RISK OF THROMBOEMBOLISM AFTER COVID-19 VACCINATION AND COVID-19 INFECTION

2.1 Data

We used Electronic Health Record (EHR) of Corewell Health East healthcare system in this study. We included patients who were registered with a primary care physician within 18 months prior to Jan 1^st 2021, and had any outcome of interest within the observation period (Dec 1st 2020 – August 31st 2022). We identified acute thrombosis events by a hospital admission or death associated with venous thromboembolism or arterial thromboembolism, see Table 4 in Appendix for a list of ICD-10 (International Classification of Diseases version 10) diagnosis codes. We further excluded individuals who were under 18 on Jan 1^st 2021, had incomplete covariate data, or had a hospital admission for thrombosis within 365 days before the first event in the observation period.

Our primary interest is to study the mRNA-based vaccines effect on the occurrence of thrombosis using the SCCS design. We included patients who had at least the first vaccine dose of mRNA-1273 (Moderna) or BNT162b2 (Pfizer) and experienced any outcome of interest. We further excluded patients who had an infection 90 days prior to any dose of vaccination. We used the EHR data system to identify vaccine exposure, which includes the date of vaccination, vaccine type and dose order.

The same SCCS was used to study the risk of thromboembolism associated with a SARS-Cov-2 infection, in which we included patients who had at least one positive PCR test and a thrombosis event during the observational period. However, as a result of misclassification, the SCCS derived an inflated IRR 23.46. Therefore, we proposed a case-control design, in which we included patients who either had a thrombosis event or hospitalized injury in the observational period, (see Table 4 in Appendix for a list of ICD-10 codes), and a SARS-CoV-2 (PCR or antigen) test or within 28 days period prior to the hospital admission date. We determined SARS-CoV-2 infection status based on the test results during these 28 days period. If an individual only had negative PCR or antigen test results, this subject was classified into noninfection. Moreover, we identified disease status by the earliest event, in which patients having thrombosis is classified into case group. Additionally, if an individual had an admission for both thrombosis and injury, then this subject would be considered as having thrombosis. In this way, each of individual were classified into one disease status (case-control).

2.2 Statistical Method

We used self-controlled case series (SCCS) design to examinate the association of acute thromboembolism and the first two doses of mRNA-based COVID-19 vaccines (mRNA-1273 or BNT162b2) from December 1^st 2020 to August 31^st 2022. The SCCS method compares the incidence of thromboembolism after and before vaccination. In this method, subjects are their own control and comparisons are made within subjects, therefore, all time invariant confounding is removed. We included subjects who had a thrombosis event and had at least the first vaccine dose in the study period. The control period (i.e., unexposed period) is defined from December 1^st 2020 to 28 days prior to the first dose of vaccination. We excluded the 28 days prior vaccination to avoid bias due to contra-indications (21). Two separate risk periods for the first and second dose were defined from the vaccination date to 28 days, the date of infection or death, whichever occurred first. See Figure 1 in the Appendix. We also excluded subjects who had COVID-19 infection within 90 days before the first vaccine dose to remove the confounding effect of infection on the outcome. To ensure independence between repeatedly occurred outcomes, we considered outcome events occurred at least one year after the previous event. We used the conditional Poisson regression including an offset for the length of risk period to obtain the incidence rate ratio (IRR) of thrombosis after and before COVID-19 vaccination.

• Download figure
• Open in new tab

Figure 1.

Direct and indirect association between mRNA vaccine and thromboembolism.

It is challenging to study the association between the adverse events and the COVID-19 infection due to inaccurate infection data in patients who did not have the thrombosis event. Patients visited emergency department or before admitted to hospital, were likely to take a COVID-19 test (antigen or PCR) especially during early pandemic. Therefore, observing event of thromboembolism is more likely to be associated with a COVID-19 infection because subjects with adverse events is more likely to take the COVID-19 test compared to subjects without any adverse event, even if there is no association between the acute events and an infection. The under-estimated infection rate in controls group leading to the overestimation of the risk of thromboembolism after a COVID-19 infection [estimated IRR is 23.46 showing an inflation]. This fact is evidenced by the analysis of the hospitalized injury as the outcome, which has the same misclassification issues, but is not associated with COVID-19 infection, we found IRR is 10.35.

The main idea is to address the misclassification problem in the control subjects (i.e., subjects without thromboembolism). We proposed a simple and efficient method to quantify the association while dealing with the misclassification issue. Early in the pandemic, everyone who visited ED or before admitted to hospital were tested for COVID-19. Therefore, we assumed subjects who had a thrombosis event had correct infection data (i.e., test has 100% accuracy), and the misclassification only occurred in subjects without thrombosis (control subjects). The idea is to select a subset of control group whose infection data is certain. To this end, we selected event of hospitalization related to injury, as injury is not associated with COVID-19 infection and hospitalized event had correct infection data, we selected it as the control group, and obtained the odd ratio measuring ratio of the odd of thrombosis and inpatient injury after and before a positive test.

In the case-control design, we compared the odds of infection (exposed) vs uninfection (unexposed) in the cases (thromboembolism) vs. controls (injury). If an individual has a thrombosis/injury (case/control) event, we considered this subject is infected by SARS-CoV-2 (exposed) if they had a positive PCR or antigen test within 28 days before the event. Moreover, to further avoid the possible systematic error caused by unreported infection, only patients who had all negative PCR test result in the period of 28 days before the occurrence of event is considered as uninfected (unexposed). Logistic regression models were adjusted for age, race, gender, Charlson comorbidity index (CCI), and prior vaccination status (Yes/No). The CCI was obtained using the R package comorbidity and categorized into 4 categories, ’0’, ’1-2’, ’3-4’, and ’≥ 5’ (22), (23). We reported Odd Ratios (OR), 95% confidence interval (CI).

