Effectiveness and Safety of Enoxaparin Versus Unfractionated Heparin as Thromboprophylaxis in Hospitalized COVID-19 Patients: Real-World Evidence

Abstract

Background:

Coronavirus 2019 (COVID-19) patients are at risk of thrombosis. Literature that compares the effectiveness of enoxaparin to unfractionated heparin (UFH) in COVID-19 patients is scarce.

Objective:

We aimed to evaluate the effectiveness and safety of enoxaparin compared with UFH when used at their standard/intermediate dosing in COVID-19 patients.

Methods:

This was a retrospective study conducted at a large COVID-19 center located in Eastern Province, Saudi Arabia. Confirmed COVID-19 cases (≥18 years old) admitted between January and December 2020 were randomly screened for inclusion. Exclusion criteria were patients receiving therapeutic anticoagulation, on chronic anticoagulation, had active bleeding, a platelet count <25 × 10⁹/L, or an incomplete electronic file. The primary endpoint was the occurrence of any thrombotic event (pulmonary embolism, deep venous thrombosis, stroke, or myocardial infarction) or mortality. Secondary endpoints were major or minor bleeding. We applied inverse propensity score weighting (IPTW) with survival analysis to analyze the primary endpoint. Logistic regression was used for the secondary endpoint.

Results:

A total of 980 patients were included (enoxaparin, n = 470 and UFH, n = 510) with a mean age (±SD) of 47.7 (± 12.3) for the enoxaparin arm and 52 (±13.9) for the UFH arm. There was a statistically significant difference in the primary endpoint with an adjusted hazard ratio (aHR) of 0.46 (95%CI: 0.22 to 0.96, P = 0.039) in favor of the enoxaparin arm. There was no statistically significant difference in major or minor bleeding rates between the two arms.

Conclusion and Relevance:

When compared with UFH, enoxaparin was associated with a significant reduction in thrombotic events or mortality among COVID-19 patients. The results need confirmation from randomized controlled trials.

Introduction

Coronavirus disease 2019 (COVID-19) is a contagious disease caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). COVID-19 is associated with a myriad of symptoms that include but are not limited to fever, shortness of breath, hemoptysis, dry cough, diarrhea, sputum production, anosmia, and upper airway congestion.¹ After SARS-CoV-2 binds to the angiotensin-converting enzyme (ACE) 2 receptor, upregulation of inflammatory cytokines (e.g. IL-6) occurs, which can complicate the clinical course of COVID-19 with complications that include pneumonia with acute respiratory distress syndrome (ARDS).² This upregulation in inflammatory cytokines can also result in vascular injury, thus inducing a hypercoagulable state which could eventually result in venous thromboembolism (VTE).^2-4

The observed hypercoagulable state in hospitalized COVID-19 patients and the poor prognosis associated with it has prompted the scientific communities to recommend prophylactic anticoagulation as part of COVID-19 treatment protocols. The current recommendations suggest that all COVID-19 patients who are not at high risk of bleeding receive an anticoagulant for VTE prophylaxis.^5,6 Although data on the preferred anticoagulant for thromboprophylaxis in COVID-19 patients is scarce, unfractionated heparin (UFH) and low-molecular-weight heparin (LMWH) are the most administered parenteral anticoagulants. Interestingly, the American College of Chest Physicians (CHEST) and the American Society of Hematology (ASH) recommend the use of LMWH rather than UFH as a thromboprophylaxis agent of choice, mostly to decrease the number of both administered injections and health care workers exposure to the virus.^5-7 For patients with severe renal insufficiency, UFH is recommended over LMWH.⁸

The current literature has focused on heparins’ dosing rather than the choice of agent. A study by Spyropoulos et al⁹ showed that a therapeutic dose of LMWH for thromboprophylaxis in high-risk COVID-19 patients (defined as D-dimer levels > 4 times the upper limit of normal or sepsis-induced coagulopathy score of ≥ 4) reduced thromboembolic events compared with intermediate and prophylactic doses of heparins (LMWH or UFH) without increased major bleeding in non–intensive care unit (ICU) patients with relative risk (RR) = 0.37, P < .001, 95% confidence interval (CI): 0.21 to 0.66.

