Introduction

Globally, according to the WHO’s weekly epidemiological updates until January 9, 2023, there have been 659,108,952 confirmed cases of COVID-19 that caused 6,684,756 deaths [1]. A total of 13,073,712,554 vaccine doses have been administered worldwide so far [1, 2] and most cases of severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) infections are frequently asymptomatic or associated with mild symptoms.

Nowadays, severe COVID-19 cases primarily affect unvaccinated patients or those with comorbidities, and are associated with progressive respiratory failure with a high rate of mortality [3]. In the most severe COVID-19 cases, mortality is related to a particular form of acute respiratory distress syndrome involving vascular inflammation and endothelial injury. The evidence supporting microvascular COVID-19 lung vessels obstructive thromboinflammatory syndrome (MicroCLOTS), mediated by an immune-mediated micro-thrombosis, is growing [4]. Thromboprophylaxis has been proposed to prevent MicroCLOTS, although the best timing, method, and dosage of anticoagulation remains to be clarified [5, 6].

International guidelines recommend anticoagulation thromboprophylaxis in hospitalized COVID-19 patients [7,8,9,10,11,12]. Whereas full-dose anticoagulation reduces major arterial and venous thrombotic events in COVID-19 patients, survival benefits were not documented in non-critically ill patients [13] and in the overall population of COVID-19 hospitalized patients [14].

Consequently, we conducted an updated systematic review and meta-analysis of all available randomized controlled trials (mRCTs) evaluating therapeutic-dose heparin-based anticoagulation in non-critically ill patients and evaluated the impact of thromboprophylaxis on mortality.

Methods

The present systematic review and meta-analysis was conducted following the Preferred Reporting for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [15] and Cochrane methodology [16]. The original protocol for this study was registered in the International Prospective Register of Systematic Reviews (PROSPERO). The registration number is CRD42022348993.

We designed the review following the PICOS framework: (1) Population: hospitalized non-critically ill COVID-19 patients, (2) Intervention: therapeutic-dose anticoagulation with heparin/LMWH, (3) Comparison: no therapeutic-dose anticoagulation (including standard treatment, placebo, or prophylactic or intermediate dose anticoagulation), (4) Outcome: the primary and secondary outcomes as listed below, (5) Study design: multicenter randomized controlled trials.

Search Strategy and Study Selection

Two reviewers independently searched PubMed/Medline, Embase, the Cochrane Central Register of clinical trials, medRxiv.org and Clinicaltrials.gov (from inception to November 2022) with no language restrictions. Studies were included if there was agreement between the two reviewers, with disagreements resolved by discussion including a third author. The search strategy (Supplementary material) sought to identify RCTs that compared full-dose heparin-based anticoagulation to prophylactic-dose or intermediate-dose heparin-based anticoagulation in hospitalized non-critically ill COVID-19 patients. We included (1) randomized trials (2) enrolling hospitalized non-critically ill adults (age ≥ 18 years) with a positive RT-PCR or antigen test for SARS-CoV-2 viral infection regardless of gender or ethnicity (3) comparing therapeutic-dose anticoagulation with a heparin product (unfractionated or low-molecular weight heparin) versus no therapeutic-dose anticoagulation. We excluded: trials enrolling critically ill patients (defined according to Authors of each individual study); non-hospitalized patients; pediatric patients; and those involving a non-parallel, non-randomized, or quasi-randomized trial design.

Study Outcomes

The primary outcome of our study was all-cause mortality at the longest follow-up available. Secondary outcomes included: the rate of major thrombotic events (as defined by each individual trial); the rate of major bleeding events (as defined by the International Society on Thrombosis and Haemostasis ISTH); the need for mechanical ventilation; and the need for intensive care unit (ICU) admission [17].

Data Abstraction and Risk of Bias Assessment

For each identified trial, two reviewers independently abstracted data about the overall sample size, treatment type and dose, control therapy and the specified outcomes. Trial investigators were contacted by e-mail for additional data if it was not available in the trial manuscript (Supplementary Material Table 1).

Two independent reviewers evaluated the risk of bias with the Cochrane risk-of-bias tool for randomized trials version 2 (RoB 2) [18]. The overall risk of bias was classified as “Low,” “Some concerns,” or “High” for each study. Small study effect and publication bias were investigated for primary outcome by visual inspection funnel plot and formally tested using the Egger’s test with STATA v.17 (STATACorp, College Station, USA).

