Autoimmune Sequelae After Delta or Omicron Variant SARS-CoV-2 Infection in a Highly Vaccinated Cohort

This cohort study examines the risk of developing an autoimmune condition after a SARS-CoV-2 Delta or Omicron variant infection among adults in Singapore, which has high COVID-19 testing, vaccination, and booster rates.


Introduction
Autoimmunity has been reported in patients with severe COVID-19, 1,2 with development of autoantibodies associated with risk of postacute sequelae. 24][5] Autoantibody prevalence after COVID-19 is associated with greater severity of initial infection. 11,12COVID-19 vaccination, predominantly with messenger RNA (mRNA) vaccines, has demonstrated some degree of protection against postacute sequelae across multiple organ systems. 13During the current era of COVID-19 endemicity dominated by milder Omicron (B.1.1.529)variants and availability of COVID-19 booster vaccines, reduced severity of acute illness attributed to milder infection and booster vaccination may potentially translate to lower long-term risk of postacute autoimmune sequelae.
However, while increased risk of autoimmune diseases after SARS-CoV-2 Omicron variant infection has been observed across several retrospective population-based cohort studies, [6][7][8] clinical evidence of the potential protection from vaccination has been mixed.In a large EHR study extending into the predominance of the Omicron variant and including almost 4 million patients, risk of autoimmune disease after infection was higher in the vaccinated group than the unvaccinated group, although fewer than 10% of the patients were vaccinated. 6In contrast, vaccination was associated with reduced risk of autoimmune disease after COVID-19 in other population-based cohort studies from the pre-Omicron and Omicron variant periods, 7,9,10 although the risk of autoimmune sequelae was not specifically evaluated in cohorts who received boosters.In this study, we aimed to estimate the 300-day risk of new-incident autoimmune sequelae after SARS-CoV-2 Delta (B.1.617.2) and Omicron BA.1 or BA.2 variant infection in adults who received COVID-19 vaccines and boosters compared with a contemporary control group without infection.

Methods
This cohort study was part of national public health research under the Infectious Diseases Act of Singapore.In accordance with the Infectious Diseases Act, individual patient consent and separate ethics review by an institutional review board were not required.We followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

Study Setting and Databases
Singapore is a multiethnic city-state in Asia with a population of 5.4 million.The national SARS-CoV-2 testing registry was used to construct cohorts of adult Singaporeans (ie, citizens or permanent residents of Singapore) with first SARS-CoV-2 infection during periods of community transmission from September 1, 2021, to March 7, 2022, predominated by the Delta and Omicron BA.1 or BA.2 variants.The Delta variant predominated community transmission by September 2021; in December 2021, the Omicron BA.1 or BA.2 variant displaced the Delta variant as the predominant strain. 14Pfizer/BioNTech) and mRNA-1273 (Moderna) were originally approved for use in a 2-dose primary mRNA vaccination series, with subsequent recommendation of boosters.15 By March 2022, 91% of the population had completed a primary vaccination series and 69% had received booster doses, 15 with 95% or more receiving mRNA vaccines.

JAMA Network Open | Infectious Diseases
Risk of prespecified new-incident autoimmune sequelae was assessed using the national health care claims database (MediClaims), encompassing all public and private inpatient and outpatient health care practices.Participation in the national government-administered medical savings scheme (MediSave) and national health insurance plan (MediShield) is mandatory for all Singaporeans. 16This environment enabled comprehensive data capture across different health care settings.][5][6][7][8][9][10] The ICD-10 codes used for outcomes of interest are listed in the eAppendix in Supplement 1.

Covariates
The following covariates were incorporated: demographics (age, sex, and ethnicity), vaccination status at the index date, comorbidity burden (Charlson Comorbidity Index [range: 0 to Ն5, with higher scores indicating more severe burden]), and socioeconomic status (SES).Information on sociodemographic characteristics (including ethnicity) was defined based on information recorded in national databases maintained by the Ministry of Health.Ethnicity was assessed in this study to evaluate if risks of autoimmune sequelae differed across ethnic groups.Housing type was used as a surrogate marker of SES. 17 The majority of Singaporeans (Ն90%) live in owner-occupied public housing under a tiered subsidy scheme, with the purchase eligibility for more highly subsidized smaller-sized flats dependent on monthly household income. 17

