Association between warfarin and COVID-19-related outcomes compared with direct oral anticoagulants: population-based cohort study

We conducted a population-based cohort study between 1 March 2020 and 28 September 2020.

Primary care records managed by the software provider TPP were linked to SARS-CoV-2 antigen testing data from the Second Generation Surveillance System, COVID-19-related hospital admissions from the secondary uses service, and Office for National Statistics death data through OpenSAFELY, a data analytics platform created by our team on behalf of NHS England [8]. The data set analysed within OpenSAFELY is based on 24 million people currently registered with primary care practices using TPP SystmOne software, representing 40% of the English population. It includes pseudonymised data such as coded diagnoses, prescribed medications and physiological parameters.

We first identified all patients with a diagnosis of AF on or before study start (1 March 2020) (Fig. 1). People with missing data for sex, Index of Multiple Deprivation, < 1 year of primary care records, or aged < 18 or > 110 and prescribed injectable anticoagulants 4 months before study start date were excluded. In addition, people with a record of mitral stenosis or prosthetic mechanical valves, chronic kidney disease stage V (estimated glomerular filtration rate < 15 mL/min or on dialysis), or antiphospholipid antibody syndrome before study start were also excluded in this study because DOACs are not recommended for use in these patient groups.

The outcomes were (1) testing positive for SARS-CoV-2, (2) COVID-19-related hospital admission, and (3) COVID-19-related death (defined as the presence of ICD-10 codes U071 (confirmed COVID-19) and U072 (suspected COVID-19) anywhere on the death certificate). Testing outcomes were obtained from the UK’s Pillar 1 (NHS and Public Health England laboratories) and Pillar 2 (commercial partners) testing strategies and included results from polymerase chain reaction swab tests used to identify symptomatic individuals [9]. As pre-specified analyses, we also conducted negative control outcome analyses to examine the presence of residual confounding between warfarin and DOAC users. First, we anticipated that, within our population of people with non-valvular AF, there were unlikely to be marked differences in the likelihood of being tested for SARS-CoV-2 infection in relation to drug treatment with warfarin or DOAC. Therefore, we included being tested for SARS-CoV-2 as a negative control outcome to test our assumption. Second, we also included non-COVID-19 death, as differences in this outcome between DOAC and warfarin users could imply that differences in health characteristics had not been successfully controlled for. We conducted additional post hoc analyses to include cause-specific deaths as outcomes (i.e. death due to myocardial infarction, ischaemic stroke, venous thromboembolism, gastrointestinal bleeding and intracranial bleeding) to aid the interpretation of our results.

Follow-up for each cohort began on 1 March 2020 and ended at the latest of the outcome of interest in each analysis, deregistration from the TPP practice, death or study end date (28 September 2020) (Fig. 1).

Covariates were pre-specified, identified from a directed acyclic graph (DAG) approach (Additional file 1: Figure S1), including age, sex, obesity, smoking status, hypertension, heart failure, myocardial infarction, peripheral arterial disease, stroke/transient ischemic attack, venous thromboembolism, diabetes, flu vaccination, current antiplatelet use, current oestrogen and oestrogen-like therapy use, Index of Multiple Deprivation and care home residence. We identified covariates that are both associated with the exposure and the risk of severe COVID-19 outcomes either directly [8], or via venous thromboembolism [10, 11]. All codelists for identifying exposures, covariates and outcomes are openly shared at https://​codelists.​opensafely.​org/​ for inspection and reuse.

Baseline characteristics in each study were summarised using descriptive statistics, stratified by exposure status. We present adjusted cumulative incidence/mortality curves using the Royston–Parmar model (Additional file 1: Figure S2). We estimated hazard ratios (HRs) with 95% confidence intervals (CIs) using Cox regression with time since cohort entry as the underlying timescale. We accounted for competing risk by modelling the cause-specific hazard (i.e. censoring other deaths for COVID-19 death analysis and censoring any death for other outcomes analysis). We used graphical methods and tests based on Schoenfeld residuals to explore violations of the proportional hazards assumption.

We performed unadjusted models, models adjusted for age (using restricted cubic splines) and sex, and DAG-adjusted models (stratified by general practice).