More pertinently, researchers postulate that the type of anticoagulant may be related to the effectiveness of thromboprophylaxis in COVID-19 patients.¹⁰ UFH, a natural product extracted from porcine intestine, has a large structure with a mean molecular weight of 15 to 19 kDa and great heterogeneity in its structure. To ensure its pharmacological activity, purity, and safety, the process of its extraction and purification must be controlled.^11,12 In contrast to UFH, LMWHs are smaller in size with a molecular weight of 3.0 to 6.5 kDa. Due to the differences in the structure compared with UFH, LMWHs are more reliable in dosing and have higher bioavailability and longer half-lives. Because of the differences in production processes, LMWHs are a class of products in which each has its unique chemical structure, polysaccharide chain, and molecular weight. The chemical structure may play a role beyond the anticoagulation effect as well. In addition, UFH is highly negatively charged, which in turn causes aggregation and therefore less interaction with the monocytes. Two key cytokines maybe implicated in COVID-19 severity: tumor necrosis factor alpha (TNF-α) and IL-6. LMWH or UFH can bind to both cytokines to prevent cytokine storm development.¹³ In theory, smaller molecules such as LMWH as opposed to UFH may have the advantage of better access to cell receptors and therefore better anti-inflammatory effect.¹⁴ Unlike LMWH, selective Xa inhibitors such as rivaroxaban may not have the desired pleiotropic effects (i.e, anti-inflammatory or immunomodulation) in COVID-19 patients.¹⁵

Comparative studies regarding the choice of the commonly used heparin agents (LMWH versus UFH) in COVID-19 patients is lacking. Moreover, due to the urgency of the pandemic and relative unfeasibility of designing a prospective randomized controlled trial (RCT) in our setting, we sought to retrospectively evaluate the two prophylactic choices in the real-world setting. Hence, this study aimed to compare the effectiveness of enoxaparin to UFH in COVID-19 patients. In addition, we evaluated bleeding risks between the two arms.

Methods

Study Design and Setting

This was a retrospective observational cohort study conducted via medical record review at Dammam Medical Complex (DMC) in the Eastern Province of Saudi Arabia. DMC is a tertiary care hospital with a capacity of 423 beds accredited by the Saudi Central Board of Accreditation of Health Care Institution (CBAHI). The center was designated for the management of COVID-19 patients during the pandemic. Of note, health care in Saudi Arabia is mainly funded by the public sector through the Saudi ministry of health (MOH).

The study was externally approved and conducted in compliance with the institutional review board at King Fahad Specialist Hospital-Dammam (IRB # PHA0320). In our report, we followed the statement checklist for Strengthening the Reporting of Observational Studies in Epidemiology (STROBE).¹⁶

Participant Selection

We obtained a list of COVID-19 patients who were admitted between January 1 and December 30, 2020. The obtained cases were randomized by creating a column of random numbers (utilizing the RAND function in excel). Through this random selection process, we screened COVID-19 patients to be included in the study. The random selection process was used to minimize selection bias as it provided an equal opportunity for patients’ files to be selected and coded for analysis.^17-19

We included patients who were 18 years of age or older with a diagnosis of SARS-CoV-2 confirmed with a real-time polymerase chain reaction (RT-PCR) test and who had received anticoagulation (either a standard or an intermediate dose) as thromboprophylaxis within 2 days of their admission. We excluded patients who were younger than 18; were on chronic anticoagulant therapy; had active bleeding; had a platelet count <25 × 10⁹/L; had heparin-induced thrombocytopenia; had already received a therapeutic dose of anticoagulation as their prophylactic regimen; had received the anticoagulant prophylaxis for less than 3 days; had been admitted to the Dammam Medical Complex and then were transferred to another hospital (lost to follow-up); or had received the COVID-19 vaccine; or had incomplete patient profile.

Data Collection

Clinical data were manually extracted from electronic health records (EHRs) and entered into the Research Electronic Data Capture (REDCap) system in a de-identified manner by a trained team.²⁰ The data managers periodically checked data quality during the collection process to assess the consistency of the collected data and resolve any discrepancies.

Anticoagulation Dosing

Standard dosing VTE prophylaxis was defined as using either enoxaparin at 40 mg a day, UFH at 5000 two or three times a day or 7500 units (U) two times a day dosing. Intermediate dosing VTE prophylaxis was defined as using VTE prophylaxis with doses above the standard dosing but not at the treatment dose. Treatment dosing was defined as using either enoxaparin 1 mg/kg every 12 hour (or 1.5 mg/kg every 24 hours) or UFH infusion.²¹

Definitions

Disease severity was defined according to World Health Organization (WHO)’s classification.²² Moderate illness was defined clinical symptoms with evidence of pneumonia and not requiring supplemental oxygen. Severe COVID-19 was defined as fever plus ≥1 of the following symptoms: respiratory rate ≥30/min, dyspnea, respiratory distress, SpO2: ≤93% on room air, PaO2/FiO2 ratio <300 or lung infiltrate >50% of lung field within 24–48 hours. Critical illness was defined by one or more of the following presentations: ARDS, septic shock, altered consciousness, or/and multi-organ failure. Major bleeding was defined according to the International Society on Thrombosis and Haemostasis (ISTH)/Scientific and Standardization Committee (SSC). Clinically major bleeding was defined as fatal bleeding or symptomatic bleeding that occurs in a critical area such as intracranial, intraspinal, intraocular, retroperitoneal, intraarticular, or pericardial bleeding or intramuscular bleeding with compartment syndrome or overt bleeding associated with a fall ≥ Hb 2 g/ or bleeding that leads to transfusion of ≥ 2 units of packed red blood cells (PRBC). A clinically relevant minor bleed is “an acute or subacute clinically overt bleed that does not meet the criteria for a major bleed but prompts a clinical response.”²³

Study Outcomes

The primary endpoint was a composite of death or any thrombotic event, including pulmonary embolism (PE), deep venous thrombosis, ischemic stroke, or myocardial infarction. The secondary endpoints were 30 days of in-hospital survival, thrombotic events (ie, PE, deep venous thrombosis, ischemic stroke, myocardial infarction), and occurrence of major bleeding or minor bleeding.