Statistical analyses were performed in R 4.3.0. OR and IRR estimates were presented with 95% CIs and p-value. We used de-identified EHR data, the use of which was approved by the Institutional Review Board (IRB) of Corewell Health.

3.1 Study population

During the study period, there were 448524 COVID-19 vaccinated subjects in Corewell Health East data base, among which 170960 (38.12%) had the first two doses of mRNA-1273 and 277564 (61.88%) took BNT162b2. Overall, the number of patients who were fully vaccinated was 427458 (95.30%), and there were 21066 (4.70%) patients having only one dose. The median age is 57 (with interquartile range [IQR]: 42 – 69), and 60.26% patients were female and 39.74% are male.

Among those vaccinated, there were 12415 patients, who had at least one event and had the first dose of either mRNA-1273 or BNT162b2 vaccine. Moreover, these patients did not have any SARS-CoV-2 infection within 90 days before vaccination or between vaccination date and date of event occurrence, and they also had a clean one-year medical history, so that the first hospital admission relating to thrombosis in observational period is the newest within one year. Demographic of patients in the study of vaccination is presented in Table 2 in Appendix. We identified 16722 events in the observational period, among which there were only 2139 thrombosis occurred in the control period, 527 and 575 events in the risk periods after the first and second dose of vaccination, see Table 1 in Appendix. In conclusion, there were only 2798 patients and total of 3241 events were considered in our conditional Poisson regression.

During observational period, there were 26052 patients receiving a positive PCR/antigen test result. The median age is 46 (with IQR: 28-62), and 58.63% are female and 41.37% are male.

There were 20249 patients who got a hospital admission relating to either thromboembolism or injury during the observational period. These patients had at least one COVID-19 test recorded in 28 days prior to the hospital admission. There were 43.04% patients in control group (injury) and the rest of 56.96% are in the case group (thromboembolism). Demographic of patients in the study of infection is presented in Table 3 in Appendix.

3.2 The association between thromboembolism and exposures

3.3 Estimate the overall effect of vaccination on thromboembolism while considering infection

We extended our investigation to evaluate the overall influence of COVID-19 vaccination on the occurrence of acute thromboembolism events. COVID-19 vaccines dramatically reduce the SARS-Cov-2 infection rate (24), (25), (26), hence potentially indirectly decreases the likelihood of experiencing an acute thromboembolism event among the vaccinated group. Figure 1 below illustrates the direct and indirect effect of the COVID-19 vaccination on thromboembolism while considering the vaccine efficacy (VE). As presented in the diagram, the association between thromboembolism and COVID-19 vaccination is described by two paths, the direct association represents the direct increased risk of thromboembolism after any dose of vaccination, and the indirect association between thrombosis and vaccination is represented by the potential reduction in the risk of thromboembolism through the decreased risk of COVID-19 infection by vaccination. We estimated the overall influence of vaccination on the occurrence of thromboembolism by considering both direct and indirect paths.

For a vaccinated subject, the total risk is P(Y|V, Ī) + P(I|V) × P(Y|I, V), where P(Y|V, I) is the direct risk of thromboembolism after vaccination given that infection has not yet occurred, and P(I|V) × P(Y|I, V) is the indirect risk calculated by multiplying the risk of COVID-19 infection of a vaccinated subject, P(I|V), and the risk of thromboembolism given a COVID-19 infection in the vaccinated group, P(Y|I, V). Similarly, the overall risk of thromboembolism for an unvaccinated subject is as . Hence the relative risk of thrombosis for a vaccinated subject compared to an unvaccinated subject is The terms RR_V is the relative risk of thromboembolism associated with COVID-19 vaccination, and is the relative risk of thromboembolism after a COVID-19 infection in the group of unvaccinated. The term RR_IV is relative risk of thromboembolism between the group having both type of exposures, vaccination, and then infection, compared to unexposed group. Our analysis with the Corewell Health EHR data in the previous section gave an IRR of 1.19 as the measure of the association between thromboembolism and COVID-19 vaccination, therefore, RR_V = 1.19. We also obtained an odd ratio and OR_IV = 1.43 from the analysis using case-control design. We therefore set and RR_IV = 1.43 as the event of thromboembolism is rare (27). Hence, the above overall relative risk becomes The overall relative risk of thromboembolism after a SARS-CoV-2 vaccination depends on the infection rate for both vaccinated and unvaccinated subjects. We defined vaccine efficacy , then with a given infection rate for an unvaccinated subject, , we derived risk of infection for a vaccinated subject as , and obtained the overall risk, RR _Overall, as a function of VE.

Figure 2 illustrates the overall relative risk of thromboembolism after a mRNA-vaccination as a function of VE. As VE increases from 0 to 1, the overall risk decreases. Moreover, infection rate, , also affects the reduction of the risk of thromboembolism. During the period with higher infection rate, the benefit of vaccination is stronger. More specifically, with VE is .8, the overall risk of thromboembolism after vaccination is decreased by 10% during the period of infection rate .2, while it decreased only 3% during the period of infection rate .15.

• Download figure
• Open in new tab

Figure 2.

Overall risk of thromboembolism, comparison between vaccinated and unvaccinated groups. Infection rate .15 and .2 is the probability of SARS-Cov-2 infection before vaccination (27).