Anticoagulants for people hospitalised with COVID‐19

Summary of findings 1. Higher‐dose anticoagulants compared to lower‐dose anticoagulants for people hospitalised with COVID‐19

Outcomes	*Anticipated absolute effects^ (95% CI)**		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
Higher‐dose anticoagulants compared to lower‐dose anticoagulants for people hospitalised with COVID‐19
Patient or population: people hospitalised with COVID‐19 Setting: hospital Intervention: higher‐dose anticoagulants (LMWH, UFH or rivaroxaban) Comparison: lower‐dose anticoagulants (LMWH or UFH)
Outcomes	Risk with lower‐dose anticoagulants (short‐term outcomes)	Risk with higher‐dose anticoagulants	Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
All‐cause mortality Follow‐up: from 28‐30 days	Study population		RR 1.03 (0.92 to 1.16)	4489 (4 RCTs)	⊕⊕⊕⊕ High^a	Higher‐dose anticoagulants results in little to no difference in all‐cause mortality
All‐cause mortality Follow‐up: from 28‐30 days	191 per 1000	196 per 1000 (175 to 221)	RR 1.03 (0.92 to 1.16)	4489 (4 RCTs)	⊕⊕⊕⊕ High^a
Necessity for additional respiratory support Follow‐up: from 28‐30 days	Study population		RR 0.54 (0.12 to 2.47)	3407 (3 RCTs)	⊕⊝⊝⊝ Very low^b,c,d	The evidence is very uncertain about the effect of higher‐dose anticoagulants on necessity for additional respiratory support.
	117 per 1000	63 per 1000 (14 to 289)	RR 0.54 (0.12 to 2.47)	3407 (3 RCTs)	⊕⊝⊝⊝ Very low^b,c,d
Mortality related to COVID‐19	No studies measured this outcome
Deep vein thrombosis Follow‐up: from 28‐30 days	Study population		RR 1.08 (0.57 to 2.03)	3422 (4 RCTs)	⊕⊕⊝⊝ Low^d	Higher‐dose anticoagulants may result in little to no difference in DVT
Deep vein thrombosis Follow‐up: from 28‐30 days	11 per 1000	12 per 1000 (6 to 22)	RR 1.08 (0.57 to 2.03)	3422 (4 RCTs)	⊕⊕⊝⊝ Low^d
Pulmonary embolism Follow‐up: from 28‐30 days	Study population		RR 0.46 (0.31 to 0.70)	4360 (4 RCTs)	⊕⊕⊕⊝ Moderate^b	Higher‐dose anticoagulants likely reduce PE
Pulmonary embolism Follow‐up: from 28‐30 days	33 per 1000	15 per 1000 (10 to 23)	RR 0.46 (0.31 to 0.70)	4360 (4 RCTs)	⊕⊕⊕⊝ Moderate^b	Higher‐dose anticoagulants likely reduce PE
Major bleeding Follow‐up: from 28‐30 days	Study population		RR 1.78 (1.13 to 2.80)	4400 (4 RCTs)	⊕⊕⊕⊝ Moderate^b	Higher‐dose anticoagulants likely increase major bleeding slightly
Major bleeding Follow‐up: from 28‐30 days	14 per 1000	24 per 1000 (15 to 38)	RR 1.78 (1.13 to 2.80)	4400 (4 RCTs)	⊕⊕⊕⊝ Moderate^b
Adverse events (minor bleeding) Follow‐up: from 28‐30 days	Study population		RR 3.28 (1.75 to 6.14)	1196 (3 RCTs)	⊕⊕⊕⊕ High	Higher‐dose anticoagulants increase adverse events (minor bleeding)
	20 per 1000	47 per 1000 (18 to 121)	RR 3.28 (1.75 to 6.14)	1196 (3 RCTs)	⊕⊕⊕⊕ High
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; COVID‐19: coronavirus disease 2019; DVT: deep vein thrombosis; LMWH: low‐molecular‐weight heparin; PE: pulmonary embolism; RCT: randomised controlled trial; RR: risk ratio; UFH: unfractionated heparin
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.
^aThe largest study in the analysis was at high risk of bias in almost all domains; however, we did not downgrade for study limitations as removing this study in the sensitivity analysis did not change the pooled estimate. ^bDowngraded one level due to study limitations. One randomised controlled trial provided high risk of bias in almost all domains leading to a different pooled estimate after sensitivity analysis. ^cDowngraded one level due to inconsistency. We identified substantial unexplained heterogeneity (I² = 60%). ^dDowngraded two levels due to imprecision. Confidence interval of the absolute difference comprises both important clinical benefit and important clinical harm.

Summary of findings 2. Anticoagulants compared to no treatment for people hospitalised with COVID‐19

Outcomes	*Anticipated absolute effects^ (95% CI)**		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
Anticoagulants compared to no treatment for people hospitalised with COVID‐19
Patient or population: people hospitalised with COVID‐19 Setting: hospital Intervention: anticoagulants (LMWH, UFH, fondaparinux, DOACs or VKA) Comparison: no treatment (no anticoagulants)
Outcomes	Risk with no treatment	Risk with anticoagulants	Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
All‐cause mortality Follow‐up: from 15‐30 days	Study population		RR 0.64 (0.55 to 0.74)	8395 (3 observational studies)	⊕⊝⊝⊝ Very low^a,b	Anticoagulants may reduce all‐cause mortality but the evidence is very uncertain due to two study results being at critical and serious risk of bias. The numerical results are very unreliable for outcomes where critical risk of bias is an issue
All‐cause mortality Follow‐up: from 15‐30 days	307 per 1000	196 per 1000 (169 to 227)	RR 0.64 (0.55 to 0.74)	8395 (3 observational studies)	⊕⊝⊝⊝ Very low^a,b
Necessity for additional respiratory support	No studies measured this outcome
Mortality related to COVID‐19	No studies measured this outcome
Deep vein thrombosis Follow‐up: up to 15 days	Study population		RR 5.67 (1.30 to 24.70)	1403 (1 observational study)	⊕⊝⊝⊝ Very low^c,d	It is uncertain if anticoagulants have any effect on DVT. The numerical results are very unreliable for outcomes where critical risk of bias is an issue.
Deep vein thrombosis Follow‐up: up to 15 days	3 per 1000	19 per 1000 (4 to 82)	RR 5.67 (1.30 to 24.70)	1403 (1 observational study)	⊕⊝⊝⊝ Very low^c,d
Pulmonary embolism Follow‐up: up to 15 days	Study population		RR 24.19 (3.31 to 176.53)	1403 (1 observational study)	⊕⊝⊝⊝ Very low^c,d	It is uncertain if anticoagulants have any effect on PE. The numerical results are very unreliable for outcomes where critical risk of bias is an issue.
Pulmonary embolism Follow‐up: up to 15 days	2 per 1000	40 per 1000 (5 to 292)	RR 24.19 (3.31 to 176.53)	1403 (1 observational study)	⊕⊝⊝⊝ Very low^c,d
Major bleeding Follow‐up: from 15‐26 days	Study population		RR 1.19 (0.66 to 2.12)	7218 (2 observational studies)	⊕⊝⊝⊝ Very low^b,c,e	It is uncertain if anticoagulants have any effect on major bleeding. The numerical results are very unreliable for outcomes where critical risk of bias is an issue.
Major bleeding Follow‐up: from 15‐26 days	19 per 1000	23 per 1000 (13 to 41)	RR 1.19 (0.66 to 2.12)	7218 (2 observational studies)	⊕⊝⊝⊝ Very low^b,c,e
Adverse events (minor bleeding)	No studies measured this outcome
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; COVID‐19: coronavirus disease 2019; DOACs: direct oral anticoagulants; DVT: deep vein thrombosis; LMWH: low‐molecular‐weight heparin; PE: pulmonary embolism; RR: risk ratio; UFH: unfractionated heparin; VKA: vitamin K antagonist
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.
^aDowngraded two levels due to study limitations. Overall critical/serious risk of bias in two studies, especially related to confounding. ^bDowngraded one level due to inconsistency. We found moderate unexplained heterogeneity (I² = 30% to 60%). ^cDowngraded one level due to study limitations. Overall critical risk of bias, especially related to confounding. ^dDowngraded two levels due to imprecision. Fewer than 300 events were included in the analysis and very large confidence interval. ^eDowngraded one level due to imprecision. Confidence interval of the absolute difference comprises both unimportant clinical harm and important clinical harm.

Background

See Table 1 for a glossary of terms.

Table 1. Glossary of terms

Term	Definition
Anticoagulants	Drugs that suppress, delay or prevent blood clots
Antiplatelet agents	Drugs that prevent blood clots by inhibiting platelet function
Arterial thrombosis	An interruption of blood flow to an organ or body part due to a blood clot blocking the flow of blood
Body mass index (BMI)	Body mass divided by the square of the body height, universally expressed in units of kg/m²
Catheters	Medical devices (tubes) that can be inserted in the body for a broad range of functions, such as to treat diseases, to perform a surgical procedure, and to provide medicine, fluids and food
COVID‐19	An infectious disease caused by SARS‐CoV‐2 virus
Deep vein thrombosis (DVT)	Coagulation or clotting of the blood in a deep vein, that is, far beneath the surface of the skin
Disseminated intravascular coagulopathy	A severe condition in which blood clots form throughout the body, blocking small blood vessels and that may lead to organ failure. As clotting factors and platelets are used up, bleeding may occur, throughout the body (e.g. in the urine, in the stool, or bleeding into the skin)
Duplex ultrasound	Non‐invasive evaluation of blood flow through the arteries and veins by ultrasound devices
Heparin (also known as unfractionated heparin (UFH))	A drug used to prevent blood clotting (anticoagulant, blood thinner)
Hypercoagulability	An abnormality of blood coagulation that increases the risk of blood clot formation in blood vessels (thrombosis)
Low‐molecular‐weight heparin	A drug used to prevent blood clotting (anticoagulant)
Obesity	Amount of body fat beyond healthy conditions (BMI > 30 kg/m²)
Placebo	Substance or treatment with no active effect, like a sugar pill
Platelet	Colourless blood cells that help blood to clot by clumping together
Pulmonary embolism (PE)	Blood clot in the lung or blood vessel leading to the lung. The clot originates in a vein (e.g. deep vein thrombosis) and travels to the lung
Quasi‐randomised controlled trial (quasi‐RCT)	A study in which participants are divided by date of birth or by hospital register number, i.e. not truly randomly divided into separate groups to compare different treatments
Randomised controlled trial (RCT)	A study in which participants are divided randomly into separate groups to compare different treatments
Respiratory failure	An abnormality that results from inadequate gas exchange by the respiratory system
SARS‐CoV‐2	The virus (coronavirus 2) that causes COVID‐19
Thrombosis	Local coagulation of blood (clot) in a part of the circulatory system
Vascular	Relating to blood vessels (arteries and veins)
Venous	Relating to a vein
Venous thromboembolism (VTE)	A condition that involves a blood clot that forms in a vein and may migrate to another location (e.g. the lung)

Description of the condition

The novel coronavirus disease strain, coronavirus disease 2019 (COVID‐19), is caused by severe acute respiratory syndrome coronavirus 2 (SARS‐CoV‐2). COVID‐19 emerged in Wuhan, China and rapidly spread worldwide (Lai 2020). SARS‐CoV‐2 is a highly transmissible virus, and up to 16% of people hospitalised may develop a severe form of the disease (Giannis 2020). Pulmonary effects are typical, but due to high inflammation, hypoxia, immobilisation and diffuse intravascular coagulation, COVID‐19 may predispose patients to both arterial and venous thromboembolism (Ackermann 2020; COVIDSurg 2021; Dolhnikoff 2020; Fox 2020; Long 2020). Venous and arterial thromboembolic complications affect 16% of people hospitalised with COVID‐19 and 31% to 49% of people with COVID‐19 in intensive care units (ICUs), with 90% of such cases being venous thromboembolism (Bilaloglu 2020; Klok 2020a; Klok 2020b). Viral infections induce an imbalance between anticoagulant and procoagulant mechanisms and raise the systemic inflammatory response. Indeed, people with COVID‐19 commonly present with both elevated D‐dimer (fibrin degradation product) and reductions of factors related to clot formation (Giannis 2020). Excessive activation of the coagulation cascade and platelets could explain these haematological findings (Giannis 2020). Coagulopathy and vascular endothelial dysfunction have been proposed as complications of COVID‐19. Emerging data support the hypothesis that asymptomatic individuals with COVID‐19 are at risk of developing pathological thrombosis. The association between large‐vessel stroke and COVID‐19 in young asymptomatic individuals requires further investigation (Oxley 2020); however, Li 2020 found the incidence of stroke among people hospitalised with COVID‐19 to be approximately 5% in a retrospective cohort. Activation of the coagulation system seems to be important in the development of acute respiratory distress syndrome, one of the most typical complications of COVID‐19 infection and it could be related to pulmonary microthrombosis (Ackermann 2020; Dolhnikoff 2020; Fox 2020; Marini 2020).

Description of the intervention

Anticoagulants are pharmacological interventions used in reducing hypercoagulability (Amaral 2020; Dias 2021). The decision to use thromboprophylaxis or not depends on the risk stratification of each patient (NHS 2020).

Anticoagulants are medications used in the prevention and treatment of venous or arterial thromboembolic events (Amaral 2020; Biagioni 2020; Clezar 2020; Dias 2021). When used for a prophylactic purpose, the dose of anticoagulants is usually half or significantly lower than that given for therapeutic purposes (Alquwaizani 2013). Even so, adverse events such as bleeding may occur and can have a significant impact on patient care (Amaral 2020; AVF 2020; Biagioni 2020; Clezar 2020).

How the intervention might work

D‐dimers are a reflection of the pathophysiology in COVID‐19, which is highly associated with increased mortality in people with COVID‐19 infection (Becker 2020). The elevated D‐dimer levels seen are most likely a reflection of the overall clot burden and critically ill people with COVID‐19 have lower levels of fibrinolytic system activation than the reference population (Panigada 2020). Tang 2020 reported decreased mortality after the use of heparin in people with COVID‐19 (40.0% versus 64.2%, P = 0.029). Long 2020 reported that anticoagulation (mainly low‐molecular‐weight heparin), may reduce mortality in people with severe COVID‐19 infection or those with higher levels of D‐dimer (e.g. greater than six times the upper limit).

Some authors had also correlated this effect with the anti‐inflammatory effect of heparinoids, for instance, binding and neutralising a wide variety of mediators released from inflammatory cells, reducing IL‐6 and as potent inhibitors of the complement system, which may have effects on the clinical evolution of people with COVID‐19 (Liu 2019; Shi 2020; Tang 2020; Young 2008). It can attenuate ongoing tissue damage (Liu 2019; Young 2008). Practical guidelines and specialist consensus are addressing the management of thromboprophylaxis and anticoagulation in people with COVID‐19 infection (Bikdeli 2020; NHS 2020; Obe 2020; Ramacciotti 2020). However, the effects of anticoagulants on people with COVID‐19 is still under debate (Sobreira 2020).

Objectives

To assess the benefits and harms of anticoagulants versus active comparator, placebo or no intervention in people hospitalised with COVID‐19.

Methods

Criteria for considering studies for this review

Types of studies

The protocol for this review was prospectively registered with the Open Science Framework on 7 August 2020 (Flumignan 2020a), and a previous version of this review was published on 02 October 2020 (Flumignan 2020b), and disseminated, including a short version published in another international journal (Flumignan 2021).

We considered parallel or cluster‐randomised controlled trials (RCTs), quasi‐RCTs, and cohort studies. Non‐randomised studies (NRS), such as cohort studies, may be useful for rare adverse events and clinical decisions if there is a lack of controlled studies. Related NRS can be developed faster than RCTs and may represent the only available evidence to guide decision making in this setting. To ensure that we captured all relevant study types, we considered a broad range of empirical studies of any size that provided a quantitative measure of impact (Reeves 2021). We did not consider studies without a comparator group or any retrospective NRS because we identified prospective NRS (better study design). We performed meta‐analyses for all of the included studies (RCTs or NRS) with available data to follow Chapter 24 of the Cochrane Handbook for Systematic Reviews of Interventions (Reeves 2021). When at least 400 participants were included from RCTs, we no longer considered NRS for inclusion. We considered all other types of studies irrelevant for this review. Please find further explanations in Appendix 1.

In order to minimise selection bias for NRS, we planned to include only studies that used statistical adjustment for baseline factors using multivariate analyses for at least these confounding factors: participants already using anticoagulants (e.g. atrial fibrillation), participants who underwent surgery during the hospitalisation, active cancer treatment, concomitant antiplatelet use and history of venous thromboembolism. We only considered studies with a minimum duration of two weeks.

Types of participants

We included all participants eligible for anticoagulation, both male and female of all ages, hospitalised with the diagnosis of COVID‐19. Any hospitalised participants with confirmed COVID‐19 infection were eligible, independent of the disease severity (e.g. patients hospitalised in ICUs or wards). We also considered participants with a previous history of venous thromboembolism for inclusion in this review. However, participants with COVID‐19 treated outside of hospital, that is, those who were not hospitalised, were not eligible for our review.

In future updates of this review, if we find studies with mixed populations, that is, hospitalised and non‐hospitalised participants, and only a subset of the participants meet our inclusion criteria, we will attempt to obtain data for the subgroup of interest from the study authors in order to include the study. For studies with mixed populations for which we cannot get data for the subgroup of interest but at least 50% of the study population are of interest, we will include all participants in our analysis. Moreover, we will explore the effect of this decision in a sensitivity analysis. We will exclude studies in which less than 50% of the population are of interest and the subgroup of interest data are not available.

Types of interventions

We considered the following pharmacological interventions.

Heparinoids, that is, both unfractionated heparin and low‐molecular‐weight heparin, and pentasaccharides (synthetic and selective anticoagulant drugs similar to low‐molecular‐weight heparin)
Vitamin K antagonists
Direct anticoagulants, both factor Xa inhibitors and direct thrombin inhibitors, that is, direct oral anticoagulants and non‐oral direct anticoagulants (e.g. bivalirudin)

We considered studies that compared different formulations, doses, and schedules of the same intervention (e.g. heparinoids).

Some commonly applicable prophylactic doses of the interventions of interest are low‐molecular‐weight heparin 30 mg twice a day or 40 mg daily, and unfractionated heparin 5000 IU three times a day. However, we considered all doses of anticoagulants when used for primary or secondary prophylaxis of thromboembolism as eligible for our review.

Types of comparisons

We included studies that compared one pharmacological intervention (agent or drug) versus another active comparator, or placebo or no treatment with any combination of interventions, provided that co‐treatments were balanced between the treatment and control arms. We allowed other potential interventions (e.g. antiplatelet agents, elastic stockings, intermittent pneumatic compression) as comparators or additional interventions. We also included studies that compared different doses of drugs. We pooled the studies that addressed the same comparisons.

Anticoagulant versus placebo or no treatment (we planned to pool all anticoagulants together – heparinoids, vitamin K antagonists, direct anticoagulants, etc. – if possible)
Anticoagulant versus a different anticoagulant
Anticoagulant versus a different dose, formulation, or schedule of the same anticoagulant
Anticoagulant versus other pharmacological interventions such as antiplatelet agents
Anticoagulant versus non‐pharmacological interventions

Types of outcome measures

We evaluated core outcomes as pre‐defined by the Core Outcome Measures in Effectiveness Trials Initiative for people with COVID‐19 (COMET 2020). We also considered the outcomes after hospital discharge. We intended to present the outcomes at two different time points following the start of the intervention if data were available: short‐term outcomes (at hospital discharge or before); and long‐term outcomes (after hospital discharge).

Our time point of primary interest is short‐term; we, therefore, intended to produce related summary of findings tables only for this time point, and also planned to report the long‐term outcomes at the longest possible time of follow‐up.

Primary

All‐cause mortality
Necessity for additional respiratory support: oxygen by non‐invasive ventilators or high‐flow intubation and mechanical ventilation or extracorporeal membrane oxygenation.

Secondary

Mortality related to COVID‐19
Deep vein thrombosis, symptomatic or asymptomatic, first episode or recurrent confirmed by ultrasonography or angiography (e.g. by computed tomography (CT), magnetic resonance imaging (MRI) or by digital subtraction) from any site (e.g. lower limbs, upper limbs, abdominal)
Pulmonary embolism (symptomatic or asymptomatic, first episode or recurrent, fatal or non‐fatal): a diagnosis had to be confirmed by angiography (e.g. by CT, MRI or digital subtraction) and ventilation‐perfusion scan, or both. We also considered post mortem examination as an objective confirmation of deep vein thrombosis and pulmonary embolism.
Major bleeding: defined by a haemoglobin concentration decrease of 2 g/dL or more, a retroperitoneal or intracranial bleed, a transfusion of two or more units of blood, or fatal haemorrhagic events, as defined by International Society on Thrombosis and Haemostasis (Schulman 2010)
Adverse events. We will consider all possible adverse events separately, as individual outcomes, such as minor bleeding, gastrointestinal adverse effects (e.g. nausea, vomiting, diarrhoea, abdominal pain), allergic reactions, renal failure and amputations.
Hospitalisation time in days
Quality of life: participant's subjective perception of improvement (yes or no) as reported by the study authors or using any validated scoring system such as the Short Form‐36 Health Survey (SF‐36) (Ware 1992)

We planned to include studies in the review irrespective of whether measured outcome data were reported in a ‘usable’ way.

Search methods for identification of studies

An information specialist (LLA) designed and conducted all searches on 20 June 2020, which were informed and verified by a content expert (RLGF) and independently peer reviewed. The search was updated on 14 April 2021.

Electronic searches

We identified eligible study references through systematic searches of the following bibliographic databases.

Cochrane Central Register of Controlled Trials (CENTRAL; 2020, Issue 6) in the Cochrane Library (searched 20 June 2020)
MEDLINE PubMed (1946 to 20 June 2020)
Embase.com Elsevier (1974 to 20 June 2020)
LILACS Virtual Health Library (Latin American and Caribbean Health Sciences Literature database; 1982 to 20 June 2020)
IBECS Virtual Health Library (Indice Bibliográfico Español de Ciencias de la Salud; 2015 to 20 June 2020)

We adapted the preliminary search strategy for MEDLINE PubMed for use in the other databases. We did not apply any RCT filters for any databases; we selected the study design manually because we also considered NRS for inclusion in this review. See Flumignan 2020b for search strategies conducted in June 2020.

For this update, we subsequently conducted systematic update searches of the following databases for relevant trials without language, publication year or publication status restrictions on 14 April 2021:

Cochrane Central Register of Controlled Trials (CENTRAL; 2021, Issue 3) in the Cochrane Library (searched from 20 June 2020 to 14 April 2021; Appendix 2)
MEDLINE PubMed (searched from 20 June 2020 to 14 April 2021; Appendix 3)
Embase.com Elsevier (searched from 1 January 2020 to 14 April 2021; Appendix 4)
LILACS Virtual Health Library (searched from 1 January 2020 to 14 April 2021; Appendix 5)
IBECS Virtual Health Library (searched from 1 January 2020 to 14 April 2021; Appendix 5)

We searched all databases from their inception to the present, and we did not restrict the language of publication or publication status. We considered the adverse effects described in the included studies only. All relevant MeSH and Emtree index terms for COVID‐19 and SARS‐CoV‐2 will be integrated into electronic search strategies in future updates.

Searching other resources

We also conducted a search of the Cochrane COVID-19 Study Register (Appendix 6), a specialised register containing both trial registry records, journal articles and preprints, and medRxiv (Appendix 7), a preprint server, for ongoing or unpublished studies (both searched 20 June 2020). The Cochrane COVID-19 Study Register is a specialised register built within the Cochrane Register of Studies (CRS) and is maintained by Cochrane Information Specialists. The register contains study reports from several sources, including:

daily searches of ClinicalTrials.gov
weekly searches of PubMed
weekly searches of Embase.com
weekly searches of the WHO International Clinical Trials Registry Platform (ICTRP)
monthly searches of the Cochrane Central Register of Controlled Trials (CENTRAL)

Complete data sources and search methods for the register are available at community.cochrane.org/about-covid-19-study-register.

For this update, we subsequently performed update searches of the following on 14 April 2021:

We checked the reference lists of all included studies and any relevant systematic reviews identified for additional references to studies. We examined any relevant retraction statements and errata for included studies. We contacted the authors of the included studies for any possible unpublished data. Furthermore, we contacted field specialists to enquire about relevant ongoing or unpublished studies.

Data collection and analysis

Inclusion of non‐English language studies

We considered abstracts and full texts in all languages for inclusion. All potentially eligible non‐English language abstracts progressed to full‐text review, with methods translated for eligibility, and full text translated for data extraction.

Selection of studies

Two review authors (JDST, LCUN) independently screened titles and abstracts of all the potential studies we identified as a result of the search and coded them as 'retrieve' (eligible or potentially eligible/unclear) or 'do not retrieve', using the Covidence tool. If there were any disagreements, we asked a third review author to arbitrate (RLGF). We retrieved the full‐text study reports/publications, and two review authors (JDST, LCUN) independently screened the full text and identified studies for inclusion, while identifying and recording reasons for the exclusion of ineligible studies. We resolved any disagreement through discussion or, if required, we consulted a third person (RLGF). We identified and excluded duplicates and collated multiple reports of the same study so that each study, rather than each report, is the unit of interest in the review. We recorded the selection process in sufficient detail to complete a PRISMA flow diagram (Page 2021), and Characteristics of excluded studies table. We considered studies reported as full text, those published as abstract only, and unpublished data. We considered abstracts and conference proceedings if they were eligible and had usable data.