Sample Size Calculation

Assuming a −10% risk difference in the primary outcome between the UFH group and the enoxaparin group with 40% of patients in the UFH group having the outcome of interest, we needed a total of 738 patients after applying continuity correction (1:1 ratio) to detect the difference between the 2 groups with 80% power. We used a 2-sided significance level (1-alpha) of 0.05%.

Statistical Analysis

Descriptive statistics were used to present the baseline characteristics. When the variable was continuous, we used mean and standard deviation (±SD) or median and interquartile ranges (IQRs) to present the data based on distributions and skewness. To compare between groups, we utilized either a t-test or a Mann–Whitney U test. If the variable was categorical, we presented it as counts and percentages. To compare between arms, we used a chi-square test for n×m tables or Fisher’s exact test for the 2x2-table group.

Inverse Propensity Score Weighting

We conducted unadjusted and adjusted analyses; for the adjusted analysis, we used inverse propensity score weighting (IPTW). All pre-treatment covariates of interest were considered in the model. Our target estimand was average treatment effect (ATE), which estimates the effect of the treatment (ie, enoxaparin) when the whole study population is treated versus a reference (ie unfractionated heparin). Diagnostics for covariance balance were visualized using love plots and mirror diagrams. The literature suggests the use of absolute standardized mean difference (SMD) (a value <20% or <0.2 is desired) to assess for adequate covariate balance.²⁴ To obtain relative treatment effects for time to event data, we fitted cox proportional model with the treatment as a covariate. In case of binary outcome, we fitted logistic regression model with the treatment as a covariate. For the absolute treatment effect, we fitted Kaplan–Meier models and tested the difference in survival using the log-rank test for the unadjusted data and the weighted log-rank test for the weighted data. Time 0 was the start of anticoagulation therapy. Subjects were censored if they had been discharged and no events had occurred.

Missing Data

There were no missing values in the outcome data, but there was some missing data in some of the baseline variables. Since our analytical plan included the estimation of propensity score weights, missing data for important confounders could be problematic. We determined that the data were missing at random (MAR); therefore, we performed the multivariate imputation by chained equations (MICE) technique ( 5 imputations with 20 iterations).²⁵

E Value Calculation

If this was a randomized controlled trial, there would be no systematic differences in the measured and the unmeasured baseline confounders. One of the assumptions for using propensity score is exchangeability. In this assumption, there are no unmeasured confounders, and the investigators have considered all the variables that can affect the treatment assignment and outcomes.^24,26 Although this assumption cannot be formally tested using the observable data, many researchers have advocated the use of the E value.²⁷ The E value is defined as “the minimum strength of association, on the risk ratio scale, that an unmeasured confounder would need to have with both the treatment and outcome, conditional on the measured covariates, to fully explain a specific treatment–outcome association.”²⁷ Large E values are desired because they imply a stronger association that the unmeasured variable would have to have with the outcome to explain the results to null. Meanwhile, small E values indicate weaker association the unmeasured confounders would have to have with the outcome to explain the results to null.

Software Used

We used R Core Team (2020) software (R Foundation for Statistical Computing, Version 4.0.1, Vienna, Austria). The following packages in the R interface were used to conduct the analyses: survminer,²⁸ survey,²⁹ mice,³⁰ matchthem,³¹ and cobalt.³² Technical details of the statistics were provided in the supplementary.

