A retrospective study of adverse drug events in anticoagulant administration with relevance to COVID-19

Abstract

Initial studies in COVID-19 patients reported lower mortality rates associated with the use of the drug heparin, a widely used anticoagulant. The objective of this analysis was to determine whether there are adverse events associated with the administration of anticoagulants, and specifically how this might apply in patients known to have COVID-19. Data for this study were obtained from the Food and Drug Administration's Adverse Event Reporting System (FAERS) public database and from the NIH's clinical trials website. Proportional Reporting Ratios (PRR) with lower 95% confidence intervals (lower CI) and empirical Bayes geometric mean (EBGM) scores with lower 95% confidence limits were calculated for data from the FAERS database where the adverse events studied mimicked COVID-19 symptoms.

1. Introduction

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), the pathogen causing the COVID-19 disease, can cause a rapid progression to acute respiratory distress syndrome (ARDS) in severe cases [1,2]. Many patients who progress to ARDS require mechanical ventilatory support. A major complication of having ARDS and being on a ventilator is an enhanced potential to develop blood clots. Ischemic stroke (due to a clot formed in the heart with subsequent travel to the brain) is frequently observed in severe COVID-19 patients [3]. According to Zakeri et al. the risk of stroke from arterial thrombosis in COVID-19 patients is high, ranging from 2.8% to 3.8% [3]. Due to this elevated risk of stroke, it is important to identify efficient and effective approaches to managing anticoagulant drugs in these patients to prevent the occurrence of adverse events such as stroke. Conducting pharmacovigilance studies is crucial to managing and monitoring drugs used in the healthcare system. Pharmacovigilance analyses help in detecting and assessing whether a drug causes any major or minor adverse events in a population. In the present study, adverse drug events (ADEs) considered are those that are similar to symptoms of COVID-19 disease. As discussed above, thrombosis is a complication caused by COVID-19 in severe cases and since anticoagulant therapies are clinically used as thromboprophylaxis, it becomes imperative to assess how different anticoagulant therapies affect patients with COVID-19 and whether they exacerbate any adverse drug event in these critically ill patients [3]. Thus, the objective of this retrospective study is to observe how anticoagulant therapies behave in a general patient population (with patients who may or may not have COVID-19) and whether there are any significant adverse events reported. To achieve this, we investigated anticoagulant therapies, including heparin, warfarin, dabigatran, rivaroxaban, edoxaban, and apixaban as our drugs of interest.

2. Relation between COVID-19 and heparin

In an article by Clausen et al. heparan sulfate, a linear glycosaminoglycan polysaccharide present on the cell surface was demonstrated to act as a coreceptor along with ACE2, to facilitate binding of the SARS-CoV-2 spike (S) protein onto the cellular surface [4]. Unique heparan sulfate chains are covalently bound to a core protein to form heparan sulfate proteoglycans (HSPG) [5]. Structural similarities between heparan sulfate glycosaminoglycan and heparin's glycosaminoglycan chains make heparin an interesting drug to study from this perspective.

The work by Clausen and team suggests that competitive binding of heparin and heparan sulfate to the binding S protein of SARS-CoV-2 might prove to be inhibitory towards viral uptake by the cell and may offer therapeutic benefits to control the transmission of SARS-COV-2 [4]. This association is consistent with heparin and other anticoagulant drugs being evaluated in clinical trials (CTs) for the treatment of COVID-19 disease. To study this association between anticoagulants and their administration as a treatment for COVID-19, we examined possible adverse events related to the aforementioned anticoagulants.

3. Methods

This research presents a retrospective analysis of data from the FDA's Adverse Event Reporting System (FAERS) database. Using the data available from the database, appropriate statistical and mathematical tools were used for the analysis. Data from FAERS were extracted from spontaneous reports that were reported for Human Drug Adverse Events on openFDA and were later integrated into 1DATA (www.1data.life). Additionally, primary outcomes for ongoing CTs were analyzed that included observational, interventional, and both observational and interventional study designs. Primary outcomes of all the studies were analyzed, except for studies that were suspended or terminated at the time. Both of these data sources, FAERS and the clinical trials website, included general patient populations [6].

