More information about text formats
To the Editor
We were interested to read the study of Filosto et al  concluding a significant link between Guillain-Barre Syndrome (GBS) and COVID-19 infection in Northern Italy at the peak of the 1st wave SARS-CoV2 pandemic. We urge caution in accepting such a causative conclusion using a retrospective observational study; causation is not conclusively proven and is drawn from potentially biased data and small case numbers of a rare condition, and a rate calculation without confidence intervals to infer uncertainty.
Only 34 cases of GBS, of whom 30 were COVID-19 positive, are reported over a 2-month period, with a denominator population of 8,400,107. We calculated the 95% confidence intervals of the incidence rates as 0.08 per 100,000 per month (95% C.I.: 0.04-0.15) in 2019 and 0.2 per 100,000 per month (95% C.I.: 0.14-0.28) in 2020. The overlapping confidence intervals do not support a statistically significant increase in GBS rates from 2019 to 2020. Furthermore, the simple multiplication of the monthly rate by 12 to create an approximate annualised incidence potentially amplifies the inaccuracy. We suggest that the 2.6-fold difference in GBS incidence from 2019 to 2019 is prone to meaningful statistical error.
During the initial stages of the pandemic the denominator of COVID-19 positive cases will have been under-reported because testing was limited to the symptomatic and presenting populations. We are told that 62,679 inhabitants of the regi...
During the initial stages of the pandemic the denominator of COVID-19 positive cases will have been under-reported because testing was limited to the symptomatic and presenting populations. We are told that 62,679 inhabitants of the region had positive COVID-19 swabs during the study, but this is unlikely to provide an accurate population-based rate of COVID-19 infection. More specifically, given that hospital admissions were the only ascertainment source for GBS cases in this study, we are not given comparative information on the rate of COVID-19 positivity in the hospital population, which was likely to have been very high as the Northern Italy healthcare system was reportedly overwhelmed. It is possible, or probable, that a large proportion of the hospitalised population were COVID-19 positive independent of other factors. It is thus likely that these data report a coincidental association between COVID-19 and GBS rather than a causative one.
The COVID-19 seroprevalence rate in the Italian Instituto Nazionale di Statistica (ISTAT) study for Lombardy was the highest in Italy at 7.5%.  From this rate, the estimated number of COVID-19 cases would be approximately 630,000 across the seven cities, compared to the 62 679 reported from swabs in the Filosto study. Using seroprevalence as the denominator, we believe the reported calculated incidence reduces from 47.9 to approximately 4.76 per 100,000 COVID-19 infections. The 7.5% seroprevalence may also be an underestimate because of selection bias (under half the planned sample size were tested) or imperfect sensitivity of the assay. It is striking that the seroprevalence reported in London from the NHS Blood Transfusion Service at a similar time was 17.5%; Northern Italy, recognised as one of the hardest hit populations worldwide, seemingly had a seroprevalence of less than half that. Without major differences in health care structure and economic status of these two European countries it is difficult to explain the discrepancy other than the recorded seroprevalence figures were lower than actual infection rates.
Importantly, no account is given for the dramatic 3.3-fold decline in non-COVID GBS to 0.29/100000 per year either. One potential reason for the apparent decline is the mis-attribution of GBS causation, where COVID-19 causation is applied to every COVID-19 positive GBS case. The authors do not explore the possibility of alternative causes of GBS and incidental COVID-19 infection. For example, four of the COVID-19 cases had gastrointestinal symptoms but causation was still attributed to COVID-19 despite a gastroenteritis being the commonest precursor of GBS and occurring less frequently as a COVID-19 specific feature. The authors have presented no data on the exclusion of other causes of GBS such as Campylobacter serology etc. and this is of relevance, as without excluding other causes as far as possible the possibility of misattribution again exists.
Part of the justification Filosto et al make for the causative link is published data to support an interaction of the SarCoV2 spike protein with ganglioside GM1, referencing in silico modelling studies.  These studies model shape and energy efficient interactions of the Receptor Binding Domain (RBD) of the spike (S-)protein of SARS-CoV2 with the ACE2 Receptor. Convincing in silico predictive data of the subsequent available N-Terminal Domain of the S-Protein also indicate energy efficient potential binding sites for two GM1 molecules. However, these are models and interactions have not been shown to occur in or ex-vivo. Furthermore, there are no data to suggest that ACE2 receptors are present in peripheral nerve axons or myelin to provide the basis of binding to a colocalised ganglioside.
GBS is a para- or more likely post-infectious autoimmune neuropathy, although acute polyradiculoneuropathies indistinguishable from classical Campylobacter jejuni enteritis associated GBS may occur by direct infection in Zika- and West Nile viruses and others. These cases of GBS occur with short latency whereas classical GBS occurs one to three (and possibly up to six) weeks after infection after an immune response induction. In the series of Filosto et al their median time from infection to GBS was 23 days (IQR 16-34), almost all within the COVID-19 symptom period (5 cases started after COVID symptoms resolved and thus are most consistent with a post-infectious timing). Thus, if it proposed that covalent binding to gangliosides is a mechanism for direct infection of peripheral nerves this would be considered too long a latency, and if an immune response is the key, then COVID-19 spike protein binding to a ganglioside is possibly irrelevant to the pathogenesis. Clearly more work is required to establish whether a definitive temporal relationship exists between COVID-19 and GBS, and more so the pathogenic mechanism which links infection to neuropathy.
The Witebsky Postulates, with some minor modifications, have stood the test of time to provide support for autoimmune linkage of a pathogen to a disease. None are really fulfilled here and we should be very cautious about drawing causative conclusions from observational studies of small numbers of patients where the prevalence of COVID-19 in the population is so high as to have affected very many people who may have been presenting to hospital anyway.
Lastly, it is reassuring that the treatment and outcomes of patients are no different between COVID-19 positive and negative GBS patients and that given the lack of any spike in GBS presentations in Northern Italy (with the total incidence rate being 2.43/100000) we do not expect an additional pandemic of acute neuromuscular paralysis.
Stephen Keddie, Julia Pakpoor, Aisling Carr, Michael P Lunn
1. Filosto M, Cotti Piccinelli S, Gazzina S, et al. Guillain-Barré syndrome and COVID-19: an observational multicentre study from two Italian hotspot regions. J Neurol Neurosurg Psychiatry 2020;0:jnnp-2020-324837. doi:10.1136/jnnp-2020-324837
2. ISTAT, Ministero della Salute. Primi risultati dell’indagine di sieroprevalenza sul SARS-CoV-2. 2020;:10.
3. Public Health England. Sero-surveillance of COVID-19. 2020.https://www.gov.uk/government/publications/national-covid-19-surveillanc... (accessed 13 Jul 2020).
4. Fantini J, Di Scala C, Chahinian H, et al. Structural and molecular modelling studies reveal a new mechanism of action of chloroquine and hydroxychloroquine against SARS-CoV-2 infection. Int J Antimicrob Agents 2020;55. doi:10.1016/j.ijantimicag.2020.105960
5. Witebsky E, Rose NR, Terplan K, et al. Chronic thyroiditis and autoimmunization. J Am Med Assoc 1957;164:1439–47. doi:10.1001/jama.1957.02980130015004