Data extraction and management

We managed and synthesised the available data using Review Manager 5 (Review Manager 2020). If there was a conflict between data reported across multiple sources for a single study (e.g. between a published article and a trial registry record), we planned to use the article published for numerical analysis, and we planned to report the differences and consider it on the certainty of evidence (GRADE approach; Schünemann 2013).

We used a data collection form, which we piloted on at least one study in the review, for study characteristics and outcome data. One review author (RLGF) extracted study characteristics from the included studies. We planned to extract the following study characteristics.

Methods: study design, total duration of the study, number of study centres and location, study setting, and date of the study
Participants: comorbidities, ventilation support, pregnancy, number randomised, number lost to follow‐up/withdrawn, number analysed, number of interest, mean age, age range, gender, the severity of the condition, inclusion criteria, and exclusion criteria
Interventions: intervention and comparison characteristics (e.g. manufacture, dosage, additional procedures, method of administration), concomitant medications, and excluded medications
Outcomes: primary and secondary outcomes specified and collected (e.g. how outcomes are measured), and time points reported. For NRS: confounding factors controlled for each relevant analysis presented
Notes: funding for the trial, and notable conflicts of interest of study authors

One review author (RLGF) extracted outcome data from included studies independently, which were verified by the other two review authors (CM, BT). We planned to resolve disagreements by discussion. One review author (RLGF) transferred data into Review Manager 5 (Review Manager 2020). We double‐checked that data were entered correctly by comparing the data presented in the systematic review with the data extraction form. Two review authors (CM, BT) spot‐checked study characteristics for accuracy against the study report.

Assessment of risk of bias in included studies

For data from RCTs we used RoB 1 to analyse the risk of bias in the underlying study results (Higgins 2017). For data from prospective NRS, we used the Risk Of Bias in Non‐randomised Studies of Interventions (ROBINS‐I) tool, version of 2016 (Sterne 2016). We also planned to use ROBINS‐I to assess the risk of bias in quasi‐RCTs or retrospective NRS.

Randomised controlled trials

We planned for one review author (RLGF) to assess the risk of bias for each study, and another review author (LCUN) to check all judgements, using the criteria outlined in the Cochrane Handbook for Systematic Reviews of Interventions for RCTs (RoB 1) (Higgins 2017). We planned to resolve any disagreements by consensus or by involving other review authors (CM, BT). For RCTs, we planned to assess the risk of bias according to the following domains.

Random sequence generation
Allocation concealment
Blinding of participants and personnel
Blinding of outcome assessment
Incomplete outcome data
Selective outcome reporting
Other bias

In cluster‐randomised trials, we planned to consider particular biases as recommended by section 8.15.1.1 of the Cochrane Handbook for Systematic Reviews of Interventions:

recruitment bias;
baseline imbalance;
loss of clusters;
incorrect analysis; and
comparability with individually randomised trials (Higgins 2017).

We planned to grade each potential source of bias as high, low or unclear and provide a quote from the study report together with a justification for our judgement in the risk of bias table. We planned to summarise the risk of bias judgements across different studies for each of the domains listed. Where information on the risk of bias relates to unpublished data or correspondence with a study author, we planned to note this in the risk of bias table.

When considering treatment effects, we planned to take into account the risk of bias for the studies that contributed to that outcome.

We planned to base the overall bias judgement of included RCTs on the following three domains of RoB 1:

adequate sequence generation;
blinding of outcome assessors; and
selective outcome reporting.

An RCT at low risk on all of these domains we planned to label as a low‐risk study. An RCT at high risk on one of these domains we planned to label as a high‐risk study. If there was no clear information on the risk of bias for one or more key domains, but the RCT was not at high risk for any domain, we planned to indicate that the risk of bias in the study was unclear.

Non‐randomised studies

Using the ROBINS‐I tool, version of 2016, we planned to assess the risk of bias of quasi‐RCTs and NRS based on the following seven domains (Sterne 2016).

Bias due to confounding
Bias in selection of participants into the study
Bias in classification of interventions
Bias due to deviations from the intended intervention
Bias due to missing data
Bias in measurement of outcomes
Bias in selection of the reported result

We planned to use our risk of bias judgements for quasi‐RCTs and NRS to label all outcomes at all time points, for each comparison, on these domains as 'critical risk', 'serious risk', 'moderate risk', 'low risk', or 'no information'. We planned to judge the overall risk of bias (across domains) as the worst judgment across all the domains. We were interested in the effect of assignment (intention to treat (ITT)) and ROBINS‐I was used to assess all outcomes at all time points.

We considered the following confounders for the assessment of ROBINS‐I domain on 'confounding' and used the Robvis tool to create the risk of bias graphs for NRS (McGuinness 2020).

Participants already using anticoagulants (e.g. atrial fibrillation)
Participants who underwent surgery during hospitalisation
Active cancer treatment
Concomitant antiplatelet use
History of venous thromboembolism

Measures of treatment effect

Please refer to Appendix 1 for information regarding how we had planned to measure the treatment effects of RCTs, quasi‐RCTs and NRS.

Unit of analysis issues

We included RCTs (for one comparison) and NRS (for another comparison) and performed meta‐analysis when appropriate.

Please refer to Appendix 1 for information regarding how we had planned to combine studies with multiple treatment groups.

Dealing with missing data

We planned to contact investigators or study sponsors in order to verify key study characteristics and obtain missing numerical outcome data where possible (e.g. when a study is identified as abstract only). Where possible, we planned to use the Review Manager 5 calculator to calculate missing standard deviations using other data from the study, such as confidence intervals. Where this was not possible, and the missing data were thought to introduce serious bias, we planned to explore the impact of including such studies in the overall assessment of results by a sensitivity analysis. For all outcomes, we planned to follow ITT principles to the highest degree possible: that is, we planned to analyse participants in their randomised group regardless of what intervention they received. We planned to use available case data for the denominator if ITT data were not available. We estimated the mean difference (MD) using the method reported by Wan 2014 to convert median and interquartile range (IQR) into MD and confidence intervals (CI). When it was not possible, we narratively described skewed data reported as medians and IQRs.

Dealing with sparse data

We planned to adjust comparisons (e.g. grouping broader categories of participants (all ages), grouping broader variations of intervention (all types of anticoagulants) accordingly, regardless of sparse data.

Assessment of heterogeneity

We included RCTs (for one comparison) and NRS (for another comparison) and performed meta‐analysis when appropriate.

Please refer to Appendix 1 for information regarding how we had planned to assess heterogeneity.

Assessment of reporting biases

If we were able to pool more than 10 studies, we planned to create and examine a funnel plot to explore possible small‐study biases for the primary outcomes.

Data synthesis

We planned to use a fixed‐effect model for meta‐analysis when included studies were homogeneous (considering population, interventions, comparators and outcomes characteristics). We planned to use a random‐effects model if we identified at least substantial heterogeneity, or if significant clinical differences regarding participants and interventions existed among included studies.

Please refer to Appendix 1 for information regarding how we had planned to synthesise data from RCTs, quasi‐RCTs and NRS. We meta‐analysed data from RCTs (one comparison) and from NRS (another comparison) when appropriate. We also reported the outcome data of each included study narratively and using tables.

Synthesis without meta‐analysis

We planned to synthesise the data using Review Manager 5 (Review Manager 2020). We planned to report data narratively if it was not appropriate to combine it in a meta‐analysis, and planned to undertake meta‐analyses only where this was meaningful, that is, if the treatments, participants and underlying clinical question were similar enough for pooling to make sense.

We aimed to analyse data from NRS separately in a spreadsheet with the exposure of the sample number and the quantitative and qualitative variables relevant to the review, and we also meta‐analysed data from NRS.

We intended to describe skewed data reported as medians and IQRs narratively.

If a meta‐analysis was not possible, we planned to explore the possibilities above to show data of all relevant outcomes considered in this review. Where there was substantial clinical, methodological, or statistical heterogeneity across studies that prevented the pooling of data, we aimed to use a narrative approach to data synthesis. We planned to describe skewed data reported as medians and IQRs narratively.

Subgroup analysis and investigation of heterogeneity

We planned to explore the following subgroups related to participants or interventions if heterogeneity was substantial.

Different doses of drugs
Duration of prophylaxis (e.g. until 30 days after the start of intervention or more)
Age (e.g. children (up to 18 years), adults (18 years to 64 years) and seniors (65 years and over))
Gender
Comorbidities
Illness severity
Type of ventilator support:
- oxygen by non‐invasive ventilators or high flow
- intubation and mechanical ventilation
- extracorporeal membrane oxygenation

Sensitivity analysis

We planned to carry out the following sensitivity analyses to test whether critical methodological factors or decisions have affected the main result. We planned to group according to study design (RCTs or cluster‐RCTs, quasi‐RCTs, NRS).

Only including studies with a low risk of bias, as previously specified ('Assessment of risk of bias in included studies').
We planned to examine both the fixed‐effect model and random‐effects model meta‐analyses, and we planned to explore the differences between the two estimates.
We planned to explore the decision to include all participants when at least 50% were of interest in a study with a mixed population.
We planned to explore the impact of missing data. If we identified studies with missing data that were unobtainable, we planned to repeat analyses excluding these studies to determine their impact on the primary analyses.

We also planned to carry out sensitivity analyses considering cluster‐RCTs and investigate the effect of variation in the intracluster correlation coefficient (ICC), as well as planning to acknowledge heterogeneity in the randomisation unit and perform a sensitivity analysis to investigate the effects of this randomisation unit. We aimed to present these results and compare them with the overall findings. We planned to justify any post hoc sensitivity analyses that arose during the review process in the final report.

Summary of findings and assessment of the certainty of the evidence

We created a summary of findings table for both short‐term and long‐term time points using the following outcomes.

All‐cause mortality
Necessity for additional respiratory support
Mortality related to COVID‐19
Deep vein thrombosis
Pulmonary embolism
Major bleeding
Minor bleeding

We used the five GRADE considerations (study limitations; consistency of effect; imprecision; indirectness; and publication bias) to assess the certainty of a body of evidence as it relates to the studies that contributed data to the analyses for the prespecified outcomes. We used methods and recommendations described in Chapter 14 of the Cochrane Handbook for Systematic Reviews of Interventions (Schünemann 2021), using GRADEpro GDT software. We made a separate summary of findings table for each of the following comparisons with available data.

Anticoagulants (higher dose) versus anticoagulants (lower dose)
Anticoagulant (all types) versus no treatment

We justified all decisions to downgrade the certainty of evidence using footnotes, and made comments to aid the reader's understanding of the review where necessary.

Two review authors (RLGF, VTC) made judgements about the certainty of the evidence, with disagreements resolved by discussion or by involving a third review author (LCUN). We justified, documented and incorporated judgements into reporting of results for each outcome. We extracted study data, formatted our comparisons in data tables and prepared a summary of findings table with meta‐analysis before writing the results and conclusions of our review.

Results

Results of the search

For this update, we identified 7322 new records in addition to the 1148 potentially relevant records from the first version (altogether 8470 records). After removing duplicates, we screened 7329 records based on their titles and abstracts, and excluded 7072 records that did not meet the prespecified inclusion criteria. We selected 257 records for full‐text reading. We excluded 129 studies after a full‐text analysis as we considered them not relevant and we excluded 59 studies for at least one reason (see Characteristics of excluded studies). Sixty‐two studies are ongoing (see Characteristics of ongoing studies).

For this review, we found seven studies with available data for inclusion; four RCTs (Lemos 2020; Lopes 2021; Sadeghipour 2021; Zarychanski 2021), and three NRS (Albani 2020; Rentsch 2020; Santoro 2020). See Figure 1 for the study flow diagram (Page 2021). As there is now evidence available from RCTs, and prospective NRS, we excluded the studies analysed in the previous version of this review because they were all retrospective NRS (Ayerbe 2020; Liu 2020; Paranjpe 2020; Russo 2020; Shi 2020; Tang 2020; Trinh 2020) (Flumignan 2020b).

Figure 1

Study flow diagram
RCT: randomised controlled trial; NRS: non‐randomised study

Included studies

See Table 2 for the summarised characteristics of included studies.

Table 2. Summary of characteristics of included studies

Study (design)	Country	Participant age (mean ± SD)	Setting	Intervention type (dose)	Comparator	All‐cause mortality	Necessity for additional respiratory support	Follow‐up time (mean days)	Total participants allocated	Intervention group participants (anticoagulant)
Albani 2020 (Prospective cohort)	Italy	68.66 ± 12.62 (experimental), 70.6 ± 15.01 (comparator)	Hospital^a	Enoxaparin (40‐80 mg once daily, duration 3‐9 days)	NA	In‐hospital mortality: aOR 0.53 (95% CI 0.10 to 0.70), in favour of intervention group	NR	Until death or hospital discharge (time in days NR)	1403	799
Lemos 2020 (RCT)	Brazil	55 ± 10 (experimental), 58 ± 16 (comparator)	Hospital^a	Therapeutic anticoagulation: heparin (SC enoxaparin, adjusted dose by age and CrCl (maximum dose allowed 140 mg twice daily)	Prophylactic anticoagulation: SC UFH 5000 IU three times/day (if weight < 120 kg) and 7500 IU 3 times/day (if weight > 120 kg) or enoxaparin 40 mg once daily (if weight < 120 kg) and 40 mg twice daily (if weight > 120 kg) according to the doctor's judgment	RR 0.33 (95% CI 0.04 to 2.69)	NR	28	20	10
Lopes 2021 (RCT)	Brazil	56.7 ± 14.1 (experimental), 56.5 ± 14.5 (comparator)	Hospital^a	Therapeutic anticoagulation: stable participants = rivaroxaban 20 mg once daily; unstable participants = enoxaparin 1 mg/kg twice daily. Followed by rivaroxaban for 30 days, irrespective of the duration of hospitalisation	Prophylactic anticoagulation: enoxaparin 40 mg once daily	RR 1.49 (95% CI 0.90 to 2.46)	RR 0.16 (95% CI 0.02 to 1.35)	30	615	310
Rentsch 2020 (Prospective cohort)	USA	67.03 ± 12.31 (experimental), 67.83 ± 13.74 (comparator)	Hospital^a	SC UFH (5000 IU twice daily or 3 times/day (1094 participants; 30.2%) LMWH (enoxaparin 40 mg once or twice daily (2506 participants; 69.1%), fondaparinux 2.5 mg once daily (4 participants; 0.1%), dalteparin 2500‐5000 IU once daily, all SC) DOACs (apixaban 2.5 mg twice daily (21 participants; 0.6%), rivaroxaban 10 mg once daily or 2.5 mg twice daily (2 participants; 0.1%), dabigatran 220 mg once daily, all orally)	NA	Inpatient mortality: aHR 0.69 (95% CI 0.61 to 0.77) 30‐day mortality: aHR 0.73 (95% CI 0.66 to 0.81)	NR	30	4297	3627
Sadeghipour 2021 (RCT)	Iran	61.23 ± 14.68 (experimental), 59.66 ± 17.88 (comparator)	Hospital^a	Higher‐dose anticoagulation: enoxaparin 1 mg/kg once daily, modified according to body weight and CrCl	Lower‐dose anticoagulation: enoxaparin 40 mg once daily, modified according to body weight and CrCl	Short‐term time point: RR 1.05 (95% CI 0.87 to 1.28) Long‐term time point: RR 1.07 (95% CI 0.89 to 1.29)	Short‐term time point: no events in both groups Long‐term time point: no events in both groups	90	562	276
Santoro 2020 (Prospective cohort)	Spain, Italy, Ecuador, Cuba, Germany, China, Canada, Serbia, USA, Chile, and Colombia	66 ± 15 (experimental), 63 ± 27 (comparator)	Hospital^a	Anticoagulant (oral, SC, or IV): 327 (12%) participants = previous anticoagulation treatment 1888 (72%) participants = prophylactic (lower‐dose) during hospitalisation 341 (13%) participants = therapeutic (higher‐dose) LMWH 23 (0.75%) oral anticoagulation with VKA 23 (0.75%) DOACs	NA	RR 0.91 (95% CI 0.89 to 0.93), in all participants (N = 3089) RR 0.58 (95% CI 0.49 to 0.67), in those non‐anticoagulated before admission (N = 2695) RR 0.50 (95% CI 0.37 to 0.70), in those undergoing invasive ventilation (N = 391) RR 0.72 (95% CI 0.51 to 1.01), in those undergoing non‐invasive ventilation (N = 583)	NR	15	5838	2601
Zarychanski 2021 (RCT)	UK, USA, Canada, Brazil, Ireland, Netherlands, Australia, Nepal, Saudi Arabia, and Mexico	Critically ill: 60.2 ± 13.1 (experimental), 61.6 ± 12.5 (comparator) Moderate‐severity illness: 59.0 ± 14.1 (experimental), 58.8 ± 13.9 (comparator)	Hospital^a	Therapeutic anticoagulation: LMWH or UFH according to local protocols used for the treatment of acute VTE for up to 14 days or until recovery (defined as hospital discharge, or liberation from supplemental oxygen for ≥ 24 h)	Prophylactic anticoagulation: LMWH or UFH according to local practice or with guidance from the trial protocol on maximum dosing, which included either standard low‐dose thromboprophylaxis or enhanced intermediate dose thromboprophylaxis	Short‐term time point: moderate‐severity RR 0.89 (95% CI 0.67 to 1.19), critically ill RR 1.03 (95% CI 0.88 to 1.21) Long‐term time point: NR	Short‐term time point: moderate‐severity RR 0.89 (95% CI 0.74 to 1.08), critically ill: NR Long‐term time point: NR	90	3450	1780
Total	Australia: 1 Brazil: 3 Canada: 2 Chile: 1 China: 1 Colombia: 1 Cuba: 1 Ecuador: 1 Germany: 1 Iran: 1 Ireland: 1 Italy: 2 Mexico: 1 Nepal: 1 Netherlands: 1 Saudi Arabia: 1 Serbia:1 Spain: 1 UK: 1 USA: 3	55 to 68.66 (mean, 7 studies)	‐	‐	‐	7 studies considered mortality	4 studies considered additional respiratory support	15 to 90 (7 studies)	16,185	9403
aHR: adjusted hazard ratio; aOR: adjusted odds ratio; twice daily: twice a day; CI: confidence interval; CrCl: creatinine clearance; DOACs: direct oral anticoagulants; GFR: glomerular filtration rate;HR: hazard ratio; ICU: intensive care units; IU: international unit;LMWH: low‐molecular‐weight heparin; NA: no anticoagulation; NR: not reported; NRS: non‐randomised study;OR: odds ratio; RCT: randomised controlled trial; RR: risk ratio; SC: subcutaneous; SD: standard deviation; SIC: sepsis‐induced coagulopathy; TID: three times a day; UFH: unfractionated heparin; VKA: vitamin K antagonist

^aHospital: includes intensive care unit, hospital wards or emergency department.
^bAnticoagulation used twice daily if glomerular filtration rate (GFR) was > 30 mL/min, or once daily if GFR was 30 mL/min or less.

We included seven studies describing 16,185 participants in this review, of whom at least 9403 received anticoagulants (Albani 2020; Lemos 2020; Lopes 2021; Rentsch 2020; Sadeghipour 2021; Santoro 2020; Zarychanski 2021). From the seven included studies, four were RCTs (Lemos 2020; Lopes 2021; Sadeghipour 2021; Zarychanski 2021) and the other three NRS of interventions (Albani 2020; Rentsch 2020; Santoro 2020), with a comparator group. Of the seven included studies, two originated from Brazil (Lemos 2020; Lopes 2021), one from Iran (Sadeghipour 2021), one from Italy (Albani 2020), and one from the USA (Rentsch 2020), while two involved several countries (Santoro 2020; Zarychanski 2021).

All included RCTs compared different doses of anticoagulant (lower versus higher) (Lemos 2020; Lopes 2021; Sadeghipour 2021; Zarychanski 2021), and all included NRS compared anticoagulation versus no anticoagulation (Albani 2020; Rentsch 2020; Santoro 2020). All included studies reported a follow‐up period that varied from 15 to 90 days. All included studies considered participants from all settings (ICU, hospital wards and emergency departments), but only Zarychanski 2021 reported data separately by disease severity (critically ill and moderate severity). All included studies reported data regarding the age of participants; the mean age varied from 55 to 68.66 years. All studies reported data on mortality, and four reported data for the necessity for additional respiratory support (Lemos 2020; Lopes 2021; Sadeghipour 2021; Zarychanski 2021).

All studies described the type and dose of anticoagulation. Four studies used only heparin in the intervention group (Albani 2020; Lemos 2020; Sadeghipour 2021; Zarychanski 2021), and the other three analysed data from heparin, vitamin K antagonist or direct anticoagulants. Only Sadeghipour 2021 compared higher‐dose anticoagulation (enoxaparin 1 mg/kg once daily, modified according to body weight and creatinine clearance) versus lower‐dose anticoagulation (enoxaparin 40 mg once daily, modified according to body weight and creatinine clearance) without a therapeutic dose. The other three RCTs compared therapeutic (higher) dose anticoagulation versus prophylactic (lower) dose anticoagulation.

Please refer to the Characteristics of included studies for detailed information.

Excluded studies

We excluded 59 studies for at least one reason (Characteristics of excluded studies). Four studies had ineligible interventions because they evaluated aspirin (NCT04365309), anticoagulants for arterial line heparinisation (Maurer 2020), or anti‐inflammatory drugs (EUCTR2020‐001748‐24‐SE; Mareev 2020), and there was no difference between the intervention groups regarding anticoagulants. Ten studies evaluated ineligible participants (CTRI/2021/01/030373; Kukin 2020; NCT04483830; NCT04492254; NCT04504032; NCT04516941; NCT04662684; NCT04673214; NCT04715295; NCT04757857), and all other excluded studies had an ineligible study design for one of the following reasons:

retrospective cases series without a consistent comparator group;
prospective cohort study without a comparator group (single‐arm study);
prospective cohort study without an intervention purpose;
prospective before‐after cohort study without a parallel comparator group;
editorial articles;
retrospective NRS (new registers and previously included studies from the first version of this review (Flumignan 2020b));
prospective cohort study without a parallel comparator group of intervention.

Ongoing studies

Sixty‐two ongoing studies met our inclusion criteria. They plan to evaluate 35,470 participants (120 participants from two NRS and 35,350 participants from 60 RCTs). From 60 RCTs, 28 are comparing different doses of anticoagulants, 24 are comparing anticoagulants versus no anticoagulants, seven are comparing different types of anticoagulants, and one did not report detail of the comparator group (Wilkinson 2020). We tried to contact the study authors and also searched by study registration number and title of the study on all databases of interest for this review. However, there are no additional data for all these ongoing studies. See the Characteristics of ongoing studies table for further details.

Six of the ongoing studies plan to include 1000 participants or more in RCTs (CTRI/2020/11/029345; EUCTR2020‐001708‐41‐IT; NCT04333407; NCT04366960; NCT04512079; Wilkinson 2020).

CTRI/2020/11/029345 plans to compare prophylactic enoxaparin, full‐dose enoxaparin and apixaban versus no anticoagulant in 3600 participants to assess the composite of all‐cause mortality, intubation requiring mechanical ventilation, systemic thromboembolism or ischaemic stroke.
EUCTR2020‐001708‐41‐IT plans to compare 40 mg enoxaparin once daily versus twice daily in 2000 participants to assess the incidence of venous thromboembolism.
NCT04333407 plans to compare aspirin, clopidogrel, rivaroxaban, atorvastatin, and omeprazole with no treatment in 3170 participants to assess mortality at 30 days.
NCT04366960 plans to compare 40 mg subcutaneous enoxaparin twice daily versus 40 mg subcutaneous enoxaparin once daily to assess venous thromboembolism in 2712 participants. NCT04512079 plans to compare apixaban versus prophylactic enoxaparin and full‐dose enoxaparin in 3600 participants to assess overt bleeding plus haemoglobin drop, cardiac tamponade, bleeding requiring surgical intervention for control, bleeding requiring vasoactive agents, or intracranial haemorrhage (time to event).
Wilkinson 2020 plans to compare several possible interventions (without details about type or dose of anticoagulants) in 1800 participants to assess time to clinical improvement.

See Table 3 for a summary of the characteristics of ongoing studies.

