Results The study included 4 376 535 people who received ChAdOx1 nCoV-19, 3 588 318 who received BNT162b2, 244 913 who received mRNA-1273, and 120 731 who received Ad26.CoV.2; 735 870 people with SARS-CoV-2 infection; and 14 330 080 people from the general population. Overall, post-vaccine rates were consistent with expected (background) rates for Bell’s palsy, encephalomyelitis, and Guillain-Barré syndrome. Self-controlled case series was conducted only for Bell’s palsy, given limited statistical power, but with no safety signal seen for those vaccinated. Rates were, however, higher than expected after SARS-CoV-2 infection. For example, in the data from the UK, the standardised incidence ratio for Bell’s palsy was 1.33 (1.02 to 1.74), for encephalomyelitis was 6.89 (3.82 to 12.44), and for Guillain-Barré syndrome was 3.53 (1.83 to 6.77). Transverse myelitis was rare (<5 events in all vaccinated cohorts) and could not be analysed.

Main outcome measures Outcomes were incidence of Bell’s palsy, encephalomyelitis, Guillain-Barré syndrome, and transverse myelitis. Incidence rates were estimated in the 21 days after the first vaccine dose, 90 days after a positive test result for SARS-CoV-2, and between 2017 and 2019 for background rates in the general population cohort. Indirectly standardised incidence ratios were estimated. Adjusted incidence rate ratios were estimated from the self-controlled case series.

Participants 8 330 497 people who received at least one dose of covid-19 vaccines ChAdOx1 nCoV-19, BNT162b2, mRNA-1273, or Ad.26.COV2.S between the rollout of the vaccination campaigns and end of data availability (UK: 9 May 2021; Spain: 30 June 2021). The study sample also comprised a cohort of 735 870 unvaccinated individuals with a first positive reverse transcription polymerase chain reaction test result for SARS-CoV-2 from 1 September 2020, and 14 330 080 participants from the general population.

We leveraged large routinely collected datasets including millions of vaccinated people in the UK and Spain to study the potential association between covid-19 vaccines and the short term risk of developing Bell’s palsy, encephalomyelitis, Guillain-Barré syndrome, and transverse myelitis. To place these risks in context, we also studied the association between SARS-CoV-2 infection and risk of these immune mediated neurological events.

More recently, concerns have been raised about immune mediated neurological disorders post-covid-19 vaccination. Owing to reports of people developing Guillain-Barré syndrome after vaccination with Ad.26.COV2.S (108 of 21 million people vaccinated as of late June 2021), 8 and ChAdOx1 nCoV-19 (833 people of 592 million doses administered as of late July 2021), 9 10 the EMA listed Guillain-Barré syndrome as a rare side effect related to these vaccines. Guillain-Barré syndrome has also been associated with mRNA vaccines in a few people. 11 12 In addition, Bell’s palsy, encephalomyelitis, and transverse myelitis events have been described in case series studies after covid-19 vaccination with both viral vector and mRNA vaccines. 13 14 15 16 17 Although these events are not necessarily due to covid-19 vaccines, the temporal association between the events and vaccination warrants robust post-vaccination surveillance. Large scale epidemiological studies are required to determine whether covid-19 vaccination increases the risks of these events above background rates in the general population.

As of 13 January 2022, the covid-19 pandemic has resulted in more than 5.5 million deaths worldwide. After the rapid development of anti-SARS-CoV-2 vaccines, 9.2 billion doses have been administered through national vaccination programmes. 1 To date, five vaccines against SARS-CoV-2 have received a conditional marketing authorisation by the European Medicines Agency. These include two mRNA vaccines: BNT162b2 (Pfizer-BioNTech) and mRNA-1273 (Moderna); two viral vector vaccines: ChAdOx1 nCoV-19 (Oxford-AstraZeneca) and Ad.26.COV2.S (Janssen/Johnson & Johnson); and one adjuvanted, recombinant spike protein nanoparticle vaccine: NVX-CoV2373 (Novavax). 2 All of these vaccines have shown high efficacy in preventing severe covid-19 and acceptable safety profiles in clinical trials. 3 4 5 6 7 However, potential adverse events related to these new vaccines have been reported, and continuous vaccine safety surveillance is needed as mass immunisation against covid-19 continues.

