Increased risk of myocarditis and pericarditis and reduced likelihood of severe clinical outcomes associated with COVID-19 vaccination: a cohort study in Lombardy, Italy

Four data sources were used to develop the Lombardy Vaccine Integrated Platform. First, a vaccine registry was established by the Regional Health Authority beginning on December 27, 2020, to collect individual data on the date, type, and dose of the dispensed vaccine (the vaccine is administered for free to all candidate to vaccination). Second, a registry of patients with a confirmed diagnosis of SARS-CoV-2 infection was established on February 21, 2020 (the date of the first confirmed diagnosis) to monitor SARS-CoV-2 infections, hospital admissions, emergency room visits, and deaths due to COVID-19. Third, a healthcare utilization database that was established in 2000 to collect a variety of information, including inpatient diagnoses supplied by public or private hospitals, and outpatient drug and services supplied by RHS departments, was used to obtain data on the health profile of the study population. Finally, the health registry reports and updates data on patients' status of the Lombardy Regional Health Service (RHS), which includes almost all Lombardy residents, including date and cause of entry (birth, immigration) or exit (death, emigration) from the database. These databases use a unique individual identification code that allows to link each database with each other. To maintain privacy, each identification code is anonymized so that individuals can only be identified by the Regional Health Authority upon request from judicial authorities.

Vaccine candidates (aged 12 years or older) who were not beneficiaries of the RHS or who became beneficiaries after January 1, 2019 (N = 24,188), were excluded from the study. The remaining 9,184,146 individuals were included in the study cohort and the date and type of vaccine administered from December 27, 2020 (index date) to November 30, 2021 (observation period) were recorded for each person. Sex, age, and clinical profile were recorded at the index date. Hospital admissions and drug prescriptions obtained within 2 years before the index date were used to investigate nine specific cardiovascular diseases. A complete list of these diseases, along with their accompanying ICD-9 and ATC codes, in addition to other conditions and exposures considered during this study, is provided in Additional file 1: Table S1.

The time of observation for each cohort member was partitioned into four exposure categories to assess the risk of vaccine harm. From the date of administration of each vaccine dose to day 28 post-administration, three periods that covered exposure to the first, second, and third dose were defined. The periods that remained uncovered by vaccine exposure were defined as no exposure to the risk of vaccine harm. Hospital admissions associated with a diagnosis of myocarditis (ICD9-CM codes 422.xx and 429.0) or pericarditis (ICD9-CM codes 420.xx) that occurred during the observation period were identified and denoted as the vaccine harm outcome of interest.

Cohort members accumulated person-months of observation from December 27, 2020 until outcome onset, death, emigration out of the region, or November 30, 2021, whichever came first. The incidence rates of myocarditis and pericarditis were separately calculated as the ratio between the number of outcomes and the number of person-months occurred during the period of interest (exposure or no exposure to the risk of vaccine harm). Where relevant, rates were stratified by sex, age category, and vaccine dose and type.

Cox proportional hazard models were fitted for estimating the hazard ratio (HR) and 95% confidence interval (CI) of the defined outcomes, associated with exposure to the first, second, and third dose of each vaccine type, compared to no exposure. As vaccine exposure changed over time, assessment of its effect required consideration of this change. This was performed by fitting the Cox model with dummy factors for exposure categories that were expressed as time-dependent covariates. Adjustments were made for sex, age, and cardiovascular comorbidities.

The time of observation of each cohort member was partitioned into two categories of exposure to assess vaccine benefit. The first category included the period 14 days after the first vaccine dose until November 30, 2021, and was based on the assumption that clinically significant immunity is achieved 2 weeks after receiving the COVID-19 vaccine [30]. The second category included time-periods uncovered by vaccine exposure and was defined as no exposure to vaccine protection.

A composite outcome that included hospital admission, including patients in intensive care units, and deaths attributed to COVID-19, whichever occurred first, was tracked for each individual and denoted as the outcome of interest. Cohort members accumulated person-months of observation from December 27, 2020, until outcome onset, death, out of region emigration, or November 30, 2021, whichever came first. The incidence rates of severe COVID-19 disease were calculated as the ratio between the number of outcomes and the number of person-months occurred during the period of interest (exposure or no exposure to vaccine). Incidence rates were stratified by sex, age category, and vaccine type.

Different metrics were used to compare the harms and benefits of vaccine exposure. In particular, the NNH was considered the number of individuals who would need to be vaccinated to generate a case of myocarditis, while the NNT was considered the number of individuals who should be vaccinated to avoid a case of severe COVID-19 disease [10]. Therefore, a desirable vaccination profile was defined as a low NNT and a high NNH [31]. NNT was calculated as the reciprocal of the absolute difference in incidence rates of myocarditis during the exposure and no exposure periods. NNH was calculated as the reciprocal of the absolute difference in incidence rates of severe COVID-19 diseases during the no exposure and the exposure periods. The corresponding 95% CI were calculated as the inverse of the CI around the absolute incidence rate difference [32]. As a secondary analysis, NNH and NNT were re-calculated by excluding cohort members who did not receive the vaccine during the entire observation period.