Table 3. Summary of characteristics of ongoing studies

Study	Country	Design	Experimental intervention	Comparator intervention	Primary outcomes	Estimated number of participants	Estimated primary completion date
ACTRN12620000517976	Australia	RCT	Nebulised heparin (UFH)	Standard care (without anticoagulants)	Time to separation from invasive ventilation	172	25 July 2021
Busani 2020	Italy	RCT	Enoxaparin	UFH	All‐cause mortality at day 28, defined as the comparison of proportions of patients' deaths for any cause at day 28 from randomisation	210	6 May 2021
Chambers 2020	USA	RCT	Intermediate‐dose enoxaparin	Standard prophylactic dose enoxaparin	Risk of all‐cause mortality (time frame: 30 days post‐intervention)	170	16 April 2021
ChiCTR2000030700	China	RCT	Enoxaparin	Standard care (without anticoagulants)	Time to virus eradication	60	30 September 2020
ChiCTR2000030701	China	RCT	Enoxaparin	Standard care (without anticoagulants)	Time to virus eradication	60	30 September 2020
ChiCTR2000030946	China	Prospective cohort	LMWH	Mechanical prevention	Biochemical indicators	120	24 April 2020
CTRI/2020/06/026220	India	RCT	Nafamostat (synthetic serine proteinase inhibitor)	Standard care (without anticoagulants)	Proportion of patients showing clinical improvement	40	27 January 2021
CTRI/2020/08/027033	India	RCT	Enoxaparin	Standard care (without anticoagulants)	Reduction in clinical symptoms and RT‐PCR test negative	100	27 January 2021
CTRI/2020/11/029175	India	RCT	Nebulised heparin	Standard care (without anticoagulants)	Time to separation from mechanical ventilation (duration of mechanical ventilation) up to day 28	58	27 January 2021
CTRI/2020/11/029345	India	RCT	Higher‐dose enoxaparin	Lower‐dose enoxaparin; apixaban	Time to first event rate within 30 days of randomisation of the composite of all‐cause mortality, intubation requiring mechanical ventilation, systemic thromboembolism (including PE) confirmed by imaging or requiring surgical intervention or ischaemic stroke confirmed by imaging	3600	27 January 2021
EUCTR2020‐001302‐30‐AT	Austria	RCT	Rivaroxaban	Standard care (without anticoagulants)	Time to sustained improvement of one category from admission	500	11 January 2021
EUCTR2020‐001708‐41‐IT	Italy	RCT	Higher‐dose enoxaparin	Lower‐dose enoxaparin	Incidence of VTE (a composite of incident asymptomatic and symptomatic proximal DVT diagnosed by serial compression ultrasonography, and symptomatic PE diagnosed by CT scan), in patients with SARS‐CoV‐2 infection	2000	30 October 2020
EUCTR2020‐001709‐21‐FR	France	RCT	Higher‐dose LMWH	Lower‐dose LMWH	VTE (causing death or not)	230	11 May 2020
EUCTR2020‐001891‐14‐ES	Spain	RCT	Enoxaparin	Standard care (without anticoagulants)	Need for oxygen therapy escalation due to oxygen saturation (Sat O₂) = 92% with FiO₂ = 0.5 and respiratory rate = 30 (IROX index = SatO₂/FiO₂)/FR < 5.5) or invasive mechanical ventilation or mortality during admission	140	16 November 2020
EUCTR2020‐002234‐32‐IT	Switzerland	RCT	Higher‐dose edoxaban	Lower‐dose edoxaban	Major vascular thrombotic events at 25 (+/‐3) days defined as a composite of: Asymptomatic proximal DVT Symptomatic proximal or distal DVT Symptomatic PE or thrombosis Myocardial infarction Ischaemic stroke Non‐CNS systemic embolism Death	420	11 January 2021
EUCTR2020‐002504‐39‐DE	Germany	RCT	Edoxaban	Fondaparinux	Composite of all‐cause mortality and/or VTE and/or arterial thromboembolism within 42 days	172	5 January 2021
EUCTR2020‐003349‐12‐IE	Ireland	RCT	Heparin	Standard care (without anticoagulants)	D‐dimer profile, with data collected on days 1, 3, 5 and 10	40	19 October 2020
Goldin 2020	USA	RCT	Higher‐dose LMWH	Lower‐dose LMWH	Composite outcome of arterial thromboembolic events, venous thromboembolic events and all‐cause mortality at day 30 ± 2 days (time frame: day 30 ± 2 days). Risk of arterial thromboembolic events (including myocardial infarction, stroke, systemic embolism), VTE (including symptomatic DVT of the upper or lower extremity, asymptomatic proximal DVT of the lower extremity, non‐fatal PE), and all‐cause mortality at day 30 ± 2 days	308	26 April 2021
IRCT20200515047456N1	Iran	RCT	UFH	Standard care (without anticoagulants)	Decrease D‐dimer level Improve compliance Improve of oxygenation Improve SOFA score	15	13 July 2020
ISRCTN14212905	UK	RCT	Nafamostat (synthetic serine proteinase inhibitor)	Standard care (without anticoagulants)	Safety of candidate agents as add‐on therapy to standard care in patients with COVID‐19 measured at 30, 60 and 90 days post‐treatment	100	3 August 2020
Kharma 2020	Qatar	RCT	Bivalirudin (DOAC)	LMWH or UFH	PaO₂/FiO₂ ratio (time frame: 3 days of intervention)	100	24 June 2020
Lasky 2021	USA	RCT	Dociparstat (heparinoid)	Placebo	Proportion of participants who are alive and free of invasive mechanical ventilation	525	17 February 2021
Lins 2020	Brazil	RCT	UFH	Standard care (without UFH)	The percentage of clotted dialysers within 72 h in each of the studied groups	90	27 July 2020
Marietta 2020	Italy	RCT	Higher‐dose LMWH	Lower‐dose LMWH	Clinical worsening (includes death and necessity for additional respiratory support)	300	June 2021
NCT04333407	UK	RCT	Rivaroxaban	Standard care (without anticoagulants)	All‐cause mortality at 30 days after admission	3170	30 March 2021
NCT04344756	France	RCT	Higher‐dose LMWH or UFH	Lower‐dose LMWH or UFH	Survival without ventilation	808	31 July 2020
NCT04345848	Switzerland	RCT	Higher‐dose LMWH or UFH	Lower‐dose LMWH or UFH	Composite outcome of arterial or venous thrombosis, disseminated intravascular coagulation and all‐cause mortality	200	30 November 2020
NCT04352400	Italy	RCT	Nafamostat (synthetic serine proteinase inhibitor)	Placebo	Time to clinical improvement	256	December 2021
NCT04366960	Italy	RCT	Higher‐dose enoxaparin	Lower‐dose enoxaparin	Incidence of VTE detected by imaging	2712	August 2020
NCT04367831	USA	RCT	Higher‐dose enoxaparin	Lower‐dose enoxaparin	Total number of patients with clinically relevant venous or arterial thrombotic events in ICU	100	November 2020
NCT04373707	France	RCT	Higher‐dose enoxaparin	Lower‐dose enoxaparin	VTE	602	September 2020
NCT04377997	USA	RCT	Higher‐dose LMWH or UFH	Lower‐dose LMWH or UFH	Risk of composite efficacy endpoint of death, cardiac arrest, symptomatic DVT, PE, arterial thromboembolism, myocardial infarction, or haemodynamic shock Risk of major bleeding event according to the ISTH definition	300	1 January 2021
NCT04397510	USA	RCT	Nebulised heparin	Placebo	Mean daily PaO₂/FiO₂	50	31 December 2020
NCT04406389	USA	RCT	Higher‐dose heparinoid or fondaparinux	Lower‐dose heparinoid or fondaparinux	30‐day mortality	186	December 2021
NCT04409834	USA	RCT	Higher‐dose heparinoid plus antiplatelet agent	Lower‐dose heparinoid without antiplatelet agent	Venous or arterial thrombotic events	750	May 2021
NCT04416048	Germany	RCT	Higher‐dose DOAC (rivaroxaban)	Lower‐dose heparinoid	Composite endpoint of VTE (DVT and/or fatal or non‐fatal PE), arterial thromboembolism, new myocardial infarction, non‐haemorrhagic stroke, all‐cause mortality or progression to intubation and invasive ventilation (time frame: 35 days post‐randomisation)	400	30 May 2021
NCT04420299	Spain	RCT	Higher‐dose heparin	Lower‐dose heparin	Combined worsening variable. Presence of any of the following will be considered worsening Death ICU admission Need for either non‐invasive or invasive mechanical ventilation Progression to moderate/severe respiratory distress syndrome according to objective criteria (Berlin definition) VTE (DVT or PE) or arterial (acute myocardial infarction or stroke) Proportion of patients that worsen	120	31 March 2021
NCT04444700	Brazil	RCT	Higher‐dose enoxaparin	Lower‐dose enoxaparin	Composite outcome of ICU admission (yes/no), non‐invasive positive pressure ventilation (yes/no), invasive mechanical ventilation (yes/no), or all‐cause death (yes/no) up to 28 days	462	31 December 2020
NCT04485429	Brazil	RCT	Higher‐dose heparin	Lower‐dose heparin	Rate of invasive mechanical ventilation	268	31 December 2020
NCT04508439	Mexico	RCT	Higher‐dose enoxaparin	Lower‐dose enoxaparin	Ventilatory support time Thrombotic complications Length of hospital stay Mortality rate	130	30 December 2020
NCT04511923	Ireland	RCT	Nebulised heparin	Standard care (without anticoagulants)	D‐dimer profile up to day 10 Frequency of severe adverse outcomes up to day 60	40	January 2022
NCT04512079	USA	RCT	Apixaban (DOAC)	Lower‐dose enoxaparin; higher‐dose enoxaparin	Time to first event (time frame: 30 days) Number of in‐hospital rate of BARC 3 or 5 (time frame: 30 days) Number of in‐hospital rate of BARC 3 or 5 bleeding (binary). BARC Type 3: a. Overt bleeding plus haemoglobin drop of 3 to < 5 g/dL (provided haemoglobin drop is related to bleed); transfusion with overt bleeding b. Overt bleeding plus haemoglobin drop < 5 g/dL (provided haemoglobin drop is related to bleed); cardiac tamponade; bleeding requiring surgical intervention for control; bleeding requiring IV vasoactive agents c. Intracranial haemorrhage confirmed by autopsy, imaging, or lumbar puncture; intraocular bleed compromising vision	3600	March 2022
NCT04530578	Argentina	RCT	Nebulised heparin	Enoxaparin	Percentage of patients requiring mechanical ventilation (time frame: 15 days)	200	1 June 2021
NCT04542408	Germany	RCT	Higher‐dose LMWH	Lower‐dose LMWH	Combined endpoint: all‐cause mortality and/or VTE and/or arterial thromboembolism (time frame: 42 days) All‐cause mortality and/or VTE and/or arterial thromboembolism during follow‐up (42 days). Thromboembolisms will be detected by duplex ultrasonography of arms and legs	172	30 September 2021
NCT04545541	USA	RCT	Nebulised heparin	Placebo	Alive and Ventilator‐Free Score (time frame: day 28)	300	June 2022
NCT04584580	Egypt	RCT	Higher‐dose LMWH	Lower‐dose LMWH	Mortality (time frame: until patient is discharged or up to 4 weeks whichever comes first) Occurrence of venous and/or arterial thrombosis (time frame: until patient is discharged or up to 4 weeks whichever comes first)	50	31 December 2020
NCT04600141	Brazil	RCT	Higher‐dose LMWH or UFH	Lower‐dose LMWH or UFH	Proportion of patients with clinical improvement (time frame: 30 days) Not hospitalised, with no limitations on activities Not hospitalised, but limited to activities Hospitalised, with no need for supplemental oxygen Hospitalised, needing supplemental oxygen Hospitalised, requiring high‐flow oxygen therapy, non‐invasive mechanical ventilation or both Hospitalised, requiring ECMO, invasive mechanical ventilation or both Death	308	31 December 2020
NCT04604327	Spain	RCT	Higher‐dose LMWH	Lower‐dose LMWH	Clinical deterioration (time frame: 10 days) Combined outcome that includes number of patients who suffer any of the following: death, ICU admission, mechanical ventilatory support, progression to moderate or severe ARDS (according to Berlin criteria) or arterial or venous thrombosis	164	31 July 2021
NCT04623177	Spain	Prospective cohort	Higher‐dose LMWH	Lower‐dose LMWH; no anticoagulation	ICU mortality rate (time frame: from admission to ICU discharge, an average of 1 month)	950	30 September 2020
NCT04640181	USA	RCT	Rivaroxaban at low, intermediate or therapeutic dose	Enoxaparin at low, intermediate or therapeutic dose	Death or 30‐day all‐cause mortality (time frame: 30 days) Mechanical ventilation, intubation (time frame: 30 days) Transfer to an ICU setting (time frame: 30 days)	150	31 July 2021
NCT04646655	Italy	RCT	Higher‐dose enoxaparin	Lower‐dose enoxaparin	Mortality rate (time frame: 30 days from enrolment ) Progression of respiratory failure (time frame: 30 days from enrolment) Progression of respiratory failure (time frame: 30 days from enrolment) Progression of respiratory failure (time frame: 30 days from enrolment) Number of major bleeding episodes (time frame: up to 6 months from randomisation)	300	31 July 2021
NCT04655586	USA	RCT	Higher‐dose heparin	Lower‐dose heparin	Change in D‐dimer level from baseline to day 8, or day of discharge if prior to day 8 Number of major or non‐major clinically relevant bleeding events within 8 days of randomisation Time to recovery within 30 days of randomisation	100	31 May 2021
NCT04723563	USA	RCT	Nebulised heparin	Placebo	Need for mechanical ventilation at day 28	50	29 May 2021
NCT04730856	Spain	RCT	Higher‐dose heparin	Lower‐dose heparin	Reduction of suspicion of systemic thrombotic symptomatic events (time frame: 30 days) Use of mechanical ventilation (time frame: 30 days) Progression on the WHO Progression Scale during follow‐up (time frame: 30 days) Overall survival at 30 days (time frame: 30 days) Length of hospital stay (days) (time frame: 30 days) Length of ICU stay (days) (time frame: 30 days)	600	31 July 2021
NCT04743011	Brazil	RCT	Nebulised heparin	Placebo	Change in aPTT > 1.5 (time frame: immediately or up to 8 days after starting treatment) Viral load in nasal swab RT‐PCR (time frame: immediately or up to 8 days after starting treatment)	50	31 December 2021
NCT04745442	Spain	RCT	Heparin	No anticoagulant	Combined variable: mortality or worsening rate with need for non‐invasive mechanical ventilation or with need for invasive mechanical ventilation (time frame: at day 31 after randomisation or hospital discharge (whichever occurs first)	48	15 January 2021
PACTR202007606032743	Egypt	RCT	Nebulised heparin	No anticoagulant	The average daily ratio of partial pressure of oxygen to FiO2 (PaO2/FiO2) while the patient is on room air for 7 days	100	22 February 2021
RBR‐7y8j2bs	Brazil	RCT	Nebulised heparin	Placebo	Efficacy: relative to the proposed treatment, through the analysis of the viral load of the SARS‐CoV‐2 virus in the participants treated by the sequential evaluation of the viral load in RT‐PCR of nasal swab. Safety: related to the use of inhalational high‐molecular‐weight heparin in patients with SARS‐CoV‐2 through the assessment of haemorrhagic events of any nature, alteration of the coagulogram that indicates an increase in aPTT > 1.5 and HIT	40	11 October 2021
Sholzberg 2021a	Canada	RCT	Higher‐dose heparinoids	Lower‐dose heparinoids	Composite outcome of ICU admission (yes/no), non‐invasive positive pressure ventilation (yes/no), invasive mechanical ventilation (yes/no), or all‐cause death (yes/no) up to 28 days	462	April 2022
Vanassche 2020	Belgium	RCT	LMWH	DOAC plus aprotinin	The overall objective of the study is to evaluate the clinical efficacy and safety of different investigational therapeutics relative to standard care in patients hospitalised with COVID‐19	210	18 August 2020
Van Haren 2020	Argentina	RCT	Nebulised heparin	No anticoagulant	Intubation rate (time frame: day 28) Proportion of patients requiring invasive mechanical ventilation	712	1 June 2021
Wilkinson 2020	UK	RCT	Anticoagulants (no details)	NA	Time to clinical improvement of at least 2 points (from randomisation) on a 9‐point category ordinal scale, live discharge from the hospital, or considered fit for discharge (a score of 0, 1, or 2 on the ordinal scale), whichever comes first, by day 29 (this will also define the 'responder' for the response rate analyses)	1800	04 September 2021
Total number of studies	Argentina: 2 Australia: 1 Austria: 1 Belgium: 1 Brazil: 6 Canada: 1 China: 3 Egypt: 2 France: 3 Germany: 3 India: 4 Iran: 1 Ireland: 2 Italy: 6 Mexico: 1 Qatar: 1 Spain: 6 Switzerland: 2 UK: 3 USA: 13	Prospective cohort: 2 RCT: 60			35 studies considered mortality 26 studies considered additional respiratory support	35,470 participants (120 from NRS; 35,350 from RCTs)	58 studies to December 2021 Four studies to July 2022
aPTT: activated partial thromboplastin time; ARDS: acute respiratory distress syndrome; BARC: Bleeding Academic Research Consortium; CNS: central nervous system; DOACs: direct oral anticoagulants; DVT: deep vein thrombosis; ECMO: extracorporeal membrane oxygenation; FiO₂: fraction of inspired oxygen; HIT: heparin‐induced thrombocytopenia; ICU: intensive care unit; ISTH: International Society on Thrombosis and Haemostasis; LMWH: low‐molecular‐weight heparin; NA: not available; NRS: non‐randomised studies; PaO₂: arterial oxygen pressure;PE: pulmonary embolism; RCT: randomised controlled trial; RT‐PCR: reverse transcription polymerase chain reaction; SOFA: sequential organ failure assessment; UFH: unfractionated heparin; VKA: vitamin K antagonist; VTE: venous thromboembolism; WHO: World Health Organization

Risk of bias in included studies

Risk of bias in randomised controlled trials

Overall judgement

We assessed the risk of bias at the study level using RoB 1 for RCTs (Higgins 2017). The specific judgements ('high risk', 'low risk' or 'unclear risk') by available studies are presented in Figure 2 and Figure 3, and the support for judgement is explained in the related risk of bias tables (Characteristics of included studies). Lopes 2021 and Sadeghipour 2021 had a low overall risk of bias. We judged the other two RCTs at a high overall risk of bias because of 'blinding of outcomes assessment' domain issues (Lemos 2020; Zarychanski 2021), and 'selective reporting' domain issues (Zarychanski 2021).

Figure 2

RoB 1.0 assessments for randomised controlled trials

Figure 3

RoB 1.0 graph: assessments for randomised controlled trials presented as percentages across studies.

Allocation (selection bias)

All four studies had a low risk of bias for random sequence generation and for allocation concealment (Lemos 2020; Lopes 2021; Sadeghipour 2021; Zarychanski 2021).

Blinding (performance bias and detection bias)

Although anticoagulation is a pharmacological intervention that allows the blinding of participants and personnel, all included studies had a high risk of bias (Lemos 2020; Lopes 2021; Sadeghipour 2021; Zarychanski 2021).

We assessed two studies to be at low risk of bias for blinding of outcome assessment (Lopes 2021; Sadeghipour 2021), and two at high risk of bias for this domain (Lemos 2020; Zarychanski 2021).

Incomplete outcome data (attrition bias)

Two studies had a high risk of bias for incomplete outcome reporting (Lemos 2020; Zarychanski 2021). Conversely, Lopes 2021 and Sadeghipour 2021 had a low risk of bias.

Selective reporting (reporting bias)

Zarychanski 2021 was at high risk of bias for selective reporting and none was at unclear risk of bias for this domain. All other included studies (3/4) had a low risk of bias for this domain (Lemos 2020; Lopes 2021; Sadeghipour 2021).

Other potential sources of bias

Zarychanski 2021 was at high risk of bias for other potential sources of bias and all other studies were at low risk of bias for this domain (Lemos 2020; Lopes 2021; Sadeghipour 2021).

Although the study authors declare that they harmonised their protocols into a "prospectively multiplatform uniformisation", Zarychanski 2021 combined the results from three different trials registries, with different 'centres' of randomisation and documentation. There is a possibility of additional heterogeneity in overall results when combining these three trials as a unique trial:

There was an imbalance of losses to follow‐up (moderate‐severity: experimental = 19 losses (1.5%), comparator = 7 losses (0.6%)).
There was a factorial randomisation for antiplatelet agent intervention in one of the considered trials (REMAP‐CAP).
There was a change in the primary outcome specified in the registered protocols compared to the unique reported primary outcome.

We contacted the study authors requesting the data separately, without success. Therefore, we considered Zarychanski 2021 data as a unique study.

Risk of bias in non‐randomised controlled trials

Overall judgement

We assessed the risk of bias at the results level using ROBINS‐I tool for all NRS (Sterne 2016). The specific judgements ('critical risk', 'serious risk', 'moderate risk', 'low risk', or 'no information') by available outcomes are presented in Figure 4, Figure 5, Figure 6, Figure 7, Figure 8, Figure 9 and Figure 10. The support for ROBINS‐I judgement is explained in the related risk of bias tables (Table 4; Table 5; Table 6; Table 7; Table 8; Table 9; Table 10). We will provide detailed risk of bias assessment data on request. The overall risk of bias for all‐cause mortality, deep vein thrombosis, pulmonary embolism, major bleeding, adverse events (stroke and myocardial infarction) in the comparison 'anticoagulants (all types) versus no treatment' was critical. The overall risk of bias for hospitalisation was serious for the same comparison.

Figure 4

ROBINS‐I assessments: anticoagulants (all types) versus no treatment for people hospitalised with COVID‐19 (all‐cause mortality)

Figure 5

ROBINS‐I assessments: anticoagulants (all types) versus no treatment for people hospitalised with COVID‐19 (deep vein thrombosis)

Figure 6

ROBINS‐I assessments: anticoagulants (all types) versus no treatment for people hospitalised with COVID‐19 (pulmonary embolism)

Figure 7

ROBINS‐I assessments: anticoagulants (all types) versus no treatment for people hospitalised with COVID‐19 (major bleeding)

Figure 8

ROBINS‐I assessments: anticoagulants (all types) versus no treatment for people hospitalised with COVID‐19 (adverse events: stroke)

Figure 9

ROBINS‐I assessments: anticoagulants (all types) versus no treatment for people hospitalised with COVID‐19 (adverse events: myocardial infarction)

Figure 10

ROBINS‐I assessments: anticoagulants (all types) versus no treatment for people hospitalised with COVID‐19 (hospitalisation)

Table 4. ROBINS‐I assessments: anticoagulants (all types) versus no treatment for people hospitalised with COVID‐19 (all‐cause mortality)

Study	Bias due to confounding	Bias in selection of participants into the study	Bias in classification of interventions	Bias due to deviations from the intended intervention	Bias due to missing data	Bias in measurement of outcomes	Bias in selection of the reported result	Overall risk of bias
Albani 2020	Serious risk	Serious risk	Low risk	Low risk	Low risk	Low risk	Low risk	Serious risk
Judgement	One or more prognostic variables are likely to be unbalanced between the compared groups. To minimise the impact of the absence of randomisation, an adjusted analysis with propensity scores was performed considering age, sex, disease severity, admission to ICU and COVID‐19 treatment. However, the essential confounding factors: 'participants already using anticoagulants', 'participants who underwent surgery during the hospitalisation', 'active cancer treatment', 'concomitant antiplatelet use' and 'history of venous thromboembolism' were not considered.	Participants included in both groups were selected from a single hospital, and the first dose of anticoagulant was administered between 0 and 3 days after hospital admission. The start of follow‐up and start of intervention possibly did not coincide for most participants, and adjustment techniques to correct the presence of selection bias were not used. It is not clear how prevalent use of anticoagulation was handled.	The intervention groups were clearly defined and recorded at the start of the intervention. Intervention status was probably not affected by knowledge of the outcome or the risk of the outcome.	No deviations from the intended intervention were reported in the study, and if any deviation occurred from usual practice, it was unlikely to impact on the outcome.	There were missing outcome data for 27 participants (1.9% of the total) and balanced between the groups. These missing data possibly could not cause an important impact on the estimate.	It is unlikely that the outcome assessment (objective outcome) was influenced by the knowledge of the intervention received by the study participants.	The study protocol was not identified but all reported results corresponded to the intended outcome.	The study has some important problems.
Rentsch 2020	Moderate risk	Low risk	Low risk	Low risk	Low risk	Low risk	Low risk	Moderate risk
Judgement	One or more prognostic variables are likely to be unbalanced between the compared groups. Essential characteristics, such as participants who underwent surgery during the hospitalisation, and history of venous thromboembolism, were not considered. However, an appropriate analysis method to control for measured confounders was used (inverse probability of treatment weighting), and all the important confounding domains for this study were probably controlled.	Participants included in both groups were selected from a nationwide cohort of patients receiving care in the Department of Veterans Affairs in the USA, and selection may have not been related to intervention and outcome. The start of follow‐up and start of intervention coincided for most participants (the first 24 h of hospitalisation).	The intervention groups were clearly defined and recorded at the start of the intervention. Intervention status was probably not affected by knowledge of the outcome or the risk of the outcome.	No deviations from the intended intervention were reported in the study, and if any deviation occurred from usual practice, it was unlikely to impact on the outcome.	No missing data were reported for the outcome.	It is unlikely that the outcome assessment (objective outcome) was influenced by the knowledge of the intervention received by the study participants.	The study protocol was not identified but all reported results corresponded to the intended outcome.	The study is sound for a non‐randomised study with regard to this domain but cannot be considered comparable to a well‐ performed randomised trial.
Santoro 2020	Critical risk	No information	Serious risk	Low risk	Low risk	Low risk	Serious risk	Critical risk
Judgement	One or more prognostic variables are likely to be unbalanced between the compared groups. Essential characteristics, such as participants who underwent surgery during the hospitalisation, and antiplatelet use were not considered. The Cox's multivariable regression analysis was performed to define independent risk factors for the mortality outcome, but only for participants with respiratory failure.	Insufficient information to judge. There was insufficient information if the start of follow‐up and the start of intervention coincided for most participants.	The intervention groups were not clearly defined and recorded at the start of the intervention. Information about frequency and dose was not provided.	No deviations from the intended intervention were reported in the study, and if any deviation occurred from usual practice, it was unlikely to impact on the outcome.	No missing data were reported for the outcome.	It is unlikely that the outcome assessment (objective outcome) was influenced by the knowledge of the intervention received by the study participants.	The study protocol was available, but it is not possible to exclude bias in selection of reported effect estimate, based on the results, from multiple outcome measurements within the outcome domain (mortality), and multiple effect estimates for different subgroups were provided, omitting varying proportions of the original cohort.	The study is too problematic to provide useful evidence.
COVID‐19: coronavirus disease 2019; ICU: intensive care unit

Table 5. ROBINS‐I assessments: anticoagulants (all types) versus no treatment for people hospitalised with COVID‐19 (deep vein thrombosis)

Study	Bias due to confounding	Bias in selection of participants into the study	Bias in classification of interventions	Bias due to deviations from the intended intervention	Bias due to missing data	Bias in measurement of outcomes	Bias in selection of the reported result	Overall risk of bias
Albani 2020	Critical risk	Serious risk	Low risk	Low risk	Low risk	Low risk	Low risk	Critical risk
Judgement	One or more prognostic variables are likely to be unbalanced between the compared groups. Essential characteristics, such as participants who underwent surgery during the hospitalisation, concomitant antiplatelet use, and history of venous thromboembolism, were not considered. The outcome was reported without any adjustment.	Participants included in both groups were selected from a single hospital, and the first dose of anticoagulant was administered between 0 and 3 days after hospital admission. The start of follow‐up and start of intervention possibly did not coincide for most participants, and adjustment techniques to correct the presence of selection bias were not used.	The intervention groups were clearly defined and recorded at the start of the intervention. Intervention status was probably not affected by knowledge of the outcome or the risk of the outcome.	No deviations from the intended intervention were reported in the study, and if any deviation occurred from usual practice, it was unlikely to impact on the outcome.	There were missing outcome data for 27 participants (1.9% of the total), balanced between the groups. These missing data would probably not have an important impact on the estimate.	It is unlikely that the outcome assessment (objective outcome) was influenced by the knowledge of the intervention received by the study participants.	The study protocol was not identified but all reported results corresponded to the intended outcome.	The study is too problematic to provide useful evidence.

