Abhishek Sharma, Maimonides Medical Center, 1016 50th Street Apt 2C, Brooklyn, NY 11219, USA. Tel.: +1 201 892 6548; fax: +1 3472447148. E-mail: email@example.com
Summary. Background: Duration of thromboprophylaxis beyond hospital discharge for medically ill patients has been controversial. Therefore an evaluation was made of the evidence currently available.
Methods: A search was made of the Pub Med, CENTRAL and EMBASE databases for randomized controlled trials from 1966 through to 2011. Interventions included thromboprophylaxis administered over an extended period in patients hospitalized for acute medical illness with decreased level of mobility. No differentiation was made for the medication used for individual studies. The comparator included standard medical therapy and/or placebo. The efficacy outcomes assessed were a composite of asymptomatic and symptomatic deep vein thromboses (DVT), pulmonary emboli (PE) and venous thromboembolism (VTE)-related deaths in the intervention group vs. the comparator group, as well as the safety outcomes evaluated with rates of bleeding events at the end of at least 30 days of follow-up. The methodological quality of the studies was assessed, as was publication bias. Event rates were compared using a forest plot of relative risk (RR; 95% confidence interval (CI)) using a random effects model (Mantel-Haenszel) between the active thromboprophylaxis and controls. Statistical analysis was carried out with Review Manager V5.1.
Results: Three recent studies were included. Extended duration thromboprophylaxis reduced the combined composite event rate, RR 0.75 (0.64, 0.88). However, individual clinical endpoints were not significantly improved with extended prophylaxis: asymptomatic proximal DVT, RR 0.85 (0.68, 1.05); symptomatic DVT, RR 0.44 (0.19, 1.00); symptomatic non-fatal PE, RR 0.80 (0.43, 1.48); VTE-related death, RR 0.64 (0.38, 1.10). However, bleeding events were far more prevalent with extended thromboprophylaxis with major bleeds, RR 2.68 (1.78, 4.05), with a number needed to harm of 194.
Conclusion: Currently available evidence does not indicate that routine administration of post-discharge prophylaxis will be beneficial to the patients admitted for medical illness.
Despite an increase in expenditure on primary care health services, the annual rate of hospitalization remains high (> 30 million)  and venous thromboembolism (including deep vein thrombosis [DVT] and pulmonary embolism [PE]) remains an important cause of preventable morbidity and mortality among hospitalized patients. Though more common in surgical patients, hospitalization due to an acute medical condition itself is associated with an eight times increased risk of venous thromboembolic events (VTE), secondary to relative immobilization and severity and acuity of illness . In high-risk medical patients who are not on prophylaxis, the prevalence of DVT on venography and ultrasonography is as high as 15% and 5%, respectively, and up to 20% of these result in fatal pulmonary embolism [3–6]. As many as > 70% cases are missed clinically because venous thrombosis is asymptomatic in the majority of patients. In a study by Anderson et al. , as many as one-third of adult hospitalized patients were at risk of various thromboembolic events and needed thromboprophylactic treatment. They further reported that more than 50% of medical patients discharged from the hospital were at risk of VTE as per American College of Chest Physician criteria .
Thromboprophylaxis is based on the premise of improving venous flow and reducing hypercoagulability. In medical patients short-term thromboprophylaxis (10 ± 4 days) has been shown to reduce the rate of various thromboembolic events, including fatal PE, but no difference has been shown in the rate of death from any cause [4–6]. Dentali et al. , in their meta-analyses, show that short-term anticoagulation prophylaxis reduces the relative risk of symptomatic DVT and fatal PE by 53% and 64%, respectively, during treatment.
The risks of VTE events increase after hospital discharge and reported event rates at 30 days are as high as 5–6% in various multicentric trials, and keep on increasing in patients receiving the standard of care (SOC) [9,10]. Thus, extended duration thromboprophylaxis treatment has been proposed to reduce the risk of VTE that persists even beyond hospital discharge. Such a regime has been shown to reduce the incidence of both DVT and PE in surgical patients, and is thus currently recommended in this group of patients [6,11–16]. However, duration of thromboprophylaxis beyond hospital discharge for medically ill patients has been controversial. So an updated evaluation was made of the evidence currently available.
