Articles |

Hemorrhagic Complications of Anticoagulant Treatment : The Seventh ACCP Conference on Antithrombotic and Thrombolytic Therapy FREE TO VIEW

Mark N. Levine, MD, MSc, Chair; Gary Raskob, PhD; Rebecca J. Beyth, MD, MSc; Clive Kearon, MD, PhD; Sam Schulman, MD, PhD
Author and Funding Information

Correspondence to: Mark N. Levine, MD, MSc, Room 104, First Floor, Henderson Research Centre, 711 Concession St, Hamilton, Ontario L8V 1C3

Chest. 2004;126(3_suppl):287S-310S. doi:10.1378/chest.126.3_suppl.287S
Text Size: A A A
Published online

This chapter about hemorrhagic complications of anticoagulant treatment is part of the seventh ACCP Conference on Antithrombotic and Thrombolytic Therapy: Evidence Based Guidelines. Bleeding is the major complication of anticoagulant therapy. The criteria for defining the severity of bleeding varies considerably between studies, accounting in part for the variation in the rates of bleeding reported. The major determinants of vitamin K antagonist-induced bleeding are the intensity of the anticoagulant effect, underlying patient characteristics, and the length of therapy. There is good evidence that vitamin K antagonist therapy, targeted international normalized ratio (INR) of 2.5 (range, 2.0 to 3.0), is associated with a lower risk of bleeding than therapy targeted at an INR > 3.0. The risk of bleeding associated with IV unfractionated heparin (UFH) in patients with acute venous thromboembolism (VTE) is < 3% in recent trials. This bleeding risk may increase with increasing heparin dosages and age (> 70 years). Low molecular weight heparin (LMWH) is associated with less major bleeding compared with UFH in acute VTE. UFH and LMWH are not associated with an increase in major bleeding in ischemic coronary syndromes, but are associated with an increase in major bleeding in ischemic stroke. Information on bleeding associated with the newer generation of antithrombotic agents has begun to emerge. In terms of treatment decision making for anticoagulant therapy, bleeding risk cannot be considered alone, ie, the potential decrease in thromboembolism must be balanced against the potential increased bleeding risk.

The major complication of anticoagulant therapy is bleeding. In this review, the incidence of hemorrhage in patients receiving oral anticoagulants or heparin and the clinical and laboratory risk factors that predispose to bleeding are discussed. The focus is on major bleeding and fatal bleeding. Details of the method used to select relevant articles can be found in the six previous symposia16 of the American College of Chest Physicians and the chapter in this Supplement by Schünemann et al. Bleeding was generally classified as major if it was intracranial or retroperitoneal, if it led directly to death, or if it resulted in hospitalization or transfusion.12 However, there was variation between studies for the definition of bleeding. Although bleeding is the major side effect of anticoagulant therapy, it should not be considered in isolation of potential benefit, ie, reduction in thromboembolism. This chapter focuses on bleeding related to vitamin K antagonists and heparins. In the section on vitamin K antagonists, risk factors for bleeding are first considered, and then bleeding rates for specific clinical conditions are presented. The same format is used for heparins. Bleeding related to new antithrombotic agents is also briefly discussed in a chapter by Weitz et al in this Supplement. The search and eligibility criteria used for our review are described in Table 1 .

1.1 Determinants of bleeding

The major determinants of oral vitamin K antagonist-induced bleeding are the intensity of the anticoagulant effect, patient characteristics, the concomitant use of drugs that interfere with hemostasis, and the length of therapy.

1.1.1 Intensity of anticoagulant effect

There is a strong relationship between the intensity of anticoagulant therapy and the risk of bleeding that has been reported in patients with deep vein thrombosis (DVT),7tissue heart valves,8mechanical heart valves,913 ischemic stroke,14and atrial fibrillation.1517 In randomized clinical trials (RCTs) for these indications,710 the frequency of major bleeding in patients randomly assigned to warfarin therapy at a targeted international normalized ratio (INR) of approximately 2.0 to 3.0 has been less than half the frequency in patients randomly assigned to warfarin therapy at a targeted INR > 3.0. The intensity of anticoagulant effect is probably the most important risk factor for intracranial hemorrhage, independent of the indication for therapy, with the risk increasing dramatically with an INR > 4.0 to 5.0.14,1819 In a case-control study, the risk of intracerebral hemorrhage doubled for each increase of approximately 1 in the INR.19