Table 1 shows the selected baseline characteristics of the 9,184,146 vaccine recipients on December 27, 2020, and their vaccination status on November 30, 2021 (41,463 (0.5%) cohort members were lost to follow-up because they emigrated out of the Region). Men were 51.3% of the overall population. The percentage of individuals aged 12–15, 16–19, 20–29 and ≥ 30 were, respectively, 4.4%, 4.2%, 10.7% and 80.7%. Almost one in four recipients had a cardiac comorbidity. Globally, more than 16 million doses were administered over the 11-month period since the first vaccine was given. About 90% and 80% of recipients received one and two vaccine doses, respectively. More than half of individuals received the Pfizer-BioNTech vaccine, 12.5% received Oxford-AstraZeneca, 10% received Moderna, and even fewer individuals received Jansen. Just over 10% of recipients received the booster dose.

Of the cohort members, 587 and 1658 experienced myocarditis and pericarditis during the observation period, respectively. Of these patients, 12 and 28 died during their hospital stay, respectively. No individuals who experienced events during person-months of exposure to the risk of vaccine harm (i.e. 28 days after vaccine administration) died during hospitalization. The incidence rates of myocarditis and pericarditis were 5.8 and 16.2 per million person-months, respectively. The incidence of myocarditis was almost double during exposure to the risk of vaccine harm than no exposure, with a corresponding number of cases and incidence rate of 124 and 9.9 every million person-months, respectively, for exposed individuals and 463 and 5.2 per million person-months, respectively, for those with no exposure. The incidence of pericarditis was 1.2-fold higher during exposure than no exposure, with the corresponding number of cases and incidence rate of 245 and 19.5 per million person-months, respectively, for exposed individuals, and 1413 and 15.9 per million person-months, respectively, for those with no exposure.

Exposure to an adenovirus vector-based vaccine was not associated with an increased risk of myocarditis (Fig. 1, top panel). In contrast, exposure to the first and second doses of mRNA-based vaccines was associated with an increased risk of myocarditis, in particular after receipt of the second dose of Moderna (adjusted HR 5.5, 95% CI 3.7 to 8.1). There was no evidence that exposure to an adenovirus vector-based vaccine or the Pfizer-BioNTech mRNA vaccine was associated with an increased risk of pericarditis (Fig. 1, bottom panel). In contrast, exposure to the first and particularly the second dose of Moderna was associated with an increased risk of pericarditis, although with lower strength than that observed for myocarditis (adjusted HR 2.2, 95% CI 1.5 to 3.1).

The timing of myocarditis and pericarditis presentation after vaccine administration is shown in Fig. 2. Myocarditis primarily occurred within a few days after receipt of the second dose of either Moderna or Pfizer-BioNTech (27 of 29 cases and 24 of 40 cases one week after receipt of the Moderna and Pfizer-BioNTech vaccine, respectively). Pericarditis also tended to occur within a few days after the second dose of Moderna (24 of 33 cases one week after vaccine receipt). In contrast, pericarditis that occurred following receipt of the Pfizer-BioNTech vaccine did not follow a temporal trend.

The highest rate of myocarditis occurred among men 16 to 19 years of age following exposure to the Moderna vaccine (Table 2). This group had a 9.5-fold higher incidence than men of the same age during time periods with no vaccine exposure. It should be emphasized that this subpopulation was the only one to have a higher NNT than NNH (9471 vs. 7213, respectively). All other subpopulations defined by sex, age category and vaccine type assessed in this study had a favourable harm-benefit profile with an NNH and NNT ratio ranging from 1.5 (men 12–15 years of age receiving Pfizer-BioNTech) to 962.5 (men > 30 years of age receiving Pfizer-BioNTech). These findings were confirmed after excluding 1,276,381 individuals from the study cohort who did not receive a vaccine during the entire observation period (Table 3).

Using the Vaccine Integrated Platform that covered the entire Lombardy population, a strong association was observed between exposure to mRNA-based vaccines and the risk of both myocarditis and pericarditis compared to periods of no exposure. The risk of myocarditis was particularly high after receipt of the second dose of the Moderna vaccine (adjusted HR 5.5, 95% CI: 3.7 to 8.1), lower after the first dose (adjusted HR 3.5, 95% CI: 2.1 to 6.1), and even less following the first and second doses of the Pfizer-BioNTech vaccine (adjusted HR 1.5, 95% CI: 1.1 to 2.2). Increased risk of pericarditis was noticed after receiving the first and second dose of Moderna (adjusted HR, and 95% CI, 1.6, 1.0 to 2.5, and 2.2, 1.5 to 3.1, respectively). These founding should be compared with risk of myocarditis and pericarditis following post-acute COVID-19, which, in a recent US cohort study, was shown to be 5.38-fold and 1.85-fold, respectively, as compared to controls [33]. Adenovirus vector-based vaccines did not increase the risk of myocarditis or pericarditis, and the Pfizer-BioNTech vaccine did not increase the risk of pericarditis. The study was underpowered to estimate the impact of booster doses on harmful outcomes.