Table 6. ROBINS‐I assessments: anticoagulants (all types) versus no treatment for people hospitalised with COVID‐19 (pulmonary embolism)

Study	Bias due to confounding	Bias in selection of participants into the study	Bias in classification of interventions	Bias due to deviations from the intended intervention	Bias due to missing data	Bias in measurement of outcomes	Bias in selection of the reported result	Overall risk of bias
Albani 2020	Critical risk	Serious risk	Low risk	Low risk	Low risk	Low risk	Low risk	Critical risk
Judgement	One or more prognostic variables are likely to be unbalanced between the compared groups. Essential characteristics, such as participants who underwent surgery during the hospitalisation, concomitant antiplatelet use, and history of venous thromboembolism, were not considered. The outcome was reported without any adjustment.	Participants included in both groups were selected from a single hospital, and the first dose of anticoagulant was administered between 0 and 3 days after hospital admission. The start of follow‐up and start of intervention possibly did not coincide for most participants, and adjustment techniques to correct the presence of selection bias were not used.	The intervention groups were clearly defined and recorded at the start of the intervention. Intervention status was probably not affected by knowledge of the outcome or the risk of the outcome.	No deviations from the intended intervention were reported in the study, and if any deviation occurred from usual practice, it was unlikely to impact on the outcome.	There were missing outcome data for 27 participants (1.9% of the total), balanced between the groups. These missing data would probably not have an important impact on the estimate.	It is unlikely that the outcome assessment (objective outcome) was influenced by the knowledge of the intervention received by the study participants.	The study protocol was not identified but all reported results corresponded to the intended outcome.	The study is too problematic to provide useful evidence.

Table 7. ROBINS‐I assessments: anticoagulants (all types) versus no treatment for people hospitalised with COVID‐19 (major bleeding)

Study	Bias due to confounding	Bias in selection of participants into the study	Bias in classification of interventions	Bias due to deviations from the intended intervention	Bias due to missing data	Bias in measurement of outcomes	Bias in selection of the reported result	Overall risk of bias
Albani 2020	Critical risk	Serious risk	Low risk	Low risk	Low risk	Low risk	Low risk	Critical risk
Judgement	One or more prognostic variables are likely to be unbalanced between the compared groups. Essential characteristics, such as participants who underwent surgery during the hospitalisation, concomitant antiplatelet use, and history of venous thromboembolism, were not considered. The outcome was reported without any adjustment.	Participants included in both groups were selected from a single hospital, and the first dose of anticoagulant was administered between 0 and 3 days after hospital admission. The start of follow‐up and start of intervention possibly did not coincide for most participants, and adjustment techniques to correct the presence of selection bias were not used.	The intervention groups were clearly defined and recorded at the start of the intervention. Intervention status was probably not affected by knowledge of the outcome or the risk of the outcome.	No deviations from the intended intervention were reported in the study, and if any deviation occurred from usual practice, it was unlikely to impact on the outcome.	There were missing outcome data for 27 participants (1.9% of the total) and balanced between the groups. These missing data would probably not have an important impact on the estimate.	It is unlikely that the outcome assessment (objective outcome) was influenced by the knowledge of the intervention received by the study participants.	The study protocol was not identified but all reported results corresponded to the intended outcome.	The study is too problematic to provide useful evidence.
Santoro 2020	Critical risk	No information	Serious risk	Low risk	Low risk	Low risk	Low risk	Critical risk
Judgement	One or more prognostic variables are likely to be unbalanced between the compared groups. Essential characteristics, such as participants who underwent surgery during the hospitalisation, and antiplatelet use were not considered. The Cox's multivariable regression analysis was performed to define independent risk factors only for the mortality outcome.	Insufficient information to judge. There was insufficient information if the start of follow‐up and the start of intervention coincided for most participants.	The intervention groups were not clearly defined and recorded at the start of the intervention. Information about frequency and dose was not provided.	No deviations from the intended intervention were reported in the study, and if any deviation occurred from usual practice, it was unlikely to impact on the outcome.	No missing data were reported for the outcome.	It is unlikely that the outcome assessment (objective outcome) was influenced by the knowledge of the intervention received by the study participants.	The study protocol was available, and the reported results corresponded to the intended outcome.	The study is too problematic to provide useful evidence.

Table 8. ROBINS‐I assessments: anticoagulants (all types) versus no treatment for people hospitalised with COVID‐19 (adverse events (stroke))

Study	Bias due to confounding	Bias in selection of participants into the study	Bias in classification of interventions	Bias due to deviations from the intended intervention	Bias due to missing data	Bias in measurement of outcomes	Bias in selection of the reported result	Overall risk of bias
Albani 2020	Critical risk	Serious risk	Low risk	Low risk	Low risk	Low risk	Low risk	Critical risk
Judgement	One or more prognostic variables are likely to be unbalanced between the compared groups. Essential characteristics, such as participants who underwent surgery during the hospitalisation, concomitant antiplatelet use, and history of venous thromboembolism, were not considered. The outcome was reported without any adjustment.	Participants included in both groups were selected from a single hospital, and the first dose of anticoagulant was administered between 0 and 3 days after hospital admission. The start of follow‐up and start of intervention possibly did not coincide for most participants, and adjustment techniques to correct the presence of selection bias were not used.	The intervention groups were clearly defined and recorded at the start of the intervention. Intervention status was probably not affected by knowledge of the outcome or the risk of the outcome.	No deviations from the intended intervention were reported in the study, and if any deviation occurred from usual practice, it was unlikely to impact on the outcome.	There were missing outcome data for 27 participants (1.9% of the total), balanced between the groups. These missing data would probably not have an important impact on the estimate.	It is unlikely that the outcome assessment (objective outcome) was influenced by the knowledge of the intervention received by the study participants.	The study protocol was not identified but all reported results corresponded to the intended outcome.	The study is too problematic to provide useful evidence.

Table 9. ROBINS‐I assessments: anticoagulants (all types) versus no treatment for people hospitalised with COVID‐19 (adverse events (myocardial infarction))

Study	Bias due to confounding	Bias in selection of participants into the study	Bias in classification of interventions	Bias due to deviations from the intended intervention	Bias due to missing data	Bias in measurement of outcomes	Bias in selection of the reported result	Overall risk of bias
Albani 2020	Critical risk	Serious risk	Low risk	Low risk	Low risk	Low risk	Low risk	Critical risk
Judgement	One or more prognostic variables are likely to be unbalanced between the compared groups. Essential characteristics, such as participants who underwent surgery during the hospitalisation, concomitant antiplatelet use, and history of venous thromboembolism, were not considered. The outcome was reported without any adjustment.	Participants included in both groups were selected from a single hospital, and the first dose of anticoagulant was administered between 0 and 3 days after hospital admission. The start of follow‐up and start of intervention possibly did not coincide for most participants, and adjustment techniques to correct the presence of selection bias were not used.	The intervention groups were clearly defined and recorded at the start of the intervention. Intervention status was probably not affected by knowledge of the outcome or the risk of the outcome.	No deviations from the intended intervention were reported in the study, and if any deviation occurred from usual practice, it was unlikely to impact on the outcome.	There were missing outcome data for 27 participants (1.9% of the total), balanced between the groups. These missing data would probably not have an important impact on the estimate.	It is unlikely that the outcome assessment (objective outcome) was influenced by the knowledge of the intervention received by the study participants.	The study protocol was not identified but all reported results corresponded to the intended outcome.	The study is too problematic to provide useful evidence.

Table 10. ROBINS‐I assessments: anticoagulants (all types) versus no treatment for people hospitalised with COVID‐19 (hospitalisation)

Study	Bias due to confounding	Bias in selection of participants into the study	Bias in classification of interventions	Bias due to deviations from the intended intervention	Bias due to missing data	Bias in measurement of outcomes	Bias in selection of the reported result	Overall risk of bias
Albani 2020	Serious risk	Serious risk	Low risk	Low risk	Low risk	Low risk	Low risk	Serious risk
Judgement	One or more prognostic variables are likely to be unbalanced between the compared groups. Essential characteristics, such as participants who underwent surgery during the hospitalisation, concomitant antiplatelet use, and history of venous thromboembolism, were not considered.	Participants included in both groups were selected from a single hospital, and the first dose of anticoagulant was administered between 0 and 3 days after hospital admission. The start of follow‐up and start of intervention possibly did not coincide for most participants, and adjustment techniques to correct the presence of selection bias were not used.	The intervention groups were clearly defined and recorded at the start of the intervention. Intervention status was probably not affected by knowledge of the outcome or the risk of the outcome.	No deviations from the intended intervention were reported in the study, and if any deviation occurred from usual practice, it was unlikely to impact on the outcome.	There were missing outcome data for 27 participants (1.9% of the total) and balanced between the groups. These missing data would probably not have an important impact on the estimate.	It is unlikely that the outcome assessment (objective outcome) was influenced by the knowledge of the intervention received by the study participants.	The study protocol was not identified but all reported results corresponded to the intended outcome.	The study has some important problems.

All‐cause mortality

'Three studies reported mortality for the comparison 'anticoagulants (all types) versus no treatment'. We rated Albani 2020 as serious risk due to confounding and selection of participants. We rated Albani 2020 as low risk for all other domains. We rated Rentsch 2020 as moderate risk for confounding and low risk for all other domains. We rated Santoro 2020 as a critical risk due to confounding, serious risk due to problems with the 'classification of interventions' and 'selection of reported results' items. There was no information about bias due to the selection of participants in Santoro 2020, and all other domains were at low risk. See Figure 4 and Table 4.

Deep vein thrombosis

Albani 2020 reported deep vein thrombosis for the comparison 'anticoagulants (all types) versus no treatment'. We rated Albani 2020 as a critical risk for confounding, serious risk for the selection of participants and low risk for all other domains. See Figure 5 and Table 5.

Pulmonary embolism

Albani 2020 reported pulmonary embolism for the comparison 'anticoagulants (all types) versus no treatment'. We rated Albani 2020 as a critical risk for confounding, serious risk for the selection of participants and low risk for all other domains. See Figure 6 and Table 6.

Major bleeding

Albani 2020 and Santoro 2020 reported major bleeding for the comparison 'anticoagulants (all types) versus no treatment'. We rated both studies as a critical risk of confounding, Albani 2020 as serious risk for the selection of participants and Santoro 2020 as a serious risk for the classification of interventions. Albani 2020 did not report information on the selection of participants and we rated all other domains as low risk for both studies. See Figure 7 and Table 7.

Adverse events (stroke)

Albani 2020 reported stroke for the comparison 'anticoagulants (all types) versus no treatment'. We rated Albani 2020 as a critical risk to confounding, serious risk to the selection of participants and low risk to all other domains. See Figure 8 and Table 8.

Adverse events (myocardial infarction)

Albani 2020 reported myocardial infarction for the comparison 'anticoagulants (all types) versus no treatment'. We rated Albani 2020 as a critical risk for confounding, serious risk for the selection of participants and low risk for all other domains. See Figure 9 and Table 9.

Hospitalisation

Albani 2020 reported hospitalisation for the comparison 'anticoagulants (all types) versus no treatment'. We rated Albani 2020 as serious risk for confounding and for the selection of participants, and as low risk for all other domains. See Figure 10 and Table 10.

Effects of interventions

We included three NRS of interventions for the comparison 'anticoagulants (all types) versus no treatment' (short‐term time point) and four RCTs for the comparison 'higher‐dose anticoagulants versus lower‐dose anticoagulants' (short‐term and long‐term time points) and performed quantitative data analysis (meta‐analysis) when appropriate. We did not perform any funnel plot analysis because there is no comparison with 10 or more studies in this review.

1. Higher‐dose anticoagulants versus lower‐dose anticoagulants (short term)

Four RCTs (Lemos 2020; Lopes 2021; Sadeghipour 2021; Zarychanski 2021), compared heparins (unfractionated heparin) or low‐molecular‐weight heparin) or direct oral anticoagulants (rivaroxaban) in higher doses (2376 participants) versus heparins (unfractionated heparin or low‐molecular‐weight heparin) or direct oral anticoagulants (rivaroxaban) in lower doses (2271 participants). Only Zarychanski 2021 reported outcomes data separately by moderate‐severity and critically ill disease, and we, therefore, included this study in both subgroups. More than 82% of participants in Lopes 2021 had moderate‐severity disease; therefore, we included them in the moderate‐severity subgroup. Lemos 2020 and Sadeghipour 2021 included only participants under invasive ventilatory support or in ICU; therefore, we analysed their data in the critically ill subgroup. See summary of findings Table 1.

Primary outcomes

All‐cause mortality

All studies reported all‐cause mortality with a follow‐up of up to 30 days. Higher‐dose anticoagulants result in little to no difference in all‐cause mortality compared to lower‐dose anticoagulants for up to 30 days (RR 1.03, 95% CI 0.92 to 1.16; I² = 5%; 4 studies, 4489 participants; high‐certainty evidence; Analysis 1.1). The test for subgroup differences suggested that the severity of the condition has no modifying effect on the all‐cause mortality (Chi² = 0.07, df = 1 (P = 0.80), I² = 0%; Analysis 1.1).

The sensitivity analysis including only trials at low risk of bias (RR 1.16, 95% CI 0.86 to 1.57; Analysis 1.2) did not substantially change the effect estimate.

Necessity for additional respiratory support

Three studies reported the necessity for additional respiratory support with a follow‐up for up to 30 days (Lopes 2021; Sadeghipour 2021; Zarychanski 2021). Lemos 2020 did not report the necessity for additional respiratory support and Zarychanski 2021 reported this outcome only for moderate‐severity participants. Lopes 2021 and Sadeghipour 2021 reported these outcomes for all participants. The evidence is very uncertain about the effect of higher‐dose anticoagulants on necessity for additional respiratory support compared to lower‐dose anticoagulants up to 30 days (RR 0.54, 95% CI 0.12 to 2.47; I² = 60%; 3 studies, 3407 participants; very low‐certainty evidence; Analysis 1.3). The test for subgroup differences was not applicable because the effect in Sadeghipour 2021 was not estimable (no events).

The sensitivity analysis including only trials at low risk of bias (RR 0.16, 95% CI 0.02 to 1.35; Analysis 1.4) substantially changed the effect estimate.

Secondary outcomes

Mortality related to COVID‐19

There were no available data for this outcome.

Deep vein thrombosis

Although Zarychanski 2021 reported this outcome only for moderate severity participants, all studies reported deep vein thrombosis with a follow‐up of up to 30 days. Higher‐dose anticoagulants may result in little to no difference in deep vein thrombosis compared to lower‐dose anticoagulants up to 30 days (RR 1.08, 95% CI 0.57 to 2.03; I² = 0%; 4 studies, 3422 participants; low‐certainty evidence; Analysis 1.5). The test for subgroup differences suggested that the severity of the condition has no modifying effect on deep vein thrombosis (Chi² = 0.82, df = 1 (P = 0.36), I² = 0%; Analysis 1.5).

The sensitivity analysis including only trials at low risk of bias (RR 1.21, 95% CI 0.53 to 2.79; Analysis 1.6) did not change the effect estimate substantially.

Pulmonary embolism

All studies reported pulmonary embolism with a follow‐up of up to 30 days. Higher‐dose anticoagulants may reduce pulmonary embolism compared to lower‐dose anticoagulants for up to 30 days (RR 0.46, 95% CI 0.31 to 0.70; I² = 0%; 4 studies, 4360 participants; moderate‐certainty evidence; Analysis 1.7). The test for subgroup differences suggested that the severity of the condition has no modifying effect on pulmonary embolism (Chi² = 0.08, df = 1 (P = 0.78), I² = 0%; Analysis 1.7).

The sensitivity analysis including only trials at low risk of bias (RR 0.50, 95% CI 0.23 to 1.10; Analysis 1.8) changed the effect estimate substantially.

Major bleeding

All studies reported major bleeding with a follow‐up of up to 30 days. Higher‐dose anticoagulants likely increase major bleeding slightly compared to lower‐dose anticoagulants up to 30 days (RR 1.78, 95% CI 1.13 to 2.80; I² = 0%; 4 studies, 4400 participants; moderate‐certainty evidence; Analysis 1.9). The test for subgroup differences suggested that the severity of the condition has no modifying effect on major bleeding (Chi² = 1.03, df = 1 (P = 0.31), I² = 2.8%; Analysis 1.9).

The sensitivity analysis including only trials at low risk of bias (RR 2.13, 95% CI 0.92 to 4.90; Analysis 1.10) substantially changed the effect estimate.

Adverse events (minor bleeding, gastrointestinal adverse effects (e.g. nausea, vomiting, diarrhoea, abdominal pain), allergic reactions, renal failure and amputations)

Minor bleeding

Three studies reported minor bleeding with a follow‐up of up to 30 days (Lemos 2020; Lopes 2021; Sadeghipour 2021). Higher‐dose anticoagulants increase adverse events (minor bleeding) compared to lower‐dose anticoagulants up to 30 days (RR 3.28, 95% CI 1.75 to 6.14; I² = 0%; 3 studies, 1196 participants; high‐certainty evidence; Analysis 1.11). The test for subgroup differences suggested that the severity of the condition has no modifying effect on minor bleeding (Chi² = 1.50, df = 1 (P = 0.22), I² = 33.5%; Analysis 1.11).

The sensitivity analysis including only trials at low risk of bias (RR 3.67, 95% CI 1.82 to 7.40; Analysis 1.12) did not change the effect estimate substantially.

Stroke

Three studies reported stroke with a follow‐up of up to 30 days (Lopes 2021; Sadeghipour 2021; Zarychanski 2021). Higher‐dose anticoagulants may result in little to no difference in adverse events (stroke) compared to lower‐dose anticoagulants for up to 30 days (RR 0.91, 95% CI 0.40 to 2.03; I² = 0%; 3 studies, 4349 participants; low‐certainty evidence; Analysis 1.13). We downgraded two levels due to imprecision (CI of the absolute difference comprises both important clinical benefit and important clinical harm). The test for subgroup differences suggested that the severity of the condition has no modifying effect on stroke (Chi² = 0.00, df = 1 (P = 0.97), I² = 0%; Analysis 1.13).

The sensitivity analysis including only trials at low risk of bias (RR 1.62, 95% CI 0.20 to 13.13; Analysis 1.14) did not substantially change the effect estimate.

Major adverse limb events

Two studies reported major adverse limb events with a follow‐up of up to 30 days (Lopes 2021; Sadeghipour 2021). Higher‐dose anticoagulants may result in little to no difference in major adverse limb events compared to lower‐dose anticoagulants for up to 30 days (RR 0.33, 95% CI 0.01 to 7.99; I² not applicable; 2 studies, 1176 participants; low‐certainty evidence; Analysis 1.15). We downgraded two levels due to imprecision (CI of the absolute difference comprises both important clinical benefit and important clinical harm). The test for subgroup differences was not applicable because Sadeghipour 2021 reported no events in both groups.

We judged both studies as low risk of bias and, therefore, no sensitivity analysis was applicable.

Myocardial infarction

Three studies reported myocardial infarction with a follow‐up of up to 30 days (Lopes 2021; Sadeghipour 2021; Zarychanski 2021). Higher‐dose anticoagulants may result in little to no difference in myocardial infarction compared to lower‐dose anticoagulants for up to 30 days (RR 0.86, 95% CI 0.48 to 1.55; I² = 0%; 3 studies, 4349 participants; low‐certainty evidence; Analysis 1.16). We downgraded two levels due to imprecision (CI of the absolute difference comprises both important clinical benefit and important clinical harm). The test for subgroup differences suggested that the severity of the condition has no modifying effect on stroke (Chi² = 0.08, df = 1 (P = 0.78), I² = 0%; Analysis 1.16).

The sensitivity analysis including only trials at low risk of bias (RR 0.91, 95% CI 0.44 to 1.91; Analysis 1.17) did not substantially change the effect estimate.

Atrial fibrillation

Sadeghipour 2021 reported atrial fibrillation with a follow‐up of up to 30 days. Higher‐dose anticoagulants may result in little to no difference in atrial fibrillation compared to lower‐dose anticoagulants for up to 30 days (RR 0.35, 95% CI 0.07 to 1.70, I² not applicable; 1 study, 562 participants; low‐certainty evidence; Analysis 1.18). We downgraded two levels due to imprecision (CI of the absolute difference comprises both important clinical benefit and important clinical harm).

Thrombocytopenia

Two studies reported thrombocytopenia with a follow‐up of up to 30 days (Sadeghipour 2021; Zarychanski 2021). Higher‐dose anticoagulants may result in little to no difference in thrombocytopenia compared to lower‐dose anticoagulants for up to 30 days (RR 0.94, 95% CI 0.71 to 1.24; I² not applicable; 2 studies, 2789 participants; low‐certainty evidence; Analysis 1.19). We downgraded two levels due to imprecision (CI of the absolute difference comprises both important clinical benefit and important clinical harm). The test for subgroup differences and the sensitivity analysis were not applicable because Zarychanski 2021 reported no events in both groups.

Hospitalisation time in days

Two studies reported the hospitalisation time in days with a follow‐up of up to 30 days for moderate‐severity (Lopes 2021), and critically ill (Lemos 2020), participants. Higher‐dose anticoagulants probably result in little to no difference in hospitalisation time compared to lower‐dose anticoagulants up to 30 days (MD 0.28 days, 95% CI −0.87 to 1.44; I² = 0%; 2 studies, 634 participants; moderate‐certainty evidence; Analysis 1.20). The test for subgroup differences (Chi² = 0.06, df = 1 (P = 0.81), I² = 0%; Analysis 1.20) did not change the effect estimate.

The sensitivity analysis including only trials at low risk of bias (MD 0.30, −0.86 to 1.46; Analysis 1.21) did not change the effect estimate substantially.

Quality of life

There were no available data for this outcome.

2. Higher‐dose anticoagulants versus lower‐dose anticoagulants (long term)

Sadeghipour 2021 compared enoxaparin (low‐molecular‐weight heparin) in higher doses (299 participants) versus enoxaparin in lower doses (299 participants) for participants in ICU and reported data at the follow‐up of up to 90 days (long term).

Primary outcomes

All‐cause mortality

Sadeghipour 2021 reported all‐cause mortality with a follow‐up of up to 90 days. Higher‐dose anticoagulants may result in little to no difference in all‐cause mortality compared to lower‐dose anticoagulants up to 90 days (RR 1.07, 95% CI 0.89 to 1.28; I² not applicable; 1 study, 590 participants; moderate‐certainty evidence; Analysis 2.1). We downgraded the evidence one level due to imprecision (fewer than 300 events were included in the analysis).

Necessity for additional respiratory support

Sadeghipour 2021 reported the necessity for additional respiratory support with a follow‐up of up to 90 days. The evidence is not estimable about the effect of higher‐dose anticoagulants on necessity for additional respiratory support compared to lower‐dose anticoagulants up to 90 days (no events in both groups; I² not applicable; 1 study, 590 participants; low‐certainty evidence; Analysis 2.2). We downgraded the evidence two levels due to imprecision (no events).

