Heparin for the prevention of venous thromboembolism in acutely ill medical patients (excluding stroke and myocardial infarction)
Abstract
Background
Venous thromboembolic disease has been extensively studied in surgical patients. The benefit of thromboprophylaxis is now generally accepted, but it is medical patients who make up the greater proportion of the hospital population. Medical patients differ from surgical patients with regard to their health and the pathogenesis of thromboembolism and the impact that preventative measures can have. The extensive experience from thromboprophylaxis studies in surgical patients is therefore not necessarily applicable to non‐surgical patients. This is an update of a review first published in 2009.
Objectives
To determine the effectiveness and safety of heparin (unfractionated heparin or low molecular weight heparin) thromboprophylaxis in acutely ill medical patients admitted to hospital, excluding those admitted to hospital with an acute myocardial infarction or stroke (ischaemic or haemorrhagic) or those requiring admission to an intensive care unit (unless the study population can be clearly defined as acute medical and not post‐surgical).
Search methods
For this update the Cochrane Peripheral Vascular Diseases Group Trials Search Co‐ordinator searched the Specialised Register (last searched November 2013) and CENTRAL (2013, Issue 10).
Selection criteria
Randomised controlled trials comparing unfractionated heparin (UFH) or low molecular weight heparin (LMWH) with placebo or no treatment, or comparing UFH with LMWH.
Data collection and analysis
One review author identified possible trials and a second review author confirmed their eligibility for inclusion in the review. Two review authors extracted the data. Disagreements were resolved by discussion. We performed the meta‐analysis using a fixed‐effect model with the results expressed as odds ratios (ORs) with 95% confidence intervals (CIs).
Main results
Sixteen studies with a combined total of 34,369 participants with an acute medical illness were included in this review. We identified 10 studies comparing heparin with placebo or no treatment and six studies comparing LMWH to UFH. Just under half of the studies had an open‐label design, putting them at a risk of performance bias. Descriptions of random sequence generation and allocation concealment were missing in most of the studies. Heparin reduced the odds of deep vein thrombosis (DVT) (OR 0.41, 95% CI 0.25 to 0.67; P = 0.0004) . The estimated reductions in symptomatic non‐fatal pulmonary embolism (PE) (OR 0.46; 95% CI 0.20 to 1.07; P = 0.07), fatal PE (OR 0.71; 95% CI 0.43 to 1.15; P = 0.16) and in combined non‐fatal PE and fatal PE (OR 0.66, 95% CI 0.43 to 1.02; P = 0.06) associated with heparin were imprecise. Heparin resulted in an increase in major haemorrhage (OR 1.65, 95% CI 1.01 to 2.71; P = 0.05). There was no clear evidence that heparin had an effect on all‐cause mortality and thrombocytopaenia. Compared with UFH, LMWH reduced the risk of DVT (OR 0.77; 95% CI 0.62 to 0.96; P = 0.02) and major bleeding (OR 0.43; 95% CI 0.22 to 0.83; P = 0.01). There was no clear evidence that the effects of LMWH and UFH differed for the PE outcomes, all‐cause mortality and thrombocytopaenia.
Authors' conclusions
The data from this review describe a reduction in the risk of DVT in patients presenting with an acute medical illness who receive heparin thromboprophylaxis. This needs to be balanced against an increase in the risk of bleeding associated with thromboprophylaxis. The analysis favoured LMWH compared with UFH, with a reduced risk of both DVT and bleeding.
PICOs
Plain language summary
Heparin to prevent deep vein thrombosis or pulmonary embolism in acutely ill medical patients (excluding those with stroke or myocardial infarction)
Blood clots may form in the veins of patients who are admitted to hospital suffering from an acute medical illness. These types of blood clots are referred to as deep vein thromboses (DVT) and they may break free from the blood vessel wall and travel to the lungs and cause death, at which point they are referred to as a pulmonary embolism (PE). These types of blood clots and their prevention have been thoroughly studied in surgical patients but not as much in non‐surgical, medical patients, who make up a greater proportion of hospital patients. Medical patients differ from surgical patients with regard to their health, the progression of clots and the impact that preventative measures can have. The extensive experience from clot prevention studies in surgical patients is therefore not necessarily applicable to non‐surgical patients.
Heparin is a blood thinning drug, which has been shown to reduce the occurrence of blood clots in patients after they have had surgery. Heparin exists in two forms, the original unfractionated (UFH) form and a newer form called low molecular weight heparin (LMWH). The aim of the current review is to determine the effectiveness and safety of heparin (UFH or LMWH) to prevent DVTs and PEs in non‐surgical, medical patients admitted to hospital, excluding those admitted to hospital with a heart attack or stroke or those requiring admission to an intensive care unit. The outcomes investigated in this review were DVT, PE that did not cause death, PE that resulted in death, combined non‐fatal and fatal PE, all‐cause death, bleeding complications and thrombocytopaenia, which is a condition that can be caused by heparin and results in decreased platelets in the blood.
This review of 16 trials in 34,369 non‐surgical patients who suffered an acute medical illness found that heparin reduced the number of patients suffering DVTs but also increased the risk of bleeding complications when compared to participants that received a placebo or no medication. We had some concerns over how reliable the results were from the unblinded studies, which made up just under half of the studies. Also, most of the studies were lacking explanations of how the allocation of the treatments was performed. The lower risk of PEs (when combining those that caused death and those that did not) with heparin could have been a chance effect. There was no clear evidence of a difference in the rate of death or thrombocytopaenia. The review also found that patients who were given LMWH developed fewer DVTs and fewer bleeding complications compared with those given UFH, leading to the conclusion that LMWH is more effective and carries a lower risk of adverse events in preventing blood clots than with UFH. There was no clear evidence of differences between LMWH and UFH for PE, death or thrombocytopaenia.
Authors' conclusions
Background
Description of the condition
Venous thromboembolism (VTE) is a major health problem, the significance and seriousness of which is often not fully appreciated. VTE is one of the most important preventable causes of morbidity and mortality in hospital patients, having an annual incidence of one per 1000 individuals (Dahlback 2008). Most clinicians, whatever their specialty, will experience patients with this condition.
Description of the intervention
During the past 40 years numerous studies have shown that UFH and LMWH are effective and safe for the prevention of VTE in surgical patients (Geerts 2008). This has led to the widespread use of these agents for thromboprophylaxis in surgical patients with a resultant reduction in the incidence of fatal PE (Cohen 1996). During the same period of time, there have been fewer trials investigating the benefits and risks of thromboprophylaxis in medical patients. Most of these studies have concentrated on specific conditions such as myocardial infarction (MI) (Collins 1996) and ischaemic stroke (Gubitz 2004).