Statistical Analysis
Risks and excess burdens of prespecified new-incident autoimmune sequelae after SARS-CoV-2 infection were estimated using controls as a comparator at 31 to 300 days from the index date.For estimation of risks for each new-incident sequelae, a subcohort of individuals without a history of the studied complication being reported was constructed.Individuals were excluded from each subcohort if they had a preexisting history of the specific sequelae being studied in the past 5 years.
Baseline sociodemographic characteristics of cases with Delta or Omicron variant infection and controls were described, along with standardized mean differences (SMDs) between groups, at baseline and after adjustment using overlap weighting 18 and incorporating all available covariates: demographic characteristics (age, sex, and ethnicity), SES (housing type), COVID-19 vaccination status at the index date (not fully vaccinated, fully vaccinated, or received boosters), and comorbidities.An SMD less than 0.1 was taken as the threshold for good covariate balance after weighting.Hazard ratios (HRs) of new-incident autoimmune sequelae between cases and controls were then estimated using Cox proportional hazards regression models, with overlap weights applied.Excess burden per 1000 persons at 300 days of follow-up was defined as the excess number of a specific new-incident autoimmune sequelae due to COVID-19 per 1000 persons in the 300-day follow-up period.Excess burden was computed based on the differences in weighted incidence rates of the prespecified postacute sequelae between cases and controls over the 300-day follow-up period.Risks and excess burden of new-incident autoimmune sequelae after SARS-CoV-2 infection were also stratified by vaccination status (fully vaccinated vs received boosters) and severity of initial infection (ambulatory care vs hospitalization).Subgroup analyses by age (18-64 years vs Ն65 years), ethnicity (Chinese, Malay, or Indian), and sex (male vs female) were also conducted.
Additional analyses were performed to investigate the robustness of the results.First, the risk of a series of negative outcomes (various malignant neoplasms) was evaluated across the entire study cohort.These outcomes were chosen because no prior knowledge, to date, supports the existence of a causal association between SARS-CoV-2-infection and cancer risk, with increased cancer risk not reported in survivors of severe acute respiratory syndrome coronavirus and Middle East respiratory syndrome coronavirus. 19Second, use of inverse probability weighting (computed as 1/propensity score for cases; 1/[1 -propensity score] for controls) as an alternative to overlap weighting in the regression model was explored.Third, although we focused on prespecified hypotheses and end points, rather than every possible comparison, results with P < .004were considered additionally robust to correction for multiple comparisons (Bonferroni correction).

Results
In Estimates of risk were robust to sensitivity analyses using inverse propensity weighting instead of overlap weighting (eTables 5-7 in Supplement 1).No significantly elevated risk of new-incident autoimmune sequelae was observed across subgroup analyses by sex, age, and ethnicity (eTables 8-14 in Supplement 1).There was also no significantly increased risk of controls with negative outcomes (eTable 15 in Supplement 1).