Secondary outcomes

Mortality related to COVID‐19

There were no available data for this outcome.

Deep vein thrombosis

Sadeghipour 2021 reported deep vein thrombosis with a follow‐up of up to 90 days. Higher‐dose anticoagulants may result in little to no difference in deep vein thrombosis compared to lower‐dose anticoagulants up to 90 days (RR 1.39, 95% CI 0.45 to 4.33; I² not applicable; 1 study, 590 participants; low‐certainty evidence; Analysis 2.3). We downgraded the evidence two levels due to imprecision (CI of the absolute difference comprises both important clinical benefit and harm, and fewer than 300 events were included in the analysis).

Pulmonary embolism

Sadeghipour 2021 reported pulmonary embolism with a follow‐up of up to 90 days. Higher‐dose anticoagulants may reduce pulmonary embolism compared to lower‐dose anticoagulants up to 90 days (RR 0.40, 95% CI 0.08 to 2.03; I² not applicable; 1 study, 590 participants; low‐certainty evidence; Analysis 2.4). We downgraded the evidence two levels due to imprecision (CI of the absolute difference comprises both important clinical benefit and harm, and fewer than 300 events were included in the analysis).

Major bleeding

Sadeghipour 2021 reported major bleeding with a follow‐up of up to 90 days. Higher‐dose anticoagulants may result in little to no difference in major bleeding compared to lower‐dose anticoagulants up to 90 days (RR 1.74, 95% CI 0.51 to 5.87; I² not applicable; 1 study, 590 participants; low‐certainty evidence; Analysis 2.5). We downgraded the evidence two levels due to imprecision (CI of the absolute difference comprises both important clinical benefit and harm, and fewer than 300 events were included in the analysis).

Adverse events (minor bleeding, gastrointestinal adverse effects (e.g. nausea, vomiting, diarrhoea, abdominal pain), allergic reactions, renal failure and amputations)

Minor bleeding

Sadeghipour 2021 reported minor bleeding with a follow‐up of up to 90 days. Higher‐dose anticoagulants may increase minor bleeding compared to lower‐dose anticoagulants up to 90 days (RR 2.32, 95% CI 0.90 to 5.95; I² not applicable; 1 study, 590 participants; low‐certainty evidence; Analysis 2.6). We downgraded the evidence two levels due to imprecision (CI of the absolute difference comprises both important clinical benefit and harm, and fewer than 300 events were included in the analysis).

Stroke

Sadeghipour 2021 reported stroke with a follow‐up of up to 90 days. Higher‐dose anticoagulants may result in no difference in stroke compared to lower‐dose anticoagulants up to 90 days (RR 0.99, 95% CI 0.06 to 15.80; I² not applicable; 1 study, 590 participants; low‐certainty evidence; Analysis 2.7). We downgraded the evidence two levels due to imprecision (CI of the absolute difference comprises both important clinical benefit and harm, and fewer than 300 events were included in the analysis).

Acute peripheral arterial thrombosis

Sadeghipour 2021 reported acute peripheral arterial thrombosis with a follow‐up of up to 90 days. The evidence is not estimable about the effect of higher‐dose anticoagulants on acute peripheral arterial thrombosis compared to lower‐dose anticoagulants up to 90 days (no events in both groups; I² not applicable; 1 study, 590 participants; low‐certainty evidence; Analysis 2.8). We downgraded the evidence two levels due to imprecision (no events).

Myocardial infarction

Sadeghipour 2021 reported myocardial infarction with a follow‐up of up to 90 days. The evidence is not estimable about the effect of higher‐dose anticoagulants on myocardial infarction compared to lower‐dose anticoagulants up to 90 days (no events in both groups; I² not applicable; 1 study, 590 participants; low‐certainty evidence; Analysis 2.9). We downgraded the evidence two levels due to imprecision (no events).

Atrial fibrillation

Sadeghipour 2021 reported atrial fibrillation with a follow‐up of up to 90 days. Higher‐dose anticoagulants may result in little to no difference in atrial fibrillation compared to lower‐dose anticoagulants up to 90 days (RR 0.50, 95% CI 0.13 to 1.97; I² not applicable; 1 study, 590 participants; low‐certainty evidence; Analysis 2.10). We downgraded two levels due to imprecision (CI of the absolute difference comprises both important clinical benefit and important clinical harm).

Thrombocytopenia

Sadeghipour 2021 reported thrombocytopenia with a follow‐up of up to 90 days. Higher‐dose anticoagulants may result in little to no difference in adverse events (thrombocytopenia) compared to lower‐dose anticoagulants up to 90 days (RR12.91, 95% CI 0.73 to 228.18; I² not applicable; 1 study, 590 participants; low‐certainty evidence; Analysis 2.11). We downgraded two levels due to imprecision (CI of the absolute difference comprises both important clinical benefit and important clinical harm).

Hospitalisation time in days

There were no available data for this outcome.

Quality of life

There were no available data for this outcome.

3. Anticoagulants (all types) versus no treatment

Three studies compared enoxaparin (low‐molecular‐weight heparin) (Albani 2020), heparinoids (unfractionated heparin, low‐molecular‐weight heparin or fondaparinux), direct anticoagulants or vitamin K antagonists (Rentsch 2020; Santoro 2020) (7027 participants) to no treatment (4511 participants). Albani 2020 and Rentsch 2020 compared "prophylactic anticoagulation" (including oral, subcutaneous, or intravenous forms) to no treatment. Santoro 2020 compared "prophylactic anticoagulation" in 83% of cases, while 15% received a full dose of low‐molecular‐weight heparin, 1% oral anticoagulation with AVK, and 1% direct anticoagulants (including oral, subcutaneous, or intravenous forms) no treatment. See summary of findings Table 2.

The Cochrane Handbook for Systematic Reviews of Interventions states that studies judged to be at critical risk of bias should be excluded from the meta‐analysis (Reeves 2021). However, given the small number of studies, there is a balance between loss of information and excluding unreliable information. Therefore, we retained all studies in the analyses, but we also stated the critical risk with the related evidence.

Primary outcomes

All‐cause mortality

Albani 2020 reported all‐cause mortality (follow‐up of up to 15 days) as the proportion of participants and as odds ratio (OR) after adjusting for some covariates (e.g. age, sex, disease severity, admission to ICU and COVID‐19 treatment). They found 200 (25%) deaths in the intervention group and 154 (25.5%) deaths in the comparator group (adjusted OR 0.53, 95% CI 0.40 to 0.70; 1403 participants), in favour of the intervention group after all adjustments (serious risk of bias).

Rentsch 2020 reported all‐cause mortality (follow‐up of up to 30 days) as the proportion of participants and as hazard ratio (HR) after adjusting for some covariates (inverse probability of treatment weighting). They found 418 (11.5%) deaths in the intervention group and 92 (13.7%) deaths in the comparator group (adjusted HR 0.69, 95% CI 0.61 to 0.77; 4297 participants), in favour of the intervention group after all adjustments (moderate risk of bias).

Santoro 2020 reported all‐cause mortality (follow‐up of up to 26 days) as the proportion of participants and as RR after the Cox's multivariable regression analysis only for participants with respiratory failure (2859 participants, 49%). They found 467 (32%) deaths in the intervention group and 588 (42%) deaths in the comparator group (adjusted RR 0.58, 95% CI 0.49 to 0.67; 2859 participants), in favour of the intervention group after all adjustments (critical risk of bias).

We combined these results in a meta‐analysis of adjusted values (Analysis 3.1). Anticoagulants may reduce all‐cause mortality but the evidence is very uncertain due to two study results being at critical and serious risk of bias (RR 0.64, 95% CI 0.55 to 0.74; I² = 53%; 3 NRS, 8395 participants; very low‐certainty evidence; Analysis 3.1). It was not possible to test for subgroup differences and carry out sensitivity analysis.

Necessity for additional respiratory support

There were no available data for this outcome.

Secondary outcomes

Mortality related to COVID‐19

There were no available data for this outcome.

Deep vein thrombosis

Albani 2020 reported deep vein thrombosis (follow‐up of up to 15 days) as the proportion of participants but without any adjustment for covariates (e.g. age, sex, disease severity, admission to ICU and COVID‐19 treatment). They found 15 (1.87%) deep vein thromboses in the intervention group and 2 (0.33%) in the comparator group (critical risk of bias). The evidence on DVTs is uncertain (RR 5.67, 95% CI 1.30 to 24.70; I² not applicable; 1 NRS, 1403 participants, very low‐certainty evidence; Analysis 3.2).

Pulmonary embolism

Albani 2020 reported pulmonary embolism (follow‐up of up to 15 days) as the proportion of participants but without any adjustment for covariates (e.g. age, sex, disease severity, admission to ICU and COVID‐19 treatment). They found 32 (4%) pulmonary embolism in the intervention group and 1 (0.1%) pulmonary embolism in the comparator group (critical risk of bias). The evidence on pulmonary embolism is uncertain (RR 24.19, 95% CI 3.31 to 176.53; I² not applicable; 1 NRS, 1403 participants; very low‐certainty evidence; Analysis 3.3).

Major bleeding

Albani 2020 reported major bleeding (follow‐up of up to 15 days) as the proportion of participants but without any adjustment for covariates (e.g. age, sex, disease severity, admission to ICU and COVID‐19 treatment). They found 16 (2%) major bleeding in the intervention group and 15 (2.4%) major bleeding in the comparator group (critical risk of bias).

Santoro 2020 reported major bleeding (follow‐up of up to 26 days) as the proportion of participants but without any adjustment for covariates. The Cox's multivariable regression analysis was performed to define independent risk factors only for the mortality outcome. They found 70 (2.7%) major bleeding in the intervention group and 58 (1.8%) major bleeding in the comparator group (critical risk of bias).

The evidence on major bleeding is uncertain (RR 1.19, 95% CI 0.66 to 2.12; I² = 58%; 2 NRS, 7218 participants; very low‐certainty evidence; Analysis 3.4). It was not possible to test for subgroup differences and carry out sensitivity analysis.

Adverse events (minor bleeding, gastrointestinal adverse effects (e.g. nausea, vomiting, diarrhoea, abdominal pain), allergic reactions, renal failure and amputations)

Stroke

Albani 2020 reported stroke (follow‐up of up to 15 days) as the proportion of participants but without any adjustment for covariates (e.g. age, sex, disease severity, admission to ICU and COVID‐19 treatment). They found 6 (0.7%) stroke events in the intervention group and 4 (0.6%) in the comparator group (critical risk of bias). The evidence on stroke is uncertain (RR 1.13, 95% CI 0.32 to 4.0; I² not applicable; 1 NRS, 1403 participants; very low‐certainty evidence; Analysis 3.5). We downgraded one level due to study limitations (overall critical risk of bias, especially related to confounding) and two levels due to imprecision (fewer than 300 events were included in the analysis and very large CI).

Myocardial infarction

Albani 2020 reported myocardial infarction (follow‐up of up to 15 days) as the proportion of participants but without any adjustment for covariates (e.g. age, sex, disease severity, admission to ICU and COVID‐19 treatment). They found 10 (1.2%) myocardial infarction events in the intervention group and no events in the comparator group (critical risk of bias). The evidence on myocardial infarction is uncertain (RR 15.88, 95% CI 0.93 to 270.48; I² not applicable; 1 NRS, 1403 participants; very low‐certainty evidence; Analysis 3.6). We downgraded one level due to study limitations (overall critical risk of bias, especially related to confounding) and two levels due to imprecision (fewer than 300 events were included in the analysis and very large CI).

Hospitalisation time in days

Albani 2020 reported hospitalisation time in days (follow‐up of up to 15 days) but without any adjustment for covariates (e.g. age, sex, disease severity, admission to ICU and COVID‐19 treatment). Anticoagulants may increase hospitalisation time compared to no anticoagulation (MD 5.00, 95% CI 4.47 to 5.53; I² not applicable; 1 NRS, 1376 participants; moderate‐certainty evidence; Analysis 3.7). We downgraded one level due to study limitations (overall serious risk of bias, especially related to confounding).

Quality of life

There were no available data for this outcome.

Discussion

This review aimed to assess the effects of anticoagulants versus active comparator, placebo or no intervention on mortality and need for additional respiratory support for people hospitalised with COVID‐19.

Summary of main results

We found no quasi‐RCTs with available data assessing the effects of anticoagulants compared to active comparator, placebo or no intervention on mortality and the need for additional respiratory support for people hospitalised with COVID‐19. Since we found better study designs for the comparisons of interest, we excluded all retrospective studies from this review.

We found four RCTs that compared higher versus lower doses of anticoagulants (unfractionated heparin, low‐molecular‐weight heparin, or direct anticoagulants (rivaroxaban)) in 4647 participants hospitalised with COVID‐19 (Table 2). Higher‐dose anticoagulants result in little to no difference in all‐cause mortality and increase minor bleeding compared to lower‐dose anticoagulants for up to 30 days. Higher‐dose anticoagulants probably reduce pulmonary embolism, slightly increase major bleeding, and result in little to no difference in hospitalisation time. They may result in little to no difference in deep vein thrombosis and in adverse events (stroke, major adverse limb event, myocardial infarction, atrial fibrillation, or thrombocytopenia). We are uncertain about the effects on necessity for additional respiratory support because the certainty of evidence is very low. There were no data regarding mortality related to COVID‐19, and quality of life. See summary of findings Table 1.

One included RCT, which compared higher versus lower doses of anticoagulants (low‐molecular‐weight heparin) in 590 participants hospitalised with COVID‐19 also provided data for the long‐term time point (after hospital discharge) of up to 90 days after intervention (Table 2). Higher‐dose anticoagulants probably result in little to no difference in all‐cause mortality, deep vein thrombosis, and major bleeding, and may reduce pulmonary embolism, increase adverse events (minor bleeding), and result in little to no difference in adverse events (stroke, atrial fibrillation, and thrombocytopenia) compared to lower‐dose anticoagulants for up to 90 days. The evidence is not estimable about the effect of higher‐dose anticoagulants on necessity for additional respiratory support, adverse events (acute peripheral arterial thrombosis, and myocardial infarction) compared to lower‐dose anticoagulants up to 90 days because there were no events. There were no data regarding mortality related to COVID‐19 and quality of life.

We also found three prospective NRS, which compared anticoagulants (heparinoids (unfractionated heparin, low‐molecular‐weight heparin or fondaparinux), direct anticoagulants or vitamin K antagonists) versus no anticoagulants in 11,538 participants hospitalised with COVID‐19 (Table 2). Anticoagulants may reduce all‐cause mortality but the evidence is very uncertain due to two study results being at critical and serious risk of bias. Anticoagulants may increase hospitalisation time compared to no anticoagulants for up to 30 days. We are uncertain about the effects on deep vein thrombosis, pulmonary embolism, major bleeding, adverse events (stroke, and myocardial infarction) because the certainty of evidence is very low. There were no data regarding need for additional respiratory support, mortality related to COVID‐19 or quality of life. See summary of findings Table 2.

We found 62 ongoing studies (from Argentina: 2, Australia: 1, Austria: 1, Belgium: 1, Brazil: 6, Canada: 1, China: 3, Egypt: 2, France: 3, Germany: 3, India: 4, Iran: 1, Ireland: 2, Italy: 6, Mexico: 1, Qatar: 1, Spain: 6, Switzerland: 2, the UK: 3, and the USA: 13) that plan to evaluate 35,470 participants in this setting, of whom 35,350 individuals are from 60 RCTs, and 120 are from two prospective NRS. Thirty‐five ongoing studies plan to report data for mortality. Twenty‐six ongoing studies plan to report data on the need for additional respiratory support. Fifty‐eight ongoing studies are expected to be completed in December 2021, and four in July 2022. Six of these ongoing studies plan to include 1000 participants or more. See Table 3.

One of the studies plans to compare prophylactic enoxaparin, full‐dose enoxaparin and apixaban versus no anticoagulant in 3600 participants to assess the composite of all‐cause mortality, intubation requiring mechanical ventilation, systemic thromboembolism or ischaemic stroke. One study plans to compare higher‐dose enoxaparin versus lower‐dose enoxaparin in 2000 participants to assess the incidence of venous thromboembolism, while another plans to compare aspirin, clopidogrel, rivaroxaban, atorvastatin, and omeprazole with no treatment in 3170 participants to assess mortality at 30 days. One study plans to compare higher‐dose enoxaparin versus lower‐dose enoxaparin to assess venous thromboembolism in 2712 participants, another plans to compare apixaban versus prophylactic enoxaparin and full‐dose enoxaparin in 3600 participants to assess overt bleeding plus haemoglobin drop, cardiac tamponade, bleeding requiring surgical intervention for control, bleeding requiring vasoactive agents, or intracranial haemorrhage, and another study plans to compare several possible interventions (without details about type or dose of anticoagulants) in 1800 participants to assess time to clinical improvement.

Overall completeness and applicability of the evidence

While all of the studies reported our primary outcome of all‐cause mortality, we found sparse data relating to the need for additional respiratory support and hospitalisation time. It is also noteworthy that none of the studies measured our secondary outcomes such as mortality related to COVID‐19 and quality of life. Furthermore, there are neither data comparing different types of anticoagulants or anticoagulants versus non‐pharmacological interventions, nor data from more than 30 days after the intervention.

There was moderate heterogeneity in the methods of the included studies and many did not provide complete and clear information about their data. This hindered the qualitative analyses and the assessment of the risk of bias of many outcomes in some studies.

The number of studies for each of the possible comparisons was small, ranging from three to four studies. However, the included studies had relatively large primary sample sizes (six studies with 562 or more participants), except for only one study that evaluated 20 participants. The largest study involved 5838 participants, 2601 of whom were treated with anticoagulation in a non‐randomised design but did not provide data regarding one of our primary outcomes (necessity for additional respiratory support).

There was considerable variation in the use of the same intervention (e.g. dosages, type, method of application). The variation in assessment for the confounding factor in NRS also impaired the results.

It is noteworthy that the studies included in this review were conducted in 21 different countries, most of which (52%) were high‐income countries. Social and cultural aspects of the evaluated interventions can also interfere with their acceptability and effectiveness for the treatment of people hospitalised with COVID‐19. Therefore, the external validity of the overall evidence presented in this review should be considered with caution.

We acknowledge that designing and conducting an appropriate study with available data for this topic is difficult. The new approach regarding prophylactic anticoagulants for people hospitalised with COVID‐19 has been used to provide high levels of anticoagulants for these people, although there is now available evidence based on RCTs against their use. This reinforces the importance of this review and serves as an incentive for further investigation.

Certainty of the evidence

We found four RCTs with data for one comparison ('higher‐dose anticoagulants versus lower‐dose anticoagulants') at two different time points and three prospective NRS with data for another comparison ('anticoagulants versus no treatment') at a short‐term time point for this review; we also excluded all retrospective studies.

The overall risk of bias was low for two and high for two RCTs included in the comparison 'higher‐dose anticoagulants versus lower‐dose anticoagulants'. We judged the bias domains due to random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective outcome reporting, and other biases from low to high. Although it did not change the effect estimate of all‐cause mortality when excluded in a sensibility analysis, there was a high risk of bias for one large RCT (the only study reporting some of the outcomes of interest). Despite the increasing number of studies on prophylactic anticoagulants for people hospitalised with COVID‐19 in the past months, the overall risk of bias for all‐cause mortality, deep vein thrombosis, pulmonary embolism, major bleeding and adverse events (stroke and myocardial infarction) in the comparison 'anticoagulants versus no treatment' was critical and for hospitalisation was serious in the same comparison. We judged the bias domains due to confounding, selection of participants into the study, classification of interventions, deviations from the intended intervention, measurement of outcomes, and selection of the reported results from low to critical risk of bias.

The certainty of evidence is high to very low. We downgraded the certainty of evidence due to the risk of bias, particularly with regard to detection, performance and attrition in two RCTs and also to selection and other bias in one of them. Although three RCTs (3407 participants) assessed the necessity for additional respiratory support, there is considerable uncertainty about this primary outcome (very low‐certainty evidence). In the NRS, we downgraded the certainty of evidence due to the risk of bias, particularly with regard to the overall critical/serious risk of bias across studies, especially related to confounding or selection bias. We downgraded the certainty of evidence due to study limitations (risk of bias), inconsistency (unexplained heterogeneity) and imprecision (few events and large CI). We decided to pool data, even in NRS, due to the clinically relevant question related to mortality, but the judgements of critical risk of bias mean that these data are particularly unreliable.

Potential biases in the review process

We performed a comprehensive search of the literature and performed study selection according to the Cochrane Handbook for Systematic Reviews of Interventions (Lefebvre 2021). We believe that we identified all of the relevant studies that met our inclusion criteria. However, the possibility remains that we may have missed some studies, particularly in the grey literature. Although we considered 'COVID‐19' and 'SARS‐CoV‐2' as 'Supplementary Concept' or 'free terms' in our search strategies, they were included as 'index terms' in 2021 for databases such as MEDLINE. Therefore, in the future versions of this review, we plan to include these relevant terms also as 'index terms' in our search strategies. We adhered to the inclusion and exclusion criteria prespecified in the protocol in order to limit subjectivity (Flumignan 2020a). We made efforts to obtain additional relevant data from study authors but were unable to do so for all of the included studies. If we can source supplementary data, we will consider them in future updates. Two review authors selected studies in duplicate, independently, to reduce the potential bias of the review process. One review author extracted data and assessed the risk of bias of the included studies while another checked the data extraction and 'Risk of bias' judgements, to accelerate the process and also to reduce the potential bias of the review process. Additional analysis (subgroups and sensitivity analysis) was performed as planned in our protocol, but the conclusions were based on the primary analysis (Flumignan 2020a). We assumed the pragmatic decision to include NRS at critical and serious risk of bias in meta‐analysis due to the relevance of the clinical question. It is perhaps reasonable to have included these in analyses given the small number of studies, but we note that this was a decision taken in the review process. We ensure that any such syntheses were presented throughout the review with unequivocal warnings about the risk of bias and note that the findings cannot rely upon this very low‐certainty evidence.

The synthesis of evidence is a field in constant transformation. Therefore, the Cochrane Reviews are periodically updated, mainly the rapid reviews. During the final process of this review, we identified at least three other trials (898 participants together) that seem to reach our inclusion criteria (Perepu 2021; Sholzberg 2021b; Spyropoulos 2021). We will consider these trials in the next update of this review, but we did not assess them for this rapid review updating.

Agreements and disagreements with other studies or reviews

Since the publication of the latest version of this review (Flumignan 2020b), a number of systematic reviews have addressed the role of anticoagulants in people with COVID‐19.

