The prognostic role of anticoagulants in COVID-19 patients: national COVID-19 cohort in South Korea

Introduction

The severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) and corresponding COVID-19 disease can result in respiratory, gastrointestinal, neurological and other atypical symptoms (1). In up to one-third of COVID-19 patients, coagulation abnormalities may occur, termed COVID-19-associated coagulopathy (CAC) (2). CAC is associated with increased venous (pulmonary embolism, deep vein thrombosis) and arterial (myocardial infarction and stroke) thromboembolic events (3).

Anticoagulants may have a role in preventing and treating CAC, according to a Cochrane systematic review reporting on hospitalized COVID-19 patients and outcomes associated with anticoagulants (4). Four studies compared heparin use to no treatment with respect to mortality (5-8), two (5,8) reported that anticoagulants decrease the odds of mortality, and two (6,7) reported there was no difference in odds of mortality between anticoagulants and lack of anticoagulants use. Trinh et al. compared therapeutic enoxaparin to prophylactic heparin, or prophylactic enoxaparin, and reported that 58% of patients receiving therapeutic anticoagulation had a significantly increased 35 days survival rate compared to just 14% in the prophylactic anticoagulation group (9).

Current clinical guidelines recommend for all acutely ill hospitalized COVID-19 patients to start a prophylactic-dose anticoagulation with low molecular weight heparin, unless contraindicated (10). Given the benefit of preventing deep vein thrombosis or pulmonary embolism in hospitalized patients for COVID-19, patients who were on anticoagulants before admission could have better outcome. Alternatively, patients who were on anticoagulants before hospitalization could have indication for anticoagulants use already, such as deep vein thrombosis, pulmonary embolisms or atrial fibrillation, which could lead to worse outcome. However, there exists limited published literature on the effects of pre-admission use of anticoagulants for hospitalized COVID-19 patients (11).

Given the paucity of data in the published literature there is a need for more rigorous investigations into the association between pre-admission anticoagulants use and COVID-19 outcomes. The aim of this study was to assess the association between pre-admission anticoagulants use and COVID-19 outcomes. We present the following article in accordance with the STROBE reporting checklist (available at https://apm.amegroups.com/article/view/10.21037/apm-21-3466/rc).

Methods

Data source and study population

The #OpenData4COVID19 project, launched on March 27, 2020 by the Ministry of Health and Welfare (MOHW) of Korea, released a nationwide, individual-level, and de-identified dataset on COVID-19 patients. This dataset is based on the health insurance claims database maintained by the Health Insurance Review and Assessment Service (HIRA) of South Korea, the sole nationwide governmental agency that operates a fee-for-service reimbursement system. The dataset covers all individuals who were tested by a reverse transcription-polymerase chain reaction (RT-PCR) method for COVID-19 (as of May 15, 2020). The information in the dataset was extracted from their claim records from the previous 3 years and includes each subject’s basic demographic information, healthcare utilization history, including diagnosis results, treatments, medications, and prescriptions from both inpatient and outpatient settings.

The HIRA COVID-19 dataset identified 234,427 consecutive individuals who received the RT-PCR test between January 1, 2020 and May 15, 2020. Seven thousand and five hundred ninety subjects were coded as positive for SARS-CoV-2 according to domestic codes (Appendix 1). We excluded individuals with age <40 years, since it is less likely that younger individuals experience severe COVID-19 adverse outcomes regardless of the anticoagulants use, and as anticoagulants use is less common in these younger patients, its use might be associated with higher risk features that are less generalizable to the general public. This left 4,610 individuals. Among them, our study cohort included 4,349 individuals for precise outcome measurement who were hospitalized for COVID-19 (Figure 1). We defined the cohort entry date as the date of admission for COVID-19 hospitalization. This cohort was followed up until May 15, 2020 or the date of discharge, whichever came first. In South Korea, most of the individuals who tested positive during this time interval were mandatorily hospitalized regardless of symptom for infection control purpose until full recovery, which is defined by two negative test results within 24 hours and the cessation of fever without medication use (12). Due to temporary unavailability of health facilities, a small number of test-positive individuals were not hospitalized.