Thirteen pulmonary ADEs were analyzed in this study including pulmonary edema, dysphonia, nasopharyngitis, pleurisy, pleural effusion, cough, sinusitis, bronchitis, oropharyngeal pain, pneumonia, pneumonia aspiration, dyspnea, and emphysema. These particular ADEs, also known as preferred terms (PT) per the Medical Dictionary for Regulatory Activities (MedDRA), were chosen for this study because they mimic COVID-19 symptoms [7]. Studying these COVID-19 resembling ADEs would help explore the potential significance of using anticoagulants as a COVID-19 therapy.

3.1. Data collection

Information regarding current clinical trials was obtained from the ClinicalTrials.gov website maintained by the National Library of Medicine (NLM) for the National Institutes of Health (NIH). Search criteria were adjusted for the fields: conditions or disease and other terms. COVID-19 was put in as the condition or disease and the individual drug's name of interest was inserted in each search separately. Information including drug indications and adverse events was obtained. Drug indication included the disease or condition for which that drug was prescribed. In this study, ten drug indications including thrombosis, deep vein thrombosis, pulmonary thrombosis, pulmonary embolism, hypercoagulation, acute myocardial infarction, embolism venous, pulmonary thromboembolism, embolism, and venous thrombosis were used. To analyze the CT data, occurrences of the above listed drug indication terms were searched for in the primary outcome's section of every CT (except suspended or terminated clinical trials).

Spontaneous reports were used to obtain curated information about the general patient population (n = 127,186) who were prescribed the drugs of interest. These data were collected from FAERS database's application programming interface (API) known as openFDA and represented information collected from 2018 to 2019 (pre-COVID) and 2020–2021 (COVID) datasets. The search query included both generic and brand-name drugs. We used regular expression techniques as a part of natural language processing (NLP) to extract information about drugs and related ADE from spontaneous reports [8]. Drug information-including brand-names, generic names, medicinal products, and active substances-was curated as the reference. Then, we implemented data scrubbing and cleaning on drug names regarding the same active substances to refile missing information and merge reports. ADE reports were mapped to MedDRA's terminology to avoid the occurrence of duplicated and unrelated cases.

3.2. Statistical analysis

Frequencies for various primary outcomes considered by researchers conducting CTs were tabulated. In addition to the occurrences of drug indication terms in the primary outcomes, we also looked at six other general primary outcomes. These generic terms included thrombotic events, mortality, ICU admission, hospitalization, ventilation, and PaO2/FiO2 ratio. We created a table to show the composition of primary outcomes with regard to the use of anticoagulant drugs in clinical trials (Table 2a). Another table includes the mentions of drug indications in the primary outcomes and general primary outcomes for six drugs being studied (Table 2b).

Table 2a.

Mention of drug indications in primary outcomes

Drug indications	Heparin	Rivaroxaban	Apixaban	Edoxaban	Dabigatran	Warfarin
Thrombosis	27	6	2	0	1	0
Deep vein thrombosis	5	0	1	0	0	0
Pulmonary thrombosis	0	0	0	0	0	0
Pulmonary embolism	12	2	2	0	1	0
Hypercoagulation	0	0	0	0	0	0
Acute myocardial infarction	6	3	1	0	1	0
Embolism venous	0	0	0	0	0	0
Pulmonary thromboembolism	0	0	0	0	0	0
Embolism	7	2	2	0	2	0
Venous thrombosis	7	2	0	0	0	0

Table 2b.