Abdel‐Maboud 2021 searched MEDLINE, Scopus, Cochrane Library, Science direct, OVID, medRxiv, bioRxiv, and Web of Science without language limits on 2 July 2020. They did not specify the inclusion criteria for study design and limited their search to eight keywords related to intervention and population for all databases and only for registers from December 2019. Abdel‐Maboud 2021 included only NRS, most retrospective cohorts or consecutive series, did not assess the risk of bias or the certainty of evidence and concluded that "current evidence is not sufficient to support the role of prophylactic heparin in reducing mortality among COVID‐19 patients."
Hasan 2020 searched PubMed, Google Scholar, medRxiv and SSRN (preprint server) up to 25 June 2020. They did not specify the inclusion criteria for study design and limited their search to some keywords related to heparin (without other anticoagulant terms) and population and only for data from 2020. Hasan 2020 combined 12 prospective and retrospective cohorts with cross‐sectional studies but did not assess the risk of bias or the certainty of evidence and concluded that prophylactic anticoagulants in higher doses may fail less than those in lower doses for people with COVID‐19 admitted to ICU."
Kamel 2021 searched Google Scholar, PubMed, Scopus, the Cochrane Library and Clinical Trials.gov up to 5 July 2020. They included case‐control and cohort studies and limited their search to English‐language studies. Kamel 2021 used an obsolete risk of bias tool (The Modified Newcastle–Ottawa Scoring System), did not assess the certainty of evidence and concluded that anticoagulants may reduce mortality in people with COVID‐19 and that higher‐dose anticoagulants might offer an advantage over lower‐dose anticoagulants in this setting.
Lazaridis 2021 searched PubMed, Ovid, Google Scholar, MEDLINE and Embase databases from December 2019 to 30 May 2020 with limited terms. They considered only randomised clinical trials, quasi‐experimental studies, case reports and case series for inclusion. Lazaridis 2021 combined four retrospective NRS without an assessment with a validated risk of bias and certainty of evidence tool and concluded that anticoagulants may reduce the mortality in severely ill people with COVID‐19.
Matli 2021 searched Ovid MEDLINE, Web of Science, PubMed and Google Scholar from March 2020 to January 2021 with limited terms related to anticoagulants and antiplatelet agents. They included only English‐language published studies and combined 12 NRS without any risk of bias or certainty of evidence assessment. Matli 2021 concluded that anticoagulants reduced mortality and reduced thromboembolic events in people hospitalised with COVID‐19, but there is a paucity of data on antiplatelet use in combination with anticoagulants in this setting.
McBane 2020 searched MEDLINE and Embase from November 2019 to May 2020. They did not specify the inclusion criteria for study design and the limits regarding study language but limited their search to studies with 100 participants or more. McBane 2020 used an obsolete risk of bias tool (The Modified Newcastle–Ottawa Scoring System), did not assess the certainty of evidence and combined 27 NRS in meta‐analyses to include in their recommendations: 1) lower‐dose anticoagulants for all people hospitalised with COVID‐19, 2) a baseline screening venous ultrasound of lower limbs upon admission in ICU, and 3) extending anticoagulation prophylaxis to 35–45 days post‐hospital discharge to reduce venous thromboembolism while it can increase bleeding, even under low‐quality available evidence.
Moonla 2021 searched PubMed, Embase, and the Cochrane Library from the inception of COVID‐19 (specific date not provided) to 22 October 2020. They included only studies reporting mortality and anticoagulant use in people hospitalised with COVID‐19 without limit regarding the study design, but they limited their inclusion to studies of 10 participants or more. Moonla 2021 reported only one of our included studies' results (Lemos 2020), used an obsolete risk of bias tool (The Modified Newcastle–Ottawa Scoring System) for NRS, and did not assess the certainty of evidence. They combined 17 studies into meta‐analyses and concluded that lower‐dose anticoagulants were associated with lower in‐hospital mortality without excess bleeding compared to no anticoagulation and that the higher‐dose anticoagulation revealed no survival benefit but a three‐fold increase in major bleeding.
Parisi 2021 searched MEDLINE, Embase, PubMed, Web of Science, CENTRAL, medRxiv, and Preprints.org on 8 January 2021. They reported following the Cochrane Handbook for Systematic Reviews of Interventions but did not provide a full search strategy and did not describe the date and language limits. Parisi 2021 considered two of our included studies (Albani 2020; Rentsch 2020), used an obsolete risk of bias tool (The Modified Newcastle–Ottawa Scoring System) for NRS, and did not assess the certainty of evidence. They combined 29 NRS in meta‐analyses, including one study with a mixed population (hospitalised and non‐hospitalised people), and concluded that both higher‐ and lower‐dose anticoagulant regimens are associated with better survival in people with COVID‐19, particularly the severely ill. However, Parisi 2021 added that in non‐critically ill individuals with COVID‐19, the lower‐dose anticoagulant is probably preferred due to the higher risk of bleeding at higher doses.
Patell 2021 searched MEDLINE, Embase, and Cochrane CENTRAL from inception to 29 August 2020. They considered RCTs, retrospective and prospective NRS, or case series of adults hospitalised with COVID‐19 for inclusion and limited their search to studies in English and with 10 or more participants. Patell 2021 used the validated methodological index for non‐randomised studies (MINORS) to assess the risk of bias in the included studies and did not assess the certainty of evidence. They combined 35 NRS of hospitalised people with COVID‐19 in meta‐analyses and concluded that hospitalised patients with COVID‐19 treated with lower‐dose anticoagulants have a decreased rate of thrombosis compared with those receiving no anticoagulant, while higher‐dose anticoagulant regimens were not associated with decreased in‐hospital thrombotic events compared with lower‐dose anticoagulants.
Talasaz 2021 systematically searched ClinicalTrials.gov and the World Health Organization (WHO) International Clinical Trials Registry Platform for ongoing RCTs regarding antithrombotic drugs for people hospitalised and non‐hospitalised with COVID‐19 and reported the results in a narrative review. They reported the results of only one of our included studies (Lemos 2020), and concluded that the "optimal thromboprophylaxis has not been established for patients with this disease".

In order to prevent thrombosis, some clinicians use higher‐dose anticoagulants rather than standard prophylactic (lower) dosing for inpatients with COVID‐19 (AVF 2020; Bikdeli 2020; Obe 2020). However, this practice is not supported by robust evidence. Although some practical guidelines address the management of prophylactic anticoagulation in people with COVID‐19, some of these recommendations are based on non‐COVID‐19 populations or low‐quality COVID‐19‐related evidence (AVF 2020; Bikdeli 2020; NHS 2020; Obe 2020; Ramacciotti 2020). Cuker 2021 searched Cochrane COVID‐19 study register, Embase, Epistemonikos COVID‐19 Evidence, MEDLINE, and WHO Global Research Database in August 2020 without time or language limitations to perform a living guideline under a GRADE approach. Cuker 2021 found very low‐certainty evidence, based mainly on an RCT that we also included in this review (Zarychanski 2021), and made two conditional recommendations in favour of lower‐dose anticoagulation over higher‐dose (intermediate or therapeutic‐intensity) anticoagulation for critical or acute illness patients with COVID‐19 who do not have confirmed or suspected venous thromboembolism.

Our review seems to be more comprehensive than the previous reviews identified here, which used limited search strategies, imposed language or date limits, searched overlapping databases (e.g. SCOPUS, Pubmed, MEDLINE and Web of Science in the same review) or searched a limited number of databases (e.g. ClinicalTrials.gov and the WHO International Clinical Trials Registry Platform only). None of the identified systematic reviews used the GRADE approach (excepting a guideline under GRADE approach (Cuker 2021)) to assess the certainty of evidence. Although some previous reviews identified the potential of anticoagulants in lower doses and no difference with higher doses, the evidence found is conflicting. Since we identified high‐certainty evidence, our conclusions are more decisive for clinical practice.

Figure 1

Study flow diagram
RCT: randomised controlled trial; NRS: non‐randomised study

Figure 2

RoB 1.0 assessments for randomised controlled trials

Figure 3

RoB 1.0 graph: assessments for randomised controlled trials presented as percentages across studies.

Figure 4

ROBINS‐I assessments: anticoagulants (all types) versus no treatment for people hospitalised with COVID‐19 (all‐cause mortality)

Figure 5

ROBINS‐I assessments: anticoagulants (all types) versus no treatment for people hospitalised with COVID‐19 (deep vein thrombosis)

Figure 6

ROBINS‐I assessments: anticoagulants (all types) versus no treatment for people hospitalised with COVID‐19 (pulmonary embolism)

Figure 7

ROBINS‐I assessments: anticoagulants (all types) versus no treatment for people hospitalised with COVID‐19 (major bleeding)

Figure 8

ROBINS‐I assessments: anticoagulants (all types) versus no treatment for people hospitalised with COVID‐19 (adverse events: stroke)

Figure 9

ROBINS‐I assessments: anticoagulants (all types) versus no treatment for people hospitalised with COVID‐19 (adverse events: myocardial infarction)

Figure 10

ROBINS‐I assessments: anticoagulants (all types) versus no treatment for people hospitalised with COVID‐19 (hospitalisation)

Analysis 1.1

Comparison 1: Higher‐dose anticoagulants versus lower‐dose anticoagulants (short term), Outcome 1: All‐cause mortality

Analysis 1.2

Comparison 1: Higher‐dose anticoagulants versus lower‐dose anticoagulants (short term), Outcome 2: All‐cause mortality ‐ trials at low risk of bias

Analysis 1.3

Comparison 1: Higher‐dose anticoagulants versus lower‐dose anticoagulants (short term), Outcome 3: Necessity for additional respiratory support

Analysis 1.4

Comparison 1: Higher‐dose anticoagulants versus lower‐dose anticoagulants (short term), Outcome 4: Necessity for additional respiratory support ‐ trials at low risk of bias

Analysis 1.5

Comparison 1: Higher‐dose anticoagulants versus lower‐dose anticoagulants (short term), Outcome 5: Deep vein thrombosis

Analysis 1.6

Comparison 1: Higher‐dose anticoagulants versus lower‐dose anticoagulants (short term), Outcome 6: Deep vein thrombosis ‐ trials at low risk of bias

Analysis 1.7

Comparison 1: Higher‐dose anticoagulants versus lower‐dose anticoagulants (short term), Outcome 7: Pulmonary embolism

Analysis 1.8

Comparison 1: Higher‐dose anticoagulants versus lower‐dose anticoagulants (short term), Outcome 8: Pulmonary embolism ‐ trial at low risk of bias

Analysis 1.9

Comparison 1: Higher‐dose anticoagulants versus lower‐dose anticoagulants (short term), Outcome 9: Major bleeding

Analysis 1.10

Comparison 1: Higher‐dose anticoagulants versus lower‐dose anticoagulants (short term), Outcome 10: Major bleeding ‐ trials at low risk of bias

Analysis 1.11

Comparison 1: Higher‐dose anticoagulants versus lower‐dose anticoagulants (short term), Outcome 11: Adverse events (minor bleeding)

Analysis 1.12

Comparison 1: Higher‐dose anticoagulants versus lower‐dose anticoagulants (short term), Outcome 12: Adverse events (minor bleeding) ‐ trials at low risk of bias

Analysis 1.13

Comparison 1: Higher‐dose anticoagulants versus lower‐dose anticoagulants (short term), Outcome 13: Adverse events (stroke)

Analysis 1.14

Comparison 1: Higher‐dose anticoagulants versus lower‐dose anticoagulants (short term), Outcome 14: Adverse events (stroke) ‐ trials at low risk of bias

Analysis 1.15

Comparison 1: Higher‐dose anticoagulants versus lower‐dose anticoagulants (short term), Outcome 15: Adverse events (major adverse limb event)

Analysis 1.16

Comparison 1: Higher‐dose anticoagulants versus lower‐dose anticoagulants (short term), Outcome 16: Adverse events (myocardial infarction)

Analysis 1.17

Comparison 1: Higher‐dose anticoagulants versus lower‐dose anticoagulants (short term), Outcome 17: Adverse events (myocardial infarction) ‐ trials at low risk of bias

Analysis 1.18

Comparison 1: Higher‐dose anticoagulants versus lower‐dose anticoagulants (short term), Outcome 18: Adverse events (atrial fibrillation)

Analysis 1.19

Comparison 1: Higher‐dose anticoagulants versus lower‐dose anticoagulants (short term), Outcome 19: Adverse events (thrombocytopenia)

Analysis 1.20

Comparison 1: Higher‐dose anticoagulants versus lower‐dose anticoagulants (short term), Outcome 20: Hospitalisation time

Analysis 1.21

Comparison 1: Higher‐dose anticoagulants versus lower‐dose anticoagulants (short term), Outcome 21: Hospitalisation time ‐ trials at low risk of bias

Analysis 2.1

Comparison 2: Higher‐dose anticoagulants versus lower‐dose anticoagulants (long term), Outcome 1: All‐cause mortality

Analysis 2.2

Comparison 2: Higher‐dose anticoagulants versus lower‐dose anticoagulants (long term), Outcome 2: Necessity for additional respiratory support

Analysis 2.3

Comparison 2: Higher‐dose anticoagulants versus lower‐dose anticoagulants (long term), Outcome 3: Deep vein thrombosis

Analysis 2.4

Comparison 2: Higher‐dose anticoagulants versus lower‐dose anticoagulants (long term), Outcome 4: Pulmonary embolism

Analysis 2.5

Comparison 2: Higher‐dose anticoagulants versus lower‐dose anticoagulants (long term), Outcome 5: Major bleeding

Analysis 2.6

Comparison 2: Higher‐dose anticoagulants versus lower‐dose anticoagulants (long term), Outcome 6: Adverse events (minor bleeding)

Analysis 2.7

Comparison 2: Higher‐dose anticoagulants versus lower‐dose anticoagulants (long term), Outcome 7: Adverse events (stroke)

Analysis 2.8

Comparison 2: Higher‐dose anticoagulants versus lower‐dose anticoagulants (long term), Outcome 8: Adverse events (acute peripheral arterial thrombosis)

Analysis 2.9

Comparison 2: Higher‐dose anticoagulants versus lower‐dose anticoagulants (long term), Outcome 9: Adverse events (myocardial infarction)

Analysis 2.10

Comparison 2: Higher‐dose anticoagulants versus lower‐dose anticoagulants (long term), Outcome 10: Adverse events (atrial fibrillation)

Analysis 2.11

Comparison 2: Higher‐dose anticoagulants versus lower‐dose anticoagulants (long term), Outcome 11: Adverse events (thrombocytopenia)

Analysis 3.1

Comparison 3: Anticoagulants versus no treatment (short term), Outcome 1: All‐cause mortality

Analysis 3.2

Comparison 3: Anticoagulants versus no treatment (short term), Outcome 2: Deep vein thrombosis

Analysis 3.3

Comparison 3: Anticoagulants versus no treatment (short term), Outcome 3: Pulmonary embolism

Analysis 3.4

Comparison 3: Anticoagulants versus no treatment (short term), Outcome 4: Major bleeding

Analysis 3.5

Comparison 3: Anticoagulants versus no treatment (short term), Outcome 5: Adverse events (stroke)

Analysis 3.6

Comparison 3: Anticoagulants versus no treatment (short term), Outcome 6: Adverse events (myocardial infarction)

Analysis 3.7

Comparison 3: Anticoagulants versus no treatment (short term), Outcome 7: Hospitalisation time

Summary of findings 1. Higher‐dose anticoagulants compared to lower‐dose anticoagulants for people hospitalised with COVID‐19

Outcomes	*Anticipated absolute effects^ (95% CI)**		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
Higher‐dose anticoagulants compared to lower‐dose anticoagulants for people hospitalised with COVID‐19
Patient or population: people hospitalised with COVID‐19 Setting: hospital Intervention: higher‐dose anticoagulants (LMWH, UFH or rivaroxaban) Comparison: lower‐dose anticoagulants (LMWH or UFH)
Outcomes	Risk with lower‐dose anticoagulants (short‐term outcomes)	Risk with higher‐dose anticoagulants	Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
All‐cause mortality Follow‐up: from 28‐30 days	Study population		RR 1.03 (0.92 to 1.16)	4489 (4 RCTs)	⊕⊕⊕⊕ High^a	Higher‐dose anticoagulants results in little to no difference in all‐cause mortality
All‐cause mortality Follow‐up: from 28‐30 days	191 per 1000	196 per 1000 (175 to 221)	RR 1.03 (0.92 to 1.16)	4489 (4 RCTs)	⊕⊕⊕⊕ High^a
Necessity for additional respiratory support Follow‐up: from 28‐30 days	Study population		RR 0.54 (0.12 to 2.47)	3407 (3 RCTs)	⊕⊝⊝⊝ Very low^b,c,d	The evidence is very uncertain about the effect of higher‐dose anticoagulants on necessity for additional respiratory support.
	117 per 1000	63 per 1000 (14 to 289)	RR 0.54 (0.12 to 2.47)	3407 (3 RCTs)	⊕⊝⊝⊝ Very low^b,c,d
Mortality related to COVID‐19	No studies measured this outcome
Deep vein thrombosis Follow‐up: from 28‐30 days	Study population		RR 1.08 (0.57 to 2.03)	3422 (4 RCTs)	⊕⊕⊝⊝ Low^d	Higher‐dose anticoagulants may result in little to no difference in DVT
Deep vein thrombosis Follow‐up: from 28‐30 days	11 per 1000	12 per 1000 (6 to 22)	RR 1.08 (0.57 to 2.03)	3422 (4 RCTs)	⊕⊕⊝⊝ Low^d
Pulmonary embolism Follow‐up: from 28‐30 days	Study population		RR 0.46 (0.31 to 0.70)	4360 (4 RCTs)	⊕⊕⊕⊝ Moderate^b	Higher‐dose anticoagulants likely reduce PE
Pulmonary embolism Follow‐up: from 28‐30 days	33 per 1000	15 per 1000 (10 to 23)	RR 0.46 (0.31 to 0.70)	4360 (4 RCTs)	⊕⊕⊕⊝ Moderate^b	Higher‐dose anticoagulants likely reduce PE
Major bleeding Follow‐up: from 28‐30 days	Study population		RR 1.78 (1.13 to 2.80)	4400 (4 RCTs)	⊕⊕⊕⊝ Moderate^b	Higher‐dose anticoagulants likely increase major bleeding slightly
Major bleeding Follow‐up: from 28‐30 days	14 per 1000	24 per 1000 (15 to 38)	RR 1.78 (1.13 to 2.80)	4400 (4 RCTs)	⊕⊕⊕⊝ Moderate^b
Adverse events (minor bleeding) Follow‐up: from 28‐30 days	Study population		RR 3.28 (1.75 to 6.14)	1196 (3 RCTs)	⊕⊕⊕⊕ High	Higher‐dose anticoagulants increase adverse events (minor bleeding)
	20 per 1000	47 per 1000 (18 to 121)	RR 3.28 (1.75 to 6.14)	1196 (3 RCTs)	⊕⊕⊕⊕ High
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; COVID‐19: coronavirus disease 2019; DVT: deep vein thrombosis; LMWH: low‐molecular‐weight heparin; PE: pulmonary embolism; RCT: randomised controlled trial; RR: risk ratio; UFH: unfractionated heparin
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.
^aThe largest study in the analysis was at high risk of bias in almost all domains; however, we did not downgrade for study limitations as removing this study in the sensitivity analysis did not change the pooled estimate. ^bDowngraded one level due to study limitations. One randomised controlled trial provided high risk of bias in almost all domains leading to a different pooled estimate after sensitivity analysis. ^cDowngraded one level due to inconsistency. We identified substantial unexplained heterogeneity (I² = 60%). ^dDowngraded two levels due to imprecision. Confidence interval of the absolute difference comprises both important clinical benefit and important clinical harm.

Summary of findings 1. Higher‐dose anticoagulants compared to lower‐dose anticoagulants for people hospitalised with COVID‐19

Summary of findings 2. Anticoagulants compared to no treatment for people hospitalised with COVID‐19

Outcomes	*Anticipated absolute effects^ (95% CI)**		Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
Anticoagulants compared to no treatment for people hospitalised with COVID‐19
Patient or population: people hospitalised with COVID‐19 Setting: hospital Intervention: anticoagulants (LMWH, UFH, fondaparinux, DOACs or VKA) Comparison: no treatment (no anticoagulants)
Outcomes	Risk with no treatment	Risk with anticoagulants	Relative effect (95% CI)	№ of participants (studies)	Certainty of the evidence (GRADE)	Comments
All‐cause mortality Follow‐up: from 15‐30 days	Study population		RR 0.64 (0.55 to 0.74)	8395 (3 observational studies)	⊕⊝⊝⊝ Very low^a,b	Anticoagulants may reduce all‐cause mortality but the evidence is very uncertain due to two study results being at critical and serious risk of bias. The numerical results are very unreliable for outcomes where critical risk of bias is an issue
All‐cause mortality Follow‐up: from 15‐30 days	307 per 1000	196 per 1000 (169 to 227)	RR 0.64 (0.55 to 0.74)	8395 (3 observational studies)	⊕⊝⊝⊝ Very low^a,b
Necessity for additional respiratory support	No studies measured this outcome
Mortality related to COVID‐19	No studies measured this outcome
Deep vein thrombosis Follow‐up: up to 15 days	Study population		RR 5.67 (1.30 to 24.70)	1403 (1 observational study)	⊕⊝⊝⊝ Very low^c,d	It is uncertain if anticoagulants have any effect on DVT. The numerical results are very unreliable for outcomes where critical risk of bias is an issue.
Deep vein thrombosis Follow‐up: up to 15 days	3 per 1000	19 per 1000 (4 to 82)	RR 5.67 (1.30 to 24.70)	1403 (1 observational study)	⊕⊝⊝⊝ Very low^c,d
Pulmonary embolism Follow‐up: up to 15 days	Study population		RR 24.19 (3.31 to 176.53)	1403 (1 observational study)	⊕⊝⊝⊝ Very low^c,d	It is uncertain if anticoagulants have any effect on PE. The numerical results are very unreliable for outcomes where critical risk of bias is an issue.
Pulmonary embolism Follow‐up: up to 15 days	2 per 1000	40 per 1000 (5 to 292)	RR 24.19 (3.31 to 176.53)	1403 (1 observational study)	⊕⊝⊝⊝ Very low^c,d
Major bleeding Follow‐up: from 15‐26 days	Study population		RR 1.19 (0.66 to 2.12)	7218 (2 observational studies)	⊕⊝⊝⊝ Very low^b,c,e	It is uncertain if anticoagulants have any effect on major bleeding. The numerical results are very unreliable for outcomes where critical risk of bias is an issue.
Major bleeding Follow‐up: from 15‐26 days	19 per 1000	23 per 1000 (13 to 41)	RR 1.19 (0.66 to 2.12)	7218 (2 observational studies)	⊕⊝⊝⊝ Very low^b,c,e
Adverse events (minor bleeding)	No studies measured this outcome
*The risk in the intervention group (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; COVID‐19: coronavirus disease 2019; DOACs: direct oral anticoagulants; DVT: deep vein thrombosis; LMWH: low‐molecular‐weight heparin; PE: pulmonary embolism; RR: risk ratio; UFH: unfractionated heparin; VKA: vitamin K antagonist
GRADE Working Group grades of evidence High certainty: we are very confident that the true effect lies close to that of the estimate of the effect. Moderate certainty: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different. Low certainty: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect. Very low certainty: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.
^aDowngraded two levels due to study limitations. Overall critical/serious risk of bias in two studies, especially related to confounding. ^bDowngraded one level due to inconsistency. We found moderate unexplained heterogeneity (I² = 30% to 60%). ^cDowngraded one level due to study limitations. Overall critical risk of bias, especially related to confounding. ^dDowngraded two levels due to imprecision. Fewer than 300 events were included in the analysis and very large confidence interval. ^eDowngraded one level due to imprecision. Confidence interval of the absolute difference comprises both unimportant clinical harm and important clinical harm.

Summary of findings 2. Anticoagulants compared to no treatment for people hospitalised with COVID‐19

Table 1. Glossary of terms

Term	Definition
Anticoagulants	Drugs that suppress, delay or prevent blood clots
Antiplatelet agents	Drugs that prevent blood clots by inhibiting platelet function
Arterial thrombosis	An interruption of blood flow to an organ or body part due to a blood clot blocking the flow of blood
Body mass index (BMI)	Body mass divided by the square of the body height, universally expressed in units of kg/m²
Catheters	Medical devices (tubes) that can be inserted in the body for a broad range of functions, such as to treat diseases, to perform a surgical procedure, and to provide medicine, fluids and food
COVID‐19	An infectious disease caused by SARS‐CoV‐2 virus
Deep vein thrombosis (DVT)	Coagulation or clotting of the blood in a deep vein, that is, far beneath the surface of the skin
Disseminated intravascular coagulopathy	A severe condition in which blood clots form throughout the body, blocking small blood vessels and that may lead to organ failure. As clotting factors and platelets are used up, bleeding may occur, throughout the body (e.g. in the urine, in the stool, or bleeding into the skin)
Duplex ultrasound	Non‐invasive evaluation of blood flow through the arteries and veins by ultrasound devices
Heparin (also known as unfractionated heparin (UFH))	A drug used to prevent blood clotting (anticoagulant, blood thinner)
Hypercoagulability	An abnormality of blood coagulation that increases the risk of blood clot formation in blood vessels (thrombosis)
Low‐molecular‐weight heparin	A drug used to prevent blood clotting (anticoagulant)
Obesity	Amount of body fat beyond healthy conditions (BMI > 30 kg/m²)
Placebo	Substance or treatment with no active effect, like a sugar pill
Platelet	Colourless blood cells that help blood to clot by clumping together
Pulmonary embolism (PE)	Blood clot in the lung or blood vessel leading to the lung. The clot originates in a vein (e.g. deep vein thrombosis) and travels to the lung
Quasi‐randomised controlled trial (quasi‐RCT)	A study in which participants are divided by date of birth or by hospital register number, i.e. not truly randomly divided into separate groups to compare different treatments
Randomised controlled trial (RCT)	A study in which participants are divided randomly into separate groups to compare different treatments
Respiratory failure	An abnormality that results from inadequate gas exchange by the respiratory system
SARS‐CoV‐2	The virus (coronavirus 2) that causes COVID‐19
Thrombosis	Local coagulation of blood (clot) in a part of the circulatory system
Vascular	Relating to blood vessels (arteries and veins)
Venous	Relating to a vein
Venous thromboembolism (VTE)	A condition that involves a blood clot that forms in a vein and may migrate to another location (e.g. the lung)

