1 Introduction
New Zealand’s national coronavirus disease 2019 (COVID-19) vaccine and immunisation programme began vaccinations in February 2021 using the two-dose BNT162b2 messenger RNA (mRNA) COVID-19 vaccine (Pfizer-BioNTech, referred to as the BNT162b2 vaccine hereafter) [
1]. The BNT162b2 vaccine demonstrated acceptable efficacy and safety in phase III clinical trials [
2] and received provisional approval for use in adults aged 16 years and older from Medsafe, New Zealand’s Medicines and Medical Devices Safety Authority, on 3 February 2021 [
3]. This provisional approval was renewed on 28 October 2021, with additional conditions for use in adolescents aged 12 years and older [
4]. Provisional consent of a paediatric formulation of the BNT162b2 vaccine for use in children aged 5–11 years was also granted by Medsafe on 16 December 2021 [
5]. However, rare and potentially serious adverse events following immunisation (AEFI) are often impossible to detect in clinical trials due to limited sample size, narrow patient selection criteria, and constraints on the duration of the study [
6]. Indeed, although no cases of myocarditis or pericarditis were reported during the Pfizer-BioNTech phase III clinical trials [
2], these events were identified as rare adverse reactions to the BNT162b2 vaccine in postmarketing studies [
7‐
13]. As a result, both myocarditis and pericarditis were included in the New Zealand product information by the sponsor, Pfizer-BioNTech, in July 2021 at the request of Medsafe [
14].
Continued robust postmarketing vaccine safety surveillance is crucial to detect rare and unexpected vaccine reactions in a timely manner and provides information for risk–benefit assessments that can inform health policy decisions. This helps minimise the risks associated with serious adverse reactions and ensures that accurate and credible information regarding adverse effects is communicated outwardly to maintain public confidence and trust [
15]. In New Zealand, the safety of BNT162b2 was monitored predominantly through a spontaneous reporting (passive) system by Medsafe, in collaboration with the Centre for Adverse Reactions Monitoring (CARM), with support from the COVID-19 Vaccine and Immunisation Programme within the Ministry of Health New Zealand [
16]. This system relies on reports being voluntarily submitted by health care professionals and the public. Although it is effective at signal detection, it can be subject to several limitations, namely underreporting, incomplete reports, limited information on cases, and reporting biases. Furthermore, it is difficult to accurately estimate the incidence rate and attributable risk of adverse events in a defined population using this system [
17].
To address some of these shortcomings, the COVID-19 Vaccine and Immunisation Programme, in collaboration with Medsafe, established an active surveillance system to monitor the BNT162b2 vaccine in a real-world setting. Unlike spontaneous reporting, the active system is not contingent upon voluntary reports and instead uses electronic health records (EHRs) to assess the risk of prespecified adverse events of special interest (AESIs) [
15] following vaccination compared with a non-vaccinated group, i.e. historical background rate data. This includes linking all national COVID-19 vaccination data to public hospitalisation records. As of 10 February 2022, approximately 95% of the eligible New Zealand population aged 12 years and above, and 43% of children aged 5–11 years have received at least one dose of the adult or paediatric BNT162b2 vaccine, respectively. The high vaccination coverage ensures representation across the entire population, including main ethnic groups. This is important as New Zealand has a unique demographic, consisting of three main minority ethnic groups, Māori (indigenous New Zealanders), Pacific people (Pacific Islanders living in New Zealand), and Asian, alongside the majority group, New Zealand Europeans [
18]. These minority groups, particularly Māori, are often not included in international clinical trials or postmarketing studies.
New Zealand has also been in a unique position globally during the pandemic. The implementation of strict public health measures such as imposed lockdowns and border controls led to the successful elimination of COVID-19 transmission for sustained periods from May 2020 [
19]. This coupled with the high vaccination coverage achieved, provides an ideal setting to carry out pharmacovigilance of the vaccine as high rates of undetected COVID-19 seen in many countries during the pandemic can confound vaccine safety studies, especially as many of the adverse events observed following immunisation are also associated with infection. We therefore aimed to calculate the incidence and excess risk of several AESIs for COVID-19 vaccines following BNT162b2 vaccination in a largely COVID-19 naïve New Zealand population.
4 Discussion
This nationwide cohort study involving more than 4 million vaccinated persons in New Zealand aged 5 years or older found no statistically significant association between BNT162b2 vaccination and the majority of the 12 selected AESIs, including AKI, acute liver injury, GBS, erythema multiforme, herpes zoster, arterial thrombosis, CVT, splanchnic thrombosis, VTE, and thrombocytopenia. To our knowledge, this is the largest ever postmarketing vaccine safety study carried out in the country and includes representation across the population (including the main ethnic groups). Our findings provide reassurance on the overall safety profile of the BNT162b2 vaccine.