Data sources and searches
We systematically searched PubMed, EMBASE and the Cochrane Central Register of Controlled Trials for trials that randomized medically ill study participants with decreased mobility to extended thromboprophylaxis vs. a comparator, through to December 2011. The following medical subject heading (MeSH) terms were included for the MEDLINE search and adapted for other databases as needed: ‘extended thromboprophylaxis’, ‘deep vein thrombosis prophylaxis’, ‘low molecular weight heparin’, ‘apixaban’, ‘rivaroxaban’ and ‘oral anticoagulation’. In addition to searching the databases, the reference lists of all included studies, meta analyses and reviews were manually searched. We also manually searched the conference abstracts of the annual scientific sessions of the American Heart Association, the American College of Cardiology and the European Society of Cardiology between 2008 and 2011. There was no language restriction for the search.
We included trials that studied adult (18 + years) patients with acute medical illness, restricted mobility and reported relevant clinical outcomes with extended thromboprophylaxis, at least beyond hospital discharge. Eligible trials had to be randomized clinical trials comparing extended thromboprophylaxis with comparators and be able to report the outcomes of interest. We included all trials that randomized extended thromboprophylaxis without regard to comparator therapy. We excluded trials of patients where extended thromboprophylaxis was used as a perioperative strategy, those in which duration of therapy was similar in the control group, and those that did not report outcomes of interest.
Data extraction and quality assessment
Two authors (AS and SC) reviewed the trials to ensure that they met inclusion criteria and abstracted the data in duplicate, and this was checked for accuracy by the other author. Disagreements were resolved by consensus (10% of the time). We performed objective assessment of the trials using the ‘risk of bias’ assessment method specified in the Cochrane Handbook of Systematic Reviews: assessing for randomization, concealment, blinding, intention to treat, baseline comparisons, concomitant interventions and completeness of follow-up .
Our primary outcomes of interest were a composite of asymptomatic and symptomatic deep vein thromboses, pulmonary emboli and thromboembolism-related deaths in the intervention group vs. the comparator group, as well as the safety outcomes evaluated with rates of bleeding events at the end of at least 30 days of follow up. The effects on the clinically relevant endpoints individually were assessed as secondary outcomes.
Data synthesis and analysis
Meta analysis was performed as per the recommendation of the Cochrane collaboration and in line with the PRISMA statement . Pooled treatment effects were estimated using RR with Mantel-Haenszel risk ratio, using a random-effects model. Heterogeneity was assessed by chi-square tests and the I2 statistic: we defined I2 < 25% to be low heterogeneity, as per the Cochrane Handbook of Systematic Reviews . Publication bias was estimated using funnel plots and the regression test of Egger.
We assessed quality for each included trial; all the included trials were randomized controlled trials and were considered high quality [17,19]. For statistical analysis we used Review Manager Version 5.1 (The Nordic Cochrane Centre, The Cochrane Collaboration, 2008, Copenhagen, Denmark).
Calculation of number needed to harm (NNH): NNH shows how many individuals would need to be treated with the drug in order for one to show the harmful effect. NNH has been calculated by dividing 1 by the absolute risk increase (difference of control and experimental event rate), and again multiplying by 100 when the absolute risk increase is expressed as a percentage.
Our MEDLINE search returned 16 studies. After elimination of duplicate results, the EMBASE, Cochrane registries and the conference proceedings did not return any additional studies. Through a review of titles and abstracts, 11 studies were rejected for relevance. The remaining five studies were reviewed and assessed for satisfaction of the inclusion or exclusion criteria. Two studies were excluded after full text review as one looked at extended thromboprophylaxis in surgical (orthopedic) patients and another one assessed event rates during prolonged in-hospital stay. The three studies that met all criteria were included in this analysis (Fig. 1).
