Most studies and recommendations on prophylaxis of venous thromboembolism have involved low-dose heparin. It is available either in the standard unfractionated form, usually given subcutaneously two or three times a day, or as low-molecular-weight heparin, a derivative of unfractionated heparin with more predictable and prolonged activity that can be given subcutaneously once a day. Although it was hoped that low-molecular-weight heparin would be more effective than unfractionated heparin for prophylaxis, four randomized trials in medical patients [2-5] and numerous studies in general surgery patients [6, 7] have shown no consistent differences between the two preparations in resulting venous thromboembolism, bleeding, or death. I therefore consider unfractionated heparin and low-molecular-weight heparin together. Intermittent pneumatic compression and graduated elastic compression stockings have been shown to reduce the rate of asymptomatic deep venous thrombosis in surgical patients and are reviewed elsewhere [8].

Heparin prophylaxis is well established for most surgical patients. A meta-analysis of randomized trials in more than 14 000 surgical patients showed a statistically significant reduction in deep venous thrombosis, pulmonary embolism, fatal pulmonary embolism, and total mortality with low-dose heparin [9]. The situation for hospitalized medical patients is far less certain, and the many differences between surgical and medical patients (for example, compared with surgical patients, medical patients have generally lower rates of thromboembolic disease, higher rates of aspirin use, a less discrete time frame of risk, and no wound-activated clotting) preclude extrapolation of study results from surgical patients to medical patients.

Low-dose heparin prophylaxis has been recommended for hospitalized medical patients with risk factors for thromboembolism, which include advanced age, prolonged immobility or paralysis, stroke, myocardial infarction, congestive heart failure, chest infections, cancer, hypercoagulable states, inflammatory bowel disease, estrogen therapy, and need for intensive care [1, 8]. More recently, two large randomized trials reported no important benefit from low-dose heparin prophylaxis in hospitalized medical patients [10, 11]. Even before the results of these trials were available, however, reports that prophylaxis was used in less than one third of high-risk medical patients suggest that this practice was not widely accepted by physicians [12-14].

Before examining the evidence on the effectiveness of prophylaxis in hospitalized medical patients, it is first necessary to consider how effectiveness should be measured. The most widely studied outcome measure has been deep venous thrombosis diagnosed by screening all participants, usually by using the labeled fibrinogen uptake test. Nearly all cases of deep venous thrombosis detected in this way are asymptomatic, and most are limited to the calf, where the risk for embolization is much lower than it is with proximal disease. Because studies directed at improving subclinical surrogate outcomes (rather than morbidity or mortality) can be misleading, the importance of preventing asymptomatic deep venous thrombosis has been questioned [6, 9]. The fibrinogen uptake test in particular has been shown to be a weak predictor of clinically important venous thromboembolism [15, 16]. Symptomatic thromboembolism is also a problematic study outcome because the diagnosis of pulmonary embolism is often missed and autopsy rates are low. The best measure for evaluating the effectiveness of prophylaxis has therefore been total mortality, although large numbers of patients are required to detect small but important differences in this outcome.

Studies of prophylaxis have separately addressed patients with acute myocardial infarction, patients with stroke, and general medical patients. In a recent meta-analysis of randomized trials of anticoagulant therapy in patients with acute myocardial infarction, most of whom also routinely received aspirin and a fibrinolytic agent, heparin therapy at any dose was associated with a 25% relative reduction (3.5% absolute reduction) in the mortality rate in patients who did not take aspirin (P = 0.002) and a 6% relative reduction (0.5% absolute reduction) in patients who took aspirin (P = 0.03) [17]. Heparin was also associated with reductions in pulmonary embolism, reinfarction, and stroke, but these diagnoses were not routinely sought and many studies were unblinded, raising the possibility of biased ascertainment. The meta-analysis had insufficient power to evaluate the effect of low-dose subcutaneous heparin on any outcome.

As for acute ischemic stroke, a meta-analysis of 8 randomized trials involving 982 patients found a statistically significant 81% relative reduction in deep venous thrombosis and a nonsignificant 18% relative reduction in total mortality with low-dose heparin [18]. More recently, the International Stroke Trial reported an unblinded, randomized comparison of 4861 patients who received unfractionated heparin, 5000 U twice daily for up to 14 days, with 9718 untreated controls [19]. Half of each group also received aspirin. There was no statistically significant difference in total mortality, which was slightly reduced at 14 days (0.6% absolute reduction) and slightly increased at 6 months (0.5% absolute increase), in the low-dose heparin group. In that study, low-dose heparin was associated with a statistically significant reduction in recurrent ischemic stroke (1.2% absolute reduction) and a significant increase in hemorrhagic stroke (0.3% absolute increase). The evidence is thus inconclusive but suggests a clinical benefit from low-dose heparin in patients with acute myocardial infarction or stroke.