How the intervention might work
Thrombosis results from a disturbance in the balance between pro‐thrombotic and antithrombotic forces that exist within the blood stream. The pro‐thrombotic forces include platelets and the formation of fibrin. There are two phases of normal haemostasis; the first phase is platelet activation and adhesion, and the forming of a 'platelet plug'. In parallel, the second phase results in the activation of a series of procoagulant clotting factors, which generate a burst of thrombin, formation of fibrin and stabilisation of the platelet plug, resulting in a thrombus. When this process becomes unregulated venous thrombosis may occur in any vein in the body, although it usually occurs in the deep veins of the lower limbs. People admitted to hospital with an acute medical illness appear to be particularly at risk of suffering from thrombosis of these lower limb deep veins.
Heparin (UFH and LMWH) binds to a naturally occurring anticoagulant, antithrombin (AT), via a pentasaccharide sequence. The heparin‐AT complex inactivates thrombin (factor IIa) as well as coagulation factors Xa, IXa and XIa. As a direct result of inactivating or reducing the generation of thrombin, fibrin formation is inhibited, as is thrombin directed activation of coagulation factors V, VIII and XI. In addition, thrombin mediated platelet activation is attenuated (Garcia 2012). Therefore, heparin affects both the primary and secondary phases of thrombus formation and this is the rationale behind its use as a thromboprophylactic agent.
Why it is important to do this review
Unlike their surgical counterparts, it wasn't until the late 1990s and early part of the 21st century that physicians caring for general medical patients had convincing evidence of the efficacy and safety of thromboprophylaxis for their patients upon which to base prescribing decisions (Geerts 2008). We believe that this has resulted in the underuse of thromboprophylaxis in the medical setting (Alikhan 2001), and this continues to be a problem (Cohen 2008). The under use may in part explain the high incidence rate of fatal PE in this patient group (Alikhan 2004; Sandler 1989). This is an update of a review first published in 2009 (Alikhan 2009).
Objectives
The aim of this review is to determine the effectiveness and safety of heparin (unfractionated heparin (UFH) or low molecular weight heparin (LMWH)) thromboprophylaxis in acutely ill medical patients admitted to hospital, excluding those admitted to hospital with an acute myocardial infarction or stroke (ischaemic or haemorrhagic) or those requiring admission to an intensive care unit (unless the study population can be clearly defined as acute medical and not post‐surgical).
Methods
Criteria for considering studies for this review
Types of studies
-
Randomised controlled trials comparing UFH with placebo or no treatment
-
Randomised controlled trials comparing LMWH with placebo or no treatment
-
Randomised controlled trials comparing UFH with LMWH
Types of participants
People over the age of 18 years admitted to hospital with an acute medical illness, for example:
-
heart failure;
-
respiratory failure;
-
cancer*;
-
acute infection;
-
episode of inflammatory bowel disease;
-
acute rheumatic disorder.
*Studies that primarily involve cancer patients not in an acute medical setting are excluded, such as receiving chemotherapy in tandem with thromboprophylaxis. This is the subject of another Cochrane review (Di Nisio 2012).
Studies involving participants with only myocardial infarction or stroke are excluded because the risk of VTE differs in this population, and therefore the need for thromboprophylaxis. Collins 1996 and Geerts 2001 address these patient populations in reference to VTE and thromboprophylaxis.
We included studies from intensive care units if they clearly defined their population as medical and not post‐surgical. Studies were excluded if they did not clearly define their intensive care unit population as suffering from acute medical illness.
Types of interventions
Participants randomised to receive UFH, LMWH, placebo or no treatment. Fondaparinux and other pentasaccharide agents were excluded from this review as they are addressed in another Cochrane review that is currently in progress (Song 2011).
Types of outcome measures
Primary outcomes
Efficacy
-
Asymptomatic or symptomatic deep vein thrombosis (DVT) of the lower limbs detected by fibrinogen uptake test, ultrasound, venography or plethysmography
-
Symptomatic non‐fatal PE detected by ventilation perfusion scan, computed tomography, pulmonary angiography, or confirmed at autopsy
Safety
-
Major haemorrhage. As it was not possible to obtain a standardised definition of major bleeding, the study author's definition, where given, was used. See Characteristics of included studies for individual study definitions
Secondary outcomes
Efficacy
-
All‐cause mortality
-
Fatal PE
-
Combined clinically symptomatic non‐fatal PE and fatal PE
Safety
-
Minor haemorrhage. As it was not possible to obtain a standardised definition of minor bleeding, the study author's definition, where given, was used. See Characteristics of included studies for individual study definitions
-
Thrombocytopaenia, as defined by individual study authors
Search methods for identification of studies
There was no restriction on the language of publication.
Electronic searches
For this update the Cochrane Peripheral Vascular Diseases Group Trials Search Co‐ordinator (TSC) searched the Specialised Register (last searched November 2013) and the Cochrane Central Register of Controlled Trials (CENTRAL) (2013, Issue 10), part of The Cochrane Library (www.thecochranelibrary.com). See Appendix 1 for details of the search strategy used to search CENTRAL. The Specialised Register is maintained by the TSC and is constructed from weekly electronic searches of MEDLINE, EMBASE, CINAHL and AMED, and through handsearching relevant journals. The full list of the databases, journals and conference proceedings which have been searched, as well as the search strategies used are described in the Specialised Register section of the Cochrane Peripheral Vascular Diseases Group module in The Cochrane Library (www.thecochranelibrary.com).
Searching other resources
In the previous version of this review we consulted with colleagues and investigators as well as the manufacturers of the various LMWH preparations to identify unpublished or missed studies. This was not done for the current version.
Data collection and analysis
Selection of studies
For this update one review author (RA) identified possible trials, and the trial reports were assessed independently by another review author (RB) to confirm eligibility for inclusion in the review. In the previous version of this review study selection was performed by RA and AC.
Data extraction and management
For this update, RA and RB individually extracted the data using the following endpoints: DVT of the lower limbs; symptomatic non‐fatal PE; fatal PE; combined symptomatic non‐fatal and fatal PE; all‐cause mortality; major and minor bleeding (as defined by individual authors); and thrombocytopaenia (reduced numbers of platelets). We recorded all information collected on data extraction forms. We resolved disagreements by discussion. In the previous version of this review data extraction was performed by RA and AC.