In five randomized trials20in patients with atrial fibrillation, the annual incidence of major bleeding averaged 1.3% in patients randomly assigned to warfarin therapy (targeted INR generally 2.0 to 3.0), compared with 1.0% in patients randomly assigned to treatment with placebo. In patients with atrial fibrillation, an INR of 2.5 (range, 2.0 to 3.0) minimizes the risk of either hemorrhage or thromboembolism.2122 Among patients with antiphospholipid antibody syndrome and prior thrombosis, the annual rate of major bleeding was similar in patients treated with warfarin at a targeted INR of 2.0 to 3.0, compared to those treated with warfarin at a targeted INR of 3.1 to 4.0 (3.0% vs 2.7%, respectively).23

Warfarin regimens (targeted INR < 2.0) have been investigated and found to be safe in the primary prevention of thrombosis in patients with malignancy. In two randomized trials in patients with malignancy, warfarin therapy, 1 mg/d of warfarin24and 1 mg/d for 6 weeks followed by adjustment to an INR of 1.3 to 1.9,25 did not increase the frequency of hemorrhage while still preventing thrombosis.

Increased variation in the anticoagulant effect, manifested by variation in the INR, is associated with an increased frequency of hemorrhage independent of the mean INR.15,2627 This effect is probably attributable to increased frequency and degree of marked elevations in the INR. Approaches to improve anticoagulant control (minimize INR fluctuations) could improve the safety and effectiveness of vitamin K antagonists. Anticoagulation management services (AMSs) or clinics and point-of-care INR testing are two such approaches. Two recent randomized trials2829 did not show a difference in quality of anticoagulant control or bleeding between AMSs and routine medical care. Results from four observational studies3033 showed AMSs were beneficial and associated with less bleeding than usual care.

Point-of-care testing with either patient self-testing or patient self-management is another model for potentially improving outcomes, as well as convenience. Patient self-testing provided better quality of anticoagulation control compared to routine medical care in one trial34(time in therapeutic range, 56% vs 32% [p < 0.001], and bleeding, 5.6% vs 12%, respectively [p = 0.05] after 6 months of follow-up), but no convincing difference compared with AMSs in two other studies.3536 Similarly, studies of patient self-management report better quality of anticoagulant control compared to routine medical care3739 vs AMSs.4042 Thus, no definite recommendations about the optimal approach for maintaining anticoagulant control can be made.

1.1.2 Patient characteristics

The risk of major bleeding during warfarin therapy can be related to specific comorbid conditions or patient characteristics. An increasing body of evidence supports age as an independent risk factor for major bleeding.4361 For example, Pengo et al60 evaluated the relationship of age and other risk factors to the incidence of major bleeding. Major bleeding occurred more frequently in patients > 75 years of age (5.1%/yr) than in younger patients (1%/yr). Multivariate analysis indicated that age > 75 years was the only variable independently related to primary bleeding (ie, bleeding unrelated to organic lesion). Also, risk for intracranial hemorrhage may be increased among older patients, especially those ≥ 75 years old when the INR is above therapeutic levels.,1920,43,62 The mechanism of how aging causes anticoagulant-related bleeding is not known.

A history of bleeding has been reported as a risk factor for subsequent bleeding,6365 but this observation has not been consistent.15,43,66 A history of nonbleeding peptic ulcer disease, however, has not been associated with subsequent GI bleeding.26,54,66

Other comorbid diseases have been associated with bleeding during warfarin therapy; these include treated hypertension,4647,49,51,67 cerebrovascular disease,43 ischemic stroke,21,44 serious heart disease,43,48 and renal insufficiency.26,43,6364,68 The presence of malignancy was a significant predictor of major bleeding in several studies.54,5657,64,66,69 In two of these studies,54,69 overanticoagulation was not the explanation for an increased risk of bleeding, whereas in one study,69 the severity of the cancer was identified as a risk factor. Another study68 did not confirm that malignancy predisposed to bleeding, while others47,70 excluded patients with malignancies.

Although many other patient characteristics have been associated with bleeding during warfarin therapy, the data supporting these findings are not compelling. For example, some studies noted an increased frequency of bleeding among women treated with warfarin,26,4748,50,53 but others have not.47,49,52,7172 Although most experienced clinicians believe that either alcoholism or liver disease increases the risk of bleeding during long-term warfarin therapy, two studies4647did not find such an association, whereas a large population-based study64 did.