Methods

Data sources For this study we used data from primary care records in both the UK and Spain. The Clinical Practice Research Datalink (CPRD) AURUM contains routinely collected data from primary care practices in the UK,1819 representing 20% of the current UK population.20 Data from Spain came from the Information System for Research in Primary Care (SIDIAP; www.sidiap.org), a primary care database that covers 80% of the population in Catalonia, Spain, and is linked at an individual level to hospital data. These hospital data included information from all public and private hospitals in Catalonia (Conjunt Mínim Bàsic de Dades d’Alta Hospitalària, CMBD-AH).21 Both databases have been mapped to the Observational Medical Outcomes Partnership (OMOP) common data model,22 which allowed the same analytical code to be run against both datasets without the need to share patient level data.

Study participants The populations of interest were individuals who had received at least one dose of a covid-19 vaccine and people with SARS-CoV-2 infection. Vaccination cohorts were constructed of people vaccinated according to the product (ChAdOx1 nCoV-19, BNT162b2, mRNA-1273, or Ad.26.COV2.S) and dose administered (first or second dose, with only a single dose cohort for Ad.26.COV2.S as this vaccine is comprises a one dose regimen). Ad26.COV2.S and mRNA-1273 cohorts were only available in SIDIAP. Vaccinated individuals were required to have received their first dose between the start of the vaccination campaign in each country (8 December 2020 in the UK, 27 December 2020 in Spain) and one week before the end of data availability of each database (9 May 2021 for CPRD AURUM, 30 June 2021 for SIDIAP), with the vaccination date used as index date. We excluded those who received more than one brand of a covid-19 vaccine. For the second dose cohorts, participants were required to have received their second dose in prespecified intervals after the first dose. For both databases, the interval allowed between doses of two dose regimens (except Ad26.COV2.S) was 14 to 180 days. The SARS-CoV-2 cohort included people with a first positive reverse transcription polymerase chain reaction (RT-PCR) test result or antigen test result between 1 September 2020 and one week before the end of data availability of each database, with the test date used as index date. Data from both RT-PCR and antigen tests were available in SIDIAP, whereas only RT-PCR test results were available in CPRD AURUM. From the SARS-CoV-2 cohort we excluded individuals vaccinated against covid-19 before infection. We also identified a background population cohort that included all individuals registered in CPRD AURUM and SIDIAP as of 1 January 2017 (index date). All study participants were required to be 18 years or older and to have at least 365 days of data availability before their index date. For each specific outcome, we excluded those who had experienced the outcome in the year before the index date. For each cohort, we followed participants from the index date until the earliest of end of follow-up (21 days for vaccinated people, 90 days for those with a diagnosis of covid-19, and 31 December 2019 for the general population cohort), first occurrence of the adverse event, end of data availability, or until transference out of the database or death. For cohorts that had received a first vaccine dose, we also censored follow-up if a second dose was observed before 21 days.

Events of interest The events of interest were four immune mediated neurological disorders prespecified as potential adverse events of special interest for covid-19 vaccine safety: Bell’s palsy, encephalomyelitis, Guillain-Barré syndrome, and transverse myelitis.2 We identified these events using previously published clinical codes from electronic health records.23 Supplementary appendix table 1 provides details of the Systematized Nomenclature of Medicine (SNOMED) codes used to define the outcomes.