Discussion
In a cohort of cases largely vaccinated against COVID-19 who also received boosters, there was no significantly increased risk of any autoimmune sequelae up to 300 days after infection during the predominance of the Delta and Omicron BA.1 or BA.2 variants, compared with controls.An exception was an increased risk of Sjögren syndrome and bullous skin disorders after Omicron variant infection, which did not cross the threshold of significance when adjusted for multiple comparisons.
These findings substantially differ from the majority of studies that have reported increased long-term incidence of a wide range of autoimmune sequelae after SARS-CoV-2 infection, albeit from waves of earlier SARS-CoV-2 variants [3][4][5]9,10 and without accounting for the potential protection from COVID-19 vaccination, [3][4][5]8 including boosters. Fo instance, in a large study that examined the risks of postacute autoimmune sequelae in the pre-Omicron variant era across an unvaccinated population drawn from 48 US-based health care organizations, elevated risks of a wide range of autoimmune conditions, including Sjögren syndrome, rheumatoid arthritis, spondyloarthropathies, and psoriasis (AHR range, 2.32-3.21)were observed after COVID-19 compared with controls.3 In a large EHR-based cohort study including approximately 4 million patients, fewer than 10% of whom were vaccinated, risk of any autoimmune disease was lower during the predominance of the Omicron variant. 6 Simlarly, in a Hong Kong population with SARS-CoV-2 infection during the Omicron variant period (81.9% of whom were vaccinated with 2 doses), the AHRs of new-incident autoimmune conditions after infection were generally lower than estimates from the pre-Omicron variant studies.7 a The SMDs at baseline and after overlap weighting of cases (individuals with SARS-CoV-2 infection) and controls (individuals with negative polymerase chain reaction/ rapid antigen test results) were weighted from original samples.An SMD less than 0.1 was the threshold for good covariate balance after weighting.
b Did not complete primary vaccination series was defined as either unvaccinated or partially vaccinated with a single dose of an mRNA COVID-19 vaccine (either BNT162b2 or mRNA-1273).Completed primary vaccination series only was defined as receiving 2 doses of an mRNA COVID-19 vaccine at least 8 weeks apart.Received 1 or more booster vaccination doses was defined as receiving at least a third dose of an mRNA COVID-19 vaccine 6 to 9 months after the second dose.
c Includes other ethnicities (eg, Eurasian and Arab) or multiple ethnicities.
d Comorbidity burden was defined using the CCI (range: 0 to Ն5, with higher scores indicating more severe burden), which consists of the following comorbidities: myocardial infarction, chronic heart failure, peripheral vascular disease, cerebrovascular accident, dementia, chronic obstructive pulmonary disease, connective tissue disease, peptic ulcer disease, diabetes, hemiplegia, liver disease, moderate to severe kidney impairment, solid tumor, leukemia, and HIV infection with AIDS.Omicron variant infection is anticipated to be low and confined to the small subset of cases with more severe disease.
Clinical evidence is mixed regarding the protective properties of vaccination against risk of autoimmune sequelae after SARS-CoV-2 infection.In a large EHR study across 74 health care organizations that extended into the period of Omicron variant predominance and included almost 4 million patients, vaccination (vs no vaccination) was associated with an 18% increased risk of a new-incident autoimmune disease after infection; however, only 8.2% of patients were vaccinated, which was likely a substantial underestimation. 6In contrast, in a cohort of over 10 million Korean adults with infection during the predominance of the pre-Delta and Delta variants, receipt of 1 or more vaccine doses was associated with reduced risk for new-incident autoimmune disease after mild but not moderate-to-severe COVID-19, although further breakdown of vaccination status by booster and nonbooster groups was unavailable. 9Similarly, completion of a primary vaccination series attenuated the risk of new-incident autoimmune sequelae after SARS-CoV-2-infection in the Omicron and pre-Omicron variant era in population-based cohorts in Hong Kong and Korea, respectively 7,10 ; however, 45% of the Korean cohort remained unvaccinated, 10 and 55% of the Hong Kong cohort had yet to receive a third vaccine dose.In the highly vaccinated population of Singapore, while elevated risk of vasculitis was observed with vaccine-breakthrough Omicron variant infections compared with controls, this outcome was attenuated by booster vaccination; no increased risk of vasculitis was observed in the subgroup who received boosters.The results support the continued enrollment of at-risk individuals for booster vaccinations during COVID-19 endemicity.Given persistent vaccine hesitancy, highlighting benefits may help increase acceptance, 30 particularly as COVID-19 vaccination is likely to be protective against autoimmunity, rather than contributory.In the Singaporean population, severe flares of autoimmune disease after COVID-19 vaccination were rare in individuals with preexisting autoimmune conditions. 31

Strengths and Limitations
The strengths of this study include use of comprehensive nationwide registries to classify SARS-CoV-2 infection and vaccination status to investigate the incidence of autoimmune diseases after COVID-19 in a highly vaccinated population.New-incident autoimmune diseases were identified from a comprehensive health care claims database with nationwide coverage, minimizing selection bias caused by loss to follow-up.A large number of potential confounders were controlled for, and the analyses took into account adjustment for multiple testing, thereby reducing the likelihood of falsepositive findings.A test-negative design was also used, reducing bias associated with confounding by health care-seeking behavior.
This study has several limitations.First, while comprehensive nationwide registries were used to classify SARS-CoV-2 infection status, control groups may still be contaminated by undiagnosed or asymptomatic infections, although mandatory testing at health care settings and provision of free testing were put in place to mitigate this issue.Misclassification of exposure would bias estimated HRs and excess burdens downward.Second, variant was imputed according to the period of predominant transmission and not individual-level sequencing.Third, administrative claims data were used to capture autoimmune diagnoses, while comprehensive, additional corroborative information, such as laboratory tests for autoantibodies, was unavailable.Fourth, for some individual autoimmune sequelae, rarity resulted in only a small number of new-incident cases; therefore, estimates could not be provided across all subgroups and could have resulted in imprecise interpretations.Further research is needed to replicate these findings in other populations and to measure autoantibody profiles in prospective cohorts of individuals with SARS-CoV-2 infection.