General primary outcomes from clinical trials

Primary outcomes	Heparin	Rivaroxaban	Apixaban	Edoxaban	Dabigatran	Warfarin
Thrombotic events	6	1	0	0	0	0
Mortality	38	7	2	3	1	1
ICU admission	11	0	0	0	0	0
Hospitalization	9	7	1	0	0	0
Ventilation	47	7	2	0	0	0
PaO2/FiO2 ratio	6	0	0	0	0	0

According to the FDA, PRR is defined as “the degree of disproportionate reporting of an adverse event for a product of interest compared to the same event for all other products in the database” [8]. An adverse event is considered reportable if the PRR is greater than 2 [7]. Once the PRR was calculated for the thirteen COVID-19 symptoms that mimicked pulmonary ADEs, the lower 95% confidence interval (CI) was calculated [9]. The final criterion for identifying a reportable adverse event is that the lower confidence interval value should be greater than 1 [7]. PhVid python package was used to calculate PRR values[10].

The same criteria as those for PRR were used to identify adverse events using the EBGM statistic. Along with the PRR statistic, we also used the EBGM statistic to observe relative reporting rate for the same thirteen COVID-19 symptoms that mimic pulmonary ADE [9]. EBGM is a robust statistical method that measures disproportionality of a particular drug, compared to all other drugs of interest [8]. The 5th percentiles from the lower bound of 95% CI of EBGM were also calculated (known as EB05) to show the significance of EBGM statistics. EBGM calculations were done using the openEBGM package in R. Data from FAERS were cleaned in python.

4. Results

4.1. Clinical trials in COVID-19 patients

Table 1 shows the breakdown of number of trials with respect to their status (recruiting, completed, not yet recruiting, or active-not recruiting) as of May 29, 2022 for each anticoagulant drug. At the beginning of this study in August 2020, there was a relatively small number of CTs relating to anticoagulant drugs as used during the treatment of COVID-19. For the drug heparin, there were only 36 trials recorded as of August 2020 compared to 87 recorded by May 29, 2022. This increase in the number of clinical trials on heparin as a drug of interest, especially factor Xa inhibitor, indicates the interest and concern for heparin use in COVID-19 patients. It reflects the lower mortality rate associated with heparin use in COVID-19 patients as shown by studies done earlier in the pandemic coupled with heparin's known safety, wide use and easy accessibility making it a preferable anticoagulant drug to study [10,11].

Table 1.

Number of COVID-19 clinical trials that includes anticoagulant drugs [6].

	Total	Recruiting	Completed	Not yet	Active, not
	trials			recruiting	recruiting
Heparin	112a	36	41	14	3
Warfarin	1	0	1	0	0
Dabigatran	3	1	2	0	0
Rivaroxaban	b19	8	6	1	0
Apixaban	11	6	2	1	1
Edoxaban	3	2	1	0	0

^a2 CTs is enrolling by invitation, 5 CTs having got terminated, and 4 CTs withdrawn.

^b3 CTs having got terminated.

For the drug heparin a majority outcomes from 112 CTs related to thrombosis (frequency = 27), embolism (19), ventilation (47), and mortality (38) (Table 2a, Table 2ba and 2b). Of the primary outcomes from 19 CTs involving the use of rivaroxaban, a majority consisted of thrombosis (frequency = 6), embolism (4), ventilation (7), mortality (7), and hospitalization (7) (Table 2a, Table 2ba and 2b). Outcomes from 11 trials on apixaban comprised thrombosis (frequency = 2), embolism (2), ventilation (2) and mortality (2). It indicates that there are four conditions (thrombosis, embolism, ventilation and mortality) that are truly vital in studying adverse events involving anticoagulants especially when researchers use anticoagulant drugs as a COVID-19 therapy.

4.2. General patient population on anticoagulant drugs

Given that primary outcomes of COVID-19 clinical trials involving anticoagulant drugs focused on complications based on thrombosis, it is important to find anticoagulant drug indications that include thrombotic events in the general population. Data from the FAERS database were used to identify the drug indications of interest.

As seen in Table 3, most drug indications for all the anticoagulant drugs of interest combined include thrombosis prophylaxis (26,520), deep vein thrombosis (21,448), and pulmonary embolism (14,948) respectively. Rivaroxaban is prescribed far more frequently compared to the other drugs followed distantly by warfarin and apixaban, except in the case of acute myocardial infarction where heparin has been used the most as shown in the table.

