3.1 High-Profile Signals

Shortly after authorisation of the vaccines, the PRAC established that a causal association between the adenoviral platform vaccines and the thrombosis with thrombocytopenia syndrome (TTS), as well as between the mRNA vaccines and myocarditis/pericarditis was likely. The particularities of the evaluation of these signals have been previously described [11]—in both, EV data as well as observed-to-expected (O/E) analyses, played a crucial role.

For myocarditis/pericarditis evaluation, two observational studies [12, 13] provided the attributable risk of vaccination. Estimates of the number of excess cases of myocarditis were reflected in the product information as follows: 0.26 (French data) and 0.57 (Nordic data) per 10,000 vaccinated for Comirnaty, 1.3 and 1.9 per 10,000 vaccinated, respectively, for Spikevax. This risk evaluation considered the second dose of mRNA vaccine in young male vaccinees compared to unexposed.

In the case of TTS, the EU network received initial notifications of suspected adverse reactions from independent teams, including from Austria and Norway [14]. In the absence of a case definition and a Medical Dictionary for Regulatory Activities (MedDRA) term, the signal was initially investigated based on case reports in which embolic and thrombotic events (Standardised MedDRA Query, SMQ) were co-reported with manually adjudicated MedDRA terms suggestive of thrombocytopenia.

On 7 April 2021, following an Ad Hoc Expert Group organised by EMA, PRAC recommended an update of the Vaxzevria product information with TTS. While the product information does not include an attributable risk of vaccination, it informs that severe and very rare cases of TTS have been reported post-marketing, that the reporting rates after the second dose are lower compared to after the first dose and that most of the cases of TTS occurred within the first three weeks following vaccination. The risk was further contextualized in a referral (Art. 5.3 of Reg 726/2004) [15, 16].

For myocarditis/pericarditis, the existence of MedDRA terms made the reactions easier to describe. The measure of disproportionate reporting (Reporting Odds Ratio, ROR) in EV increased significantly around the time of EU labelling in July 2021 (Fig. 1), after findings from large population studies in Israel and the USA were published [17, 18]. The need to better characterise the adverse reactions’ frequency, severity or gender distribution resulted in the conduct of two studies, one in the Nordic countries and one in France. Both study consortia made their preliminary findings available for PRAC assessment (e.g., in the case of the Nordic consortium, these findings were assessed even before their publication on 20 April 2022) [12]. The recommendation for labelling was issued on 2 December 2021, confirming that the risk was more significant after the second vaccine dose in younger men.

3.2 Ad hoc Safety Studies by the European Medicines Agency (EMA)

To support the assessment of the signals of vulval ulceration with Comirnaty [19] and pemphigus and pemphigoid with the mRNA vaccines and Vaxzevria [20], EMA conducted two studies (using data sources to which the Agency has direct access). Both signals were initiated by the network following case series of vulval ulceration in adolescent girls [21] or bullous conditions [22, 23] involving several COVID-19 vaccines.

The EU network capability to (a) describe the use of Comirnaty in the general population and (b) estimate incidence rates of vulval ulceration in the general and exposed female population was paramount. A self-controlled case series (SCCS) study was conducted to further investigate a causal association. No differences were found in post-vaccination incidence rates of vulval ulceration compared to incidence rates from time not-at-risk, nor in the 30 or 90 days after receiving the respective vaccines (Comirnaty, Spikevax or Vaxzevria).

In the case of pemphigus and pemphigoid signal, EMA calculated the incidence rates using a cohort design in the UK general population and patients visiting general practices in Spain. A SCCS was also performed as a sensitivity analysis. No consistent associations were seen based on these studies. The studies are publicly available in the EU PAS Register [19, 20].

3.3 New Aspects of Known Adverse Reactions

While hypersensitivity reactions were added to the product information of COVID-19 vaccines [24, 25] based on pre-authorisation studies, new aspects of hypersensitivity were labelled following authorisation, based only on the review of clusters of spontaneous post-marketing case reports. These new aspects include anaphylactic reactions (Vaxzevria), erythema multiforme (EM) (mRNA COVID-19 vaccines), and localised swelling in persons with a history of dermal filler injections (Comirnaty). Published, medically verified reports of EM [26, 27] including skin biopsies and topical corticosteroid treatments contributed to the understanding of these reactions. Based on cumulative data assessed in SSRs, the product information of Spikevax was updated to include the median number of days to onset of injection site reactions, as well as urticaria with either acute onset (within a few days) or delayed onset (up to approximately 2 weeks), respectively [28].

3.4 Challenges of Drawing Conclusions from Consumer Data

Changes to menstruation are not routinely collected in clinical trials. Following vaccines' authorisation, cases of heavy menstrual bleeding (HMB) have been flagged as disproportionate in EV [29] since mid-2021, mainly based on consumer reports. These cases, while often lacking case definition details, and often containing solely the event in relation to time after the vaccination played an important role in the detection of this signal. The scale of reporting to national spontaneous databases [30, 31] was crucial to draw the attention of the EU network and direct resources to process the EV data to further investigate menstruation irregularities (i.e., heavy bleeding and lack of menstruation). About a tenth of the cases were reported by healthcare professionals.

In January 2022, the Norwegian Public Institute [32] issued a survey of the occurrence of menstrual disturbances in 18- to 30-year-old women after COVID-19 vaccination, which was the basis for the PRAC to initiate the signal procedure of HMB. As in the case of the myocarditis/pericarditis signal, the PRAC benefited from collaboration with authors of independent studies who made preliminary data available for assessment (the relevant article was published in January 2023 [33]).

Ultimately, HMB was added to the EU labelling on 27 October 2022, with unknown frequency. Several independent observational studies on the association were published at the time of the PRAC recommendation and later on [34,35,36]. Both the spontaneous reports and the observational research pointed to a transient and non-serious nature of changes to the flow of menstruation.

3.5 Labelling in Specific Patient Groups

In specific patient groups, namely with relapse of capillary leak syndrome (CLS) or experiencing corneal graft rejection (CGR) following COVID-19 vaccination, the signal assessment relied solely on data from EV case reports or published in scientific journals. Subsequently, PRAC recommended restriction of the use of adenoviral vector vaccines in patients with a history of CLS [37], as at that time, review of available data showed that severe cases of CLS following vaccination occurred, including with fatal outcome. No restriction of use was recommended for mRNA platform vaccines, as in their case, the PRAC considered newly available data, including unique experiences from vaccination in specific patient registries. The EurêClark Study Group showed that in patients with CLS/Clarkson disease, the burden of COVID-19 infection leads to severe disease flares, with a high fatality rate. The researchers concluded that the benefit/risk ratio favours COVID-19 vaccination in these patients, under pre-medication with intravenous immunoglobulins [38].

No product information update was recommended for the CGR signals, as the assessed evidence did not suggest that a causal association was at least a reasonable possibility. While the cases were reported in close temporal association with vaccination, with several providing good documentation (i.e., literature cases), there was no clear mechanism of action, no signal of disproportionality observed in EV, no imbalance in the O/E analysis, as well as no increased rates of CGR observed during the pandemic [39]. The challenges with the assessment of this signal have been described elsewhere [40].