Figure 1 Population-based cohort study design using the HIRA and KDCA database of South Korea. HIRA, Health Insurance Review and Assessment Service; CDC, Centers for Disease Control and Prevention; KDCA, Korean Disease Control and Prevention Agency.

This study was approved by the Human Investigation Review Board of Public Institutional Bioethics Committee designated by the MOHW. The requirement of informed consent was waived due to the retrospective study design and anonymity of the HIRA database (IRB # P01-2020-1262-001). The study was conducted in accordance with the Declaration of Helsinki (as revised in 2013).

Endpoints

Our primary endpoint was a composite endpoint of all-cause death, intensive care unit (ICU) admission, and mechanical ventilation use. We set secondary endpoints as the individual components of the composite endpoint. Each endpoint was defined from in-hospital ICD-10 codes and national procedures codes (Table S1).

Exposure

Individuals who had inpatient or outpatient prescription records of anticoagulants within 120 days of cohort entry (i.e., ascertainment window set to −120 to 0 d) were classified as anticoagulants users. We classified other individuals as non-users. We identified anticoagulants from anatomical therapeutic chemical (ATC) codes in claim records (Table S1).

Statistical analysis

The baseline characteristics for anticoagulants users and non-users were summarized by means with standard deviations (for continuous variables) and counts with percentage (for categorical variables). To investigate imbalances of covariate distributions between the two user groups, we calculated absolute standard difference (Asd) for each variable. Typically, a value with Asd ≤0.1 was preferred for indicating balance; Asd ≤0.2 was deemed acceptable for this analysis.

We conducted outcome analysis by employing the standardized mortality ratio weighting (SMRW) approach (13,14) as a main analysis, where the weight uses propensity scores (PSs). There was only a small number of anticoagulants users (128/4,349=3%) in our study population, and the SMRW approach is optimally suitable to attain covariate balance in such rare-exposure studies. The PS, a probability of a user to receive anticoagulants, was estimated by the multivariable logistic regression model that includes all potential confounders as predictors. The potential confounders we considered included age (in years), age square, health insurance type at cohort entry, 20 pre-exposure comorbidities, and 12 pre-exposure co-medications (Appendix 1). A balanced pseudo-population was created by weighting the anticoagulants users by 1 and the non-users by PS/(1-PS). Then, we fitted a univariate weighted logistic regression and estimated odds ratio (OR) with 95% confidence intervals (CI) to measure the effect size of anticoagulants use on each endpoint. In addition, to compare with methods with fewer adjustments, we estimated the ORs by unweighted univariate logistic regression models and unweighted multivariate logistic regression models that adjust for sex, age, insurance type, history of hypertension, and history of diabetes. For descriptive purposes, we reported the counts of the primary and secondary endpoints for each group.

We considered a subgroup analysis for the risk of the primary endpoint, stratified by (I) age, classified into two groups (<65 and ≥65 years), (II) sex, (III) history of atrial fibrillation and thromboembolism, and (IV) history of coronary artery disease, transient ischemic attack (TIA), stroke, and peripheral vascular disease. For each stratification, we conducted an SMRW-weighted trivariate logistic regression model that included the exposure variable, stratification indicator, and their interaction as predictors. We obtained the P value of the interaction term (P for interaction) to evaluate the significance of effect modification.

We examined the sensitivity of results against the choice of working definitions in two settings. First, we relaxed the study population from the hospitalized COVID-19 patients with age ≥40 years to confirmed COVID-19 positive patients with age ≥40 years. Second, to exclude short-term prescription and discontinuation of anticoagulants use before hospitalization, we narrowed down ascertaining window of the anticoagulants exposure, from 120 to 90 days. We repeated the outcome analysis for each change of the settings.