Occult pathologic lesions may also precipitate warfarin-related bleeding. In one study,45 73% of adequately evaluated patients with a prothrombin time ratio < 1.5 at the time of bleeding had an underlying pathologic lesion, compared to 16% of patients with a prothrombin time ratio ≥ 1.5 (p < 0.05). However, pathologic lesions were found to be associated with GI or genitourinary bleeding in 30% of patients with prothrombin time ratio ≥ 2.5.

1.1.3 Concomitant drugs

Concomitant use of aspirin has been associated with a higher frequency of bleeding, even in patients treated with warfarin therapy with a mean INR of 1.5.7376 In a large randomized trial74comparing the combination of low-dose warfarin therapy and aspirin, 80 mg/d, to aspirin, 160 mg/d, in patients with a history of myocardial infarction, the frequency of spontaneous major hemorrhage during the first year of therapy was increased to 1.4% in patients treated with 3 mg of warfarin (INR < 2.0) and aspirin, 80 mg/d, compared with 0.7% in patients treated with aspirin, 160 mg/d (p = 0.01). In a large trial75of primary prevention in persons at high risk for ischemic heart disease, the rate of hemorrhagic stroke was 0.09%/yr in those treated with low-dose warfarin (target INR 1.5) plus aspirin, 75 mg/d, 0.01%/yr with low-dose warfarin alone, and 0.02%/yr with aspirin, and none in the placebo group. Similarly, in a trial76 that compared warfarin (INR, 1.5 to 2.5) plus 81 mg of aspirin to 162 mg of aspirin alone in patients after myocardial infarction, major bleeding was more common in the warfarin and aspirin group than in the aspirin-only group (1.28 events vs 0.72 events/100 person-years of follow-up, respectively; p < 0.001).

A number of medications can influence the anticoagulant response to vitamin K antagonists.7778 A detailed discussion of such drug interactions is beyond the scope of this chapter. However, we will briefly consider the influence of acetaminophen and nonsteroidal anti-inflammatory drugs (NSAIDs) on vitamin K antagonist anticoagulation.

Hylek et al79 reported that acetaminophen was associated with excessive warfarin anticoagulation in a nested case-control study. Patients receiving warfarin with an INR > 6.0 were matched with control subjects with INRs between 1.7 and 3.3. On univariate analysis, acetaminophen ingestion was associated with having an INR > 6.0, but no increase in bleeding. However, there were marked differences between the two populations. Patients were more likely to have recently commenced a new medication, eg, antibiotics; more likely to have an acute illness; more likely to have cancer; and had a decreased intake of vitamin K-rich foods. On multivariable analyses, acetaminophen was no longer an independent predictor of excessive anticoagulation.

Johnsen et al80 conducted a population-based study in Denmark. Patients receiving oral anticoagulants were identified through a population-based prescription database, and were linked to a hospital discharge registry. The incidence of upper GI bleeding with oral anticoagulants was compared to the incidence in the general population not receiving oral anticoagulants. The standardized incidence ratio was 2.8 for oral anticoagulant alone, and 4.4 for oral anticoagulant combined with acetaminophen. However, this type of design could not control for important confounders such as intercurrent illnesses such as infection.

Gadisseur et al81 conducted a randomized trial in which 31 patients receiving stable phenprocoumon therapy for 8 weeks were randomized to paracetamol, 1,500 mg/d, 3,000 mg/d, or placebo.81 These patients had no concurrent illnesses. After 1 week, there was a mean rise of 0.46 in the INR in both paracetamol groups compared to placebo. At day 15, there was no difference between placebo and the low dose of paracetamol and a mean rise of 0.22 INR in the higher-dose paracetamol group. Hence, in this study,81 acetaminophen did not provoke clinically relevant changes in patients treated with low doses of paracetamol, and produced very small changes at higher dose. Thus, at this juncture, the weight of evidence would indicate that any important INR rise in patients receiving acetaminophen is likely as a result of concurrent illness necessitating the intake of this medication, and there is little evidence that acetaminophen increases bleeding due to vitamin K antagonists.