Results

A total of 2164 patients were identified, and through random selection we screened 1243 records of patients admitted between January 1 and December 30, 2020. Reasons for exclusions are shown in Figure 1. The LMWH (enoxaparin) arm had 470 patients, and the UFH arm had 510 patients. Table 1 presents baseline characteristics for each arm. There were statistical differences in the raw data for mean age (P < 0.001), sex (P < 0.001), ethnicity (P < 0.001), eGFR (P = 0.009), CKD stage (P = 0.004), D-dimer (P = 0.020), median INR (P < 0.001), WHO severity classification (P < 0.001), oxygen need (<0.001), ICU admission (P = 0.046), diabetes (P = 0.039), hypertension (P = 0.025), respiratory diseases (P = 0.002), established cardiovascular disease (P = 0.002), history of coagulopathy (P = 0.004), and history of antiplatelet use (P < 0.001), steroid use (P < 0.001), and other medications used during hospitalization as shown in Table 1. After IPTW, covariate imbalance was greatly minimized between the two arms, as is evident from the SMD of <20% for all the covariates as shown in Table 1 and illustrated in Figure 2. Standard dosing was used in 79.1% of patients for enoxaparin and in 96.7% of patients in the UFH group. The average (SD) daily dose was 54.6 mg (22.1) for enoxaparin and 12261.8 units (2800.6) for UFH. There was no change in dosing in 79.4% of the patients for the enoxaparin arm and in 73.3% of patients in the UFH arm. A small proportion of patients had their dosage adjusted during their hospital stay (7.9% in the enoxaparin arm and 10.6% in the UFH arm). See Table 2 for details. After IPTW, the number of subjects in the weighted analysis slightly increased in the enoxaparin arm (n = 482) and decreased in the UFH arm (n = 479). All the measured baseline characteristics were adequately balanced between the two arms with an absolute SMD < 20%.

Figure 1.

Patient flowchart.

Abbreviations: COVID-19, coronavirus disease 2019; LMWH, low molecular weight heparin; UFH, unfractionated heparin.

Table 1.

Baseline Characteristics and Balance Statistics of the Weighted Covariates (Prior and After Inverse Propensity Score Weighting).

Characteristic	Enoxaparin (n = 470)	UFH (n = 510)	P value	Prior weighting (absolute SMD%)	After weighting (absolute SMD%)	Improvement %
Age (years), mean (SD)	47.7 (12.3)	52.2 (13.9)	<0.001	34.5%	1.5%	67%
Female, n (%)	103 (21.9)	200 (39.2)	<0.001	17.3%	0.8%	83.5%
Ethnicity, n (%)			<0.001
Middle Eastern	196 (41.7)	313 (61.4)		19.7%	0.9%	81.2%
Southeast/South Asian	233 (49.6)	153 (30.0)		19.6%	0.8%	81.2%
African	39 (8.3)	42 (8.2)		0.06%	0.2%	-0.14%
Other/Unknown	2 (0.4)	2 (0.4)		0.0%	0.1%	-0.1%
Weight (kg), median (IQR)	70.0 (70 to 80.0)	70.0 (66.0 to 78.0)	0.313	-	-	-
BMI (kg/m2), median (IQR)	24.7 (24.2 to 27.7)	23.0 (24.22 to 27.7)	0.309	9.7%	3.4%	93.7%
Serum creatinine (mg), median (IQR)	0.91 (0.78 to 1.08)	0.92 (0.75 to 1.13)	0.766	-	-	-
eGFR (mL/min/1.73 m²), median (IQR)	98.8 (80.4 to 110.6)	94.5 (94.4 to 109.0)	0.009	22.4%	1.0%	78.6%
CKD stage, n (%)			0.004	-	-	-
Normal	302 (64.3)	293 (57.5)
Stage 2	130 (27.7)	150 (29.4)
Stage 3A	24 (5.1)	32 (6.3)
Stage 3B	7 (1.5)	12 (2.4)
Stage 4	2 (0.4)	2 (0.4)
Stage 5	2 (0.4)	14 (2.7)
Unknown	3 (0.6)	0 (0.0)
D-dimer (μg /mL), median (IQR)	0.79 (0.46 to 1.45)	0.95 (0.64 to 2.40)	0.020	17.3%	4%	86.7%
Platelets count (109/L), median (IQR)	228.0 (185.1 to 295.0)	224.0 (180.0 to 293.0)	0.434	-	-	-
INR, median (IQR)	0.97 (0.91 to 1.02)	0.94 (0.90 to 1.0)	<0.001	23.6%	0.9%	77.3%
aPTT (s), median (IQR)	31.9 (29.0 to 35.6)	32.3 (29.0 to 35.4)	0.838	-	-	-
WHO severity, classification, n (%)			<0.001
Moderate	299 (63.6)	317 (62.2)		1.4%	0.2%	98.8%
Severe	134 (28.5)	185 (36.3)		7.8%	0.9%	93.1%
Critical	37 (7.9)	8 (1.6)		6.3%	0.8%	94.5%
Oxygen need, n (%)			<0.001
No oxygen therapy	165 (35.1)	123 (24.1)		10.9%	1.1%	90.2%
Non-invasive oxygen therapy	284 (60.4)	379 (74.3)		13.9%	0.7%	86.8%
Invasive oxygen therapy	21 (4.5)	8 (1.6)		2.9%	0.3%	97.4%
ICU admission, n (%)	118 (25.1)	100 (19.6)	0.046	5.5%	0.1%	94.6%
Diabetes, n (%)	150 (31.9)	196 (38.4)	0.039	6.5%	0.3%	93.8%
Hypertension, n (%)	124 (26.4)	169 (33.1)	0.025	6.7%	0.5%	93.8%
Dyslipidemia, n (%)	24 (5.1)	29 (5.7)	0.795	0.5%	0.2%	99.7%
Respiratory diseases, n (%)	19 (4.0)	47 (9.2)	0.002	5.1%	0.9%	95.8%
Established cardiovascular disease, n (%)	24 (5.1)	54 (10.6)	0.002	5.4%	0.4%	95%
Psychiatric disorder, n (%)	8 (1.7)	5 (1.0)	0.479	0.7%	0.0%	99.3%
Gastrointestinal disorder, n (%)	8 (1.7)	11 (2.2)	0.776	0.4%	0.2%	99.8%
Anemia diagnosis, n (%)	14 (3.0)	14 (2.7)	0.978	0.2%	0.1%	99.9%
Neurological disorder, n (%)	8 (1.7)	7 (1.4)	0.873	0.3%	0.1%	99.8%
History of coagulopathy, n (%)	21 (4.5)	48 (9.4)	0.004	4.9%	0.9%	96%
History of major bleeding, n (%)	0 (0.0)	2 (0.4)	0.515	0.3%	0.2%	99.9%
History of antiplatelet use, n(%)	32 (6.8)	71 (13.9)	<0.001	7.1%	0.2%	93.1%
Medications during hospital stay, n(%)
Steroid	140 (29.8)	358 (70.2)	<0.001	40.4%	0.3%	59.9%
Tocilizumab	14 (3.0)	6 (1.2)	0.077	1.8%	0.7%	98.9%
Favipiravir	10 (2.1)	19 (3.7)	0.198	1.6%	0.1%	98.5%
Remdesivir	1 (0.2)	3 (0.6)	0.675	0.4%	0.1%	99.7%
Hydroxychloroquine	133 (28.3)	30 (5.9)	<0.001	22.4%	2.8%	80.4%
Hydroxychloroquine + Azithromycin	29 (6.2)	5 (1.0)	<0.001	5.1%	1.0%	95.9%
Darunavir	1 (0.2)	0 (0.0)	0.967	0.2%	0.1%	99.9%
Interferon beta	3 (0.6)	1 (0.2)	0.560	0.4%	0.2%	99.8%
Lopinavir/Ritonavir	100 (21.3)	35 (6.9)	<0.001	14.4%	5.0%	90.6%
Ribavirin	15 (3.2)	31 (6.1)	0.047	2.9%	0.4%	97.5%
Lopinavir/Ritonavir + Interferon beta + Ribavirin	28 (6.0)	37 (7.3)	0.492	1.3%	0.3%	99%