Table 1. Glossary of terms

Table 2. Summary of characteristics of included studies

Study (design)	Country	Participant age (mean ± SD)	Setting	Intervention type (dose)	Comparator	All‐cause mortality	Necessity for additional respiratory support	Follow‐up time (mean days)	Total participants allocated	Intervention group participants (anticoagulant)
Albani 2020 (Prospective cohort)	Italy	68.66 ± 12.62 (experimental), 70.6 ± 15.01 (comparator)	Hospital^a	Enoxaparin (40‐80 mg once daily, duration 3‐9 days)	NA	In‐hospital mortality: aOR 0.53 (95% CI 0.10 to 0.70), in favour of intervention group	NR	Until death or hospital discharge (time in days NR)	1403	799
Lemos 2020 (RCT)	Brazil	55 ± 10 (experimental), 58 ± 16 (comparator)	Hospital^a	Therapeutic anticoagulation: heparin (SC enoxaparin, adjusted dose by age and CrCl (maximum dose allowed 140 mg twice daily)	Prophylactic anticoagulation: SC UFH 5000 IU three times/day (if weight < 120 kg) and 7500 IU 3 times/day (if weight > 120 kg) or enoxaparin 40 mg once daily (if weight < 120 kg) and 40 mg twice daily (if weight > 120 kg) according to the doctor's judgment	RR 0.33 (95% CI 0.04 to 2.69)	NR	28	20	10
Lopes 2021 (RCT)	Brazil	56.7 ± 14.1 (experimental), 56.5 ± 14.5 (comparator)	Hospital^a	Therapeutic anticoagulation: stable participants = rivaroxaban 20 mg once daily; unstable participants = enoxaparin 1 mg/kg twice daily. Followed by rivaroxaban for 30 days, irrespective of the duration of hospitalisation	Prophylactic anticoagulation: enoxaparin 40 mg once daily	RR 1.49 (95% CI 0.90 to 2.46)	RR 0.16 (95% CI 0.02 to 1.35)	30	615	310
Rentsch 2020 (Prospective cohort)	USA	67.03 ± 12.31 (experimental), 67.83 ± 13.74 (comparator)	Hospital^a	SC UFH (5000 IU twice daily or 3 times/day (1094 participants; 30.2%) LMWH (enoxaparin 40 mg once or twice daily (2506 participants; 69.1%), fondaparinux 2.5 mg once daily (4 participants; 0.1%), dalteparin 2500‐5000 IU once daily, all SC) DOACs (apixaban 2.5 mg twice daily (21 participants; 0.6%), rivaroxaban 10 mg once daily or 2.5 mg twice daily (2 participants; 0.1%), dabigatran 220 mg once daily, all orally)	NA	Inpatient mortality: aHR 0.69 (95% CI 0.61 to 0.77) 30‐day mortality: aHR 0.73 (95% CI 0.66 to 0.81)	NR	30	4297	3627
Sadeghipour 2021 (RCT)	Iran	61.23 ± 14.68 (experimental), 59.66 ± 17.88 (comparator)	Hospital^a	Higher‐dose anticoagulation: enoxaparin 1 mg/kg once daily, modified according to body weight and CrCl	Lower‐dose anticoagulation: enoxaparin 40 mg once daily, modified according to body weight and CrCl	Short‐term time point: RR 1.05 (95% CI 0.87 to 1.28) Long‐term time point: RR 1.07 (95% CI 0.89 to 1.29)	Short‐term time point: no events in both groups Long‐term time point: no events in both groups	90	562	276
Santoro 2020 (Prospective cohort)	Spain, Italy, Ecuador, Cuba, Germany, China, Canada, Serbia, USA, Chile, and Colombia	66 ± 15 (experimental), 63 ± 27 (comparator)	Hospital^a	Anticoagulant (oral, SC, or IV): 327 (12%) participants = previous anticoagulation treatment 1888 (72%) participants = prophylactic (lower‐dose) during hospitalisation 341 (13%) participants = therapeutic (higher‐dose) LMWH 23 (0.75%) oral anticoagulation with VKA 23 (0.75%) DOACs	NA	RR 0.91 (95% CI 0.89 to 0.93), in all participants (N = 3089) RR 0.58 (95% CI 0.49 to 0.67), in those non‐anticoagulated before admission (N = 2695) RR 0.50 (95% CI 0.37 to 0.70), in those undergoing invasive ventilation (N = 391) RR 0.72 (95% CI 0.51 to 1.01), in those undergoing non‐invasive ventilation (N = 583)	NR	15	5838	2601
Zarychanski 2021 (RCT)	UK, USA, Canada, Brazil, Ireland, Netherlands, Australia, Nepal, Saudi Arabia, and Mexico	Critically ill: 60.2 ± 13.1 (experimental), 61.6 ± 12.5 (comparator) Moderate‐severity illness: 59.0 ± 14.1 (experimental), 58.8 ± 13.9 (comparator)	Hospital^a	Therapeutic anticoagulation: LMWH or UFH according to local protocols used for the treatment of acute VTE for up to 14 days or until recovery (defined as hospital discharge, or liberation from supplemental oxygen for ≥ 24 h)	Prophylactic anticoagulation: LMWH or UFH according to local practice or with guidance from the trial protocol on maximum dosing, which included either standard low‐dose thromboprophylaxis or enhanced intermediate dose thromboprophylaxis	Short‐term time point: moderate‐severity RR 0.89 (95% CI 0.67 to 1.19), critically ill RR 1.03 (95% CI 0.88 to 1.21) Long‐term time point: NR	Short‐term time point: moderate‐severity RR 0.89 (95% CI 0.74 to 1.08), critically ill: NR Long‐term time point: NR	90	3450	1780
Total	Australia: 1 Brazil: 3 Canada: 2 Chile: 1 China: 1 Colombia: 1 Cuba: 1 Ecuador: 1 Germany: 1 Iran: 1 Ireland: 1 Italy: 2 Mexico: 1 Nepal: 1 Netherlands: 1 Saudi Arabia: 1 Serbia:1 Spain: 1 UK: 1 USA: 3	55 to 68.66 (mean, 7 studies)	‐	‐	‐	7 studies considered mortality	4 studies considered additional respiratory support	15 to 90 (7 studies)	16,185	9403
aHR: adjusted hazard ratio; aOR: adjusted odds ratio; twice daily: twice a day; CI: confidence interval; CrCl: creatinine clearance; DOACs: direct oral anticoagulants; GFR: glomerular filtration rate;HR: hazard ratio; ICU: intensive care units; IU: international unit;LMWH: low‐molecular‐weight heparin; NA: no anticoagulation; NR: not reported; NRS: non‐randomised study;OR: odds ratio; RCT: randomised controlled trial; RR: risk ratio; SC: subcutaneous; SD: standard deviation; SIC: sepsis‐induced coagulopathy; TID: three times a day; UFH: unfractionated heparin; VKA: vitamin K antagonist
^aHospital: includes intensive care unit, hospital wards or emergency department. ^bAnticoagulation used twice daily if glomerular filtration rate (GFR) was > 30 mL/min, or once daily if GFR was 30 mL/min or less.

Table 2. Summary of characteristics of included studies

Table 3. Summary of characteristics of ongoing studies

Study	Country	Design	Experimental intervention	Comparator intervention	Primary outcomes	Estimated number of participants	Estimated primary completion date
ACTRN12620000517976	Australia	RCT	Nebulised heparin (UFH)	Standard care (without anticoagulants)	Time to separation from invasive ventilation	172	25 July 2021
Busani 2020	Italy	RCT	Enoxaparin	UFH	All‐cause mortality at day 28, defined as the comparison of proportions of patients' deaths for any cause at day 28 from randomisation	210	6 May 2021
Chambers 2020	USA	RCT	Intermediate‐dose enoxaparin	Standard prophylactic dose enoxaparin	Risk of all‐cause mortality (time frame: 30 days post‐intervention)	170	16 April 2021
ChiCTR2000030700	China	RCT	Enoxaparin	Standard care (without anticoagulants)	Time to virus eradication	60	30 September 2020
ChiCTR2000030701	China	RCT	Enoxaparin	Standard care (without anticoagulants)	Time to virus eradication	60	30 September 2020
ChiCTR2000030946	China	Prospective cohort	LMWH	Mechanical prevention	Biochemical indicators	120	24 April 2020
CTRI/2020/06/026220	India	RCT	Nafamostat (synthetic serine proteinase inhibitor)	Standard care (without anticoagulants)	Proportion of patients showing clinical improvement	40	27 January 2021
CTRI/2020/08/027033	India	RCT	Enoxaparin	Standard care (without anticoagulants)	Reduction in clinical symptoms and RT‐PCR test negative	100	27 January 2021
CTRI/2020/11/029175	India	RCT	Nebulised heparin	Standard care (without anticoagulants)	Time to separation from mechanical ventilation (duration of mechanical ventilation) up to day 28	58	27 January 2021
CTRI/2020/11/029345	India	RCT	Higher‐dose enoxaparin	Lower‐dose enoxaparin; apixaban	Time to first event rate within 30 days of randomisation of the composite of all‐cause mortality, intubation requiring mechanical ventilation, systemic thromboembolism (including PE) confirmed by imaging or requiring surgical intervention or ischaemic stroke confirmed by imaging	3600	27 January 2021
EUCTR2020‐001302‐30‐AT	Austria	RCT	Rivaroxaban	Standard care (without anticoagulants)	Time to sustained improvement of one category from admission	500	11 January 2021
EUCTR2020‐001708‐41‐IT	Italy	RCT	Higher‐dose enoxaparin	Lower‐dose enoxaparin	Incidence of VTE (a composite of incident asymptomatic and symptomatic proximal DVT diagnosed by serial compression ultrasonography, and symptomatic PE diagnosed by CT scan), in patients with SARS‐CoV‐2 infection	2000	30 October 2020
EUCTR2020‐001709‐21‐FR	France	RCT	Higher‐dose LMWH	Lower‐dose LMWH	VTE (causing death or not)	230	11 May 2020
EUCTR2020‐001891‐14‐ES	Spain	RCT	Enoxaparin	Standard care (without anticoagulants)	Need for oxygen therapy escalation due to oxygen saturation (Sat O₂) = 92% with FiO₂ = 0.5 and respiratory rate = 30 (IROX index = SatO₂/FiO₂)/FR < 5.5) or invasive mechanical ventilation or mortality during admission	140	16 November 2020
EUCTR2020‐002234‐32‐IT	Switzerland	RCT	Higher‐dose edoxaban	Lower‐dose edoxaban	Major vascular thrombotic events at 25 (+/‐3) days defined as a composite of: Asymptomatic proximal DVT Symptomatic proximal or distal DVT Symptomatic PE or thrombosis Myocardial infarction Ischaemic stroke Non‐CNS systemic embolism Death	420	11 January 2021
EUCTR2020‐002504‐39‐DE	Germany	RCT	Edoxaban	Fondaparinux	Composite of all‐cause mortality and/or VTE and/or arterial thromboembolism within 42 days	172	5 January 2021
EUCTR2020‐003349‐12‐IE	Ireland	RCT	Heparin	Standard care (without anticoagulants)	D‐dimer profile, with data collected on days 1, 3, 5 and 10	40	19 October 2020
Goldin 2020	USA	RCT	Higher‐dose LMWH	Lower‐dose LMWH	Composite outcome of arterial thromboembolic events, venous thromboembolic events and all‐cause mortality at day 30 ± 2 days (time frame: day 30 ± 2 days). Risk of arterial thromboembolic events (including myocardial infarction, stroke, systemic embolism), VTE (including symptomatic DVT of the upper or lower extremity, asymptomatic proximal DVT of the lower extremity, non‐fatal PE), and all‐cause mortality at day 30 ± 2 days	308	26 April 2021
IRCT20200515047456N1	Iran	RCT	UFH	Standard care (without anticoagulants)	Decrease D‐dimer level Improve compliance Improve of oxygenation Improve SOFA score	15	13 July 2020
ISRCTN14212905	UK	RCT	Nafamostat (synthetic serine proteinase inhibitor)	Standard care (without anticoagulants)	Safety of candidate agents as add‐on therapy to standard care in patients with COVID‐19 measured at 30, 60 and 90 days post‐treatment	100	3 August 2020
Kharma 2020	Qatar	RCT	Bivalirudin (DOAC)	LMWH or UFH	PaO₂/FiO₂ ratio (time frame: 3 days of intervention)	100	24 June 2020
Lasky 2021	USA	RCT	Dociparstat (heparinoid)	Placebo	Proportion of participants who are alive and free of invasive mechanical ventilation	525	17 February 2021
Lins 2020	Brazil	RCT	UFH	Standard care (without UFH)	The percentage of clotted dialysers within 72 h in each of the studied groups	90	27 July 2020
Marietta 2020	Italy	RCT	Higher‐dose LMWH	Lower‐dose LMWH	Clinical worsening (includes death and necessity for additional respiratory support)	300	June 2021
NCT04333407	UK	RCT	Rivaroxaban	Standard care (without anticoagulants)	All‐cause mortality at 30 days after admission	3170	30 March 2021
NCT04344756	France	RCT	Higher‐dose LMWH or UFH	Lower‐dose LMWH or UFH	Survival without ventilation	808	31 July 2020
NCT04345848	Switzerland	RCT	Higher‐dose LMWH or UFH	Lower‐dose LMWH or UFH	Composite outcome of arterial or venous thrombosis, disseminated intravascular coagulation and all‐cause mortality	200	30 November 2020
NCT04352400	Italy	RCT	Nafamostat (synthetic serine proteinase inhibitor)	Placebo	Time to clinical improvement	256	December 2021
NCT04366960	Italy	RCT	Higher‐dose enoxaparin	Lower‐dose enoxaparin	Incidence of VTE detected by imaging	2712	August 2020
NCT04367831	USA	RCT	Higher‐dose enoxaparin	Lower‐dose enoxaparin	Total number of patients with clinically relevant venous or arterial thrombotic events in ICU	100	November 2020
NCT04373707	France	RCT	Higher‐dose enoxaparin	Lower‐dose enoxaparin	VTE	602	September 2020
NCT04377997	USA	RCT	Higher‐dose LMWH or UFH	Lower‐dose LMWH or UFH	Risk of composite efficacy endpoint of death, cardiac arrest, symptomatic DVT, PE, arterial thromboembolism, myocardial infarction, or haemodynamic shock Risk of major bleeding event according to the ISTH definition	300	1 January 2021
NCT04397510	USA	RCT	Nebulised heparin	Placebo	Mean daily PaO₂/FiO₂	50	31 December 2020
NCT04406389	USA	RCT	Higher‐dose heparinoid or fondaparinux	Lower‐dose heparinoid or fondaparinux	30‐day mortality	186	December 2021
NCT04409834	USA	RCT	Higher‐dose heparinoid plus antiplatelet agent	Lower‐dose heparinoid without antiplatelet agent	Venous or arterial thrombotic events	750	May 2021
NCT04416048	Germany	RCT	Higher‐dose DOAC (rivaroxaban)	Lower‐dose heparinoid	Composite endpoint of VTE (DVT and/or fatal or non‐fatal PE), arterial thromboembolism, new myocardial infarction, non‐haemorrhagic stroke, all‐cause mortality or progression to intubation and invasive ventilation (time frame: 35 days post‐randomisation)	400	30 May 2021
NCT04420299	Spain	RCT	Higher‐dose heparin	Lower‐dose heparin	Combined worsening variable. Presence of any of the following will be considered worsening Death ICU admission Need for either non‐invasive or invasive mechanical ventilation Progression to moderate/severe respiratory distress syndrome according to objective criteria (Berlin definition) VTE (DVT or PE) or arterial (acute myocardial infarction or stroke) Proportion of patients that worsen	120	31 March 2021
NCT04444700	Brazil	RCT	Higher‐dose enoxaparin	Lower‐dose enoxaparin	Composite outcome of ICU admission (yes/no), non‐invasive positive pressure ventilation (yes/no), invasive mechanical ventilation (yes/no), or all‐cause death (yes/no) up to 28 days	462	31 December 2020
NCT04485429	Brazil	RCT	Higher‐dose heparin	Lower‐dose heparin	Rate of invasive mechanical ventilation	268	31 December 2020
NCT04508439	Mexico	RCT	Higher‐dose enoxaparin	Lower‐dose enoxaparin	Ventilatory support time Thrombotic complications Length of hospital stay Mortality rate	130	30 December 2020
NCT04511923	Ireland	RCT	Nebulised heparin	Standard care (without anticoagulants)	D‐dimer profile up to day 10 Frequency of severe adverse outcomes up to day 60	40	January 2022
NCT04512079	USA	RCT	Apixaban (DOAC)	Lower‐dose enoxaparin; higher‐dose enoxaparin	Time to first event (time frame: 30 days) Number of in‐hospital rate of BARC 3 or 5 (time frame: 30 days) Number of in‐hospital rate of BARC 3 or 5 bleeding (binary). BARC Type 3: a. Overt bleeding plus haemoglobin drop of 3 to < 5 g/dL (provided haemoglobin drop is related to bleed); transfusion with overt bleeding b. Overt bleeding plus haemoglobin drop < 5 g/dL (provided haemoglobin drop is related to bleed); cardiac tamponade; bleeding requiring surgical intervention for control; bleeding requiring IV vasoactive agents c. Intracranial haemorrhage confirmed by autopsy, imaging, or lumbar puncture; intraocular bleed compromising vision	3600	March 2022
NCT04530578	Argentina	RCT	Nebulised heparin	Enoxaparin	Percentage of patients requiring mechanical ventilation (time frame: 15 days)	200	1 June 2021
NCT04542408	Germany	RCT	Higher‐dose LMWH	Lower‐dose LMWH	Combined endpoint: all‐cause mortality and/or VTE and/or arterial thromboembolism (time frame: 42 days) All‐cause mortality and/or VTE and/or arterial thromboembolism during follow‐up (42 days). Thromboembolisms will be detected by duplex ultrasonography of arms and legs	172	30 September 2021
NCT04545541	USA	RCT	Nebulised heparin	Placebo	Alive and Ventilator‐Free Score (time frame: day 28)	300	June 2022
NCT04584580	Egypt	RCT	Higher‐dose LMWH	Lower‐dose LMWH	Mortality (time frame: until patient is discharged or up to 4 weeks whichever comes first) Occurrence of venous and/or arterial thrombosis (time frame: until patient is discharged or up to 4 weeks whichever comes first)	50	31 December 2020
NCT04600141	Brazil	RCT	Higher‐dose LMWH or UFH	Lower‐dose LMWH or UFH	Proportion of patients with clinical improvement (time frame: 30 days) Not hospitalised, with no limitations on activities Not hospitalised, but limited to activities Hospitalised, with no need for supplemental oxygen Hospitalised, needing supplemental oxygen Hospitalised, requiring high‐flow oxygen therapy, non‐invasive mechanical ventilation or both Hospitalised, requiring ECMO, invasive mechanical ventilation or both Death	308	31 December 2020
NCT04604327	Spain	RCT	Higher‐dose LMWH	Lower‐dose LMWH	Clinical deterioration (time frame: 10 days) Combined outcome that includes number of patients who suffer any of the following: death, ICU admission, mechanical ventilatory support, progression to moderate or severe ARDS (according to Berlin criteria) or arterial or venous thrombosis	164	31 July 2021
NCT04623177	Spain	Prospective cohort	Higher‐dose LMWH	Lower‐dose LMWH; no anticoagulation	ICU mortality rate (time frame: from admission to ICU discharge, an average of 1 month)	950	30 September 2020
NCT04640181	USA	RCT	Rivaroxaban at low, intermediate or therapeutic dose	Enoxaparin at low, intermediate or therapeutic dose	Death or 30‐day all‐cause mortality (time frame: 30 days) Mechanical ventilation, intubation (time frame: 30 days) Transfer to an ICU setting (time frame: 30 days)	150	31 July 2021
NCT04646655	Italy	RCT	Higher‐dose enoxaparin	Lower‐dose enoxaparin	Mortality rate (time frame: 30 days from enrolment ) Progression of respiratory failure (time frame: 30 days from enrolment) Progression of respiratory failure (time frame: 30 days from enrolment) Progression of respiratory failure (time frame: 30 days from enrolment) Number of major bleeding episodes (time frame: up to 6 months from randomisation)	300	31 July 2021
NCT04655586	USA	RCT	Higher‐dose heparin	Lower‐dose heparin	Change in D‐dimer level from baseline to day 8, or day of discharge if prior to day 8 Number of major or non‐major clinically relevant bleeding events within 8 days of randomisation Time to recovery within 30 days of randomisation	100	31 May 2021
NCT04723563	USA	RCT	Nebulised heparin	Placebo	Need for mechanical ventilation at day 28	50	29 May 2021
NCT04730856	Spain	RCT	Higher‐dose heparin	Lower‐dose heparin	Reduction of suspicion of systemic thrombotic symptomatic events (time frame: 30 days) Use of mechanical ventilation (time frame: 30 days) Progression on the WHO Progression Scale during follow‐up (time frame: 30 days) Overall survival at 30 days (time frame: 30 days) Length of hospital stay (days) (time frame: 30 days) Length of ICU stay (days) (time frame: 30 days)	600	31 July 2021
NCT04743011	Brazil	RCT	Nebulised heparin	Placebo	Change in aPTT > 1.5 (time frame: immediately or up to 8 days after starting treatment) Viral load in nasal swab RT‐PCR (time frame: immediately or up to 8 days after starting treatment)	50	31 December 2021
NCT04745442	Spain	RCT	Heparin	No anticoagulant	Combined variable: mortality or worsening rate with need for non‐invasive mechanical ventilation or with need for invasive mechanical ventilation (time frame: at day 31 after randomisation or hospital discharge (whichever occurs first)	48	15 January 2021
PACTR202007606032743	Egypt	RCT	Nebulised heparin	No anticoagulant	The average daily ratio of partial pressure of oxygen to FiO2 (PaO2/FiO2) while the patient is on room air for 7 days	100	22 February 2021
RBR‐7y8j2bs	Brazil	RCT	Nebulised heparin	Placebo	Efficacy: relative to the proposed treatment, through the analysis of the viral load of the SARS‐CoV‐2 virus in the participants treated by the sequential evaluation of the viral load in RT‐PCR of nasal swab. Safety: related to the use of inhalational high‐molecular‐weight heparin in patients with SARS‐CoV‐2 through the assessment of haemorrhagic events of any nature, alteration of the coagulogram that indicates an increase in aPTT > 1.5 and HIT	40	11 October 2021
Sholzberg 2021a	Canada	RCT	Higher‐dose heparinoids	Lower‐dose heparinoids	Composite outcome of ICU admission (yes/no), non‐invasive positive pressure ventilation (yes/no), invasive mechanical ventilation (yes/no), or all‐cause death (yes/no) up to 28 days	462	April 2022
Vanassche 2020	Belgium	RCT	LMWH	DOAC plus aprotinin	The overall objective of the study is to evaluate the clinical efficacy and safety of different investigational therapeutics relative to standard care in patients hospitalised with COVID‐19	210	18 August 2020
Van Haren 2020	Argentina	RCT	Nebulised heparin	No anticoagulant	Intubation rate (time frame: day 28) Proportion of patients requiring invasive mechanical ventilation	712	1 June 2021
Wilkinson 2020	UK	RCT	Anticoagulants (no details)	NA	Time to clinical improvement of at least 2 points (from randomisation) on a 9‐point category ordinal scale, live discharge from the hospital, or considered fit for discharge (a score of 0, 1, or 2 on the ordinal scale), whichever comes first, by day 29 (this will also define the 'responder' for the response rate analyses)	1800	04 September 2021
Total number of studies	Argentina: 2 Australia: 1 Austria: 1 Belgium: 1 Brazil: 6 Canada: 1 China: 3 Egypt: 2 France: 3 Germany: 3 India: 4 Iran: 1 Ireland: 2 Italy: 6 Mexico: 1 Qatar: 1 Spain: 6 Switzerland: 2 UK: 3 USA: 13	Prospective cohort: 2 RCT: 60			35 studies considered mortality 26 studies considered additional respiratory support	35,470 participants (120 from NRS; 35,350 from RCTs)	58 studies to December 2021 Four studies to July 2022
aPTT: activated partial thromboplastin time; ARDS: acute respiratory distress syndrome; BARC: Bleeding Academic Research Consortium; CNS: central nervous system; DOACs: direct oral anticoagulants; DVT: deep vein thrombosis; ECMO: extracorporeal membrane oxygenation; FiO₂: fraction of inspired oxygen; HIT: heparin‐induced thrombocytopenia; ICU: intensive care unit; ISTH: International Society on Thrombosis and Haemostasis; LMWH: low‐molecular‐weight heparin; NA: not available; NRS: non‐randomised studies; PaO₂: arterial oxygen pressure;PE: pulmonary embolism; RCT: randomised controlled trial; RT‐PCR: reverse transcription polymerase chain reaction; SOFA: sequential organ failure assessment; UFH: unfractionated heparin; VKA: vitamin K antagonist; VTE: venous thromboembolism; WHO: World Health Organization

Table 3. Summary of characteristics of ongoing studies

Table 4. ROBINS‐I assessments: anticoagulants (all types) versus no treatment for people hospitalised with COVID‐19 (all‐cause mortality)