We considered the possibility that if warfarin users had worse baseline health status they might act to lower their risk of SARS-CoV-2 infection through more risk-averse health behaviours (e.g. wearing face masks, avoiding close proximity to others) than DOAC users. Given that health behaviour is not captured in medical records, we conducted quantitative bias analyses to assess the sensitivity of our results to this potential unmeasured confounder.

We calculated the minimum strength of association required between an unmeasured confounder and one of exposure or outcome to move from the observed HR to a null bias-adjusted HR (aHR) (i.e. the E value) [12]. We also calculated the minimum strength of association required between unmeasured confounder and both of exposure and outcome to move from the observed HR to a null bias-aHR (i.e. the Cornfield condition) [12]. Furthermore, we calculated the minimum strength of association required to move from the observed protective associations to a bias-aHR of 1.2 because we hypothesised a harmful effect of warfarin in COVID-19-related outcomes.

Table 1 shows the list of other sensitivity analyses.

Data management was performed using Python 3.8 and SQL, with analysis carried out using Stata 16.1. All study analyses were pre-planned unless otherwise stated. All code for data management and analyses in addition to the pre-specified protocol (https://​github.​com/​opensafely/​anticoagulants-research/​blob/​master/​protocol/​Protocol_​%20​Anticoag%20​OpenSAFELY_​v3.​docx) are archived at: https://​github.​com/​opensafely/​anticoagulants-research.

The funder of the study had no role in the study design, data collection, data analysis, data interpretation, or writing of the report. AJW, CEM, SB, WH, CB, JC, LS and BG had access to the raw data. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.

To potentially fully explain the observed non-null inverse associations, either DOAC use would need to be associated with at least 1.7–1.85 times increased risk of unmeasured risk-prone behaviour or risk-prone behaviour would need to be associated with 1.7–1.85 times increased risk of each COVID-19 outcome. Alternatively, both DOAC use and each COVID-19 outcome would need to be associated with the unmeasured risk-prone behaviour by associations ranging from 1.20 to 1.27 (Additional file 1: Table S7).

When we assumed an association between warfarin and each COVID-19 outcome of 1.2 (i.e. a harmful effect of warfarin compared with DOACs), either DOAC use would need to be associated with 2.25–2.41 times higher risk of unmeasured risk-prone behaviour, or unmeasured risk-prone behaviour would need to be associated with 2.25–2.41 times higher risk of each outcome to potentially fully explain the observed inverse associations. Alternatively, both DOAC use and each outcome would need to be associated with the unmeasured risk-prone behaviour by associations ranging from 1.45 to 1.52.

In this large cohort study of people with non-valvular AF, we observed a lower risk of developing SARS-CoV-2 infection and hospital admission or death from COVID-19 among warfarin users compared to those prescribed DOACs. Surprisingly, we also observed that warfarin users were less likely to be tested for SARS-CoV-2.

The protective associations seen for warfarin versus DOACs in all patients with AF are surprising given the hypothesis of a possible harm with warfarin and generally more comorbidities among warfarin users than DOAC users. Also, our findings were non-specific, including an inverse association with being tested for SARS-CoV-2, and death from non-COVID causes. This might be partly explained by the behavioural differences between warfarin and DOAC users during COVID-19 pandemic. Although we cannot fully capture the behavioural differences between exposure groups in the database, we observed that warfarin users are less likely to be current smokers, had less hazardous alcohol use and more likely to have had flu vaccination than DOAC users, but the differences were small. Importantly, we adjusted for a range of confounders that had minimal impact to the estimates. We further performed quantitative bias analyses and found that an unmeasured risk-averse behaviour of moderate strength could potentially explain the observed inverse associations. Further studies that can account for behavioural differences between groups are required to confirm the findings as we cannot rule this out as a possible contributor to our findings.