Data Analysis

Analyses were conducted using STATA v.17 (STATACorp, College Station, USA) and RevMan 5.4.1 (Review Manager, The Nordic Cochrane Centre, The Cochrane Collaboration, Copenhagen, Denmark). Pooled risk ratios (RRs) were calculated for dichotomous outcomes using the fixed-effects Mantel–Haenszel method. The intention-to-treat principle was followed for all analyses.

To contextualize and visualize the findings, we used the RR and 95% CI for each outcome and simulated 100,000 trials on the log scale, generated a representative probability density function on the RR scale using kernel density estimation, and estimated the probability of any benefit (RR < 1; bleeding: any harm RR > 1) and of a number needed to treat ≤ 100 based on a 1% reduction vs. the overall control event rate on the RR scale (bleeding: number needed to harm based on a 1% increased risk) using STATA v.17 (STATACorp, College Station, USA). Sensitivity analyses were performed by sequentially removing each study and re-assessing the pooled estimates and by changing of summary statistics (RR to odds ratio [OR] or risk difference [RD]).

Heterogeneity Analysis

At first, heterogeneity was evaluated by visual inspection of the forest plots, and then formally assessed with the I2 statistic. We used STATA v.17 (STATACorp, College Station, USA) and RevMan 5.4.1 (Review Manager, The Nordic Cochrane Centre, The Cochrane Collaboration, Copenhagen, Denmark) to test the heterogeneity hypothesis. We considered statistical heterogeneity for an I2 ˂ 25% as low, 26–50% as moderate, and > 50% as high. Fixed-effects models were used in case of low-to-moderate heterogeneity, while random-effects models were to be used in case of high heterogeneity.

Strength of the Evidence Assessment

The GRADE (Grading of Recommendations Assessment, Development and Evaluation) was applied to rate the strength of evidence. The primary and secondary outcomes were evaluated taking into account the following items: study limitations (risk of bias), inconsistency, indirectness, and imprecision [19]. These items were scored as “not serious concerns,” “serious concerns,” and “very serious concerns.” Publication bias, the number of individuals in each group, and the effect measure of the meta-analysis were also evaluated. At the end, the strength of the evidence was evaluated as very low, low, moderate, or high.

Results

Study Characteristics

The search strategy in online databases identified potential studies in Pubmed/Medline (N = 80), Embase (N = 171), Cochrane Central Register of clinical trials (N = 194), ClinicalTrials.gov (N = 26), and MedRxiv (N = 71). After removing duplicates and applying our eligibility criteria, we identified 6 studies for inclusion in the analysis [20,21,22,23,24,25] (Fig. 1).

Fig. 1
figure 1

PRISMA diagram

Full size image

The 6 trials included were multicenter and all included prophylactic or intermediate dose anticoagulation (administered according to local clinical practice guidelines) as control treatment. The studies included a total of 3297 patients (1662 therapeutic-dose heparin-based anticoagulation vs 1635 with prophylactic or intermediate dose anticoagulation) from 13 different countries across 4 continents. (Table 1). The included 6 trials were considered at overall low risk of bias [26] (Fig. 2).

Table 1 Main characteristics of the included studies
Full size table
Fig. 2
figure 2

Traffic plot of mRCTs included in the meta-analysis

Full size image

Quantitative Data Synthesis

Figure 3 shows the forest plot of the rate of all-cause mortality at the longest follow-up available according to the six included randomized studies. All-cause mortality of patients treated with heparin full-dose was 6.2% (103/1662) vs 7.7% (126/1635) in those who received prophylactic or intermediate dose (risk ratio [RR] = 0.76; 95% confidence interval [CI] = 0.59–0.98; P = 0.037; I2 = 43%; Egger’s test P = 0.52). This corresponded to a 98.2% probability of any benefit and to an 84.5% probability that the number needed to treat was ≤ 100 (Supplementary material Fig. 3). Sensitivity analyses showed that the magnitude and direction of the effect size was not modified by sequentially removing each study and re-assessing the pooled estimates (lowest RR = 0.47; 95% CI = 0.27–0.81; P ˂ 0.01; I2 = 22%; removing ATTACC; ACTIV-4a; REMAP-CAP and highest RR = 0.85; 95% CI = 0.65–1.10; P = 0.22; I2 = 0%; removing RAPID). Sensitivity analyses by changing RR to odds ratio (OR = 0.75; CI = 0.57–0.98; P = 0.04; I2 = 44%) or risk difference (RD = − 0.02; CI − 0.04, − 0.00; P = 0.04; I2 = 38%) did not result in a change in significance of study findings.