Assessment of risk of bias in included studies
The methodological quality of included trials was assessed independently by RA and RB using the 'Risk of bias' tool from The Cochrane Collaboration (Higgins 2011). The following domains were assessed: selection bias (random sequence generation, allocation concealment), performance bias (blinding of participants and personnel, and blinding of outcome assessment), attrition bias (incomplete outcome data), reporting bias (selective reporting), and other bias. The domains were classified as low risk of bias, high risk of bias, or unclear risk of bias according to the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). Disagreements were resolved by discussion between the two review authors assessing bias.
Measures of treatment effect
We performed statistical analyses according to the statistical guidelines for authors recommended by the Cochrane Peripheral Vascular Diseases Group.
We divided the studies into two groups and analysed them separately:
-
heparin (UFH or LMWH) prophylaxis versus placebo or no treatment;
-
LMWH versus UFH.
For each of the two groups we pooled data from each study on DVT, non‐fatal PE, fatal PE, combined non‐fatal and fatal PE, all‐cause mortality, major bleeding, minor bleeding, and thrombocytopaenia in order to arrive at an overall estimate of efficacy and safety of heparin versus no treatment or placebo and LMWH versus UFH. We summarised the results of each trial on an intention‐to‐treat basis in 2 x 2 tables for each outcome measurement. Where possible, all randomised participants were included, even if the original trial authors excluded them. The results obtained from different methods were similar (risk ratios, Mantel‐Haenzel and Peto odds ratios), therefore the meta‐analysis was performed using odds ratios (ORs) with 95% confidence intervals (CIs).
Unit of analysis issues
The individual participant was the unit of analysis in all included studies.
Dealing with missing data
Analysis was performed on an intention‐to‐treat basis, including all participants randomised.
Assessment of heterogeneity
A test for heterogeneity examines the null hypothesis that all studies are evaluating the same effect. We obtained P values comparing the test statistic with a Chi2 distribution. To help readers assess the consistency of results of studies in a meta‐analysis, RevMan 5 software includes a method (I2 statistic) that describes the percentage of total variation across studies due to heterogeneity rather than by chance. A value of 0% indicates no observed heterogeneity, and larger values show increasing heterogeneity (Higgins 2003).
Assessment of reporting biases
To detect reporting bias we planned to construct funnel plots for meta‐analyses that included at least 10 studies, as funnel plots with less than 10 studies lack the power to distinguish chance from real asymmetry.
Data synthesis
An heterogeneity test was performed and we planned to use a random‐effects model if the test was positive (I2 > 50%); unless otherwise stated the meta‐analysis was performed using a fixed‐effect model.
Subgroup analysis and investigation of heterogeneity
Where data were available, subgroup analysis was performed to evaluate outcomes based on medical diagnosis at hospital admission.
Sensitivity analysis
Four studies included in the review had a loss of ≥ 20% of the study population between randomisation and evaluation for DVT or PE (CERTAIN 2010; CERTIFY 2010; Fraisse 2000; MEDENOX 1999). Although the majority of the missing participants were accounted for with suitable reasoning, it was still possible these losses could alter the outcome. Sensitivity analysis was performed by removing these studies from the DVT and PE outcomes and evaluating their affect on the meta‐analysis.
DVT diagnosis using a fibrinogen uptake test and plethysmography is often less accurate than by ultrasound or venography. In order to establish that the meta‐analysis results were not dependent on studies using these low‐accuracy tests, sensitivity analysis was conducted by removing three studies using a fibrinogen uptake test or plethysmography to detect DVT (Belch 1981; Dahan 1986; EMSG 1996) and the effect evaluated.
The safety data for the CERTAIN 2010 study was collected after a three month follow‐up period, which was different to the other studies that collected the efficacy and safety data during the same study time period. With this additional time allotted for these endpoints it was possible this study unduly altered the meta‐analysis, therefore sensitivity analysis was performed by removing this study from the meta‐analyses for major and minor bleeding, as well as thrombocytopaenia, and evaluating the effect.
In two studies (Belch 1981; Dahan 1986) major bleeding was not defined, and was only described as 'major bleeding'. In order to determine if these studies with inadequate definitions of major bleeding were not having an overt effect on the meta‐analysis they were removed in a sensitivity analysis to evaluate the effect.
Two studies (EMSG 1996; MEDENOX 1999) included a low‐dosage (20 mg) of the LMWH enoxaparin. For EMSG 1996 this was the only LMWH dosage, but for MEDENOX 1999 there was also a higher dosage of 40 mg. Sensitivity analysis was performed by removing the 20 mg dosages and the effect was evaluated.
Results
Description of studies
Results of the search
See Figure 1.
Study flow diagram.
Included studies
For this update there were an additional three studies included (CERTAIN 2010; CERTIFY 2010; LIFENOX 2011). For details of the included studies see Characteristics of included studies.
We included 16 studies (45 published articles) with 34,369 participants. Ten of these studies compared heparin prophylaxis with no treatment or placebo (Belch 1981; Bergmann 1996; Dahan 1986; Fraisse 2000; Gallus 1973; Gardlund 1996; Ibarra‐Perez 1988; LIFENOX 2011; MEDENOX 1999; PREVENT 2004) and six studies compared LMWH with UFH (CERTAIN 2010; CERTIFY 2010; EMSG 1996; Forette 1995; PRIME 1996; THE‐PRINCE 2003). Fifteen of the included studies were written in English, and one in French (Forette 1995) that we had translated.
Four trials took place in a single centre (Belch 1981; Dahan 1986; Gallus 1973; Ibarra‐Perez 1988). Nine trials were European multi‐centre trials: four in France (Bergmann 1996; EMSG 1996; Forette 1995; Fraisse 2000), one in Sweden (Gardlund 1996), three in Germany (CERTAIN 2010; CERTIFY 2010; THE‐PRINCE 2003), and one in Germany and Austria (PRIME 1996). One trial was performed in multiple centres across Europe and Canada (MEDENOX 1999) and two trials were truly multi‐national, one performed in centres across Europe, North and South America, Canada, North and Southern Africa, Israel, Lebanon and Australia (PREVENT 2004) and the other in China, India, Korea, Malaysia, Mexico, Philippines and Tunisia (LIFENOX 2011).
Heparin versus placebo or no treatment
Four trials compared UFH with no treatment (Belch 1981; Gallus 1973; Gardlund 1996; Ibarra‐Perez 1988). One study compared UFH 5000 IU twice daily (Gardlund 1996) and the other two studies compared UFH 5000 IU three times daily with no treatment (Belch 1981; Gallus 1973).