We investigated the sensitivity of the results against the choice of statistical methods by considering the following alternative statistical approaches. First, to improve comparability between the two groups, we excluded individuals with extreme PS values (PS <0.01 or PS >0.99) (SMR weighting with trimming). Second, to further adjust the imbalance after SMRW adjustment and obtain doubly robust estimates, we fitted a weighted multivariable regression where the weight is given by the SMRW and the predictors include confounders with SMRW-adjusted Asd more than 0.1 [SMR weighting with doubly robust method (15)]. Third, we fitted our main analysis where the weight was replaced with the standard inverse probability weighting (IPT weighting), 1/PS for the user group and 1/(1-PS) for the non-user group. Fourth, in our unweighted multivariable logistic regression model in the main analysis, we additionally included the estimated PSs to other covariates (outcome adjustment model). Finally, we used propensity score matching (PS matching). All statistical analyses were conducted using R 3.5.2.

Results

Four thousand and three hundred forty-nine hospitalized adults for COVID-19 were identified and included in this analysis. After SMR weighting, there were acceptable balances in all covariates except antiplatelet medication use which still has slight imbalance between users and non-users (Table 1).

For the primary endpoint of mortality, mechanical ventilation and ICU admission, no significant difference was noted between anticoagulants users and non-users (SMRW OR 1.11, 95% CI: 0.60–2.05). No differences were noted, among individual components (Table 2).

In stratified analyses of the primary endpoint by age, sex, history of atrial fibrillation and thromboembolism, and history of coronary artery disease, TIA, stroke, and peripheral vascular disease, there existed no effect modification (Table 3).

The results from the sensitivity analyses showed significant associations of adverse outcomes with anticoagulants use in some analysis, which were not seen in the main analysis (Tables S2-S5). When applying IPT weighting, anticoagulants use was associated with increased odds of the primary endpoint (OR 2.56, 95% CI: 1.01–6.11) and ICU admission (OR 2.94, 95% CI: 1.05–8.22). With SMR weighting with doubly robust methods, anticoagulants use was associated with increased odds of the primary endpoint (OR 1.34, 95% CI: 1.10–1.53), all cause death (OR 1.55, 95% CI: 1.26–1.90), and mechanical ventilation (OR 1.84, 95% CI: 1.33–2.54). Anticoagulants use was also associated with increased odds of mechanical ventilation when including all confirmed COVID-19 patients (OR 2.16, 95% CI: 1.03–4.51), redefining exposure windows to 90 days before and including the date of cohort entry in all confirmed COVID-19 patients (OR 2.54, 95% CI: 1.16–5.56) and in hospitalized COVID-19 patients (OR 2.42, 95% CI: 1.09–5.35), applying SMR weighting with iscrepa (OR 2.19, 95% CI: 1.01–4.77), or PS matching (OR 3.31, 95% CI: 1.47–7.45).

Discussion

This study reports on one of the largest South Korean datasets of COVID-19 patients for analysis of association between anticoagulants and COVID-19 outcomes. Our dataset reports on a nationwide study that was completely enumerated, which include most of the COVID-19 confirmed patients across the entire spectrum of COVID-19 severity (from asymptomatic to critical COVID-19 infection). This paper therefore presents on a unique cohort, as other previously-published papers typically report on hospitalized COVID-19 patients with moderate to critically ill COVID-19 infection. When considered along with the fact that PSs were used to control for confounding, this may be one of the most robust claims datasets published in the world.

Our main analysis finds that hospitalized COVID-19 patients who have been administered anticoagulants before admission experience a similar odds of suffering from the endpoints of all-cause mortality, mechanical ventilation and ICU admissions. These results are in line with those previously published by Klok et al. (16), Russo et al. (17) and Sivaloganathan et al. (18), which reported no association between baseline oral anticoagulants and adverse COVID-19 outcomes, including mortality (16-18) and ICU admission (18).