It is well recognized that NSAIDs can be associated with upper GI bleeding.82However, an important clinical question is whether NSAIDs can increase the risk of bleeding in patients receiving vitamin K antagonists. Although there are case reports8384 that NSAIDs can prolong the INR in patients receiving vitamin K antagonists, prospective studies8587 have shown that NSAIDs do not interact with vitamin K antagonists to prolong the INR excessively.

Several studies have attempted to examine the relationship between NSAIDs and vitamin K antagonist-related bleeding. In one study,88 patients who bled while receiving coumarin were identified from records of an outpatient anticoagulant clinic and hospital records. These patients were sent a questionnaire on prior use of NSAIDs.88 Twelve percent of patients who bled had been receiving NSAIDs prior to the bleed. Although there was no control population of patients receiving coumarin who did not bleed, the authors concluded that the relative risk of NSAID use with regard to bleeding complications was 5.8. Because of the limitations with the study design, the validity of the results is uncertain.

These same authors subsequently performed a nested case-control study89 in the same community to examine whether Cox-2–selective NSAIDs are associated with less bleeding complications in coumarin users compared with nonselective NSAIDs. They reported that Cox-2–selective NSAIDs are associated with less bleeding complications than nonselective NSAIDs. However, the study89 had a number of methodologic limitations, particularly in the selection of cases and control subjects.

In another study from Denmark, a pharmacologic database identified all prescriptions for NSAIDs in a specific community from 1991 to 1995.90 This data were linked to a hospital database on admissions for GI bleeds. The number of bleeds on NSAIDs was 3.6 times higher than expected in the general population not exposed to NSAIDs. Concurrent anticoagulant use increased the risk of bleeding to 11 times that expected. This type of study design is limited, however, by the retrospective nature and the inability to identify important confounders that could have contributed to the bleeding.

Shorr et al91 performed a retrospective cohort study of Tennessee Medicaid enrollees aged ≥ 65 years from 1984 through 1986.91 The incidence of hospitalization for hemorrhagic peptic ulcer disease in patients receiving anticoagulants was threefold the incidence for nonusers. In patients receiving oral anticoagulants, the risk of hospitalization for hemorrhagic peptic ulcer disease was increased a further threefold by the current use of NSAIDs. This study design is limited by a lack of adjustment for such confounders as duration of anticoagulant, intensity of anticoagulant control, and concurrent medical illnesses.

In summary, the ideal design to address the question of whether NSAIDs increase bleeding on vitamin K antagonists is a randomized trial. No such study has been done. To date, a number of observational studies have examined the question. Such studies, however, are subject to a number of important biases. Hence, it is concluded that the quality of evidence supporting any relationship between NSAID use and bleeding on vitamin K antagonists is weak.

1.1.4 Risk of bleeding and the length of time relative to when anticoagulant therapy started

Four studies reported higher frequencies of bleeding early in the course of therapy.26,4647,54 In one of these studies,46 for example, the frequency of major bleeding decreased from 3.0%/mo the first month of outpatient warfarin therapy to 0.8%/mo during the rest of the first year of therapy, and to 0.3%/mo thereafter. Other descriptive studies9295 have supported this observation, although some studies67,96 have not.

1.1.5 Estimating bleeding risk

Models have been developed for estimating the risk for major bleeding during vitamin K antagonist anticoagulant therapy. These models are based on the identification of independent risk factors for warfarin-related bleeding, such as a history of stroke, history of GI bleeding, age ≥ 65 years, and higher levels of anticoagulation.26,50,5354,56,63,97 Such prediction rules can be useful in clinical practice because although physicians’ estimates of risk for anticoagulant-related bleeding are reasonably accurate during hospitalization, they are inaccurate during long-term outpatient therapy.63,97

