This is a very interesting study – the authors have made use of the self-controlled-case-series methodology (SCCS) which was developed by Paddy Farrington and colleagues to evaluate adverse outcomes of vaccines.  In SCCS studies you evaluate WHEN an individual experience an adverse event rather than WHO (see reference below).

The study team is aware of the assumptions and limitations of the study design and conduct a number of sensitivity analysis to evaluate the robustness of their study.  We should, however, be aware that the SCCS is very sensitive to accuracy of the recording of the exposure and events.  While it appears that there are strong associations between positive PRC tests for Sars-Cov-2 and a number of events it is important to be aware that a positive PCR test result is likely to occur for several weeks after the onset of covid infection.  Thus, it is likely that many (if not most) of the cases in this study have been tested at the point that they have been diagnosed with the event(s) and have had covid for some time prior to the positive test.  Therefore, the SCCS results does not provide meaningful information about when the event occurred in relation to the onset of the covid infection and these results should be interpreted with a pinch of salt.

I noted that the team provided a sensitivity analysis excluding people with a prior positive PCR test which provided similar results to their main analysis.  Had I conducted the study I would probably have used this analysis as my main analysis.

Please see our paper on SCCS for further details of the methodology https://www.bmj.com/content/354/bmj.i4515.

‍

This interesting observational analysis by Pantone et al of neurological events occurring after vaccination for COVID-19 or the occurrence of a positive COVID-19 test poses an important question regarding the risk of rare events with such widespread use of an intervention to prevent a common disease.  This is particularly important as vaccination is extended to groups at a lower risk of severe complications of COVID-19 infection.  The authors concluded there was evidence for an increased risk of Guillain-Barre syndrome or Bell’s palsy in the period after vaccination with the AstraZeneca COVID-19 vaccines, and a smaller, increased risk of haemorrhagic stroke in the period after vaccination with the Pfizer BNT162b2, although the absolute increased risk is still small.  However, they also concluded that the risk of these same events occurring were significantly greater around the time of a positive test for COVID-19, implying that the risk:benefit of vaccination is still likely to favour vaccination even for these rare neurological outcomes alone, and before considering the much greater risk of disability and death that may result from non-neurological complications of COVID-19 infection.

The press releases for this article reasonably reflect the science, and reach valid conclusions from the data about the balance of risk after vaccination versus COVID infection.  However, they don’t necessarily reflect any uncertainty in the conclusions, particularly relating to the risk of intracerebral haemorrhage.

Specifically, the SCCS design suffers from the same inherent difficulties as with all case-control studies with a high risk of bias, especially compared to randomised designs or even prospective cohort studies.  These persist even when the rate of an event occurring for an individual after vaccination is compared with the rate of these events occurring in that individual before vaccination (or a longer time after), when any event would not be expected to be related to the vaccine.  In particular, if there is a potential relationship between the outcome (the neurological event) and the exposure (vaccination) then this can systematically affect the estimated risk.  For example, having a haemorrhagic stroke often results in significant disability for a prolonged period, or death, and therefore these patients are far less likely to have received a vaccination.  As such, people included in the study were much less likely to have had an event during their baseline period.  Even if their risk of a haemorrhagic stroke is unchanged by vaccination and remains constant after vaccination, as they have a lower measured risk during their baseline period this therefore falsely inflates the relative risk in the post-exposure period.  The authors demonstrated this problem themselves, showing a signficantly lower risk of many events in the 28 day ‘pre-risk’ period particularly for haemorrhagic stroke (i.e. having a vaccine was associated with a reduced risk of events before the vaccine was given).  However a 28 day period is unlikely to adjust for this risk given the long-term effects of a severe haemorrhagic stroke (including death) and therefore the relative weak association between Pfizer vaccination and haemorrhagic stroke is very uncertain, especially as it was reduced by sensitivity analyses adjusting for death and could not be replicated in the Scottish cohort.  Although the GBS analysis is prone to this same bias, the strength of the association, and a temporal pattern that fits our understanding of the biology of GBS (i.e. occurring ~10 days after the exposure) means that this is less likely to explain all of the association demonstrated.  In contrast, the authors do not discuss how age interacts with the risk of neurological complications after vaccination versus complications after COVID-19 infection, which could be vital in understanding the risk:benefit relationship in younger patients being vaccinated.  It is also surprising that the well-documented association between AstraZeneca vaccination and the condition known as VITT (vaccine-induced thrombosis with thrombocytopaenia) is not apparent in the data (although they didn’t look for it explicitly in this paper) as a raised rate of intracerebral hemorrhage reported after vaccination, despite often presenting with intracerebral haemorrhage in more than 100 patients in the UK.  This association with VITT and AZ vaccination has been much more strongly demonstrated by in depth analyses of individual cases, the occurrence of a very rare combination of clinical problems and a very tightly associated biological mechanism (anti-PF4 antibodies).