We included three multicenter clinical trials in our meta-analysis. The first one is a double-blind, double-dummy, placebo-controlled ADOPT trial (Apixaban vs. Enoxaparin for Thromboprophylaxis in Medically Ill Patients, ClinicalTrials.gov number, NCT00457002) , in which extended thromboprophylaxis with apixaban in medically ill patients was not found to be superior to short-term prophylaxis with enoxaparin. The rate of major bleeding was higher in the group receiving extended thromboprophylaxis. The second trial is a randomized, parallel, placebo-controlled trial, EXCLAIM (Extended-Duration Venous Thromboembolism Prophylaxis in Acutely Ill Medical Patients with Recently Reduced Mobility; ClinicalTrials.gov registration number: NCT00077753), which showed the benefit of extended thromboprophylaxis in elderly patients (> 75 years) . In women and in patients with level 1 immobility extended prophylaxis reduced the rate of VTE but at the expense of an increase in the rate of major bleeding. The third trial is a multi-national, randomized, double-blind, placebo controlled Phase III study, MAGELLAN (ClinicalTrials.gov registration number: NCT00571649), comparing rivaroxaban for the prevention of venous thromboembolism (VTE) in acutely ill medical patients . Patients receiving extended thromboprophylaxis with rivaroxaban had lower rates of VTE but at the cost of significantly higher rates of major bleeding, hence bringing into question the overall benefit of selection of such a strategy (Tables 1,2).
Enoxaparin, administered subcutaneously at a dose of 40 mg once daily during their stay in the hospital, for a minimum of 6 days
Male and female patients, 40 years of age or older, were considered for participation in the study if they were hospitalized for congestive heart failure, acute respiratory failure, infection (without septic shock), acute rheumatic disorder, or inflammatory bowel disease and had an expected hospital stay of at least 3 days. Except for patients with congestive heart failure or respiratory failure, eligible patients had to have at least one of the following additional risk factors: an age of 75 years or older, previous documented venous thromboembolism or a history of venous thromboembolism for which they received anticoagulation for at least 6 weeks, cancer, a body-mass index (the weight in kilograms divided by the square of the height in meters) of 30 or more, receipt of estrogenic hormone therapy, or chronic heart failure or respiratory failure. In addition, all patients had to be moderately or severely restricted in their mobility. Moderately restricted mobility allowed for walking within the hospital room or to the bathroom. Severely restricted mobility was defined as being confined to bed or to a chair at the bedside
Patients were excluded if they had confirmed venous thromboembolism, a disease requiring ongoing treatment with a parenteral or oral anticoagulant agent, active liver disease, anemia or thrombocytopenia, severe renal disease (creatinine clearance of < 30 mL per minute as estimated by the method of Cockcroft and Gault), a known or suspected allergy to enoxaparin, or prior heparin-induced thrombocytopenia, or if they were taking two or more antiplatelet agents or aspirin at a dose higher than 165 mg per day. Patients were also excluded if they had undergone a surgical procedure in the previous 30 days that might be associated with a risk of bleeding, had received anticoagulant prophylaxis for venous thromboembolism in the previous 14 days, were actively bleeding or were at high risk of bleeding, or had invasive procedures planned or scheduled during the treatment period. In addition, patients were excluded if they had one of the following abnormal laboratory findings: a hemoglobin level of less than 9 g per deciliter, a platelet count of less than 100 000 per cubic millimeter, an alanine or aspartate aminotransferase level more than twice the upper limit of the normal range, or direct or total bilirubin levels more than 1.5 times the upper limit of the normal range. Finally, women who might become pregnant, were pregnant, were breast-feeding, or were unwilling or unable to use an acceptable method of contraception were not eligible
Subcutaneous enoxaparin, 40 mg per day, or placebo for an additional 28 ± 4 days
40 years of age, had a life expectancy of at least 6 months, and had recently reduced mobility for up to 3 days. In addition, they had to be considered by the enrolling investigator as likely to have reduced mobility for at least 3 days after enrollment. We defined ‘reduced mobility’ as requiring total bed rest or being sedentary without bathroom privileges (level 1 immobility) or with bathroom privileges (level 2 immobility)
We used the published standard criteria for reporting of randomized clinical trials studies (PRISMA)  using Cochrane metrics to evaluate the quality of the studies selected for this review.