However, a statistically significant association between BNT162b2 and myo/pericarditis was observed in the 21 days following both doses of the vaccine. The association was found to be highest in the youngest recipients, i.e. under 39 years of age, and following the second dose, with an estimated five additional myo/pericarditis cases per 100,000 persons vaccinated. Importantly, this association was not limited to younger age groups, and we observed an increased SIR for myo/pericarditis following both doses of the vaccine in individuals between the ages of 40 and 59 years. We observed no statistically significant increased SIR for myo/pericarditis following either dose of the vaccine in individuals aged ≥60 years. In addition to myo/pericarditis, a statistically significant association between BNT162b2 and SOCV was observed following the first dose of the vaccine in the 20–39 years age group only.
Our findings align with international postmarketing studies, case series reports, and cases detected through reports to New Zealand’s spontaneous system that identify an association between the BNT162b2 vaccine and myo/pericarditis [
7‐
13], especially in younger people and after the second dose [
9]. Consistent with our results (one and three excess myo/pericarditis hospitalisation events per 100,000 persons after the first and second dose, respectively), a population-based cohort study in Israel of 884,828 vaccinees using health care data estimated the excess risk of myocarditis to be three events per 100,000 persons vaccinated with the BNT162b2 vaccine [
8]. Another Israeli study provides further evidence, with the addition of clinical review assessments, and found that the risk of myocarditis was highest in young males after the second dose [
9]. They estimated an SIR of 13.60 (95% CI 9.30–19.20), which translates to 14 additional events of myocarditis per 100,000 persons in vaccinated recipients aged 16–19 years.
In this study, the higher SIR for myo/pericarditis observed following the second dose of BNT162b2 in the youngest age group (25.6, 95% CI 15.5–37.5) is most likely due to differences in the way age groups were stratified in the observed versus expected datasets. The background rate data used to calculate the expected rate includes information on persons between the ages of 0 and 19 years, while our observed rate includes persons between the ages of 5 and 19 years. The incidence of myocarditis has been found to increase with age in children [
36], and the inclusion of children under 5 years of age may have led to an underestimate of our expected rates and thereby an overestimate of the SIR for myo/pericarditis after vaccination in this age group.
Importantly, the risk difference of myo/pericarditis is still low in individuals under 19 years of age, with an excess of two and five events per 100,000 persons after the first and second dose of the vaccine, respectively. Furthermore, given the increased public and medical awareness around myo/pericarditis as a rare adverse reaction of COVID-19 mRNA vaccines, BNT162b2 vaccination might lead to increased hospitalisations and over-identification of the event compared with pre-pandemic years. Most importantly, studies have found that the risk of myocarditis following SARS-CoV-2 infection is substantially greater than after COVID-19 mRNA vaccination [
8,
12,
37]. It is generally considered that the benefits of vaccination with the BNT162b2 vaccine against COVID-19 continue to outweigh the risks from the disease [
38].
Unlike myo/pericarditis, SOCV has not been identified as an adverse reaction to the BNT162b2 vaccine. We observed a statistically significant increased SIR for SOCV following the first dose of the BNT162b2 vaccine in the 20–39 years age group only; however, both the observed and expected numbers were extremely low (fewer than six events) and should be interpreted with caution. The number of large real-world studies investigating the incidence of SOCV following COVID-19 vaccination is limited and there have only been a few case reports and reviews in the literature. Cases that were reported occurred after both doses, with no differences in sex or age [
39‐
42]. Given the low incidence rates of SOCV observed following vaccination, as well as the lack of evidence from international studies, further research and continued safety monitoring are required to understand the association between this AESI and the BNT162b2 vaccine.
Our finding of no association between BNT162b2 and the other ten selected AESIs examined is consistent with other population-based studies. Several large real-world studies observed no increased risk of developing the selected thrombotic events following BNT162b2 vaccination [
8,
31,
43‐
46], including a self-controlled case series (SCCS) study conducted in New Zealand during the same observation period as our study (February 2021–February 2022) [
47]. Conversely, an observational study in the UK found an increased risk of CVT and arterial thrombosis [
15] following BNT162b2 vaccination, while another study in the UK, as well as one study in Spain, found an increased risk of VTE following BNT162b2 vaccination [
48,
49]. However, these statistical safety signals have not been confirmed or added to the vaccine’s product information by the sponsor, Pfizer-BioNTech, in New Zealand.
Although several case reports of acute liver injury [
50], AKI [
51], GBS [
52,
53] and erythema multiforme [
54] have been reported, no association has been confirmed in the available real-world studies using electronic records [
8,
55‐
57]. Our finding of no increased risk of herpes zoster following BNT162b2 vaccination aligns with several studies [
58,
59]. However, there have also been a number of large studies that have identified an association between herpes zoster and BNT162b2 [
8,
60], including a US study using the Vaccine Adverse Event Reporting System (VAERS) data [
13]. The differences with our results may be attributed to our outcome focusing on hospitalisation events only. As such, diagnoses made in the primary care setting, where herpes zoster can be treated, or cases reported through the spontaneous reporting system are not included in our analysis.