Six trials compared LMWH with placebo, three trials were with enoxaparin (Dahan 1986; LIFENOX 2011; MEDENOX 1999), two trials with nadroparin (Bergmann 1996; Fraisse 2000), and one trial with dalteparin (PREVENT 2004). One of the enoxaparin trials (MEDENOX 1999) compared two doses (20 mg and 40 mg) of enoxaparin with placebo. For assessment of major and minor bleeding the maximum licensed duration of treatment was used.
LMWH versus UFH
Six trials compared LMWH with UFH (CERTAIN 2010; CERTIFY 2010; EMSG 1996; Forette 1995; PRIME 1996; THE‐PRINCE 2003). Three of these trials compared enoxaparin with UFH (EMSG 1996; PRIME 1996; THE‐PRINCE 2003). Two trials compared enoxaparin 40 mg once daily with UFH 5000 IU three times daily (PRIME 1996; THE‐PRINCE 2003); one trial compared enoxaparin 20 mg once daily with UFH 5000 IU twice daily (EMSG 1996). One trial compared nadroparin 3075 antiXa units once daily with UFH 5000 to 7500 IU three times daily (Forette 1995). One trial compared certoparin 3000 antiXa units with UFH 5000 IU three times daily (CERTIFY 2010).
Excluded studies
For this update one additional study was excluded (EXCLAIM) and there were 10 (Aquino 1990; Cade 1982a; Cade 1982b; Halkin 1982; Harenberg 1990; HESIM; Manciet 1990; Mottier 1993; Poniewierski 1988; PROMPT) from the original review. The reasons for exclusion are detailed in Characteristics of excluded studies. In brief, one study did not strictly fit the criteria for a randomised control trial (Halkin 1982), six studies included patients meeting the exclusion criteria of MI (or acute coronary disease), recent surgery, stroke (or cerebrovascular disease) that could not be removed from the analysis (Cade 1982a; Cade 1982b; Harenberg 1990; HESIM; Mottier 1993; PROMPT), two studies included patients receiving orthopaedic rehabilitation (Aquino 1990; Manciet 1990), one study had all participants receiving the study medication before randomisation to continue the medication or move to placebo (EXCLAIM), and one study's method for identifying DVT was not considered sensitive enough (PROMPT).
Risk of bias in included studies
For details see Figure 2; Figure 3.
Methodological quality graph: review authors' judgements about each methodological quality item presented as percentages across all included studies.
Methodological quality summary: review authors' judgements about each methodological quality item for each included study.
Allocation
For random sequence generation five studies were considered low risk (CERTAIN 2010; CERTIFY 2010; EMSG 1996; Forette 1995; LIFENOX 2011) and the remaining 11 studies did not give enough information to determine how sequence generation was performed (Belch 1981; Bergmann 1996; Dahan 1986; Fraisse 2000; Gallus 1973; Gardlund 1996; Ibarra‐Perez 1988; MEDENOX 1999; PREVENT 2004; PRIME 1996; THE‐PRINCE 2003).
For allocation concealment, seven studies were considered to be low risk (CERTAIN 2010; CERTIFY 2010; EMSG 1996; Forette 1995; Gallus 1973; LIFENOX 2011; MEDENOX 1999) and the remaining nine studies did not give enough information to understand if allocation was properly concealed (Belch 1981; Bergmann 1996; Dahan 1986; Fraisse 2000; Gardlund 1996; Ibarra‐Perez 1988; PREVENT 2004; PRIME 1996; THE‐PRINCE 2003).
Blinding
Adequate blinding is important in reducing the chance of bias in the detection of DVT and PE. Nine trials used double blinding and were considered low risk for performance bias (Bergmann 1996; CERTIFY 2010; Dahan 1986; EMSG 1996; Fraisse 2000; LIFENOX 2011; MEDENOX 1999; PREVENT 2004; PRIME 1996). Seven trials were described as open or unblinded, and were therefore at high risk of performance bias (Belch 1981; CERTAIN 2010; Forette 1995; Gallus 1973; Gardlund 1996; Ibarra‐Perez 1988; THE‐PRINCE 2003).
In nine trials evaluation of the outcomes was undertaken by blinded assessors, and therefore had little risk of detection bias (Belch 1981; CERTAIN 2010; CERTIFY 2010; EMSG 1996; Fraisse 2000; Gardlund 1996; LIFENOX 2011; MEDENOX 1999; PREVENT 2004). Six trials did not indicate if the assessors were blinded and therefore had unclear detection bias (Bergmann 1996; Dahan 1986; Forette 1995; Gallus 1973; Ibarra‐Perez 1988; PRIME 1996). For THE‐PRINCE 2003 study the authors clearly stated that the efficacy endpoints were evaluated by blinded assessors and was therefore at low risk of detection bias, but the safety endpoints were not included in this description so they had an unclear risk of detection bias.
Incomplete outcome data
Twelve studies included all randomised participants in the analysis or clearly described how many participants were excluded from each group and why, and that the treatment groups remained similar after the exclusions (Belch 1981; Bergmann 1996; CERTIFY 2010; Dahan 1986; EMSG 1996; Forette 1995; Fraisse 2000; Gallus 1973; LIFENOX 2011; MEDENOX 1999; PREVENT 2004; PRIME 1996). These 12 studies were considered low risk for attrition bias. Two studies did not include enough information about incomplete outcome data to determine whether or not there was a risk of attrition bias (Gardlund 1996; Ibarra‐Perez 1988). In the CERTAIN 2010 study the efficacy endpoints had unclear risk as there were missing participants unaccounted for in the analysis, but there was a low risk of attrition bias for the safety endpoints as all participants were considered in this analysis. For THE‐PRINCE 2003 study the study authors described inequality between the study groups for the reasons why participants withdrew and were not included in the analysis, leading to high risk of attrition bias for these outcomes. However, the safety endpoints in this study were at a low risk of attrition bias because all participants were reported on.
Selective reporting
Fifteen of the 16 studies have low risk of reporting bias because they either followed the provided protocol or there was no evidence of selective outcome reporting. The Bergmann 1996 study was considered to have unclear risk of reporting bias because in the limited report available it was unclear whether the mortality reported was all‐cause or specifically due to thromboembolism.
Other potential sources of bias
Seven studies reported support by large pharmaceutical companies, which could have potentially introduced bias (CERTAIN 2010; CERTIFY 2010; Fraisse 2000; LIFENOX 2011; MEDENOX 1999; PREVENT 2004; THE‐PRINCE 2003).