Fig. 3
figure 3

Forest plot of the rate of all-cause mortality at the longest follow-up available

Full size image

The rate of major arterial and venous thrombotic events was significantly lower for treatment-dose anticoagulation 1.5% (25/1662) vs 3.9% (63/1635) with prophylactic or intermediate dose (RR = 0.41; 95% CI = 0.26–0.64; P ˂ 0.001; I2 = 0%; Fig. 4; Supplementary material Fig. 4). The magnitude and direction of findings did not change with the sequential removal of each trial (lowest RR = 0.29; 95% CI = 0.15–0.57; P ˂ 0.001; I2 = 0%; removing ATTACC; ACTIV-4a; REMAP-CAP and highest RR = 0.47; 95% CI = 0.26–0.83; P ˂ 0.01; I2 = 0%; removing HEP). The rate of major bleeding with treatment-dose anticoagulation was 1.7% (29/1662) vs. 1.3% (22/1636) in those receiving prophylactic or intermediate dose (RR = 1.42; 95% CI = 0.80–2.51; P = 0.23; I2 = 9%; Fig. 5; Supplementary material Fig. 5). This corresponded to an 88.4% probability of increased bleeding with only a 24.4% probability that the number needed to harm would be ≤ 100. The other secondary outcomes are presented in (Table 2 and Supplementary Material Figs. 1, 2, 6, 7).

Fig. 4
figure 4

Forest plot of the rate of major arterial and venous thrombotic events

Full size image
Fig. 5
figure 5

Forest plot of the rate of major bleeding events

Full size image
Table 2 Pooled analysis of studies comparing full-dose heparin anticoagulation to prophylactic or intermediate dose anticoagulation
Full size table

Strength of the Evidence Assessment

According to the GRADE approach (Supplementary Material Table 2) there were not “very serious concerns” about the risk of bias, inconsistency, indirectness, and imprecision [19]. The strength of evidence was moderate for the primary outcome (all-cause mortality), for major bleeding events and for need for mechanical ventilation, while it was high for major arterial and venous thrombotic events and for need for ICU admission.

Discussion

Key Findings

Our systematic review and meta-analysis shows that full-dose thromboprophylaxis (predominantly with LMWH such as enoxaparin, tinzaparin, bemiparin, or dalteparin) might improve survival in hospitalized non-critically ill patients with COVID-19.

Relationship to Previous Studies

There have been a few prior meta-analyses which studied the effect of anticoagulants in hospitalized patients with Covid-19. Important limitations of these previous studies include the following: (1) the combination of direct oral anticoagulants (DOACs) together with heparin-based anticoagulation, (2) the inclusion of patients from different settings (e.g., the inclusion of critically ill patients or outpatients), (3) the inclusion of intermediate doses of anticoagulation in the treatment group, and (4) the inclusion of observational studies which are subject to both have selection and immortal time biases [27,28,29]. By contrast, our meta-analysis focused only on full-dose heparin-based thromboprophylaxis involving non-critically ill hospitalized COVID-19 patients.

In 2021, Giossi et al. did not find a difference in overall all-cause mortality between prophylactic-dose and full-dose of anticoagulation. However, they included trials with heterogeneous treatment (e.g., apixaban or heparin) and high risk of bias studies (e.g., observational) in their analysis [30].

In 2022, Jorda A. et al. included patients receiving heterogeneous treatment (i.e., rivaroxaban or heparin) in different settings of care (critically ill and non-critically ill patients or outpatients) [31]. In addition, the intervention group of their meta-analysis included patients who were not subjected only to full-dose thromboprophylaxis due to the inclusion of the INSPIRATION trial and the X-COVID trial [32, 33].

Significance of Study Findings and What this Study Adds to Our Knowledge

SARS-CoV-2 induced infection, compared with other respiratory infections, seems to be associated with higher rates and severity of thrombotic complications which are associated with worse patient outcomes [10]. Consequently, several international guidelines recommend heparin-based thromboprophylaxis in all COVID-19 hospitalized patients [9, 11, 12]. The suitable dosage (full-dose or intermediate or prophylactic dose) of thromboprophylaxis is still debated [5, 12]. However, practical position statements already suggest that in high thrombotic risk, non-critically ill patients, a full-dose thromboprophylaxis is recommended, taking into consideration the individual patient’s bleeding risk [34].