SMD values for variables that included in the propensity scores model was reported. Respiratory disease: Including asthma and chronic obstructive lung disease, pulmonary tuberculosis, and bronchiectasis. Cardiovascular disease: Coronary artery disease, heart failure, coronary artery bypass surgery, percutaneous coronary intervention, stable angina, stroke or transient ischemic stroke, rheumatic heart disease, cardiomyopathy, myocardial hypertrophy, or unspecified cardiac disease. Psychiatric disorders: Schizophrenia, drug adiction, depression, or unspecified psychiatric disorders. Coagulopathy: Venous thromboembolism or antiphospholipid syndrome. Anemias: Sickle cell disease, glucose 6 phosphate disorders deficiency (G6PD), iron deficiency anemia, thalassemia, or unspecified anemia. Gastrointestinal disorder: Gastritis, hepatitis, inflammatory bowel disease, Crohn’s disease or liver disease. Neurology disorders: Epilepsy, Alzheimer’s disease, Parkinson’s disease or spinal degenerative disease. Missing data: Hemoglobin, platelet count, eGFR, serum creatinine: <1%. INR: 13.5%, aPTT 13.6%. D-dimer:72.6%.

Abbreviations: aPTT, activated partial thromboplastin time; s, seconds; BMI, body mass index; CKD, chronic kidney disease; COVID-19, coronavirus disease 2019; eGFR, estimated glomerular filtration rate estimated by the CKD Epidemiology Collaboration (CKD-EPI) formula; ICU, intensive care unit; INR, international normalized ratio; IQR, interquartile range; Kg, kilogram; SD, standard deviation; SMD, standardized mean difference; UFH, unfractionated heparin; WHO, World Health Organization.

Figure 2.

Love plot of the covariate balance after multiple imputation and inverse propensity score weighting.

Abbreviations: BMI, body mass index; eGFR, estimated glomerular filtration rate; HCQ, hydroxychloroquine; INR, international normalized ratio; WHO, World Health Organization.

Table 2.

Thromboprophylaxis Dosing and Change in Dosing During Admission.