We identified three studies, accounting for 20 362 subjects that met our inclusion and exclusion criteria. Extended thromboprophylaxis was associated with a reduced composite event rate comprised of mortality, symptomatic and asymptomatic PE and DVTs, RR 0.75 (0.64, 0.88) (Fig. 2), and the Number Needed to Treat (NNT) of 320. We also assessed publication bias by constructing a funnel plot and by the regression test of Egger (Fig. 4). Symmetrical funnel plot and two-tailed P-value of 0.17 on Egger’s test ruled out any publication bias. However, extended thromboprophylaxis was associated with significant numbers of major bleeds, with an RR of 2.68 (1.78, 4.05) (Fig. 5) and number needed to harm (NNH) of 194 (95% CI, 139.2–324.2).
[ Forest plot of VTE-related mortality. ]
Extended thromboprophylaxis therapy also showed no significant improvement in VTE-related deaths, RR 0.64 (0.38, 1.10), (Fig. 3) or non-fatal PE, RR 0.80 (0.43, 1.48). Also interestingly, it did not show a decreased risk of symptomatic (RR 0.44 [0.19, 1.00]) or asymptomatic DVTs (RR 0.85 [0.68, 1.05]). On the other hand, rates of fatal hemorrhage and transfusion requiring ≥ 2 units were quite high.
On sub-analyses using only symptomatic thromboembolic events (DVT and PE), mortality and major bleeding, there was no overall benefit of long-term thromboprophylaxis, RR 0.69 (95% CI, −0.25 to 1.91) (Fig. 6). The weighted mean incidence of symptomatic thromboembolic events in the extended thromboprophylaxis (ET) and control group were 7.726 and 16.79, respectively. Weighted mean incidence for major bleed in the ET and control groups was 28.95 and 10.69, respectively.
The first double-blind, double-dummy, placebo-controlled trial included in the meta-analyses (ADOPT) reported that extended thromboprophylaxis with apixaban in medically ill patients was not superior to short-term prophylaxis with enoxaparin and the rate of major bleeding was higher in the group receiving extended thromboprophylaxis . Another randomized, parallel, placebo-controlled trial (EXCLAIM) showed the benefit of extended thromboprophylaxis in elderly patients (> 75 years) . On sub-analyses of women and patients with level 1 immobility, extended prophylaxis was found to reduce the rate of VTE but at the expense of an increase in the rate of major bleeding. The third randomized controlled trial (MAGELLAN) reported that patients receiving extended thromboprophylaxis with rivaroxaban had lower rates of VTE but at the cost of significantly higher rates of major bleeding, hence bringing into question the overall benefit of selection of such a strategy (Table 1).
Medical patients who are at higher risk of VTE and may be hypothesized to benefit from long-term thromboprophylaxis include those aged > 75 years, those with cancer, stroke with paraplegia, heart failure, COPD exacerbation and sepsis, those requiring long-term/permanent respiratory assistance, patients with limited mobility and those who need long-term hospitalization due to some other chronic medical condition [6,21–23]. Many patients have multiple risk factors and due to the cumulative nature of these risk factors they are more prone to develop VTE .
The result of our meta-analysis shows that extended thromboprophylaxis (cumulatively with apixaban, enoxaparin with rivaroxaban) is not superior to in-hospital prophylaxis that is the current SOC in acutely ill medical patients, and is associated with a higher risk of major bleeding. No statistically significant reduction was seen in VTE-related mortality, PE or DVT rates.
The primary efficacy endpoint in these studies was taken as a combination of surrogate endpoints (asymptomatic DVT) and clinical endpoints (symptomatic DVT, death related to VTE, PE). Looking at the outcomes obtained with our meta-analysis, we suggest that rather than combining both endpoints (clinical and surrogate) to escalate event rates, they should be analyzed and reported separately for similar future studies (especially if used surrogate marker is not ideal like asymptomatic DVT) for two reasons. First, to evaluate if capture criteria (effect of treatment on clinical outcome must be explained entirely by effect of treatment on surrogate) hold true for surrogate markers , and second, to assess whether the results of one endpoint are masking another .