Table 3.

Drug indications for ADEs reported from FAERS database for anticoagulant drugs

Column 1	Heparin	Warfarin	Dabigatran	Rivaroxaban	Edoxaban	Apixaban
Thrombosis prophylaxis	2692	2185	1612	17183	138	2710
Thrombosis	432	2428	255	3823	50	840
Deep vein thrombosis	803	4201	528	12407	303	3206
Pulmonary thrombosis	9	152	9	323	2	56
Pulmonary embolism	782	2880	404	8090	221	2571
Coagulopathy	178	456	72	185	3	116
Acute myocardial infarction	455	71	12	28	0	21
Pulmonary thrombosis	9	152	9	323	2	56
Venous thromboembolism	0	1	0	1	0	0
Embolism	51	258	59	579	10	146
Venous thrombosis	40	119	14	192	14	52
Total	5451	12903	2974	43134	743	9774

Next, PRRs were calculated for 13 ADEs that mimic COVID-19 symptoms in this patient population cohort (Table 4). There were a total of 5,672 adverse events reported for all six drugs of our interest.

Table 4.

Proportional Reporting Ratios and corresponding CIs for preferred terms from MedDRA

	PRR	95% CI	PRR	95% CI	PRR	95% CI	PRR	95% CI	PRR	95% CI	PRR	95% CI
Bronchitis	Apixaban		Dabigatran		Edoxaban		Heparin		Rivaroxaban		Warfarin
Cough	1.93	1.46	1.2	0.75	1.28	0.48	0.61	0.40	0.46	0.34	1.5	1.17
Dysphonia	2.06	1.76	0.7	0.50	2.07	1.32	0.43	0.33	0.62	0.53	1.35	1.17
Dyspnoea	2.01	1.42	0.19	0.05	0.52	0.07	0.34	0.17	0.51	0.36	2.07	1.53
Emphysema	1.4	1.27	0.95	0.81	1.52	1.14	1.02	0.92	0.7	0.64	1.12	1.04
Nasopharyngitis	2.09	1.05	NA	NA	NA	NA	0.8	0.31	0.48	0.23	1.73	0.95
Oropharyngeal pain	1.89	1.48	0.56	0.32	0.49	0.12	0.18	0.09	0.65	0.51	1.83	1.49
Pleural effusion	1.52	1.09	0.75	0.40	1.97	0.81	0.34	0.19	0.56	0.41	1.93	1.48
Pleurisy	1.06	0.83	1.4	1.02	0.51	0.16	1.72	1.40	0.58	0.47	1.05	0.87
Pneumonia	0.42	0.10	1.11	0.26	2.89	0.39	0.9	0.31	0.75	0.34	1.74	0.85
Pneumonia aspiration	1.4	1.23	1.03	0.83	1.57	1.07	1.31	1.14	0.65	0.58	1	0.90
Pulmonary edema	0.97	0.63	1.09	0.59	4.17	2.05	1.37	0.93	0.72	0.51	0.97	0.70
Sinusitis	1.2	0.86	0.59	0.30	0.68	0.17	2.58	1.97	0.46	0.34	1	0.77
	1.66	1.16	0.78	0.39	0.51	0.07	0.3	0.15	0.63	0.45	1.82	1.35

For any adverse event to be reportable, three criteria are needed: there should be more than 3 reported incidences, the PRR value should be greater than 2, and a PRR that is greater than the lower 95% confidence interval (CI) boundary, with the lower CI itself being over 1 [6]. As per the PRRs and the corresponding CI values presented in Table 4, reportable adverse events include cough, dysphonia, and emphysema for the drug apixaban, cough, and pneumonia aspiration for edoxaban, dysphonia for warfarin and pulmonary edema for heparin. Hence, the issue lies in how the pathophysiology of a COVID-19 patient given anticoagulants is affected. Using the same data, empirical Bayesian geometric mean (EGBM) scores were worked out (Table 5) in order to be able to identify an adverse event for a particular drug compared to all other events in respect of all the drugs being examined.