Agent	Enoxaparin (n = 470)	UFH (n = 510)	P value
Dosing, n(%)	-	-	<0.001
Standard	372 (79.4)	493 (96.7)
Intermediate dosing	97 (20.6)	17 (3.3)
Change in agent or dosing, n(%)	-	-	<0.001
No change	373 (79.4)	374 (73.3)
Change from enoxaparin to UFH	27 (5.7)	0 (0.0)
Change from UFH to enoxaparin	0 (0.0)	30 (5.9)
Change to therapeutic enoxaparin	32 (6.8)	47 (9.2)
Change to therapeutic UFH	0 (0.0)	3 (0.6)
Change to therapeutic apixaban	1 (0.2)	2 (0.4)
Average daily dose, mean (SD)a	54.59 mg (22.1)	12261.8 units (2800.6)	-

Abbreviations: SD, standard deviation; UFH, unfractionated heparin.

^aStatistical test was conducted after converting both doses to similar units.

Clinical Outcomes

Table 3 presents the clinical outcome results. In the unadjusted analysis, there was a significant reduction in the primary endpoint in favor of the enoxaparin arm (15.1%) compared with UFH (19.6%) with HR of 0.66 (95% CI: 0.48 to 0.89, P = 0.007). This significance remains after the IPTW adjustment with aHR (adjusted hazard ratio) of 0.46 (95% CI: 0.22 to 0.96, P = 0.039). Mortality was not statistically significant between the two arms in the unadjusted analysis (13.2% in the enoxaparin arm and 13.5% in the UFH arm) with HR = 0.93 (95% CI: 0.66 to 1.31, P = 0.659). The IPTW analysis confirmed lack of significance for the mortality outcome with aHR of 0.49 (95% CI: 0.21 to 1.81, P = 0.112). The unadjusted analysis showed fewer thrombotic events in the enoxaparin arm (5.1%) compared with the UFH arm (9.4%) with HR = 0.47 (95% CI: 0.29 to 0.77, P = 0.003). However, this significance was not observed in the adjusted analysis with aHR 0.67 (95% CI 0.35 to 1.29, P = 0.236. All components of thrombotic events showed lack of significance in the unadjusted analysis except for the rate of PE with HR = 0.47 (95% CI: 0.27 to 0.84, P = 0.010). The PE reduction was not observed in the IPTW analysis. Using the log-rank test, the absolute 30-day survival probability was not significant (P = 0.603) between the two arms in the unadjusted analysis (61% for the enoxaparin arm vs 55.6% for the UFH arm) and the IPTW analysis (62.4% vs 52.6%, P = 0.214). Unadjusted and adjusted Kaplan Meier curves for the primary endpoint, mortality and composite thrombotic events are illustrated in Figures 3-5.

Table 3.

Clinical Outcomes (Unadjusted and Trimmed IPTW Analyses).

Unadjusted analysis
Outcome	Enoxaparin (n = 470)	UFH (n = 510)	HR (95% CI), P valuea
Primary outcomes
Composite primary endpoint ,b n (%)	71 (15.1)	100 (19.6)	0.66 (0.48 to 0.89), P = 0.007
Overall mortality, n (%)	62 (13.2)	69 (13.5)	0.93 (0.66 to 1.31), P = 0.659
Composite thrombotic events,c n (%)	24 (5.1)	48 (9.4)	0.47 (0.29 to 0.77), P = 0.003
Secondary outcomes
30-day survival probability, (%)	61.4%	55.6%	P = 0.603
Pulmonary embolism, n (%)	18 (3.8)	38 (7.5)	0.47 (0.27 to 0.84), P = 0.010
Deep venous thrombosis, n (%)	0 (0.0)	5 (1.0)	-
Ischemic stroke, n (%)	1 (0.2)	5 (1.0)	0.20 (0.02 to 1.73), P = 0.144
Myocardial infarction, n (%)	1 (0.2)	2 (0.4)	0.49 (0.04 to 5.45), P = 0.566
Minor bleeding, n (%)	9 (1.9)	12 (2.4)	0.81 (0.34 to 1.94), P = 0.637
Major bleeding, n (%)	24 (5.1)	20 (3.9)	1.32 (0.72 to 2.42), P = 0.372
Trimmed IPTW analysis
Outcome	Enoxaparin (n = 482)	UFH (n = 479)	HR (95%CI), P valuea
Primary outcomes
Composite primary endpoint,b n (%)	72 (14.9)	90 (18.9)	0.46 (0.22 to 0.96), P = 0.039
Overall mortality, n (%)	61 (12.7)	67 (14.0)	0.49 (0.21 to 1.18), P = 0.112
Composite thrombotic events,c n (%)	33 (6.8)	42 (8.8)	0.67 (0.35 to 1.29), P = 0.236
Secondary outcomes
30-day survival probability, %	62.4%	52.6%	P = 0.214
Pulmonary embolism, n (%)	27(5.6)	29 (6.1)	0.87 (0.44 to 1.71), P = 0.691
Deep venous thrombosis, n (%)	0 (0)	9 (1.9)	-
Ischemic stroke, n (%)	2 (0.4)	4 (0.8)	0.53 (0.05 to 5.97), P = 0.607
Myocardial infarction, n (%)	1 (0.2)	1 (0.2)	0.64 (0.07 to 6.30), P = 0.703
Minor bleeding, n (%)	13 (2.7)	10 (2.9)	1.37 (0.58 to 3.24), P = 0.467
Major bleeding, n (%)	26 (5.4)	16 (3.4)	1.89 (0.99 to 3.62), P = 0.052

Abbreviations: CI, confidence interval; HR, hazard ratio; IPTW, inverse propensity score weighting; UFH, unfractionated heparin.