There was a discrepancy between this main primary endpoint analysis and corresponding analyses by alternative statistical approaches, especially the IPT weighting approaches and the SMR weighting with doubly robust method. However, Hajage et al. (13) and Ross et al. (14) noted that, in rare exposure regime such as our study, the IPT weighting as well as regression adjustment approaches could possibly be biased with inflated variance. In addition, since the SMR weighting with doubly robust method was first proposed in Moodie et al. (15), its empirical properties in the rare exposure regime is yet to be explored in the statistics literature and further statistical research is needed to understand its behavior, which is out of the scope of this study. Thus, to our knowledge, the main SMR weighting approach appears to have produced the most robust and reliable result.

Our study population noticeably differs from previously published trials. The study population were ICU patients for Klok et al. (16), emergency room patients for Russo et al. (17), and hospitalized patients for Sivaloganathan et al. (18), whereas our analysis reports on a nationwide cohort of all patients admitted for COVID-19 to South Korean hospitals. As much as 94.3% of all confirmed COVID-19 patients ≥40 years of age (n=4,610) were hospitalized (n=4,349) regardless of presence of symptom or disease severity in our study. This manifestation is a result of the South Korean government’s mandate for admitting all COVID-19 patients during the time interval studied to limit spread of the COVID-19 infection, including asymptomatic infections. As a result, this study provides a more complete picture, as it reports on the entire spectrum of COVID-19 patients, from asymptomatic to critical COVID-19 patients. This may explain why some sensitivity analyses, using different statistical methods, exposure windows and including all confirmed COVID-19 patients, showed an association between adverse COVID-19 outcomes and preadmissions anticoagulants use. It is possible that our main results may have been skewed towards the null due to the inclusion of patients with non-severe COVID-19 infection in our population, as the benefit of preventing thromboembolisms with pre-admission anticoagulants is more likely to have a beneficial effect in patients with more severe COVID-19 (and therefore more prone to have said thromboembolic complications). On the other hand, patients who were already on anticoagulants prior to hospitalization may be at higher risk for poorer outcomes, due to their elevated pre-existing risk for deep vein thrombosis, pulmonary embolisms and atrial fibrillations (and therefore the use of anticoagulants). As well, patients with atrial fibrillation commonly present underlying comorbidities such as obesity, diabetes melilites, hypertension, heart failure and condition. When all considered, these comorbidities could have lead to a net worse result in sensitivity analysis.

It is important to mention that our study was not a randomized controlled study, and hence residual confounding may still exist. Additionally, as this study employs an administrative claims database, there are intrinsic limitations such as lack of baseline demographics, smoking history, and body weight data. Additionally, while the study cohort is large relative to other reports, the size did not allow for meaningful assessments of the primary endpoint effects according to different types and doses of anticoagulation medications.

In summary, we report on a South Korean nationwide claims database and find that pre-admissions anticoagulants use may not have a protective role against severe COVID-19 outcomes.

Acknowledgments

The authors thank healthcare professionals dedicated to treating COVID-19 patients in South Korea, the Ministry of Health and Welfare, the Health Insurance Review & Assessment Service (HIRA) and Ye-Jin Sohn (HIRA) for sharing invaluable national health insurance claims data.

Funding: This work was supported by 2020R1G1A1A01006229 awarded by the National Research Foundation of Korea.

Footnote

Reporting Checklist: The authors have completed the STROBE reporting checklist. Available at https://apm.amegroups.com/article/view/10.21037/apm-21-3466/rc

Data Sharing Statement: Available at https://apm.amegroups.com/article/view/10.21037/apm-21-3466/dss

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://apm.amegroups.com/article/view/10.21037/apm-21-3466/coif). CBS serves as an unpaid Editor-in-Chief of Annals of Palliative Medicine. YGC reports that his work was supported, in part, by 2020R1G1A1A01006229 awarded by the National Research Foundation of Korea. The other authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study was approved by the Human Investigation Review Board of Public Institutional Bioethics Committee designated by the MOHW. The requirement of informed consent was waived due to the retrospective study design and anonymity of the HIRA database (IRB # P01-2020-1262-001). The study was conducted in accordance with the Declaration of Helsinki (as revised in 2013).

