Adverse event reporting and Bell's palsy risk after COVID-19 vaccination

Although the incidence of Bell's palsy in the general population is low (15–30 cases per 100 000 person-years),¹ Bell's palsy following exposure to SARS-CoV-2 vaccines has attracted attention. In line with clinical trial data that suggested a substantial but non-significant risk of Bell's palsy following exposure to mRNA SARS-CoV-2 vaccines (rate ratio 7·0, p=0·07),¹ a case series and nested case-control study reported a non-significantly increased risk of Bell's palsy following BNT162b2 vaccination. However, this population-based study found a significantly increased risk of Bell's palsy following use of an inactivated (CoronaVac) SARS-CoV-2 vaccine (odds ratio 2·385, 95% CI 1·415–4·022).²

Although a number of limitations have been considered, Eric Yuk Fai Wan and colleagues² might have overlooked possible selection bias, which was partly due to their method of selecting study participants and partly due to substantially different COVID-19 vaccination rates between different age groups (appendix). The very low vaccination rate among those aged 70 years or older was attributable to widespread concerns about adverse events following vaccination.³Although a nested case-control study is an efficient method for conducting a cohort study, selection bias can occur when people in the cohort do not have equal chance of being selected for case-control analysis. In the nested case-control study by Wan and colleagues,² cases and controls were selected from patients admitted to emergency rooms or hospital wards rather than all the people who were eligible for vaccination, probably because of the robustness of clinical data.² Using published local statistics,4, 5 it can be shown that the proportion of people aged 65 years and older attending emergency rooms from 2020 to 2021 was significantly higher than that of the counterpart in the rest of the general population (35·0% vs 14·4%).

We show how selection bias can overestimate Bell's palsy risk in cohort analyses (table ). Assuming that (1) Bell's palsy occurs at equal rates among vaccinated and unvaccinated people, (2) there is a higher proportion of older people (≥65 years) with a lower overall vaccination rate among eligible people who are attending emergency rooms or hospitals wards, and (3) all cases of Bell's palsy are captured in emergency rooms or hospital wards owing to its acute and disabling symptoms, selecting cases and controls from emergency rooms and hospital wards rather than all people who are eligible for vaccination would overestimate the risk of Bell's palsy. The bigger the difference in vaccination rates between selected and non-selected people, the more severe the bias.

Table.

Cohort analyses with hypothetical figures to show the effect of selection bias

		People with Bell's palsy	People without Bell's palsy	Subtotal	Bell's palsy rate	Rate ratio
All people eligible for vaccination (n=200000)
Subgroup A* (n=100 000)
	Vaccinated	9	29 991	30 000	0·03%	1·00
	Unvaccinated	21	69 979	70 000	0·03%	..
Subgroup B† (n=100 000)
	Vaccinated	21	69 979	70 000	0·03%	1·00
	Unvaccinated	9	29 991	30 000	0·03%	..
Subgroups A and B combined (n=200 000)
	Vaccinated	30	99 970	100 000	0·03%	1·00
	Unvaccinated	30	99 970	100 000	0·03%	..
Subgroup A (all patients with Bell's palsy captured)
Vaccinated		30	29 991	30 021	0·10%	2·33
Unvaccinated		30	69 979	70 009	0·04%	..

^*Subgroup A: patients who presented to emergency rooms or hospital wards, comprising a higher proportion of older people (≥65 years) with a lower overall vaccination rate of 30%.

^†Subgroup B: other eligible people, who are relatively younger, with a higher overall vaccination rate of 70%.

We hope our views regarding possible selection bias in observational studies of Bell's palsy following COVID-19 vaccination could put things in perspective and ease concerns.

We declare no competing interests.