^aOdds ratios were estimated for bleeding outcomes.

^bDeath or a thrombotic event.

^cPulmonary embolism, deep venous thrombosis, ischemic stroke, or myocardial infarction.

Figure 3.

Primary composite endpoint (mortality or a thrombotic event). (A) Unadjusted. (B) Adjustment with inverse propensity score weighting.

Figure 4.

Overall mortality in the enoxaparin versus the unfractionated heparin arm. (A) Unadjusted. (B) Adjustment with inverse propensity score weighting.

Figure 5.

Thrombotic events in enoxaparin versus unfractionated heparin arm. (A) Unadjusted. (B) Adjustment with inverse propensity score weighting.

No differences were noted in terms of minor bleeding between the two arms with unadjusted odds ratio (OR) = 0.81 (95% CI 0.34 to 1.94, P = 0.637 and adjusted OR = 1.37 (95% CI: 0.58 to 3.24), P = 0.467 in the IPTW analysis. In addition, there was no statistical significance of major bleeding between the two arms in both the unadjusted and adjusted analyses. Lastly, the calculated E value for the primary outcome was 3.77 with a lower bound of CI of 1.25.

Discussion

This observational real-world data study evaluated the effectiveness of two commonly used thromboprophylaxis agents (enoxaparin vs UFH) in a large cohort of COVID-19 patients. The principal findings are as follows. First, with regard to the primary composite endpoint of thrombotic events or mortality, the unadjusted analysis showed overall a greater reduction in these outcomes, which was driven by the statistical significance in reduction of thrombotic events. Second, with regard to the secondary endpoint, the unadjusted analysis revealed that patients treated with enoxaparin were 53% less likely to develop PE compared with UFH. Third, after statistical adjustment using propensity score methods, the results still indicated a statistically significant reduction in the primary composite endpoint with the enoxaparin regimen compared with the UFH regimen without increasing the risk of minor or major bleeding. We noted that the majority of our cohort received the standard dose for enoxaparin (79.4%) and UFH (96.7%).

To our knowledge, there are no recent randomized controlled trial (RCTs) that explicitly compare the two agents (enoxaparin vs UFH) in COVID-19 population. In a scoping review of the planned RCTs (around 20 RCTs worldwide) that would examine anticoagulation therapy in COVID-19 patients, none planned to compare LMWHs (ie, enoxaparin) against UFH.³³ The ISTH³⁴ and National Institutes of Health (NIH)⁸ recommend LMWH over UFH depending on its availability. This recommendation is not supported by solid evidence in COVID-19 patients, but rather is based on convenience for dosing and reliability of administration. However, an individual data meta-analysis of 4 clinical trials with 3600 patients in total was conducted in the non-COVID-19 population for the prevention of VTE. The meta-analysis found that enoxaparin prophylactic treatment significantly reduced rates of VTE and all-cause mortality compared with the UFH arm without increased rates of major bleeding.³⁵ This suggests the advantage of choosing LMWH over UFH in general.

Drawing a parallel from the existing literature to our study is difficult. In our study, majorit of our patients were non-critically ill (74.9% in the enoxaparin and 80% in the UFH arm). In addition, the focus of the study was the choice of thromboprophylaxis agent when used at their standard or intermediate dose rather than therapeutic dosing. On the contrary, the recently published HEP-COVID RCT compared efficacy of therapeutic LMWH with the standard or intermediate dosing of UFH in high risk (D-dimer >4, or critically ill patients). The HEP-COVID RCT defined standard or intermediate dosing of heparins as receiving a UFH dose of 22,500 IU subcutaneously divided into 2 or 3 times a day or use of enoxaparin 30 or 40 mg once or twice a day. However, upon randomization, the majority of patients received a standard or intermediate dose of enoxaparin rather than UFH in the non-ICU stratum.⁹ Hence, making a comparison of their results to ours may not be possible. Nonetheless, the RCT showed the advantage of using therapeutic doses of LMWH with a RR = 0.46 with 95% CI: 0.27 to 0.81, P = 0.004 in the primary outcome (a composite of death and venous thromboembolism or arterial thromboembolism). The same effect was not seen in the ICU stratum, with RR = 0.92 with 95% CI: 0.62 to 1.39, P = 710.⁹ Owing to the anti-inflammatory effect, the choice of the anticoagulant can play a critical role in managing COVID-19 patients.^11,36 The ACTION trial in Brazil compared therapeutic rivaroxaban to usual care in COVID-19 patients with elevated D-dimer. The study found no statistical significance difference in the primary composite of time to death, hospital duration, and oxygen use duration between the two arms with an increased risk of clinically relevant bleeding in the rivaroxaban arm.¹⁵ However, our study showed a difference in our primary endpoint in patients who received enoxaparin versus UFH. The extent by which the anti-inflammatory effect differs by the choice of the anticoagulant is not fully elicited.³⁷