Table 5.

EBGM scores and corresponding 95% CIs for preferred terms from MedDRA

	EBGM	EB05	EBGM	EB05	EBGM	EB05	EBGM	EB05	EBGM	EB05	EBGM	EB05
	Apixaban		Dabigatran		Edoxaban		Heparin		Rivaroxaban		Warfarin
Bronchitis	1.68	1.37	1.13	0.78	1.13	0.55	0.84	0.6	0.46	0.37	1.71	1.45
Cough	1.75	1.56	0.69	0.52	1.85	1.27	0.6	0.48	0.56	0.5	1.58	1.43
Dysphonia	1.67	1.3	0.29	0.1	0.69	0.23	0.52	0.31	0.49	0.38	1.99	1.65
Dyspnoea	1.31	1.22	0.92	0.81	1.49	1.17	1.3	1.19	0.61	0.57	1.41	1.34
Emphysema	1.52	0.94	NA	NA	NA	NA	0.97	0.5	0.49	0.29	1.65	1.12
Nasopharyngitis	1.61	1.34	0.58	0.37	0.62	0.25	0.28	0.17	0.58	0.48	1.87	1.63
Oropharyngeal pain	1.4	1.1	0.77	0.47	1.5	0.76	0.52	0.33	0.54	0.43	1.98	1.68
Pleural effusion	1.04	0.86	1.29	1	0.61	0.27	1.95	1.66	0.53	0.46	1.34	1.17
Pleurisy	0.58	0.23	0.95	0.38	1.18	0.4	1.05	0.51	0.68	0.4	1.64	1.04
Pneumonia	1.32	1.2	1	0.84	1.52	1.1	1.6	1.44	0.59	0.54	1.32	1.22
Pneumonia aspiration	0.95	0.68	1.01	0.62	2.72	1.44	1.57	1.17	0.63	0.5	1.25	0.99
Pulmonary edema	1.15	0.9	0.62	0.37	0.77	0.31	2.61	2.15	0.46	0.36	1.31	1.08
Sinusitis	1.45	1.1	0.78	0.45	0.68	0.22	0.46	0.26	0.57	0.44	1.85	1.53

The relative values of EBGM (Table 5) are in general agreement with the PRR values (Table 4), above; however, we see that ADEs associated with apixaban and warfarin in Table 4 drop out in Table 5. However, pneumonia aspiration in edoxaban and pulmonary edema in heparin remain. For the drug warfarin, even though the EBGM values are not reportable, the EB05 values are relatively high. The reportable incidents found in this study are pulmonary edema in heparin and pneumonia aspiration in edoxaban. However, it is important to note here that from the FAERS data discussed above, about 57% of the six drugs were suspected as one of the 13 pulmonary ADEs regarding the characterization of drug role, 42% were concomitant, and the rest (∼1%) of drugs were interacting.

5. Discussion

Studies done early in 2020 by Tang et al. and Ayerbe et al. suggested that heparin may provide therapeutic benefit to COVID-19 patients [10,11]. Given the increased interest in anticoagulants such as heparin for use as a therapy for COVID-19 not only in the studies mentioned above but also in clinical trials, it seems extremely important to look at any adverse drug events arising from the use of this class of drugs. As anticoagulants are used to treat conditions like thrombosis, a known cause of COVID-19 complication, studying ADEs that mimic COVID-19 symptoms becomes imperative [3]. Examining such ADEs may help in identifying a pharmacovigilant signal associated with COVID-19.

Based on our findings, studies using data pulled directly from the Electronic Health Records (EHRs) should be undertaken to examine anticoagulant drug therapies and COVID-19. The more granular EHR data could be used to look for various thrombotic events for specific anticoagulant drugs when prescribed as prophylaxis for such events. Distinctly inspecting EHR data for the use of anticoagulants in a patient population with COVID-19 could imply if these drugs are truly exacerbating COVID-19 related ADEs in this population.