In hospitalized general medical patients, several small randomized trials have shown statistically significant reductions in deep venous thrombosis associated with low-dose heparin [1]. Four larger trials [10, 11, 20-22] reported on mortality. In the first of these trials [20], 1358 patients older than 40 years of age who were admitted to the medicine service from an emergency department were allocated by hospital number to receive open-label unfractionated heparin, 5000 U twice daily, or no treatment until hospital discharge or full mobilization. Eligibility was determined after allocation, but all allocated patients were analyzed regardless of eligibility. There was a statistically significant reduction in total mortality in the heparin group-from 10.9% to 7.8% (29% relative reduction, 3.1% absolute reduction)-that was apparent from the first day of therapy, even though only 61% of the group actually received heparin. This reduction in mortality is larger and more rapid in onset than can easily be attributed to prevention of fatal pulmonary embolism. Possible alternative explanations include failure of the quasi-randomization scheme to produce equivalent groups (although patient characteristics were well matched between groups) or a beneficial effect of low-dose heparin on multiple causes of death rather than on pulmonary embolism alone.

The second study involved 270 hospitalized medical patients older than 65 years of age, most of whom had risk factors for thromboembolism, who received low-molecular-weight heparin (enoxaparin, 60 mg/d) for up to 10 days or placebo [21]. The incidence of deep venous thrombosis was significantly reduced by low-molecular-weight heparin from 9% to 3%. However, the overall mortality rates did not differ significantly (5.2% in the treatment group and 4.4% in the placebo group; 14% relative increase, 0.7% absolute increase).

The third study was a multicenter, randomized, double-blind, placebo-controlled trial of low-molecular-weight heparin prophylaxis (nadroparin, 7500 U/d for a maximum of 21 days) in 2474 hospitalized medical patients who were older than 40 years of age and could not walk alone for more than 10 meters. The study was completed in 1987 but, in a worrisome example of publication bias, the results were never published as a peer-reviewed article [11, 22]. No statistically significant differences were found between the two groups in any outcome, and mortality rates were nearly identical (10.1% with heparin and 10.3% with placebo; 2.0% relative reduction; 0.2% absolute reduction).

In the fourth study, 11 693 patients older than 55 years of age admitted to infectious disease wards were randomly allocated to receive open-label unfractionated heparin, 5000 U twice daily for a maximum of 21 days, or no treatment [10]. The heparin group experienced a small, statistically insignificant reduction in total mortality-from 5.6% to 5.3% (6.5% relative reduction; 0.4% absolute reduction). Deaths from pulmonary embolism were significantly delayed in the heparin group, resulting in a fourfold reduction in death from pulmonary embolism at 3 weeks that disappeared by 6 weeks; this result may reflect the 21-day limit on prophylaxis. A statistically significant decrease in nonfatal venous thromboembolism was also observed, but, as the authors point out, this finding is not reliable because the unblinded design may have biased ascertainment and because information on this outcome was not collected systematically.

Despite the shortcomings of these four randomized trials and the considerable differences among them, it may be useful to consider a combined estimate of their effect (calculated by using Review Manager 3.0 [Cochrane Collaboration, Oxford, United Kingdom]). The resulting odds ratio for the effect of low-dose heparin on mortality in general medical patients is 0.91 (95% CI, 0.80 to 1.04). Thus, the data currently available in general medical patients exclude neither the possibility that low-dose heparin is without clinical benefit nor the possibility that it may achieve a very substantial 20% relative reduction in total mortality.

Even if 100 to 200 patients must be treated to prevent 1 death (representing a 10% relative reduction in a control mortality rate of 5% to 10%, as suggested by the combined estimate), the potential benefit is substantial. The average cost of prophylaxis is less than $100 per patient; therefore, 1 death can be averted for less than $10 000 to $20 000. Eight million nonsurgical patients 60 years of age or older are discharged each year from acute care hospitals in the United States [23]; therefore, heparin prophylaxis of medical patients could save tens of thousands of lives every year if it is beneficial or waste hundreds of millions of dollars if it is not. Despite the large number of patients already studied, an even larger study is needed. A randomized trial of the effect of low-dose heparin on mortality in hospitalized general medical patients is currently being planned by the Veterans Affairs Cooperative Studies Program, but the results will not be available for at least 5 years. For now, clinicians must make their own decisions about heparin prophylaxis for medical patients-neither a comfortable position to be in nor an unusual one, unfortunately.