Two prediction models have been developed and validated in outpatients treated with warfarin. Beyth et al63 identified four independent risk factors for bleeding: age ≥ 65 years, history of GI bleeding, history of stroke, and one or more of four specific comorbid conditions. This model was validated in another cohort of patients treated in another city; the cumulative incidence of major bleeding at 48 months was 53% in high-risk patients (three or four risk factors), 12% in middle-risk patients (one or two risk factors), and 3% in low-risk patients (no risk factors). Kuijer et al56 developed another prediction model based on age, gender, and the presence of malignancy. In patients classified at high, middle, and low risk, the frequency of major bleeding was 7%, 4%, and 1%, respectively, after 3 months of therapy. These prediction models should not be the sole criterion for deciding whether to initiate therapy, but should be used in conjunction with other assessments, such as the patient’s functional and cognitive status, likelihood of compliance to therapy, risk of thrombosis, and personal preference.98Clinicians can use these prediction models to help weigh the risks and benefits of coumarin therapy, potentially adjusting the intensity, type, or length of therapy or the frequency of INR monitoring. These assessments can be reviewed at the initiation of therapy and periodically assessed throughout the course of coumarin treatment. In the future, tests that assess the hepatic metabolism of coumarin, such as genotyping for cytochrome P450 polymorphisms, may help identify patients predisposed to bleeding during coumarin initiation.99

1.2 Risk of hemorrhage and clinical disorders
1.2.1 Ischemic cerebral vascular disease

Randomized trials have compared vitamin K antagonists with a placebo or nontreatment group,100105 a very-low-dose vitamin K antagonist group,106107 or an antiplatelet group,14,108112 after an acute episode of ischemic cerebrovascular disease of presumed arterial origin (for details of earlier studies see Fourth ACCP Consensus Conference on Antithrombotic Therapy).4 In all but four of these studies,106109,111 the intensity of vitamin K antagonist was high (middle of prothrombin time target corresponded to an INR of > 4). Vitamin K antagonists were associated with increased bleeding in all of these studies, with a frequency of major bleeding (often intracerebral) varying from 2 to 13% during a mean duration of follow-up of 6 to 30 months. In addition to use of high intensities of anticoagulation, unsuspected initial intracerebral hemorrhage (pre-CT era), suboptimal control of hypertension, and initiation of anticoagulation in the setting of acute stroke may have contributed to high rates of bleeding in early studies. However, there is recent evidence (see below) that ischemic stroke not due to cardioembolism is associated with a much higher risk of anticoagulant-induced intracranial bleeding than strokes that are due to embolism (eg, with atrial fibrillation).,108,113

Algra and colleagues112 combined the findings of five studies (approximately 4,000 patients) that compared vitamin K antagonists with antiplatelet therapy after transient ischemic attack or minor stroke of presumed arterial origin (approximately 4,000 patients) in a Cochrane systematic review (updated 2002). The authors estimated a risk of major bleeding with vitamin K antagonists compared with antiplatelet therapy of 1.3 (95% confidence interval [CI], 0.8 to 2.0) for INR 1.4 to 2.8; 1.2 (95% CI, 0.6 to 2.4) for INR 2.1 to 3.6; and 9.0 (95% CI, 3.9 to 21.0) for INR 3.0 to 4.5. As two studies14,109 (Stroke Prevention in Reversible Ischemia Trial [SPIRIT] and Warfarin-Aspirin Recurrent Stroke Study [WARSS]) accounted for 86% of the patients in this review and were recently published, they will be considered further.

In SPIRIT, 1,316 patients with a transient ischemia attack or minor ischemic stroke were randomized to aspirin, 30 mg/d, or warfarin therapy at a targeted INR of 3.0 to 4.5.14 There was a statistically significant increase in major bleeding associated with warfarin; 53 major bleeding complications (8.1%; 27 intracranial and 17 fatal) vs 6 major bleeding complications (0.9%) with aspirin (3 intracranial and 1 fatal) during a mean follow-up of 14 months. Bleeding increased by a factor of 1.4 for each 0.5-U increase of the INR. Previous stroke has not been identified as a risk factor for intracerebral bleeding in vitamin K antagonist-treated patients with atrial fibrillation (INR, 1.4 to 4.5)20,4344,114 (see related section of this chapter and the chapter on “Antithrombotic Therapy in Atrial Fibrillation”). An analysis of data from individual patients who were allocated to vitamin K antagonist in the European Atrial Fibrillation Trial44 (stroke with atrial fibrillation; INR, 2.5 to 4.0) and in SPIRIT14 (stroke without atrial fibrillation; INR, 3.0 to 4.5) found that, independent of other risk factors and achieved intensity of anticoagulation, risk of intracranial bleeding was at least 10 times higher if stroke was not due to atrial fibrillation (hazard ratio of 19 for intracranial, and only 1.9 for major extracranial bleeding).113