Although we did not anticipate there was a difference in non-COVID-19 mortality between warfarin and DOACs, others have also observed a similar association using another UK primary care data set [13]. We therefore tried to explore non-COVID deaths further to understand whether this finding was driven by a possibly superior therapeutic effect of warfarin, but there were too few outcomes to draw conclusions. Whilst the aetiology of the protective association we observed between warfarin and COVID-19-related outcomes is unclear, importantly our results do not suggest that warfarin is associated with COVID-19-related harm compared with DOACs.

Some early studies reported that elevated circulating D-dimer levels and prolonged prothrombin time were associated with mortality in people with COVID-19 disease [14, 15], suggesting COVID-19 coagulopathy [16]. The exact mechanism of hypercoagulability in COVID-19 disease is still not fully understood [17]. Therefore, anticoagulants have been proposed to be one of the investigational therapies in the management of COVID-19 disease, but the role of oral anticoagulants in severe COVID-19 outcomes remains unclear [18]. Further, effects of different types of oral anticoagulants in COVID-19 outcomes might be different due to the difference in mechanisms of action. Desphospho-uncarboxylated matrix Gla protein (dp-ucMGP) is an indirect marker of extrahepatic vitamin K status where high dp-ucMGP levels indicate low extrahepatic vitamin K status [19]. In a study of 135 hospitalised COVID-19 patients and 184 historical controls, it shows that dp-ucMGP was elevated in patients with COVID-19 compared with controls [3]. An even higher dp-ucMGP was observed in patients requiring intubation and mechanical ventilation or all-cause death. However, this study was of small sample size which required larger studies to confirm the association. In particular, the number of regular vitamin K antagonist users was very small (n = 15) to elucidate any role of vitamin K antagonists in severe COVID-19 outcomes. Another three cohort studies showed a trend of a lower risk of all-cause mortality associated with either prior use of warfarin or DOACs versus non-use of anticoagulants in patients with COVID-19 disease [4, 5, 7]. Another German cohort study reported no evidence of an association between DOACs and severe COVID-19 outcomes without including warfarin users into the analysis in people with non-valvular AF [6]. Notably, none of these studies directly compared the effects of warfarin with DOACs on COVID-19-related outcomes.

The greatest strength of this study was the power enabling us to examine the association between warfarin and various COVID-19-related outcomes versus DOACs as our data set included medical records from 24 million individuals. To our knowledge, this is the first population-based study comparing both the risk of COVID-related outcomes and cause-specific deaths between warfarin and DOACs. We also conducted quantitative bias analyses to explore the impact of unmeasured confounding to our observed results. The breadth of data available in primary care allows us to account for a wide range of potential confounders.

We recognise possible limitations. First, we could not eliminate residual confounding. Whilst differences in health behaviours and shielding between groups may partly explain our results, more studies are required to confirm these findings. Second, we are not able to capture any anticoagulant use during hospitalisation. Low molecular weight heparin or unfractionated heparin might be given for both warfarin and DOAC users with severe COVID-19 disease during hospitalisation to prevent venous thromboembolism. As the use of heparin during hospitalisation between warfarin and DOACs users is unlikely to be differential, this would merely lead to an underestimation of the effect without accounting for the anticoagulation use during hospitalisation in the analysis. Therefore, a lack of data on heparin during hospitalisation would not affect our interpretation that there is no evidence of harmful effects of warfarin on COVID-19 outcomes. Third, we do not know whether patients took the medications as prescribed. However, if there is a non-differential misclassification bias of exposure between warfarin and DOAC users, the estimates would be biased towards null. If DOAC users were more likely to adhere to their medication than warfarin users, the estimates would again tend to be biased towards null, with a hypothesis that warfarin has a harmful effect on COVID-19 outcomes. Moreover, we do not have data on inflammatory parameters to further explore the specific mechanism of severe COVID-19 disease due to coagulopathy in our cohort. It is recommended to incorporate inflammatory parameters into the outcome definition, so as to specifically investigate the impact of oral anticoagulants in severe COVID-19 diseases due to coagulopathy in future research. Lastly, we adjusted for the use of oestrogen and antiplatelets in our analysis that are both potentially associated with the oral anticoagulant type being prescribed and COVID-19 outcomes. However, we did not rule out the possibility of residual confounding due to other drug combinations with oral anticoagulants that were not adjusted for.