Our finding that 79.4% of patients treated with enoxaparin received the standard dose is important as clinicians try to understand the use of this agent in patients with COVID-19. The standard dose is commonly used in non-COVID-19 hospitalized patients, and several protocols have been developed with a robust body of evidence for safety and efficacy.³⁸ This may streamline and support the approval for use in COVID-19 patients. Arterial and venous thrombotic events are emerging as important issues in patients with COVID-19.³⁸ In many cases, the development of life threatening thromboembolic events occurred despite prophylactic doses of anticoagulation.³⁹ There has been significant variability in medical decision making with regard to anticoagulation among patients with COVID-19. It is well known that anticoagulation is not without its risks of bleeding and is therefore a therapy that requires close clinical monitoring.⁴⁰

Our study has several strengths. It was conducted at one of the main COVID-19 facilities in the Eastern Province of Saudi Arabia during the pandemic with a relatively large sample size. The study reflects real-world practice. Differences in baseline characteristics in the raw data may explain practitioners’ preferences for prescribing enoxaparin versus UFH to their patients. Patients who were older and or had reduced kidney function were prescribed UFH at higher frequencies. However, because of the feasibility of treating all eligible patients with either treatment, we considered the ATE as our target estimand. The use of this target estimand mimics the situation in RCTs, as weights were calculated directly from propensity scores.⁴¹ Our IPTW model considered many of the confounders measured at baseline and greatly reduced bias for our estimate. The E value of 3.77 suggests strong results as the unmeasured covariate would have a large association with the outcome and treatment to sway the results to null. The observed hazard reduction of 0.46 could be explained away (meaning can become null or no effect) if there was an unmeasured confounder associated with the treatment and the outcome by risk ratio of 3.77-fold each and any weaker associations could not do so. Since we considered many confounders in our model, finding such unmeasured variable with this strong level of association maybe unlikely. Moreover, this study was not dependent on administrative data, which are heavily reliant on ICD-10 codes and prone to coding errors. We reviewed patients’ medical records manually with a double-check process to ensure data consistency.

Our study has several limitations. The study design was observational and, unlike the situation of a clinical trial, where patients are routinely undergoing investigation for the presence of thrombotic events, we only considered events confirmed by imaging as prompted by the medical team. Therefore, suspected PE events, for example, that had not been confirmed by imaging were not considered. Thus, we could have missed some PE events due to the difficulty associated with obtaining images in some patients. In addition, missing data were a source of bias in our study. In our cohort, around 73% of baseline D-dimer data were missing. The inclusion of this variable was essential in our IPTW modeling. We overcame this limitation by applying a robust technique (mice equations) assuming D-dimer was missing at random. Convergence and density plots were examined for adequacy of the imputation.⁴² Also, since this study was conducted mainly in the medical wards with a median D-dimer in our study was <1 μg /L, this can limit generalizability of our findings to high-risk individuals. In addition, the study was conducted prior to the availability of effective vaccines for COIVD-19 in the country.⁴³ As the study was conducted in the early phase of the pandemic, the change in treatment protocols (such as the addition of steroids) may not have been fully captured in this study; however, we included many of the used treatments at that time in our model. In addition, this study was conducted in the first wave of the pandemic with majority of cases being in Riyadh and Eastern region where earliest cases belongs to clade GH/20C of the SARS-CoV-2 virus.^44,45 The magnitude to which the treatment effects can change with different variants of the SARS-CoV-2 is unknown.⁴⁶ While many of the described limitations are associated with the study designs, the use of real-world data from a large, geographically diverse hospital database reflects the current state of clinical and prophylactic prescription practices with enoxaparin and UFH to prevent VTE in hospitalized COVID-19 ill patients. Studies that compare the two agents are scarce. Therefore, our findings should be confirmed with randomized control trials.

Conclusion and Relevance

UFH and LMWH are the parenteral anticoagulants most frequently administered as thromboprophylaxis agents in COVID-19 patients. This observational study found that enoxaparin at standard or intermediate dosing was associated with a significant reduction in thrombotic events or mortality compared with UFH at either standard or intermediate dosing. There was no statistically significant reduction in major or minor bleeding. Our findings provide a foundation for randomized controlled clinical trials of anticoagulation regimens.