Overall, although there are inevitable limitations in analysis of such observational data, the paper is supportive of an association between AstraZeneca COVID-19 vaccination and Guillain-Barre syndrome, but the authors have commendably compared the equivalent risk after COVID-19 infection, demonstrating that the risk of COVID-19 infection is still likely to exceed that of vaccination.  This is consistent with previous reports.  The weaker, and non-replicated, potential association with intracerebral haemorrhage after Pfizer vaccination raises an important question that merits further investigation, but this analysis provides only very weak evidence for the existing of such a relationship.  As such, the paper is reassurring that the benefits of vaccination for COVID-19 are likely to exceed those of COVID-19 infection, even when only considering neurological outcomes and prior to any consideration of the much more common, life-threatening non-neurological complications of COVID-19 infection, and as such, these findings should not negatively influence people’s decision to be vaccinated.

‍

Another thorough review comparing safety of the SARS CoV2 vaccines to risks following COVID infection/illness confirms that the risks from infection far outweigh the risk from the vaccines, based on a retrospective analysis of linked healthcare data from England and Scotland.  The authors used health service information to investigate diagnoses of several CNS disorders following the first dose of either the Pfizer/Astra Zeneca vaccines with the incidence of the same diagnoses within the same period – 28 days – following a SARS CoV2 positive test.

In this analysis being infected carried a far greater risk for CNS disorders than receipt of either of the two vaccines.  While the type of event varied between vaccinations, with Guillain-Barré syndrome being associated with the viral vector rather than mRNA vaccine, none the less, far more patients, almost three times as many people, developed this disorder following SARS CoV2 infection than after vaccination.

A possible signal of risk of haemorrhagic stroke was noted following receipt of the Pfizer vaccine based on data from England, but this finding was not confirmed in the data from Scotland.  In addition, this risk was outranked by the greater risk of subarachnoid haemorrhage, found most frequently following COVID infection.

From any perspective therefore, and particularly with the community spread of infection on the rise, unvaccinated adults should come forward for vaccination to reduce their own risk of serious CNS effects should they become infected.

‍

This very large study of 32 million adults in England investigated associations between vaccinations and SARS-CoV-2 infection and the development of neurological disorders.  They investigated hospital admissions for neurological complications in the 28 days following a first dose of the Oxford–AstraZeneca or Pfizer–BioNTech vaccine, or a positive test for SARS-CoV-2, by linking data from the English National Immunisation (NIMS) Database of COVID-19 vaccination to national data for hospital admissions (NHS Hospital Episode Statistics), mortality (from the Office for National Statistics, ONS), and SARS-CoV-2 infection data (Public Health England).

After the Oxford–AstraZeneca vaccine the authors found an increased risk of Guillain–Barré syndrome within 28 days and an increased risk of Bell’s palsy 15–21 days following vaccination (but the latter risk was not significant over the whole 28-day study period).

Perhaps more unexpectedly the authors found a very small increased risk of “haemorrhagic stroke” within 28 days of vaccination with the Pfizer-BioNTech vaccine, estimated as only 60 extra cases per 10 million people who received the vaccine.  However, there are several reasons to treat this finding with caution.  First, although the term “haemorrhagic stroke” presumably means bleeding into the brain (intracerebral haemorrhage) this was not clearly specified, and the diagnosis was based on routine hospital coding without validation by stroke experts (and we know from previous studies that these do not always agree).  This means that the accuracy of the diagnosis of intracerebral haemorrhage is not entirely clear, and some of the outcome events might have been due to other types of bleeding inside the skull; for example, cerebral venous thrombosis (a known complication associated with the Oxford-AstraZeneca vaccine) can also cause bleeding in the brain, which could lead to misclassification.  Second, because haemorrhagic stroke was so rare after Pfizer-BioNTech vaccination, the estimate of any increased risk is imprecise.  Third, the finding was not replicated using Scottish cohort data since there were too few events to obtain reliable estimates.  So, although a low platelet count (which could be a risk for intracerebral haemorrhage) has previously been reported after the Pfizer-BioNTech vaccine, any possible association with intracerebral haemorrhage needs further independent confirmation in similarly large datasets internationally, ideally with validation of the diagnostic coding.

Importantly, following a positive SARS-CoV-2 test, the authors found a substantially higher risk of all the neurological diseases studied, emphasising that the benefits of ongoing vaccination efforts worldwide outweigh these potential risks.

‍

In this study of 32 million people in the UK, neurological complications of COVID vaccination were compared with those seen in people who had tested positive for SARS-CoV-2.

There was a small increased risk of Guillain–Barré syndrome (a progressive but often reversible weakness, sometimes seen after viral infection or vaccination) or Bell’s palsy (facial weakness) 15–21 days after vaccination.  However, these adverse events were dwarfed by the neurological disorders seen after testing positive for COVID: Guillain–Barré syndrome, myasthenia-like disease, subarachnoid haemorrhage, encephalitis, and Bell’s palsy were all quite common, especially in the first 2 weeks after testing positive for SARS-CoV-2.

The authors estimate that there were 38 excess cases of Guillain–Barré syndrome per 10 million people given ChAdOx1nCoV-19, compared to 145 cases per 10 million after testing positive for SARS-CoV-2.

The authors report an association between the Pfizer vaccine and haemorrhagic stroke – but this was only seen in one cohort, it wasn’t also found in the Scottish data, and it looks to me a very small signal and possibly not very significant.

The neurological complications of SARS-CoV-2 vaccines are much rarer than the neurological complications of COVID-19, showing the vital importance of getting vaccinated.

‍