The novel approach used in this study for evaluating the balance between harms and benefits of anti-COVID-19 vaccines was previously used in a similar setting, for evaluating a potential increased risk of thromboembolic events in individuals who received the adenovirus-vectored vaccines [10]. In the current study, an unfavourable balance between harms and benefits for men 16 to 19 years of age who received the Moderna vaccine was highlighted. In this subpopulation, NNT was higher than NNH with corresponding values of 9471 and 7213, respectively, considering the entire study cohort, and 10,317 and 6747, respectively, excluding those who did not receive the vaccine during the entire observation period. It should be emphasized, however, that for this “highest risk” category: (i) the incidence rate of myocarditis was only 1.5 new cases every 10,000 vaccinated individuals; (ii) 7213 citizens (6747 in the secondary analysis) should be vaccinated to generate a case of myocarditis; (iii) although the NNT was slightly higher (from 9471 in the primary analysis to 10,317 in the secondary analysis), suggesting an unfavourable balance between harm and benefit, we cannot disregard that harm and benefit outcomes had substantial difference in clinical profile, resolving the former in few days hospital stay without any fatal outcome, while avoiding severe outcomes, including severe acute respiratory syndrome, that sometimes may result in invasive intubation or death; (iv) the NNT for avoiding a case of SARS-CoV-2 infection was only 215 (data not shown), confirming the potential of a vaccine to prevent infection spread, even when there is a small risk of myocarditis. For all other combinations of sex, age and vaccine type, even lower rates and higher NNH, and a more favourable balance toward vaccine benefits was observed. Taken together, these findings suggest that the balance between harm and benefit is largely skewed towards vaccination, even in men 16 to 19 years of age, and vaccines, in particular those that are mRNA based, remain the primary method for fighting the pandemic. Careful monitoring of the harm-benefit profile, however, is strongly recommended. In fact, due to the expected reduction of viral spread, increasing NNT values are expected in the next few months, suggesting that the investigated balance may become less favourable.

As concern the underlying disease mechanism, myocarditis and pericarditis do not represent novel adverse events of vaccination. Previously, other vaccinations have been associated with an increased risk of myocarditis [34]. These findings suggest that the disease mechanism is specific neither for the newly developed mRNA vaccines, nor for the exposure to the SARS-CoV-2 spike protein. Although rare, clinicians should however be aware of the risk of myocarditis and pericarditis, which should be considered in individuals presenting with chest pain within a week after vaccination, in particular among younger individuals [35].

Previous studies in Israel, Denmark, England, and United States have also shown higher myocarditis incidence after vaccination with the SARS-CoV-2 mRNA vaccine, indicating that this adverse effect is impacting diverse populations [7, 12, 14, 18, 19, 26‐28]. In contrast to a recent study out of Denmark [27], but consistent with others, the current study confirmed (i) a higher risk of myocarditis after Pfizer-BioNTech vaccination, although less than that observed after Moderna, (ii) that the second dose of Moderna was associated with a higher risk of myocarditis than the first dose, and (iii) men were at higher risk of post-vaccine myocarditis than women. In contrast to a recent study out of England [28], the adenovirus vector-based vaccines, Oxford-AstraZeneca and Janssen, did not impact the risk of either myocarditis or pericarditis.

In contrast with prior reports, this population-based study had the advantage of using prospectively collected information on vaccination and hospital admissions for an entire population, virtually eliminating recall and selection bias. Moreover, our study design compared periods of exposure and no exposure within the same subject, thus eliminating the variability between subjects that may arise when a group of exposed is compared to a control group of unexposed. However, this approach did not allow to evaluated the post-acute COVID-19 incidence of myocarditis and pericarditis. Finally, the use of registry data ensured a systematic evaluation of exposures, covariates, and outcomes. However, valid information on electrocardiography or cardiac imaging could not be obtained, so outcome misclassification could not be excluded. Moreover, information on myocarditis and pericarditis treatments were not available.

A potential bias in this observational study is that the decision to become vaccinated is an active personal choice, potentially confounding the association between vaccination and clinical outcomes [27]. Citizens who choose vaccination may be more health-conscious, leading to a healthcare-seeking bias that could overestimate the risk of myocarditis or pericarditis after vaccination if people with mild disease are more likely to be identified among vaccinated individuals [27]. In addition, the inability to compare those who chose to receive the vaccine and those who chose not to be vaccinated, may overestimate vaccine effectiveness and reduce the NNT value. However, the secondary analysis which excluded individuals who did not receive vaccination during the observation period, did not result in substantially different findings. Confounding may also be due to time-invariant characteristics, like socioeconomic status, that may differ between vaccinated and unvaccinated people. However, the Danish study showed that estimates generated from a self-controlled case series (SCCS) design, which took time-invariant characteristics into account, did not differ from those obtained from the usual cohort design, suggesting that unmeasured covariates were unlikely to impact the estimates. The secondary analysis conducted by the current study allowed for a comparison between exposure and no exposure periods within the vaccinated cohorts, avoiding the inclusion of unvaccinated individuals. The latter approach has some elements in common with the SCCS design but includes individuals who both experienced and did not experience the outcome, and observations were censored when the outcome occurred. This avoided the introduction of any biases resulting from the modification of the vaccine propensity due to the onset of myocarditis or pericarditis.