Use of asymptomatic DVT as an ideal surrogate marker is debatable. Due to immobility and associated multiple co-morbidities (associated with impaired level of consciousness) DVT might not be manifested in its typical way in hospitalized patients and preliminary symptoms may be absent and initial manifestation could be fatal PE. Thus rate of asymptomatic DVT could be included in composite endpoints. The rationale for inclusion of asymptomatic DVT as a surrogate marker is its potential causal relationship with PE. Multiple studies have indicated that patients with asymptomatic DVT are at higher risk of developing PE, symptomatic DVT and post-thrombotic syndrome [27–30], but the reference standard technique is not always used while studying this relationship. Hence a causal relationship is yet to be confirmed, thus questioning its inclusion in primary endpoints . Though 34% of sudden fatal PEs may occur in patients with asymptomatic DVT , contrary is not true especially in surgical patients, where asymptomatic DVT remains clinically silent during the follow-up period, raising further questions over inclusion of asymptomatic DVT in primary endpoint analysis. However, in medical patients, further studies are needed before arriving at a final conclusion.
Rather than using a subclinical surrogate marker for assessment, efficacy of thromboprophylaxis should be measured in terms of definitive endpoints such as mortality benefit. Hence, to evaluate the trade-off between the efficacy and the safety of a prolonged antithrombotic prophylaxis, we also carried out sub-analyses by including only symptomatic thromboembolic events (DVT and PE), mortality and major bleeding and found no overall benefit of long-term thromboprophylaxis, RR 0.69 (95% CI −0.25 to 1.91) (Fig. 6). We further evaluated weighted mean incidence of mortality and symptomatic thromboembolic and major bleeding events in study and control arms and found no statistical significant difference in risk between the two arms (weighted difference of 0.00; 95% CI −0.01 to 0.01). However, as a smaller number of deaths occurred due to PE as compared with other medical causes in acutely ill medical patients, so higher powered studies with larger sample size or longer follow-up are needed to demonstrate a statistically significant reduction in mortality.
Risk stratification and proper selection of patients are also an important prerequisite for identifying subgroups of medically ill patients in whom extended thromboprophylaxis can result in a favorable risk to benefit ratio. Sex (female) and age (elderly > 75 years) have already been shown to be important factors when selecting patients who can maximally benefit from extended thromboprophylaxis [32,33], but more data are needed for further stratification and identification of a subset of patients who actually need extended prophylaxis.
One of the potential strengths of our meta-analysis is increased statistical power compared with individual studies, as smaller numbers of death occur due to VTE as compared with other medical causes in acutely ill medical patients, so our data account for a more precise estimation of clinical outcomes with extended thromboprophylaxis.
Use of composite endpoints and inclusion of an asymptomatic surrogate marker remain a significant limitation, especially if the used surrogate marker is not an ideal one, such as asymptomatic DVT. When using composite endpoints it should be noted that there is a significant imbalance between asymptomatic and symptomatic thromboembolic events reported in various studies, with the majority of reported events being asymptomatic and detected only on screening. This, along with the lack of causal relationship between asymptomatic DVT and symptomatic thromboembolic events, should be taken into consideration when interpreting the results with composite endpoints.
Another potential limitation is lack of data regarding longer term follow-up. Considering the trend shown in the ADOPT trial , it would be interesting to know the impact of extended thromboprophylaxis during a longer follow-up period. Another point that should be considered is that in the ADOPT and MAGELLAN trials systematic compression ultrasound examination was performed at about 14 days after randomization and detection of asymptomatic events might have influenced the future treatment of these patients. Also, in two of the three included studies two different antithrombotic drugs for different durations were compared. Though analysis was adjusted for duration of study, based on studies in orthopedic patients and in patients with atrial fibrillation it can still be speculated that different drugs may have a different efficacy and safety independently from treatment duration, hence results should be generalized with caution.
On the basis of additional risk of major bleeding without any significant mortality benefit, we suggest avoidance of routine use of extended thromboprophylaxis in medical patients for now.
Disclosure of Conflict of Interests
The authors state that they have no conflict of interest.