Severe forms of COVID-19 seem to be linked to the formation in situ of micro-clots caused by an abnormal activation of the immune system in the lungs [4]. It has been hypothesized that anticoagulant therapy may exert a prophylactic role against these immune-thrombotic events by preventing thrombi formation and progression and, furthermore, that this benefit may best be demonstrated early [7] because anticoagulants may have a lesser ability to act on already formed thrombi [8]. In our updated systematic review and meta-analysis of randomized controlled trials involving non-critically ill hospitalized patients, we demonstrated that thromboprophylaxis with therapeutic-dose heparins (primarily LMWH) was associated with a statistically significant reduction of all-cause mortality and thrombotic events balanced with a modest probability of increased major bleeding. We suggest that the overall benefits outweigh the disadvantages and that for most non-critically ill patients at low risk of bleeding, thromboprophylaxis with LMWH should be recommended. Whether or not patients receiving other forms of thromboprophylaxis at admission should be switched to therapeutic-dose heparin-based thromboprophylaxis is an important question for future research.

The reason of the differences in results of previous meta-analyses grouping critically ill and non-critically ill patients is likely related to this aspect. Final results of our meta-analysis show overall benefits of full-dose anticoagulation in hospitalized non-critically ill COVID-19 patients rather than disadvantages. In addition, because the most favorable thromboprophylactic approach during hospitalization is still not defined, our results may help physicians in the decision-making process when treating non-critically ill COVID-19 patients. Our results can help in reducing COVID-19 all-cause in hospital mortality over time and the consequences of venous thromboembolism (e.g., deep vein thrombosis, pulmonary embolism) and arterial thromboembolism (e.g., myocardial infarction and ischemic stroke) in hospitalized non-critically ill COVID-19 patients [35,36,37,38].

Furthermore, there are numerous plausible biological explanations for ancillary beneficial properties of heparin in terms of antiviral effects [39, 40]. Several mechanisms were suggested and are still a research object: (1) the capability of LMWH to bind with high affinity IFNγ, (2) the capability of heparin to influence the biological activity of IL-6 by binding either IL-6 or IL-6/IL-6Rα, (3) the heparin capability to inhibit multiple coagulation proteases, some of which might be specific to COVID-19, determining an overall anti-inflammatory action [41]. Indoubitably more clinical evidence is required to clarify heparin antiviral mechanisms of action and its capability to influence infectious conditions such as COVID-19.

Strengths and Limitations of the Study

The main strength of this meta-analysis is the high quality of the low risk of bias multicenter RCTs included [20,21,22,23,24,25]. Generalizability seems likely as the included patients were not considered critically ill and were recruited from 13 different countries across 4 continents. Disease progression may be too advanced or bleeding complications too high-risk in critically ill patients for them to benefit from full-dose anticoagulation. Another aspect to note is that the use of heparins was heterogeneous across the 6 trials and it is not possible to attribute clinical benefits to a specific molecule. Importantly, few patients were treated with therapeutic dose unfractionated heparin and so the implications for patients with advanced renal dysfunction, for whom LMWH are relative contraindicated, are less clear. Anyway, following Cochrane indications, meta-regression analysis was not included in our investigation because the pooled analysis included fewer than ten studies [42]. To note, meta-analyses should be considered hypothesis-generating, and hence our results remain speculative. Studies included in our analysis included data from the earlier phases of the pandemic when multiple reinfections were uncommon and vaccination was not widely implemented. Our findings may not apply to current COVID-19 epidemiology. However, coagulopathy in COVID-19 patients is still a field of research and there is a need for more structured clinical trials to make physicians able to take precise decisions regarding thromboprophylaxis in COVID-19 patients [43, 44]. In addition, the role of intermediate dose heparin thromboprophylaxis in non-critically ill patients has not been adequately studied in the available RCTs [34]. We searched also for preprint studies, which may be at high risk of bias [45]. However, all of the identified and included trials were published on peer-reviewed journals.

Conclusion

Pooled data from high-quality multicenter randomized trials suggest a reduction in all-cause mortality at the longest follow-up available in hospitalized non-critically ill COVID-19 patients receiving full-dose heparin-based thromboprophylaxis. This is associated with a significant reduction in thrombotic events with the benefits being partly offset by a modest probability of increased bleeding. Overall, the benefits suggest that therapeutic-dose heparin-based thromboprophylaxis should be recommended in most non-critically ill patients who are not at high risk of bleeding.