Effects of interventions
Heparin (LMWH and UFH) versus placebo or no treatment
Efficacy
Deep vein thrombosis (DVT)
Data were available for DVT in seven trials (Belch 1981; Dahan 1986; Fraisse 2000; Gallus 1973; Ibarra‐Perez 1988; MEDENOX 1999; PREVENT 2004). The heparin treatment group had reduced odds of DVT (OR 0.41; 95% CI 0.25 to 0.67; P = 0.0004). When sensitivity analysis was performed to assess the effects of a loss of ≥ 20% of the study population, the Fraisse 2000 and MEDENOX 1999 studies were removed and there was little effect on the results with an OR of 0.26 (95% CI 0.12 to 0.54; P = 0.0004). By removing Belch 1981 and Dahan 1986 for sensitivity analysis for low‐accuracy testing of DVT there was little effect on the overall outcome with an OR of 0.48 (95% CI 0.29 to 0.81; P = 0.006). Removing the low‐dose enoxaparin group from the MEDENOX 1999 study, there was only a slight variation in the analyses (OR 0.37; 95% CI 0.25 to 0.55; P < 0.00001).
Pulmonary embolism (PE)
Symptomatic, non‐fatal PE was measured in seven trials (Belch 1981; Dahan 1986; Fraisse 2000; Gardlund 1996; Ibarra‐Perez 1988; MEDENOX 1999; PREVENT 2004). Meta‐analysis was only conducted for six, excluding Gardlund 1996 as non‐fatal PE was only assessed through necropsy data and not in all participants. Non‐fatal PE had an OR of 0.46 (95% CI 0.20 to 1.07; P = 0.07). Fatal PE was recorded in six trials (Bergmann 1996; Dahan 1986; Gardlund 1996; LIFENOX 2011; MEDENOX 1999; PREVENT 2004) with an OR of 0.71 (95% CI 0.43 to 1.15; P = 0.16). Combined non‐fatal PE and fatal PE was reported for nine studies (Belch 1981; Bergmann 1996; Dahan 1986; Fraisse 2000; Gardlund 1996; Ibarra‐Perez 1988; LIFENOX 2011; MEDENOX 1999; PREVENT 2004) with an OR of 0.66 (95% CI 0.43 to 1.02; P = 0.06). When the Fraisse 2000 and MEDENOX 1999 studies were removed for sensitivity analysis to assess the effects of a loss of ≥ 20% of the study population, there was little change in results for non‐fatal PE (OR 0.55; 95% CI 0.22 to 1.39; P = 0.20), no change for fatal PE as neither study was contributing, and there was no difference between the treatment groups for combined non‐fatal and fatal PE (OR 0.68 95% CI 0.44 to 1.07; P = 0.09). When the 20 mg enoxaparin treatment group in the MEDENOX 1999 study was removed there was a small change on non‐fatal PE (OR 0.50; 95% CI 0.21 to 1.20; P = 0.12), no change for fatal PE, and almost no change for combined non‐fatal and fatal PE (OR 0.66; 95% CI 0.43 to 1.03; P = 0.07).
All‐cause mortality
All‐cause mortality was assessed in eight trials (Bergmann 1996; Dahan 1986; Fraisse 2000; Gardlund 1996; Ibarra‐Perez 1988; LIFENOX 2011; MEDENOX 1999; PREVENT 2004) but meta‐analysis was only performed for seven, excluding Ibarra‐Perez 1988 because not enough information was given to fully understand which treatment groups the deaths occurred in. There was no clear evidence of a difference in mortality between the two treatment groups, with an OR of 0.97 (95% CI 0.87 to 1.08; P = 0.57). When the MEDENOX 1999 study's 20 mg enoxaparin treatment group was removed in a sensitivity analysis there was no change to the results of the analyses.
Safety
Bleeding
Major bleeding was evaluated in eight studies (Belch 1981; Dahan 1986; Fraisse 2000; Gardlund 1996; Ibarra‐Perez 1988; LIFENOX 2011; MEDENOX 1999; PREVENT 2004) but meta‐analysis was conducted using only seven trials, excluding Gardlund 1996. This was for the same reason as for the outcome of non‐fatal PE: major bleeding was only evaluated through necropsy data and was not evaluated in all patients. Heparin was associated with an increase in major bleeding (OR 1.65; 95% CI 1.01 to 2.71; P = 0.05). When sensitivity analysis was performed by removing studies with inadequate definitions of the major bleeding, Belch 1981 and Dahan 1986, the association was more precise (OR 1.83; 95% CI 1.09 to 3.07; P = 0.02). Minor bleeding was evaluated in five studies (Fraisse 2000; Ibarra‐Perez 1988; LIFENOX 2011; MEDENOX 1999; PREVENT 2004) and heparin was associated with significantly increased odds of 1.61 (95% CI 1.26 to 2.08; P = 0.0002). When the low dose of enoxaparin in the MEDENOX 1999 study was removed, the major bleeding outcome had stronger statistical significance (OR 1.81; 95% CI 1.09 to 3.03; P = 0.02) but there was very little change in the minor bleeding outcome (OR 1.61; 95% CI 1.23 to 2.12; P = 0.0006).
Thrombocytopaenia
Thrombocytopaenia was measured in four studies (Fraisse 2000; LIFENOX 2011; MEDENOX 1999; PREVENT 2004). There was no clear evidence of a difference between the two groups for thrombocytopaenia (OR 1.05; 95% CI 0.64 to 1.74; P = 0.85). Removing the low‐dose enoxaparin treatment group from the MEDENOX 1999 study had little effect on the overall outcome (OR 1.12; 95% CI 0.63 to 2.00; P = 0.69).
Heterogeneity and reporting bias
For the comparison between heparin and placebo or no treatment there was little or no heterogeneity between the studies, therefore all meta‐analyses, except for the DVT outcome, were conducted using a fixed‐effect model. The DVT outcome was analysed using a random‐effects model but when the 20 mg enoxaparin treatment group in the MEDENOX 1999 study was removed for sensitivity analysis a fixed‐effect model was able to be utilised. None of the comparisons had 10 or more studies so an assessment for reporting bias by constructing funnel plots could not be performed.
LMWH versus UFH
Efficacy
Deep vein thrombosis (DVT)
Six studies evaluated DVT (CERTAIN 2010; CERTIFY 2010; EMSG 1996; Forette 1995; PRIME 1996; THE‐PRINCE 2003). For the LMWH treatment group there was a decreased odds compared to the UFH group (OR 0.77; 95% CI 0.62 to 0.96; P = 0.02). When the CERTAIN 2010 and CERTIFY 2010 were removed for sensitivity analysis to assess the effects of a loss of ≥ 20% of the study population, there was no longer any evidence of a difference between the treatment groups (OR 0.80; 95% CI 0.50 to 1.29; P = 0.37). Removing EMSG 1996 for sensitivity analysis for both a low‐accuracy test for DVT and low dosage of enoxaparin, there was very little effect on the overall outcome (OR 0.76; 95% CI 0.61 to 0.96; P = 0.02).
