RR:C19 Evidence Scale rating by reviewer:
Reliable. The main study claims are generally justified by its methods and data. The results and conclusions are likely to be similar to the hypothetical ideal study. There are some minor caveats or limitations, but they would/do not change the major claims of the study. The study provides sufficient strength of evidence on its own that its main claims should be considered actionable, with some room for future revision.
This study aimed to examine the risk of incident autoimmune disease following acute SARS-CoV-2 infection. A matched cohort design, and secondary analysis of routine records generated through health insurance, was used.
The study population comprised individuals continuously registered with specific German health insurance schemes- collectively covering around half the population from January 2019 to June 2021 inclusive.
Exposed individuals were those with a documented outpatient contact or inpatient admission with virologically confirmed COVID-19 as the coded diagnosis during 2020 (ie prior to the start of the vaccination programme). Results of community-based test results do not appear to have been available/considered.
Unexposed controls were those with no such documented infection Jan 2020 to June 2021.
Both groups were followed to June 2021 to determine incident cases of autoimmune disease with onset at least 3 months following the infection/matching (with look back to January 2019 to identify preexisting/prevalent cases), with 15 months being the maximum duration of follow up.
Exposed and unexposed individuals were matched 1:3 on age, sex, presence of any prevalent autoimmune condition, and propensity score reflecting other co-morbidities. When considering specific autoimmune conditions, exposed/control matched groups were excluded if any individual had the specific condition of interest as a pre-existing/prevalent condition.
64 specific outcomes reflecting 41 different autoimmune diseases were considered. These were identified using diagnostic coding on outpatient and inpatient records, supplemented by prescription records as appropriate.
Poisson regression was used to identify the incidence of autoimmune conditions following infection/matching in exposed and control groups, and hence incidence rate ratios. Due to data access restrictions, patient level data from different insurers was analysed separately by the responsible organisations. Aggregate results were then pooled for further Poisson regression to provide overall results.
In general the results are well presented.
Clear information on included participants (inclusion flow chart and descriptive characteristics of exposed and control groups) is provided.
Information is provided on the absolute and relative risk of the different outcomes in exposed and control groups. Findings show increased risk in the exposed group, and this is very consistent across different conditions and when more and less stringent case definitions (diagnostic code +/- associated prescription record) for specific condition are used. The excess relative risk appears similar in men and women, and across age groups, but appears to increase with increasing severity of the index SARS-CoV-2 infection (outpatient, inpatient, critical care).
The authors correctly avoid a causal interpretation of their findings. There remains a possibility of residual confounding or ‘surveillance bias’, with patients with prior COVID – or their doctors – more alert to symptoms that may lead to diagnosis of an autoimmune disease.
The authors also helpfully note that individuals with more (compared to less) severe COVID are likely to have been identified as exposed, and there may be some contamination of the control group with infected individuals.
This is a well conducted and well presented study that provides robust evidence demonstrating an association between SARS-CoV-2 infection and increased risk of incident autoimmune disease 3-15 months following infection.
I have no specific requests for improvement as key limitations are well covered in the discussion.
The paper would benefit from further review from colleagues with specific expertise in immunology. Statistical review would also be helpful given the relatively subtle methods involving Poisson regression on pooled aggregate data.