Conclusions
In this retrospective cohort study of vaccinated adults in Singapore who also received boosters against COVID-19, no significantly elevated long-term risk of autoimmune sequelae was observed

JAMA Network Open | Infectious Diseases
Abbreviations: AHR, adjusted hazard ratio; NA, not available.a SARS-CoV-2 cases were stratified by severity of initial infection (mild cases managed in ambulatory care alone vs cases requiring hospitalization).Results for the subgroup of individuals with severe infection (requiring oxygen or intensive care unit/high-dependency admission) were not presented because risks of individual autoimmune diagnoses could not be estimated due to too few cases; however, there was no increased risk of any autoimmune diagnosis (composite outcome) in severe cases compared with controls during Delta (AHR, 1.40; 95% CI, 0.86- 12toimmune Sequelae After Delta or Omicron Variant Infection in a Vaccinated Cohort CoV-2 infection was defined as a positive result on either a polymerase chain reaction (PCR) test or rapid antigen test (RAT) recorded in the national registry.During the study period, all Singaporeans with acute respiratory illness were strongly encouraged to seek free confirmatory testing for SARS-CoV-2 infection at a health care facility or practitioner.Testing was mandatory for individuals presenting with acute respiratory illness symptoms, and all positive results were reported to the local Ministry of Health.12Outcomeswere compared across vaccination status and severity of JAMA Network Open.2024;7(8):e2430983.doi:10.1001/jamanetworkopen.2024.30983(Reprinted) August 30, 2024 2/14 Downloaded from jamanetwork.comby guest on 09/10/2024 SARS- PCR indicates polymerase chain reaction; RAT, rapid antigen test; T 0 , the date of the first positive PCR or RAT test result in cases and the date of the negative test results in controls.

Table 1 .
Baseline Sociodemographic and Clinical Characteristics of Study Population With Standardized Mean Differences Before and After Overlap Weighting

Table 2 .
Risks and Excess Burdens of Prespecified New-Incident Autoimmune Diagnoses in SARS-CoV-2 Cases and Controls During Delta and Omicron BA.1 or BA.2 Variant Periods overlap weighted and regression adjusted based on demographic characteristics (age, sex, ethnicity), socioeconomic status (housing type), vaccination status (not fully vaccinated, fully vaccinated, fully vaccinated and received boosters), and comorbidities.
a Numbers in each subcohort for each specific autoimmune diagnosis do not add up to the original number of SARS-CoV-2 cases and controls because, for estimation of risks for each new-incident autoimmune diagnosis, a subcohort of individuals without history of the diagnosis in the past 5 years was constructed.b Excess burden greater than 0 denotes excess burden in a respective autoimmune diagnosis among infected cases vs controls.c AHR greater than 1 denotes higher risk of a respective autoimmune diagnosis among cases vs controls.d Each model was e Risks could not be estimated due to too few numbers of a new-incident autoimmune diagnosis for that subcategory.f Included mixed connective tissue disease, Behçet disease, and polymyalgia rheumatica.

Table 3 .
Risks and Excess Burdens of Prespecified New-Incident Autoimmune Diagnoses in SARS-CoV-2 Cases, Stratified by Initial Infection Severity, vs Controls During Delta and Omicron BA.1 or BA.2 Variant Periods Risks could not be estimated due to too few numbers of a new-incident autoimmune diagnosis for that subcategory.Included mixed connective tissue disease, Behçet disease, and polymyalgia rheumatica.
e f

Table 4 .
Risks and Excess Burdens of Prespecified New-Incident Autoimmune Diagnoses in SARS-CoV-2 Cases vs Controls, Stratified by Vaccination Status, During Delta and Omicron BA.1 or BA.2 Variant Periods (ie, completed primary vaccination series only) was defined as receiving 2 doses of an mRNA COVID-19 vaccine at least 8 weeks apart.Booster status (ie, received 1 or more booster vaccination doses) was defined as receiving at least a third dose of an mRNA COVID-19 vaccine 6 to 9 months after the second dose.Results for the subgroup of individuals who did not complete primary vaccination series (ie, either unvaccinated or partially vaccinated with a single vaccine dose of an mRNA COVID-19 vaccine) were not presented because risks of individual autoimmune diagnoses could not be estimated due to too few cases; however, there was no increased risk of any autoimmune diagnosis (composite outcome) in unvaccinated or partially vaccinated cases, overlap weighted and regression adjusted based on demographic characteristics (age, sex, ethnicity), socioeconomic status (housing type), and comorbidities.
a Fully vaccinated status b AHR greater than 1 denotes higher risk of a respective autoimmune diagnosis among cases vs controls.c Each model was d Excess burden greater than 0 denotes excess burden in a respective autoimmune diagnosis among infected cases vs controls.e Risks could not be estimated due to too few numbers of new-incident autoimmune diagnosis for that subcategory.f Included mixed connective tissue disease, Behcet disease, and polymyalgia rheumatica.