Pulmonary embolism (PE)
Symptomatic, non‐fatal PE was measured in six studies (CERTAIN 2010; CERTIFY 2010; EMSG 1996; Forette 1995; PRIME 1996; THE‐PRINCE 2003). There was no clear evidence of a difference between LMWH and UFH for non‐fatal PE (OR 0.93; 95% CI 0.42 to 2.08; P = 0.86). Fatal PE was measured in only two studies (CERTAIN 2010; CERTIFY 2010) for which there was no difference between the two treatment groups (OR 0.33; 95% CI 0.01 to 8.13; P = 0.50). Combined non‐fatal and fatal PE was evaluated in the same six studies as non‐fatal PE, with no evidence of a difference between UFH and LMWH (OR 0.86; 95% CI 0.39 to 1.90; P = 0.71). When CERTAIN 2010 and CERTIFY 2010 were removed for sensitivity analysis to assess the effects of a loss of ≥ 20% of the study population, the non‐fatal PE OR was lowered to 0.47, but the association remained non‐significant (95% CI 0.13 to 1.68; P = 0.25). As CERTAIN 2010 and CERTIFY 2010 were the only two studies evaluated for fatal PE, this outcome was no longer estimable and the combined non‐fatal and fatal PE was identical to non‐fatal PE.
All‐cause mortality
All‐cause mortality was assessed in five studies (CERTIFY 2010; EMSG 1996; Forette 1995; PRIME 1996; THE‐PRINCE 2003). There was no clear evidence for a difference in mortality (OR 0.79; 95% CI 0.54 to 1.16; P = 0.23).
Safety
Bleeding
Six studies measured major bleeding as an outcome (CERTAIN 2010; CERTIFY 2010; EMSG 1996; Forette 1995; PRIME 1996; THE‐PRINCE 2003). There was a reduced odds found in the LMWH group compared with the UFH group (OR 0.43; 95% CI 0.22 to 0.83; P = 0.01). Minor bleeding was assessed in three studies (CERTAIN 2010; CERTIFY 2010; Forette 1995) and was associated with an imprecise decreased odds with LMWH (OR 0.70; 95% CI 0.48 to 1.00; P = 0.05), possibly a chance effect. When the CERTAIN 2010 study was removed for sensitivity analysis to assess the effect of the longer follow‐up time period, there was little change for major bleeding (OR 0.43; 95% CI 0.22 to 0.87; P = 0.02) and minor bleeding moved from a possible chance effect to statistically significant (OR 0.66; 95% CI 0.45 to 0.96; P = 0.03).
Thrombocytopaenia
Thrombocytopaenia was evaluated in three studies (CERTAIN 2010; CERTIFY 2010; Forette 1995). There was no difference between LMWH and UFH (OR 0.41; 95% CI 0.08 to 2.11; P = 0.28). When sensitivity analysis was performed by removing the CERTAIN 2010 study to assess the effects of the longer follow‐up period, there was very little change (OR 0.43; 95% CI 0.06 to 2.92; P = 0.39).
Heterogeneity and reporting bias
For the comparison between UFH and LMWH there was little or no heterogeneity between the studies, therefore all meta‐analyses were conducted using a fixed‐effect model. None of the comparisons had 10 or more studies so an assessment for reporting bias by using a funnel plot could not be performed.
Subgroup analysis
Subgroup analysis based on medical diagnosis at hospital admission could not be performed as there are currently insufficient data within published studies on outcomes within subgroups of interest.
Discussion
Summary of main results
Heparin resulted in a reduction in DVT but no difference in non‐fatal and/or fatal PE (separately and combined) when compared with placebo or no treatment. The reduction in VTE risk is comparable to that previously reported in prophylaxis studies of patients following acute myocardial infarction (Collins 1996), acute ischaemic stroke (Gubitz 2004), colorectal surgery (Wille‐Jørgensen 2004) and orthopaedic surgery (Collins 1988).
The analysis found no clear difference in all‐cause mortality in patients receiving heparin prophylaxis. However, these studies are not powered to show a difference in mortality, which would require over 200,000 patients (assuming an overall 5% mortality, 10% of deaths due to VTE, and a 50% relative risk (RR) reduction in events).
In the presence of heparin prophylaxis the rate of major haemorrhage (0.64%) was higher than with no prophylaxis (0.36%). LMWH resulted in less major haemorrhage than UFH. This suggests a favourable benefit‐risk ratio for LMWH therapy in the prevention of VTE in general medical patients.
In summary, patients hospitalised with an acute medical illness who receive UFH or LMWH thromboprophylaxis have a reduced risk of DVT, however there is an increase in major haemorrhage when compared to those who do not receive chemical thromboprophylaxis. LMWH reduced the odds of DVT compared with UFH as well as reduced the odds of major bleeding, suggesting LMWH has a better efficacy profile and carries a lower risk of adverse events compared with UFH. There are currently insufficient published data to allow analysis of outcomes of interest based on patient medical diagnosis at hospital admission.
Overall completeness and applicability of evidence
The conclusions provided in this analysis have a strong external application as the evidence was generated using 16 good quality studies that were sufficient to address the study outcomes. The 16 studies included 34,369 participants from at least 17 different countries, with the majority of studies being multi‐centre, and four being multi‐national.
Although this review does not identify heparin as preventing all‐cause mortality in patients with an acute medical illness, it does appear to show a benefit in reducing DVT and possibly the combined outcome of symptomatic PE and fatal PE. Therefore, in patients with an acute medical illness heparin appears to reduce the risk of hospital acquired thrombosis (HAT), supporting the current medical practice of increasing awareness and advocating VTE prevention (NICE).
Currently, published studies are lacking outcome data specific to medical diagnosis at hospital admission. These data would help us to understand if certain patients, based on disease, would benefit more or less from thromboprophylaxis or are at a higher or lower risk of bleeding. Understanding these differences would help develop a stronger profile for the use of thromboprophylaxis in acutely ill medical patients.
Patients who develop VTE following an acute medical illness have a more severe presentation and significantly worse outcome than patients who develop VTE following surgery (Monreal 2004). Venous thromboembolic disease in hospitalised medical patients is a preventable occurrence and is associated with unacceptably high levels of morbidity and mortality.