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Singhania N, Bansal S, Singhania G. An Atypical Presentation of Novel Coronavirus Disease 2019 (COVID-19). Am J Med 2020;133:e365-6. [Crossref] [PubMed]
2. Hematology ASo. COVID-19 and Coagulopathy: Frequently Asked Questions: American Society of Hematology; 2020. Available online: https://www.hematology.org/covid-19/covid-19-and-coagulopathy
3. Kander T. Coagulation disorder in COVID-19. Lancet Haematol 2020;7:e630-2. [Crossref] [PubMed]
4. Flumignan RLG, Tinôco JD, Pascoal PIF, et al. Prophylactic anticoagulants for people hospitalised with COVID-19. Cochrane Database Syst Rev 2020;10:CD013739. [PubMed]
5. Ayerbe L, Risco C, Ayis S. The association between treatment with heparin and survival in patients with Covid-19. J Thromb Thrombolysis 2020;50:298-301. [Crossref] [PubMed]
6. Liu X, Zhang X, Xiao Y, et al. Heparin-induced thrombocytopenia is associated with a high risk of mortality in critical COVID-19 patients receiving heparin-involved treatment. Available online: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3582758
7. Tang N, Bai H, Chen X, et al. Anticoagulant treatment is associated with decreased mortality in severe coronavirus disease 2019 patients with coagulopathy. J Thromb Haemost 2020;18:1094-9. [Crossref] [PubMed]
8. Shi C, Wang C, Wang H, et al. The Potential of Low Molecular Weight Heparin to Mitigate Cytokine Storm in Severe COVID-19 Patients: A Retrospective Cohort Study. Clin Transl Sci 2020;13:1087-95. [Crossref] [PubMed]
9. TrinhMChangDRGovindarajuluUSTherapeutic Anticoagulation Is Associated with Decreased Mortality in Mechanically Ventilated COVID-19 Patients.medRxiv 2020. doi: 10.1101/2020.05.30.20117929
10. Moores LK, Tritschler T, Brosnahan S, et al. Prevention, Diagnosis, and Treatment of VTE in Patients With Coronavirus Disease 2019: CHEST Guideline and Expert Panel Report. Chest 2020;158:1143-63. [Crossref] [PubMed]
11. Tieleman RG, Klok FA, Belfroid E, et al. Effect of anticoagulant therapy in COVID-19 patients. Neth Heart J 2021;29:35-44. [Crossref] [PubMed]
12. Republic of Korea Ministry of Health and Welfare. COVID-19 Response: Korean Government’s Response System 2021. Available online: http://ncov.mohw.go.kr/en/baroView.do?brdId=11&brdGubun=111
13. Hajage D, Tubach F, Steg PG, et al. On the use of propensity scores in case of rare exposure. BMC Med Res Methodol 2016;16:38. [Crossref] [PubMed]
14. Ross ME, Kreider AR, Huang YS, et al. Propensity Score Methods for Analyzing Observational Data Like Randomized Experiments: Challenges and Solutions for Rare Outcomes and Exposures. Am J Epidemiol 2015;181:989-95. [Crossref] [PubMed]
15. Moodie EEM, Saarela O, Stephens DA. A doubly robust weighting estimator of the average treatment effect on the treated. Stat 2018;7:e205. [Crossref]
16. Klok FA, Kruip MJHA, van der Meer NJM, et al. Confirmation of the high cumulative incidence of thrombotic complications in critically ill ICU patients with COVID-19: An updated analysis. Thromb Res 2020;191:148-50. [Crossref] [PubMed]
17. Russo V, Di Maio M, Attena E, et al. Clinical impact of pre-admission antithrombotic therapy in hospitalized patients with COVID-19: A multicenter observational study. Pharmacol Res 2020;159:104965. [Crossref] [PubMed]
18. Sivaloganathan H, Ladikou EE, Chevassut T. COVID-19 mortality in patients on anticoagulants and antiplatelet agents. Br J Haematol 2020;190:e192-5. [Crossref] [PubMed]