A randomized study by Cabral and Ansell showed that factor Xa inhibiting drugs such as rivaroxaban and edoxaban are more effective and may have more advantages over regular anticoagulants in treating conditions like venous thromboembolism [12]. From the FAERS database it was observed that rivaroxaban, a direct oral anticoagulant and a factor Xa inhibitor, was the anticoagulant that was most often prescribed including the years 2020 and 2021 to treat various thrombotic events. This implies that more clinical studies should be undertaken to test the safety and efficacy of factor Xa inhibitors in COVID-19 patients, as opposed to the current plethora of clinical trials using heparin in treating Covid-19.

In a clinical review article by Bailly et al. the authors discuss the risk of developing heparin-induced thrombocytopenia (HIT), a clinically serious complication associated with the increased use of heparin during COVID-19 [12]. Knowing that use of heparin may result in HIT, it is important to consider shifting the focus of drug treatment to factor Xa inhibitors or other treatments [13]. Based on the PRRs and EBGMs calculations, this study suggests that the factor Xa inhibitors and other widely used anticoagulants, especially rivaroxaban, apixaban, and dabigatran have relatively fewer ADEs. A 2020 study by Billett et al. aimed at investigating whether anticoagulant therapy would have any impact on mortality in a COVID-19-positive patient population [14]. In that study, the researchers focused on enoxaparin, heparin, and apixaban drugs to learn if there is a difference between fractionated, unfractionated, and direct oral anticoagulants, when given as a therapy to COIVD-19 patients [14]. This shows that it is important to know about any possible ADE associations that the anticoagulants may have in relation to COVID-19 symptoms and therapy. Lastly, it is imperative to note that the ADE profile shown by warfarin and heparin may suggest complications to COVID-19 treatment in outpatient or hospital settings.

This study has some limitations. The ADE data obtained from FAERS lacked information regarding the COVID-19 status for the reports presented. As reporting adverse events on FAERS is voluntary, the complete number of true incidents of adverse drug events remains unknown [15]. Due to this, exposure to a particular drug does not prove causal correlation between the drug and the ADE. The proportional reporting ratio does not include relative risk associated with the use of a specific drug and may report small number of occurring events as ADEs, which is why we also included the EBGM statistic. By using a robust statistical tool such as EBGM, we have reduced the number of false-positive safety signals, as the methodology diminishes the effect of spuriously high PRR values due to the statistical differences between EBGM and PRR calculations [7]. Additionally, background noise was removed when using the machine learning and data mining techniques to avoid the impact of missing data, duplications, and other systematic errors. We used NLP for matching drugs with the same active substances and minimizing the exclusion of similar ADEs. EBGM is a machine learning framework with the ability to apply Bayesian approaches to capture characteristics in PRR and minimize PRR significance when it is spuriously high. Overall, the results found in this study are meant to enable broader discussion and study of specific ADEs, potentially with complete sets of patient data that would overcome the limitations noted above.

Author contribution statement

Gerald Wyckoff: Conceived and designed the experiments; Analyzed and interpreted the data; Wrote the paper.

Purva Patel: Conceived and designed the experiments; Performed the experiments; Analyzed and interpreted the data; Wrote the paper.

Monica Gaddis; Emma G Meyer: Analyzed and interpreted the data.

Xuan Xu: Performed the experiments; Analyzed and interpreted the data; Contributed reagents, materials, analysis tools or data.

Jim E Riviere: Conceived and designed the experiments; Wrote the paper.

Jessica Kawakami: Analyzed and interpreted the data; Wrote the paper.

Majid Jaberi-Douraki: Conceived and designed the experiments; Contributed reagents, materials, analysis tools or data; Wrote the paper.

Data availability statement

Data included in article/supplementary material/referenced in article.

Declaration of interest's statement

The authors declare the following conflict of interests: Dr. Wyckoff owns Zorilla Research and is a board member of Eir Pharma.

Dr. Kawakami has a current affiliation with KCAS but was not part of that organization while working on this manuscript.