In the WARSS,109 2,206 patients were randomized to 325 mg/d of aspirin or warfarin targeted to an INR of 1.4 to 2.8 (mean achieved INR was 2.1) within 30 days of an ischemic stroke (not including cardioembolism). There was a nonsignificant trend for an increase in major bleeding with warfarin (44 events vs 30 events; 7 events vs 4 events associated with death), and a significant increase in minor bleeding (21% vs 13%) during 2 years of follow-up (the number of intracranial bleeds were not reported).109 The relative safety of oral anticoagulant therapy in the WARSS compared with the SPIRIT is consistent with intensity of anticoagulation having a major influence on risk of bleeding in patients with ischemic stroke.

1.2.2 Prosthetic heart valves

Three meta-analyses115117 on studies including only patients with prosthetic heart valves receiving long-term vitamin K antagonist therapy have been performed. Two of the analyses aimed at evaluating risks and benefits of a combination of vitamin K antagonist and antiplatelet agent with anticoagulant alone and included 5 studies115and 10 studies,116 respectively. In the meta-analysis by Cappelleri et al,115the combined regimen increased the risk of any hemorrhage by 65% and of major hemorrhage by 49%, but only the former difference was statistically significant. In the meta-analysis by Massel and Little,116the risk of major hemorrhage was increased by the addition of antiplatelet regimens (p = 0.033), but there was sufficient evidence that 100 mg of aspirin compared to higher doses provided a safer combination. A meta-analysis by Pouleur and Buyse,117 based on six trials evaluating the efficacy and safety of adding dipyridamole to anticoagulant therapy, showed that the risk of hemorrhage was identical in the two groups. The studies included in these analyses differed regarding the targeted intensity of anticoagulation (in many of the studies based on a best guess of the reagents used to perform the prothrombin time), the antiplatelet agent (aspirin or dipyridamole), and the dose of aspirin (100 mg, 500 mg, or 1,000 mg), which increases the heterogeneity of the findings. These trials are described in greater detail below.

Bleeding rates in patients receiving different regimens with long-term vitamin K antagonist therapy for prosthetic heart valves have been reported from 14 randomized clinical trials912,118127 The targeted intensity of oral anticoagulant therapy was not defined with the INR in the first five trials.118122

In all nine randomized trials of long-term vitamin K antagonist therapy in patients with mechanical heart valves since 1990, the targeted intensity of anticoagulation was reported using the INR.912,123127 The rates of major bleeding reported in these trials (Table 2 ) is, however, based on somewhat different definitions, and the time within the target INR range varied from 35 to 86%. In the study by Altman et al,,10 major bleeding was not defined; in the study by Pruefer et al,127 numerical data are not provided by treatment group. In three trials,9,11,124 different intensities of vitamin K antagonists were compared. Saour et al9 randomized patients to either warfarin therapy at a targeted INR of 2.65 or very-high-intensity warfarin therapy (targeted INR 9.0). The rate of major bleeding in the former treatment arm was 3.3%, compared with 7.2% in the latter study arm during 3.5 years of follow-up (p = 0.27). In the trial conducted by Acar et al,124 380 patients were randomized to treatment with acenocoumarol at a targeted INR of 2.0 to 3.0, or the same medication at a targeted INR of 3.0 to 4.5. The rate of major bleeding in the lower-intensity group was 9.0% compared with 12.0% in the higher-intensity group over 2.2 years (p = 0.29). In a trial conducted by Pengo and colleagues,11 205 patients were randomized to treatment with either warfarin or acenocoumarol at a targeted INR of 2.5 to 3.5 or the same medications at a targeted INR of 3.5 to 4.5, and followed up for a mean of 3 years. The rate of major bleeding was 3.8% in the former group, compared with 11% in the latter group (p = 0.019). Thus, the pooled analysis of these trials indicates that by lowering the intensity of anticoagulation, a rate difference for major hemorrhage of 1.06 per 100 patient-years (95% CI, − 0.23 to + 2.34) is achieved, when the results of these three trials are pooled.