In a report by the House of Commons Health Committee, it was recommended that "VTE and its prevention, including the implementation of, and adherence to guidelines relating to thromboprophylaxis, counselling and risk assessment, be given more prominence in undergraduate medical education, continuing professional development, and other relevant aspects of medical and paramedical training" (Health Committee 2005).
The current 'Prevention of VTE' guidelines presented by the American College of Chest Physicians also stress the importance of all hospitals having formal, written guidelines for thromboprophylaxis and that all patients undergo VTE risk assessment (Kahn 2012).
Current medical practice concerning VTE emphasizes awareness of HAT, with appreciation for the risk of VTE increasing dramatically within the past decade. All patients admitted to hospital with an acute medical illness should be risk assessed and thromboprophylaxis offered to those perceived to be at increased risk of VTE. In fact, patients are actively encouraged to question their admitting physicians as to their individual VTE risk and need for thromboprophylaxis (NICE) in an attempt to reduce the risk of HAT.
Quality of the evidence
Sixteen studies were used to compile evidence, 10 comparing heparin prophylaxis with no treatment or placebo and six which compared LMWH to UFH, with a total of 34,369 participants. For the majority of the outcomes there was general agreement between the studies. Overall quality of the studies was good, with concern for performance bias in the studies that were open label, which made up just under half of the studies. Also, most of the studies were lacking in an explanation of how random sequence generation and allocation concealment were performed, although this is most likely due to the fact that many included studies are older and were not held to as high reporting standards when published.
Potential biases in the review process
All data used were dichotomous, therefore little data manipulation that could introduce bias was required for analysis. However, for the Bergmann 1996 study all‐cause mortality data were calculated using provided percentages and were transformed into dichotomous values for the purposes of the meta‐analysis.
In order to reduce bias, both the 20 mg and 40 mg enoxaparin groups from MEDENOX 1999 were included in the meta‐analysis. To avoid double‐counting of participants, the participants and events within the comparison (placebo) group were split evenly between the two treatment groups, rounding down for odd numbers.
While all attempts to include relevant studies were made by the review authors, it is possible pertinent data were overlooked.
Agreements and disagreements with other studies or reviews
Several meta‐analysis are available on this topic.
The findings of this review are in keeping with the meta‐analysis published by Mismeti et al, which assessed seven trials comparing prophylactic heparin to control and nine trials comparing UFH with LMWH (Mismetti 2000). The meta‐analysis found a statistically significant decrease in DVT (RR 0.44; 95% CI 0.29 to 0.64; P < 0.001) and clinical PE (RR 0.48; 95% CI 0.34 to 0.68; P < 0.001) when heparin was compared to control, which are consistent with our review, although Mismeti identified a stronger relationship with PE as our findings for non‐fatal and fatal PE were imprecise and possibly due to chance effect. The Mismeti review found no difference in mortality, the same as our review, but they also found no difference in major bleeding, which is inconsistent with our review. No difference was observed in efficacy between UFH and LMWH in the Mismeti review, but our review found a decrease in DVT favouring LMWH. Also, our review found a decrease in major bleeding favouring LMWH, which is consistent with the Mismeti review findings.
The Mismetti review used six of the same studies as we did for the comparison of heparin to control (Belch 1981; Bergmann 1996; Dahan 1986; Fraisse 2000; Gardlund 1996; Ibarra‐Perez 1988) and one study which was excluded from our review. For the comparison between UFH and LMWH, the Mismeti study analysed four studies used in our review (EMSG 1996; Forette 1995; PRIME 1996; THE‐PRINCE 2003) and five studies that we excluded. While the authors described that they excluded cases of acute ischaemic stroke and acute myocardial infarction some of the studies included patients with acute coronary disease that could not be removed for analysis. Other criteria that warranted exclusion in our review were not upheld in the Mismeti review, such as detection method and inclusion of surgical patients, which led to differences in the profile of the included studies and the minor differences seen in the results.
A meta‐analysis by Dentali et al assessed anticoagulant prophylaxis (UFH, LMWH and fondaparinux) to prevent symptomatic VTE in hospitalised medical patients (Dentali 2007). Nine trials were included in this analysis. Anticoagulant prophylaxis was associated with a borderline risk reduction in DVT (RR 0.47; 95% CI 0.22 to 1.00), which is consistent with this review. The Dentali review found a reduction in all PE (RR 0.43; 95% CI 0.26 to 0.71) for the prophylaxis group, but our review found no difference in non‐fatal and fatal PE. It was also noted in the Dentali review that anticoagulant prophylaxis was associated with a RR 1.32 (95% CI 0.73 to 2.37) for major bleeding, where our review found a statistically significant increase in major bleeding (OR 1.81; 95% CI 1.10 to 2.98; P = 0.02).
The Dentali review considered seven of the same studies that we included in this review for the comparison between anticoagulant prophylaxis and control (Belch 1981; Bergmann 1996; Dahan 1986; Fraisse 2000; Gardlund 1996; MEDENOX 1999; PREVENT 2004) and two studies considered not relevant or excluded from our review. Although study inclusion and exclusion criteria were similar for the Dentali review and this review, the minor differences led to the different study profile and variations in results. A major difference is that the Dentali review included studies that observed effects of fondaparinux, and not solely heparin. Also, one included study did contain participants with a recent myocardial infarction, which excluded it from the current review.
A meta‐analysis published in 2009 by Bump et al compared heparin prophylaxis to control in seven studies, and LMWH to UFH in six studies (Bump 2009). Comparing heparin versus control, the authors found a decreased risk of all DVT, with a RR of 0.55 (95% CI 0.36 to 0.83), and all PE, with a RR of 0.70 (95% CI 0.53 to 0.93) in the heparin treatment group, but no difference in mortality between the groups (RR 0.92; 95% CI 0.82 to 1.03). The findings on all DVT and all‐cause mortality are consistent with our review, but our review found did not find a difference in non‐fatal and fatal PE. The Bump review did not find a difference between heparin and control for major bleeding, but in our review we found major bleeding to be significantly increased in the heparin group. For the studies comparing LMWH to UFH there were no statistically significant differences between the two treatment groups for DVT, PE, death or bleeding. The data on PE and death are consistent with our review, but we found a decrease in DVT and major bleeding favouring LMWH.
For the comparison between heparin and control the Bump et al review included six of the same studies as in our review (Belch 1981; Bergmann 1996; Dahan 1986; Gardlund 1996; MEDENOX 1999; PREVENT 2004) with the addition of two studies that we excluded or considered not relevant. Comparing UFH to LMWH the Bump review included three of the same studies as in our review (EMSG 1996; PRIME 1996; THE‐PRINCE 2003) with two extra studies which were excluded from our review. Although the Bump review sought to exclude trials predominantly made up of stroke or acute myocardial infarction patients, it did include studies that had some of these types of participants that could not be separated out for statistical purposes. This difference, along with the review being conducted before some of the big trials included in our review were published, has lead to a different profile of included studies and variations in results.