Study design Firstly, we used the historical rate comparison method (see fig 1). Incidence rates of each outcome in the vaccinated and SARS-CoV-2 cohorts were used as observed rates and compared with the expected background incidence rates estimated from the general population cohort. For the vaccinated cohorts, we estimated the rates during 1-21 days after a first vaccine dose (day 0). For the SARS-CoV-2 cohort we used a post-test period of 90 days. Fig 1 Study design. Potential risk period (dark blue) for vaccination cohorts was defined as time between the start of the vaccination campaign and one week before the end of data availability for each database (CPRD AURUM: 8 December 2020 to 2 May 2021; SIDIAP: 27 December 2020 to 23 June 2021). For the SARS-CoV-2 infected cohort, the potential risk period started on 1 September 2020. The baseline period for the self-controlled case series analysis (light blue) was defined from 1 January 2017 to 21 days before the day of vaccination or SARS-CoV-2 positive test result. The pre-risk period (pink) was defined as −21 to −1 days before vaccination or SARS-CoV-2 positive test result, and the risk period (orange) was defined as 1 to 21 days after vaccination and 1-90 days after a SARS-CoV-2 positive test result Secondly, we used a self-controlled case series method. In this approach, only individuals who experience the outcome are included, with participants acting as their own controls and thereby eliminating time fixed confounding.2 Within person comparisons of event rates were made between the baseline period before vaccination and the period at risk of an outcome. We defined the at risk period as the 0-21 days after a first vaccine dose or after a SARS-CoV-2 positive test result. The 0-21 day period was subdivided into several prespecified time periods: 0, 1-7, 8-14, 15-21 days, as well as for the 1-21 day period overall. We considered events at day 0 separately, as these events might precipitate hospital admissions and subsequent covid-19 testing, which could result in a positive association between SARS-CoV-2 infection and the studied events.25 The risk period for the SARS-CoV-2 cohort was also extended to 90 days. The study period of the self-controlled case series analysis was defined from 1 January 2017 to 21 days after the first vaccine dose or 90 days after a positive test result.

Statistical analysis We characterised the participants in each cohort by personal characteristics, such as age and sex; comorbidities any time before vaccination; and recent drug use during the six months before the index date. Supplementary appendix table 2 shows the codes used for definitions of comorbidities and drug use. We estimated the observed rates during the 21 days after immunisation for the vaccinated cohorts and 90 days after testing for the SARS-CoV-2 cohort. Similarly, background rates were estimated for the general population from 1 January 2017 to 31 December 2019. We calculated crude incidence rates as the total number of events divided by the person time at risk per 100 000 person years. Indirect standardisation is used to account for differences between the age structure of the vaccinated or SARS-CoV-2 cohorts and the general population.26 Observed and expected rates were compared using standardised incidence ratios with corresponding 95% confidence intervals. For the self-controlled case series analysis, only the first event was considered for each participant. Because events could potentially decrease the probability of being vaccinated, we removed a 21 day pre-risk period from the baseline period and reported separately. To assess potential for event dependent observation periods, we plotted a histogram of the time between the occurrence of the event and the end of observation for individuals censored and uncensored, and we calculated the number of deaths that occurred after each event (day 0 to day 7). Conditional Poisson regression models were fitted to estimate incidence rate ratios and 95% confidence intervals for each outcome,27 comparing the risk period with the baseline period. These models were estimated separately for each cohort of interest and were adjusted for age (in five year bands) and seasonality (four seasons). Self-controlled case series analyses were only conducted for comparisons with a ≤2 minimum detectable relative risk.28 We applied two sensitivity analyses to assess the impact of study design choices. Firstly, to exclude events potentially related to covid-19, in the vaccinated cohorts we excluded individuals infected with SARS-CoV-2 before the index date. For this, as RT-PCR tests were not routinely performed during the first wave of the pandemic, the covid-19 definition was broadened to include positive SARS-CoV-2 results (RT-PCR or antigen test) or a compatible covid-19 clinical code (see supplementary appendix table 3). Secondly, to include those with little previous use of healthcare, we also replicated the analyses after removing the one year before observation requirement for participants. Any subgroups with fewer than five people were blinded and reported as less than five, following the requirements of information governance.