Study	Bias due to confounding	Bias in selection of participants into the study	Bias in classification of interventions	Bias due to deviations from the intended intervention	Bias due to missing data	Bias in measurement of outcomes	Bias in selection of the reported result	Overall risk of bias
Albani 2020	Serious risk	Serious risk	Low risk	Low risk	Low risk	Low risk	Low risk	Serious risk
Judgement	One or more prognostic variables are likely to be unbalanced between the compared groups. To minimise the impact of the absence of randomisation, an adjusted analysis with propensity scores was performed considering age, sex, disease severity, admission to ICU and COVID‐19 treatment. However, the essential confounding factors: 'participants already using anticoagulants', 'participants who underwent surgery during the hospitalisation', 'active cancer treatment', 'concomitant antiplatelet use' and 'history of venous thromboembolism' were not considered.	Participants included in both groups were selected from a single hospital, and the first dose of anticoagulant was administered between 0 and 3 days after hospital admission. The start of follow‐up and start of intervention possibly did not coincide for most participants, and adjustment techniques to correct the presence of selection bias were not used. It is not clear how prevalent use of anticoagulation was handled.	The intervention groups were clearly defined and recorded at the start of the intervention. Intervention status was probably not affected by knowledge of the outcome or the risk of the outcome.	No deviations from the intended intervention were reported in the study, and if any deviation occurred from usual practice, it was unlikely to impact on the outcome.	There were missing outcome data for 27 participants (1.9% of the total) and balanced between the groups. These missing data possibly could not cause an important impact on the estimate.	It is unlikely that the outcome assessment (objective outcome) was influenced by the knowledge of the intervention received by the study participants.	The study protocol was not identified but all reported results corresponded to the intended outcome.	The study has some important problems.
Rentsch 2020	Moderate risk	Low risk	Low risk	Low risk	Low risk	Low risk	Low risk	Moderate risk
Judgement	One or more prognostic variables are likely to be unbalanced between the compared groups. Essential characteristics, such as participants who underwent surgery during the hospitalisation, and history of venous thromboembolism, were not considered. However, an appropriate analysis method to control for measured confounders was used (inverse probability of treatment weighting), and all the important confounding domains for this study were probably controlled.	Participants included in both groups were selected from a nationwide cohort of patients receiving care in the Department of Veterans Affairs in the USA, and selection may have not been related to intervention and outcome. The start of follow‐up and start of intervention coincided for most participants (the first 24 h of hospitalisation).	The intervention groups were clearly defined and recorded at the start of the intervention. Intervention status was probably not affected by knowledge of the outcome or the risk of the outcome.	No deviations from the intended intervention were reported in the study, and if any deviation occurred from usual practice, it was unlikely to impact on the outcome.	No missing data were reported for the outcome.	It is unlikely that the outcome assessment (objective outcome) was influenced by the knowledge of the intervention received by the study participants.	The study protocol was not identified but all reported results corresponded to the intended outcome.	The study is sound for a non‐randomised study with regard to this domain but cannot be considered comparable to a well‐ performed randomised trial.
Santoro 2020	Critical risk	No information	Serious risk	Low risk	Low risk	Low risk	Serious risk	Critical risk
Judgement	One or more prognostic variables are likely to be unbalanced between the compared groups. Essential characteristics, such as participants who underwent surgery during the hospitalisation, and antiplatelet use were not considered. The Cox's multivariable regression analysis was performed to define independent risk factors for the mortality outcome, but only for participants with respiratory failure.	Insufficient information to judge. There was insufficient information if the start of follow‐up and the start of intervention coincided for most participants.	The intervention groups were not clearly defined and recorded at the start of the intervention. Information about frequency and dose was not provided.	No deviations from the intended intervention were reported in the study, and if any deviation occurred from usual practice, it was unlikely to impact on the outcome.	No missing data were reported for the outcome.	It is unlikely that the outcome assessment (objective outcome) was influenced by the knowledge of the intervention received by the study participants.	The study protocol was available, but it is not possible to exclude bias in selection of reported effect estimate, based on the results, from multiple outcome measurements within the outcome domain (mortality), and multiple effect estimates for different subgroups were provided, omitting varying proportions of the original cohort.	The study is too problematic to provide useful evidence.
COVID‐19: coronavirus disease 2019; ICU: intensive care unit

Table 4. ROBINS‐I assessments: anticoagulants (all types) versus no treatment for people hospitalised with COVID‐19 (all‐cause mortality)

Table 5. ROBINS‐I assessments: anticoagulants (all types) versus no treatment for people hospitalised with COVID‐19 (deep vein thrombosis)

Study	Bias due to confounding	Bias in selection of participants into the study	Bias in classification of interventions	Bias due to deviations from the intended intervention	Bias due to missing data	Bias in measurement of outcomes	Bias in selection of the reported result	Overall risk of bias
Albani 2020	Critical risk	Serious risk	Low risk	Low risk	Low risk	Low risk	Low risk	Critical risk
Judgement	One or more prognostic variables are likely to be unbalanced between the compared groups. Essential characteristics, such as participants who underwent surgery during the hospitalisation, concomitant antiplatelet use, and history of venous thromboembolism, were not considered. The outcome was reported without any adjustment.	Participants included in both groups were selected from a single hospital, and the first dose of anticoagulant was administered between 0 and 3 days after hospital admission. The start of follow‐up and start of intervention possibly did not coincide for most participants, and adjustment techniques to correct the presence of selection bias were not used.	The intervention groups were clearly defined and recorded at the start of the intervention. Intervention status was probably not affected by knowledge of the outcome or the risk of the outcome.	No deviations from the intended intervention were reported in the study, and if any deviation occurred from usual practice, it was unlikely to impact on the outcome.	There were missing outcome data for 27 participants (1.9% of the total), balanced between the groups. These missing data would probably not have an important impact on the estimate.	It is unlikely that the outcome assessment (objective outcome) was influenced by the knowledge of the intervention received by the study participants.	The study protocol was not identified but all reported results corresponded to the intended outcome.	The study is too problematic to provide useful evidence.

Table 5. ROBINS‐I assessments: anticoagulants (all types) versus no treatment for people hospitalised with COVID‐19 (deep vein thrombosis)

Table 6. ROBINS‐I assessments: anticoagulants (all types) versus no treatment for people hospitalised with COVID‐19 (pulmonary embolism)

Study	Bias due to confounding	Bias in selection of participants into the study	Bias in classification of interventions	Bias due to deviations from the intended intervention	Bias due to missing data	Bias in measurement of outcomes	Bias in selection of the reported result	Overall risk of bias
Albani 2020	Critical risk	Serious risk	Low risk	Low risk	Low risk	Low risk	Low risk	Critical risk
Judgement	One or more prognostic variables are likely to be unbalanced between the compared groups. Essential characteristics, such as participants who underwent surgery during the hospitalisation, concomitant antiplatelet use, and history of venous thromboembolism, were not considered. The outcome was reported without any adjustment.	Participants included in both groups were selected from a single hospital, and the first dose of anticoagulant was administered between 0 and 3 days after hospital admission. The start of follow‐up and start of intervention possibly did not coincide for most participants, and adjustment techniques to correct the presence of selection bias were not used.	The intervention groups were clearly defined and recorded at the start of the intervention. Intervention status was probably not affected by knowledge of the outcome or the risk of the outcome.	No deviations from the intended intervention were reported in the study, and if any deviation occurred from usual practice, it was unlikely to impact on the outcome.	There were missing outcome data for 27 participants (1.9% of the total), balanced between the groups. These missing data would probably not have an important impact on the estimate.	It is unlikely that the outcome assessment (objective outcome) was influenced by the knowledge of the intervention received by the study participants.	The study protocol was not identified but all reported results corresponded to the intended outcome.	The study is too problematic to provide useful evidence.

Table 6. ROBINS‐I assessments: anticoagulants (all types) versus no treatment for people hospitalised with COVID‐19 (pulmonary embolism)

Table 7. ROBINS‐I assessments: anticoagulants (all types) versus no treatment for people hospitalised with COVID‐19 (major bleeding)

Study	Bias due to confounding	Bias in selection of participants into the study	Bias in classification of interventions	Bias due to deviations from the intended intervention	Bias due to missing data	Bias in measurement of outcomes	Bias in selection of the reported result	Overall risk of bias
Albani 2020	Critical risk	Serious risk	Low risk	Low risk	Low risk	Low risk	Low risk	Critical risk
Judgement	One or more prognostic variables are likely to be unbalanced between the compared groups. Essential characteristics, such as participants who underwent surgery during the hospitalisation, concomitant antiplatelet use, and history of venous thromboembolism, were not considered. The outcome was reported without any adjustment.	Participants included in both groups were selected from a single hospital, and the first dose of anticoagulant was administered between 0 and 3 days after hospital admission. The start of follow‐up and start of intervention possibly did not coincide for most participants, and adjustment techniques to correct the presence of selection bias were not used.	The intervention groups were clearly defined and recorded at the start of the intervention. Intervention status was probably not affected by knowledge of the outcome or the risk of the outcome.	No deviations from the intended intervention were reported in the study, and if any deviation occurred from usual practice, it was unlikely to impact on the outcome.	There were missing outcome data for 27 participants (1.9% of the total) and balanced between the groups. These missing data would probably not have an important impact on the estimate.	It is unlikely that the outcome assessment (objective outcome) was influenced by the knowledge of the intervention received by the study participants.	The study protocol was not identified but all reported results corresponded to the intended outcome.	The study is too problematic to provide useful evidence.
Santoro 2020	Critical risk	No information	Serious risk	Low risk	Low risk	Low risk	Low risk	Critical risk
Judgement	One or more prognostic variables are likely to be unbalanced between the compared groups. Essential characteristics, such as participants who underwent surgery during the hospitalisation, and antiplatelet use were not considered. The Cox's multivariable regression analysis was performed to define independent risk factors only for the mortality outcome.	Insufficient information to judge. There was insufficient information if the start of follow‐up and the start of intervention coincided for most participants.	The intervention groups were not clearly defined and recorded at the start of the intervention. Information about frequency and dose was not provided.	No deviations from the intended intervention were reported in the study, and if any deviation occurred from usual practice, it was unlikely to impact on the outcome.	No missing data were reported for the outcome.	It is unlikely that the outcome assessment (objective outcome) was influenced by the knowledge of the intervention received by the study participants.	The study protocol was available, and the reported results corresponded to the intended outcome.	The study is too problematic to provide useful evidence.

Table 7. ROBINS‐I assessments: anticoagulants (all types) versus no treatment for people hospitalised with COVID‐19 (major bleeding)

Table 8. ROBINS‐I assessments: anticoagulants (all types) versus no treatment for people hospitalised with COVID‐19 (adverse events (stroke))

Study	Bias due to confounding	Bias in selection of participants into the study	Bias in classification of interventions	Bias due to deviations from the intended intervention	Bias due to missing data	Bias in measurement of outcomes	Bias in selection of the reported result	Overall risk of bias
Albani 2020	Critical risk	Serious risk	Low risk	Low risk	Low risk	Low risk	Low risk	Critical risk
Judgement	One or more prognostic variables are likely to be unbalanced between the compared groups. Essential characteristics, such as participants who underwent surgery during the hospitalisation, concomitant antiplatelet use, and history of venous thromboembolism, were not considered. The outcome was reported without any adjustment.	Participants included in both groups were selected from a single hospital, and the first dose of anticoagulant was administered between 0 and 3 days after hospital admission. The start of follow‐up and start of intervention possibly did not coincide for most participants, and adjustment techniques to correct the presence of selection bias were not used.	The intervention groups were clearly defined and recorded at the start of the intervention. Intervention status was probably not affected by knowledge of the outcome or the risk of the outcome.	No deviations from the intended intervention were reported in the study, and if any deviation occurred from usual practice, it was unlikely to impact on the outcome.	There were missing outcome data for 27 participants (1.9% of the total), balanced between the groups. These missing data would probably not have an important impact on the estimate.	It is unlikely that the outcome assessment (objective outcome) was influenced by the knowledge of the intervention received by the study participants.	The study protocol was not identified but all reported results corresponded to the intended outcome.	The study is too problematic to provide useful evidence.

Table 8. ROBINS‐I assessments: anticoagulants (all types) versus no treatment for people hospitalised with COVID‐19 (adverse events (stroke))

Table 9. ROBINS‐I assessments: anticoagulants (all types) versus no treatment for people hospitalised with COVID‐19 (adverse events (myocardial infarction))

Study	Bias due to confounding	Bias in selection of participants into the study	Bias in classification of interventions	Bias due to deviations from the intended intervention	Bias due to missing data	Bias in measurement of outcomes	Bias in selection of the reported result	Overall risk of bias
Albani 2020	Critical risk	Serious risk	Low risk	Low risk	Low risk	Low risk	Low risk	Critical risk
Judgement	One or more prognostic variables are likely to be unbalanced between the compared groups. Essential characteristics, such as participants who underwent surgery during the hospitalisation, concomitant antiplatelet use, and history of venous thromboembolism, were not considered. The outcome was reported without any adjustment.	Participants included in both groups were selected from a single hospital, and the first dose of anticoagulant was administered between 0 and 3 days after hospital admission. The start of follow‐up and start of intervention possibly did not coincide for most participants, and adjustment techniques to correct the presence of selection bias were not used.	The intervention groups were clearly defined and recorded at the start of the intervention. Intervention status was probably not affected by knowledge of the outcome or the risk of the outcome.	No deviations from the intended intervention were reported in the study, and if any deviation occurred from usual practice, it was unlikely to impact on the outcome.	There were missing outcome data for 27 participants (1.9% of the total), balanced between the groups. These missing data would probably not have an important impact on the estimate.	It is unlikely that the outcome assessment (objective outcome) was influenced by the knowledge of the intervention received by the study participants.	The study protocol was not identified but all reported results corresponded to the intended outcome.	The study is too problematic to provide useful evidence.

Table 9. ROBINS‐I assessments: anticoagulants (all types) versus no treatment for people hospitalised with COVID‐19 (adverse events (myocardial infarction))

Table 10. ROBINS‐I assessments: anticoagulants (all types) versus no treatment for people hospitalised with COVID‐19 (hospitalisation)

Study	Bias due to confounding	Bias in selection of participants into the study	Bias in classification of interventions	Bias due to deviations from the intended intervention	Bias due to missing data	Bias in measurement of outcomes	Bias in selection of the reported result	Overall risk of bias
Albani 2020	Serious risk	Serious risk	Low risk	Low risk	Low risk	Low risk	Low risk	Serious risk
Judgement	One or more prognostic variables are likely to be unbalanced between the compared groups. Essential characteristics, such as participants who underwent surgery during the hospitalisation, concomitant antiplatelet use, and history of venous thromboembolism, were not considered.	Participants included in both groups were selected from a single hospital, and the first dose of anticoagulant was administered between 0 and 3 days after hospital admission. The start of follow‐up and start of intervention possibly did not coincide for most participants, and adjustment techniques to correct the presence of selection bias were not used.	The intervention groups were clearly defined and recorded at the start of the intervention. Intervention status was probably not affected by knowledge of the outcome or the risk of the outcome.	No deviations from the intended intervention were reported in the study, and if any deviation occurred from usual practice, it was unlikely to impact on the outcome.	There were missing outcome data for 27 participants (1.9% of the total) and balanced between the groups. These missing data would probably not have an important impact on the estimate.	It is unlikely that the outcome assessment (objective outcome) was influenced by the knowledge of the intervention received by the study participants.	The study protocol was not identified but all reported results corresponded to the intended outcome.	The study has some important problems.

Table 10. ROBINS‐I assessments: anticoagulants (all types) versus no treatment for people hospitalised with COVID‐19 (hospitalisation)

Comparison 1. Higher‐dose anticoagulants versus lower‐dose anticoagulants (short term)

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1.1 All‐cause mortality Show forest plot	4	4489	Risk Ratio (IV, Random, 95% CI)	1.03 [0.92, 1.16]

1.1.1 Moderate severity	2	2833	Risk Ratio (IV, Random, 95% CI)	1.11 [0.68, 1.81]
1.1.2 Critical ill	3	1656	Risk Ratio (IV, Random, 95% CI)	1.04 [0.91, 1.17]
1.2 All‐cause mortality ‐ trials at low risk of bias Show forest plot	2	1176	Risk Ratio (IV, Random, 95% CI)	1.16 [0.86, 1.57]

1.2.1 Moderate severity	1	614	Risk Ratio (IV, Random, 95% CI)	1.49 [0.90, 2.46]
1.2.2 Critically ill	1	562	Risk Ratio (IV, Random, 95% CI)	1.05 [0.87, 1.28]
1.3 Necessity for additional respiratory support Show forest plot	3	3407	Risk Ratio (IV, Random, 95% CI)	0.54 [0.12, 2.47]

1.3.1 Moderate severity	2	2845	Risk Ratio (IV, Random, 95% CI)	0.54 [0.12, 2.47]
1.3.2 Critically ill	1	562	Risk Ratio (IV, Random, 95% CI)	Not estimable
1.4 Necessity for additional respiratory support ‐ trials at low risk of bias Show forest plot	2	1176	Risk Ratio (IV, Random, 95% CI)	0.16 [0.02, 1.35]

1.4.1 Moderate severity	1	614	Risk Ratio (IV, Random, 95% CI)	0.16 [0.02, 1.35]
1.4.2 Critically ill	1	562	Risk Ratio (IV, Random, 95% CI)	Not estimable
1.5 Deep vein thrombosis Show forest plot	4	3422	Risk Ratio (IV, Random, 95% CI)	1.08 [0.57, 2.03]

1.5.1 Moderate severity	2	2840	Risk Ratio (IV, Random, 95% CI)	0.85 [0.38, 1.92]
1.5.2 Critically ill	2	582	Risk Ratio (IV, Random, 95% CI)	1.55 [0.56, 4.26]
1.6 Deep vein thrombosis ‐ trials at low risk of bias Show forest plot	2	1176	Risk Ratio (IV, Random, 95% CI)	1.21 [0.53, 2.79]

1.6.1 Moderate severity	1	614	Risk Ratio (IV, Random, 95% CI)	0.98 [0.29, 3.35]
1.6.2 Critically ill	1	562	Risk Ratio (IV, Random, 95% CI)	1.45 [0.47, 4.52]
1.7 Pulmonary embolism Show forest plot	4	4360	Risk Ratio (IV, Random, 95% CI)	0.46 [0.31, 0.70]

1.7.1 Moderate severity	2	2840	Risk Ratio (IV, Random, 95% CI)	0.49 [0.27, 0.88]
1.7.2 Critically ill	3	1520	Risk Ratio (IV, Random, 95% CI)	0.44 [0.25, 0.78]
1.8 Pulmonary embolism ‐ trial at low risk of bias Show forest plot	2	1176	Risk Ratio (IV, Random, 95% CI)	0.50 [0.23, 1.10]

1.8.1 Moderate severity	1	614	Risk Ratio (IV, Random, 95% CI)	0.53 [0.21, 1.31]
1.8.2 Critically ill	1	562	Risk Ratio (IV, Random, 95% CI)	0.41 [0.08, 2.12]
1.9 Major bleeding Show forest plot	4	4400	Risk Ratio (IV, Random, 95% CI)	1.78 [1.13, 2.80]

1.9.1 Moderate severity	2	2841	Risk Ratio (IV, Random, 95% CI)	2.25 [1.19, 4.27]
1.9.2 Critically ill	3	1559	Risk Ratio (IV, Random, 95% CI)	1.41 [0.75, 2.67]
1.10 Major bleeding ‐ trials at low risk of bias Show forest plot	2	1176	Risk Ratio (IV, Random, 95% CI)	2.13 [0.92, 4.90]

1.10.1 Moderate severity	1	614	Risk Ratio (IV, Random, 95% CI)	2.45 [0.78, 7.73]
1.10.2 Critically ill	1	562	Risk Ratio (IV, Random, 95% CI)	1.81 [0.54, 6.13]
1.11 Adverse events (minor bleeding) Show forest plot	3	1196	Risk Ratio (IV, Random, 95% CI)	3.28 [1.75, 6.14]

1.11.1 Moderate severity	1	614	Risk Ratio (IV, Random, 95% CI)	5.10 [1.98, 13.11]
1.11.2 Critically ill	2	582	Risk Ratio (IV, Random, 95% CI)	2.31 [1.00, 5.36]
1.12 Adverse events (minor bleeding) ‐ trials at low risk of bias Show forest plot	2	1176	Risk Ratio (IV, Random, 95% CI)	3.67 [1.82, 7.40]

1.12.1 Moderate severity	1	614	Risk Ratio (IV, Random, 95% CI)	5.10 [1.98, 13.11]
1.12.2 Critical ill	1	562	Risk Ratio (IV, Random, 95% CI)	2.49 [0.89, 6.97]
1.13 Adverse events (stroke) Show forest plot	3	4349	Risk Ratio (IV, Random, 95% CI)	0.91 [0.40, 2.03]

1.13.1 Moderate severity	2	2840	Risk Ratio (IV, Random, 95% CI)	0.88 [0.13, 5.97]
1.13.2 Critical ill	2	1509	Risk Ratio (IV, Random, 95% CI)	0.91 [0.37, 2.23]
1.14 Adverse events (stroke) ‐ trials at low risk of bias Show forest plot	2	1176	Risk Ratio (IV, Random, 95% CI)	1.62 [0.20, 13.13]

1.14.1 Moderate severity	1	614	Risk Ratio (IV, Random, 95% CI)	2.94 [0.12, 71.94]
1.14.2 Critical ill	1	562	Risk Ratio (IV, Random, 95% CI)	1.04 [0.07, 16.49]
1.15 Adverse events (major adverse limb event) Show forest plot	2	1176	Risk Ratio (IV, Random, 95% CI)	0.33 [0.01, 7.99]

1.15.1 Moderate severity	1	614	Risk Ratio (IV, Random, 95% CI)	0.33 [0.01, 7.99]
1.15.2 Critically ill	1	562	Risk Ratio (IV, Random, 95% CI)	Not estimable
1.16 Adverse events (myocardial infarction) Show forest plot	3	4349	Risk Ratio (IV, Random, 95% CI)	0.86 [0.48, 1.55]

1.16.1 Moderate severity	2	2840	Risk Ratio (IV, Random, 95% CI)	0.91 [0.45, 1.85]
1.16.2 Critically ill	2	1509	Risk Ratio (IV, Random, 95% CI)	0.76 [0.27, 2.17]
1.17 Adverse events (myocardial infarction) ‐ trials at low risk of bias Show forest plot	2	1176	Risk Ratio (IV, Random, 95% CI)	0.91 [0.44, 1.91]

1.17.1 Moderate severity	1	614	Risk Ratio (IV, Random, 95% CI)	0.91 [0.44, 1.91]
1.17.2 Critically ill	1	562	Risk Ratio (IV, Random, 95% CI)	Not estimable
1.18 Adverse events (atrial fibrillation) Show forest plot	1	562	Risk Ratio (IV, Random, 95% CI)	0.35 [0.07, 1.70]

1.19 Adverse events (thrombocytopenia) Show forest plot	2	2789	Risk Ratio (IV, Random, 95% CI)	0.94 [0.71, 1.24]

1.19.1 Moderate severity	1	2227	Risk Ratio (IV, Random, 95% CI)	Not estimable
1.19.2 Critically ill	1	562	Risk Ratio (IV, Random, 95% CI)	0.94 [0.71, 1.24]
1.20 Hospitalisation time Show forest plot	2	634	Mean Difference (IV, Random, 95% CI)	0.28 [‐0.87, 1.44]

1.20.1 Moderate severity	1	614	Mean Difference (IV, Random, 95% CI)	0.30 [‐0.86, 1.46]
1.20.2 Critically ill	1	20	Mean Difference (IV, Random, 95% CI)	‐1.00 [‐11.58, 9.58]
1.21 Hospitalisation time ‐ trials at low risk of bias Show forest plot	1	614	Mean Difference (IV, Random, 95% CI)	0.30 [‐0.86, 1.46]

Comparison 1. Higher‐dose anticoagulants versus lower‐dose anticoagulants (short term)

Comparison 2. Higher‐dose anticoagulants versus lower‐dose anticoagulants (long term)

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
2.1 All‐cause mortality Show forest plot	1	590	Risk Ratio (IV, Random, 95% CI)	1.07 [0.89, 1.28]

2.2 Necessity for additional respiratory support Show forest plot	1	590	Risk Ratio (IV, Random, 95% CI)	Not estimable

2.3 Deep vein thrombosis Show forest plot	1	590	Risk Ratio (IV, Random, 95% CI)	1.39 [0.45, 4.33]

2.4 Pulmonary embolism Show forest plot	1	590	Risk Ratio (IV, Random, 95% CI)	0.40 [0.08, 2.03]

2.5 Major bleeding Show forest plot	1	590	Risk Ratio (IV, Random, 95% CI)	1.74 [0.51, 5.87]

2.6 Adverse events (minor bleeding) Show forest plot	1	590	Risk Ratio (IV, Random, 95% CI)	2.32 [0.90, 5.95]

2.7 Adverse events (stroke) Show forest plot	1	590	Risk Ratio (IV, Random, 95% CI)	0.99 [0.06, 15.80]

2.8 Adverse events (acute peripheral arterial thrombosis) Show forest plot	1	590	Risk Ratio (IV, Random, 95% CI)	Not estimable

2.9 Adverse events (myocardial infarction) Show forest plot	1	590	Risk Ratio (IV, Random, 95% CI)	Not estimable

2.10 Adverse events (atrial fibrillation) Show forest plot	1	590	Risk Ratio (IV, Random, 95% CI)	0.50 [0.13, 1.97]

2.11 Adverse events (thrombocytopenia) Show forest plot	1	590	Risk Ratio (IV, Random, 95% CI)	12.91 [0.73, 228.18]

Comparison 2. Higher‐dose anticoagulants versus lower‐dose anticoagulants (long term)

Comparison 3. Anticoagulants versus no treatment (short term)

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
3.1 All‐cause mortality Show forest plot	3		Risk Ratio (IV, Random, 95% CI)	0.64 [0.55, 0.74]

3.2 Deep vein thrombosis Show forest plot	1	1403	Risk Ratio (IV, Random, 95% CI)	5.67 [1.30, 24.70]

3.3 Pulmonary embolism Show forest plot	1	1403	Risk Ratio (IV, Random, 95% CI)	24.19 [3.31, 176.53]

3.4 Major bleeding Show forest plot	2	7218	Risk Ratio (IV, Random, 95% CI)	1.19 [0.66, 2.12]

3.5 Adverse events (stroke) Show forest plot	1	1403	Risk Ratio (IV, Random, 95% CI)	1.13 [0.32, 4.00]

3.6 Adverse events (myocardial infarction) Show forest plot	1	1403	Risk Ratio (IV, Random, 95% CI)	15.88 [0.93, 270.48]

3.7 Hospitalisation time Show forest plot	1	1376	Mean Difference (IV, Random, 95% CI)	5.00 [4.47, 5.53]

Comparison 3. Anticoagulants versus no treatment (short term)