A more recent meta‐analysis published by Lederle et al analysed 10 studies that compared prophylaxis with heparin or related agents to no heparin and nine studies comparing LMWH with UFH, all in hospitalised medical patients (Lederle 2011). The review also analysed stroke patients, but this was performed separately and is not considered here. For studies that compared heparin prophylaxis with no heparin, a statistically significant decrease in PE was found in the heparin group with an OR of 0.69 (95% CI 0.52 to 0.90; P = 0.006), but our review found no difference in non‐fatal and fatal PE. In the Lederle review there were no differences found for symptomatic DVT, fatal PE or all‐cause mortality. The fatal PE and all‐cause mortality results are consistent with our review, but DVTs were found to be statistically decreased in the heparin group compared to the placebo or no treatment group. No difference in major bleeding was observed in the Lederle review (OR 1.49; 95% CI 0.91 to 2.43; P = 0.110), whereas in our review the increase in major bleeding in the heparin arm was statistically significant. No differences were found between LMWH and UFH for efficacy or safety outcomes, but in our review we found a statistically significant decrease in both DVT and major bleeding favouring LMWH.
The Lederle review considered seven of the same studies as in the current review for the comparison between heparin and placebo or no treatment (Belch 1981; Bergmann 1996; Dahan 1986; Fraisse 2000; Gardlund 1996; MEDENOX 1999; PREVENT 2004), with the addition of three studies which were either considered not relevant or excluded from the current review. For the comparison between UFH and LMWH the Lederle review used five of the same studies as in our review (CERTAIN 2010; CERTIFY 2010; EMSG 1996; PRIME 1996; THE‐PRINCE 2003), with the addition of four studies that we either considered not relevant or excluded. The Lederle review extended inclusion to studies with common treatments for VTE, and not strictly heparin (which led to the inclusion of a study evaluating fondaparinux), and while studies including only acute myocardial infarction patients were excluded in the Lederle review it did include studies that had some acute coronary syndrome patients, which would have been excluded from our review. Also, the Lederle review was conducted before several newer studies were published that have been included in the current review. These differences can account for the variations in findings regarding both the heparin versus no heparin and UFH versus LMWH comparisons.
Methodological quality graph: review authors' judgements about each methodological quality item presented as percentages across all included studies.
Methodological quality summary: review authors' judgements about each methodological quality item for each included study.
Comparison 1 Heparin versus placebo or no treatment, Outcome 1 Deep vein thrombosis.
Comparison 1 Heparin versus placebo or no treatment, Outcome 2 Non‐fatal pulmonary embolism.
Comparison 1 Heparin versus placebo or no treatment, Outcome 3 Fatal pulmonary embolism.
Comparison 1 Heparin versus placebo or no treatment, Outcome 4 Combined non‐fatal and/or fatal pulmonary embolism.
Comparison 1 Heparin versus placebo or no treatment, Outcome 5 All cause mortality.
Comparison 1 Heparin versus placebo or no treatment, Outcome 6 Major bleeding.
Comparison 1 Heparin versus placebo or no treatment, Outcome 7 Minor bleeding.
Comparison 1 Heparin versus placebo or no treatment, Outcome 8 Thrombocytopaenia.
Comparison 2 Low molecular weight heparin versus unfractionated heparin, Outcome 1 Deep vein thrombosis.
Comparison 2 Low molecular weight heparin versus unfractionated heparin, Outcome 2 Non‐fatal pulmonary embolism.
Comparison 2 Low molecular weight heparin versus unfractionated heparin, Outcome 3 Fatal pulmonary embolism.
Comparison 2 Low molecular weight heparin versus unfractionated heparin, Outcome 4 Combined non‐fatal and/or fatal pulmonary embolism.
Comparison 2 Low molecular weight heparin versus unfractionated heparin, Outcome 5 All cause mortality.
Comparison 2 Low molecular weight heparin versus unfractionated heparin, Outcome 6 Major bleeding.
Comparison 2 Low molecular weight heparin versus unfractionated heparin, Outcome 7 Minor bleeding.
Comparison 2 Low molecular weight heparin versus unfractionated heparin, Outcome 8 Thrombocytopaenia.
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Deep vein thrombosis Show forest plot | 7 | 5511 | Odds Ratio (M‐H, Random, 95% CI) | 0.41 [0.25, 0.67] |
| 2 Non‐fatal pulmonary embolism Show forest plot | 6 | 5485 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.46 [0.20, 1.07] |
| 3 Fatal pulmonary embolism Show forest plot | 6 | 27563 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.71 [0.43, 1.15] |
| 4 Combined non‐fatal and/or fatal pulmonary embolism Show forest plot | 9 | 27971 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.66 [0.43, 1.02] |
| 5 All cause mortality Show forest plot | 7 | 27786 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.97 [0.87, 1.08] |
| 6 Major bleeding Show forest plot | 7 | 13804 | Odds Ratio (M‐H, Fixed, 95% CI) | 1.65 [1.01, 2.71] |
| 7 Minor bleeding Show forest plot | 5 | 13434 | Odds Ratio (M‐H, Fixed, 95% CI) | 1.61 [1.26, 2.08] |
| 8 Thrombocytopaenia Show forest plot | 4 | 13349 | Odds Ratio (M‐H, Fixed, 95% CI) | 1.05 [0.64, 1.74] |
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Deep vein thrombosis Show forest plot | 6 | 5942 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.77 [0.62, 0.96] |
| 2 Non‐fatal pulmonary embolism Show forest plot | 6 | 5942 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.93 [0.42, 2.08] |
| 3 Fatal pulmonary embolism Show forest plot | 2 | 3581 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.33 [0.01, 8.14] |
| 4 Combined non‐fatal and/or fatal pulmonary embolism Show forest plot | 6 | 5942 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.86 [0.39, 1.90] |
| 5 All cause mortality Show forest plot | 5 | 5605 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.79 [0.54, 1.16] |
| 6 Major bleeding Show forest plot | 6 | 5942 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.43 [0.22, 0.83] |
| 7 Minor bleeding Show forest plot | 3 | 3876 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.70 [0.48, 1.00] |
| 8 Thrombocytopaenia Show forest plot | 3 | 3876 | Odds Ratio (M‐H, Fixed, 95% CI) | 0.41 [0.08, 2.11] |