Our study has several strengths, including the robustness and completeness of the datasets used. We linked data on all individuals vaccinated with a first or second dose of BNT162b2 to national hospitalisation records for all persons who use public health and disability services in New Zealand. Importantly, BNT162b2 is freely available to all eligible individuals in New Zealand (as of 10 February 2022, this is individuals aged ≥5 years), regardless of eligibility to public health and disability services. This enabled us to rapidly assess, analyse, and contextualise the risk for all vaccine-related outcomes of interest in patients who were hospitalised in the public setting from 19 February 2021 through 10 February 2022. The same source of hospitalisation data and diagnosis codes as the SAFE background rate study were also used to determine the observed and expected number of AESIs, making our study less susceptible to misclassification bias. Real-world population-based studies using EHRs also offer some advantages over studies using passive reporting. Although the passive reporting system can detect unexpected patterns of adverse event reporting, it can be subject to several limitations [
17] and does not allow for the calculation of incident rates or attributable risk of AESIs in a defined population. This is a strength of our study, especially in the context of a pandemic and large-scale immunisation programme using a novel vaccine where population-based information is needed rapidly. Furthermore, during the study period, New Zealand suppressed community transmission of SARS-CoV-2 [
19] and experienced one of the lowest incidences of infection and COVID-19-related mortality among the Organisation for Economic Co-Operation and Development (OECD) countries [
61,
62]. High community transmission of the virus in other countries can introduce bias to pharmacovigilance studies, especially as many of the adverse events observed following immunisation, such as myocarditis, are also associated with SARS-CoV-2 infection. New Zealand also achieved extremely high vaccination coverage, particularly in individuals over 12 years of age, with 95% of this population vaccinated with at least one dose of the BNT162b2 vaccine. To put this in context, approximately 86% of Māori, 88% of Pacific people, and 90% of NZ European and Asian aged 12 years and older received at least one dose of the vaccine during the study period. This allowed us to study the true effects of the vaccine in nearly an entire population, which includes representation across the main ethnic groups in New Zealand. We therefore believe that our analysis includes the largest number of Māori and Pacific people living in New Zealand in any vaccine surveillance study undertaken globally to date.
Our study is subject to several limitations. First, only hospital discharge information was used to identify the outcomes of interest in the vaccinated and historical comparator cohorts. Although many of the AESIs analysed in this study resulted in hospitalisation, less serious conditions such as herpes zoster are commonly treated in primary care settings. Therefore, diagnoses made in general practice are not included in our analyses and the rate for certain AESIs following vaccination could be underestimated. Second, ICD-10-AM codes were used to identify outcomes of interest. There is potential for misclassification as clinical record assessments were not conducted to validate the diagnoses or codes used. Third, although a historical comparison cohort design (i.e., observed versus expected analysis) is a sensitive signal detection method, differences between the historical background population and the vaccinated cohort can lead to false positives (type 1 errors) [
63]. For example, the healthy vaccinee effect can occur, where, on average, people that are healthier are more likely to get vaccinated [
64]. Conversely, medically compromised individuals are often prioritised for vaccination. Comparisons between pre-pandemic years 2014–2019 and 2021–2022 might also be limited due to secular trends in disease, seasonal variations in outcomes, and changes to viral circulation, especially in the context of the pandemic. For example, the influenza virus circulation in New Zealand was almost non-existent during the 2020 winter [
65], with a 99.9% reduction from previous years. This trend continued, with no cases of influenza reported during the 2021 winter season [
66,
67]. Additionally, variation in diagnostic or coding practices from 2014 through 2022 can lead to an under- or overestimate of risks for certain AESIs. Fourth, although the study population included more than 4 million people, for extremely rare outcomes (e.g., CVT and GBS), too few events were observed, particularly in specific age groups, to draw any conclusions from the estimated SIR. Fifth, we used a risk period of 1–21 days, and the SIRs may be under- or over-estimated if the real risk period was longer or shorter. Furthermore, the study population might not be fully representative of those under 12 years of age and eligible for the paediatric BNT162b2 vaccine, as vaccinations for this age group started on 17 January 2022. Sixth, we only included the first event experienced by an individual since 2008 in our observed count for each AESI to align with the exclusion criteria applied in the SAFE background rate study. As such, our estimated SIR might differ from the SIR estimated for individuals with a history of certain conditions. Finally, although we adjusted for age in our analysis, we could not adjust for other factors such as sex, ethnicity, or comorbidities. Stratification by both age and another factor, e.g., sex and ethnicity, was not provided in our background rate data, making further subgroup analysis impossible. Additional observational studies that are not reliant on background rates, such as the SCCS or revised background rates that stratify by multiple factors, are needed to allow for this. However, given that this study is representative of nearly the entire eligible New Zealand population, including 86% of Māori and 88% of Pacific people aged ≥12 years, we are confident that the overall safety profile of the vaccine in these groups is understood.