vaccines with respect to rare adverse events is essential. Such assessments rarely rely on cohort studies, which are difficult to use for investigating rare adverse effects linked to vaccination. Traditional case-control studies can be difficult to set up in countries with high vaccine coverage and are at risk for bias because the rare unvaccinated children can be so because of multiple reasons.³⁷ 

Among the several methods available based on case-only designs,^38,39 we chose case-crossover analysis, which has been previously used to investigate the link between vaccination and the occurrence of diseases.^40,41 As compared with self-controlled case series, case-crossover analysis has the advantage of avoiding “reverse bias,”³⁹ which would have occurred in the likely scenario of a diagnosis of IgAV negatively influencing the probability of vaccine exposure in the postevent period. The main weakness of case-crossover analysis is that it implicitly supposes that vaccine exposure remains constant over time during control periods. Thus, although we did not anticipate such a “time trend bias” in light of the vaccination schedules for children aged 2 to 16 years, we observed an increasing trend in vaccine exposure over time (Fig 2). In turn, this unexplained scenario implies that the overall OR value of 1.6 might even represent an overestimation of the true association between vaccine exposure and IgAV onset. Whether the use of a different case-only design might have produced different results is hypothetical, and in simulation studies, it was suggested that either technique may result in similar estimate ranges.⁴² The risk of misclassification or memory bias seems highly unlikely because both vaccination and IgAV onset are datable and because vaccines administered were recorded in each child’s vaccination booklet.

The interpretation of our findings also hinges on the choice of the interval during which a vaccine could induce an immune-mediated disease. That difficulty is reflected by the variability of the at-risk “windows” of exposure considered in the studies.⁴³ For example, a 42-day index period was chosen in a study of MMR vaccination and risk of aseptic meningitis,³⁷ a 3-month period was chosen for a case-control study of vaccination and IgAV,²² and a 2-month period was chosen for a study of vaccination and risk of relapse in multiple sclerosis, with sensitivity analyses for 1- and 3-month periods.⁴⁰ In contrast, a 2-year period was used in a longitudinal exposed and unexposed study of human papillomavirus vaccination and onset of 14 immune-mediated diseases in female adolescents.⁴⁴ For IgAV, an acute immune complex–mediated condition, the choice of intervals of a maximum of 3 months for onset after vaccination seemed appropriate.

Our results cannot exclude a small increase in IgAV risk because the study was designed to have 80% power to detect an OR ≥2.5 of risk of IgAV onset within 3 months after vaccination. However, if any, small excess risks are difficult to separate from potential confounding factors.⁴⁵ In the same line, our observations of the moderately higher OR in children <5 years old, in IgAV with onset in the winter, in 2015 and thereafter, and in boys ought to be viewed with caution because they refer to subgroup analyses with at best borderline statistical significance. Also, these IgAV characteristics were rather under- than overrepresented among those vaccinated as compared with unvaccinated children. In addition, low vaccine-attributable risk of IgAV would have a minor impact at the population scale. Under the hypothesis of a causal relationship and considering the result of the main analysis (OR: 1.6), the attributable risk of IgAV to any vaccines performed within the 3 months would be 5% in our population. This rate, related to an annual disease incidence of 20 to 30 out of 100 000 children,²³ would indicate an absolute risk of IgAV developing in the 3 months after a vaccination of 1 to 1.5 out of 100 000 children vaccinated per year. This result puts into perspective the risk/benefit ratio of vaccination compared with a potential risk of vaccine-triggered IgAV, which is in most cases a self-limiting disease.

Because we investigated vaccination in general, it was not designed to address the potential role of specific vaccines. As in the study published by Da Dalt et al²² (involving 288 cases and 617 controls), we found a higher OR for MMR vaccination, a live vaccine that has been linked to immune thrombocytopenic purpura.^46,47 Thus, the potential association of MMR vaccination with IgAV needs cautious interpretation because our corresponding estimate did not reach statistical significance and because in both studies by Da Dalt et al²² and us, the children who were vaccinated against MMR within defined risk periods accounted for <3% of the analyzed sample. Our study also does not provide insight into the risks associated with vaccines (eg, influenza or human papillomavirus) that are not commonly administered to children in the age range of the highest IgAV incidence. Altogether, the small numbers of children exposed to such vaccines challenge whether they may play an important role in the etiology of pediatric IgAV.

Our study’s strengths include the large number of children with well-characterized IgAV and the prospective, population-based recruitment of most cases, independent of their vaccination status. Although most children were recruited from hospitals, which in our opinion reflects the management paths for IgAV in France, we believe our results ensure generalizability to overall childhood IgAV because patients followed in community settings versus hospitals do not likely differ in terms of pre-IgAV vaccination. Its limitations are the too-small number of children to estimate the risk of IgAV for a given vaccine and for robust subgroup analyses, although we think strong indications were not generated in our study for potential vaccine- or subgroup-specific safety issues.

The results of our study are reassuring with regard to the risk of IgAV onset during the 3 months after vaccination in children. The importance of vaccine safety in terms of public health may justify further monitoring of the relation between vaccines and IgAV risk. Also, more studies are needed to enhance our understanding of the role of other environmental factors in the development of IgAV.

CI

confidence interval

IgAV

immunoglobulin A vasculitis

MMR

measles-mumps-rubella

OR

odds ratio

Td-IPV

tetanus-diphtheria–inactivated polio vaccine