The addition of aspirin to vitamin K antagonists has been studied in four trials. In a blinded trial, Turpie et al123 compared warfarin (INR, 3.0 to 4.5) with warfarin plus 100 mg of aspirin. The rate of major bleeding was 10.3% in the warfarin-alone group compared with 12.9% in the warfarin-plus-aspirin group after 2.5 years of follow-up (p = 0.43). Altman et al125 compared two different doses of aspirin (100 mg/d vs 650 mg/d) in patients receiving acenocoumarol at an INR of 2.0 to 3.0 followed up for an average of 24.1 months and 21.7 months, respectively. The rate of major bleeding in the lower-dose aspirin group was 7.2%, compared with 9.4% in the higher-dose group (p = 0.4). Meschengieser et al12 compared acenocoumarol alone (INR, 3.5 to 4.5) with a combination of acenocoumarol at a lower intensity (INR, 2.5 to 3.5) plus 100 mg of aspirin. The rate of major bleeding was 4.5% in the monotherapy group, compared with 2.3% in the combination therapy group after a median follow-up of 23 months (p = 0.27). The trial of Laffort et al126 is unique in its homogeneity, since only patients with St. Jude Medical (St. Paul, MN) valve prosthesis in the mitral position were included. Treatment with vitamin K antagonists alone (INR, 2.5 to 3.5) was compared with a combination of vitamin K antagonist and aspirin (200 mg/d) for 1 year, starting immediately after surgery. This may explain the high rate of major bleeding: 8.3% with monotherapy, and 19.2% with the combination (p = 0.02).

The annual bleeding rates (percentage per year) regarding major, fatal, or intracranial hemorrhage are shown in Table 2. Cannegieter et al128reported the results of a retrospective study in 1,608 patients who received vitamin K antagonist therapy for mechanical heart valves. The rate of intracranial and spinal bleeding was 0.57%/yr, and the rate of major extracranial bleeding was 2.1%/yr. Cannegieter et al129 also published the results of a meta-analysis of 46 studies (randomized trials and case series) of patients who received vitamin K antagonists for mechanical valves. The incidence of major bleeding was 1.4%/yr.

1.2.3 Atrial fibrillation

The efficacy of warfarin in preventing stroke in patients with nonvalvular atrial fibrillation has been consistently demonstrated in a number of RCTs and in meta-analyses of randomized trials.1517,29,44,114,130145 Overall, the rates of warfarin-related bleeding in these studies has been low (Table 3 ). Hart et al,139 conducted a meta-analysis of six trials that compared adjusted-dose warfarin to placebo or control. The rate of intracranial hemorrhage was 0.3%/yr with warfarin and 0.1%/yr with placebo. This difference was not statistically significant. The relative risk for major extracranial hemorrhage was 2.4 (95% CI, 1.2 to 4.6), an absolute increase of 0.3%/yr for warfarin patients. In the analysis by Hart et al,139 five trials that compared adjusted-dose warfarin with aspirin were also examined. The relative risk of intracranial hemorrhage for warfarin vs aspirin was 2.1 (95% CI, 1.0 to 4.6).139

Segal et al138 conducted a meta-analysis of five trials that compared warfarin to placebo. The odds ratio (OR) for major hemorrhage for patients receiving warfarin compared to placebo was 2.35 (95% CI, 1.20 to 4.24). The rate of major hemorrhage on placebo was 7 per 1,000 person-years, compared to 13 per 1,000 person-years receiving warfarin.

More recently, two meta-analyses142143 have evaluated the relative risks and benefits of oral anticoagulant therapy vs antiplatelet therapy in patients with atrial fibrillation. Oral anticoagulant therapy was associated with increased major bleeding (1.45 OR,143 and 1.71 hazard ratio142); the increased risk of oral anticoagulant therapy was offset by reduced nonfatal stroke (OR, 0.68),143 all strokes (hazard ratio, 0.55),142 and all cardiovascular events (hazard ratio, 0.71).142The analysis by Taylor et al143 did not include the Stroke Prevention in Atrial Fibrillation (SPAF) I and SPAF III studies15,131 or the European Atrial Fibrillation Trial study.21 The analysis by van Walraven et al142 used the pooled individual patient data from all published trials comparing oral anticoagulants with aspirin for atrial fibrillation. Oral anticoagulant therapy was associated with increased major bleeding (hazard ratio, 1.71; absolute increase, 0.9 events per 100 patient-years, p = 0.02); the corresponding hazard ratios and absolute increases in rates per 100 patient-years for all hemorrhagic stroke were 1.84 and 0.2, (p = 0.19) and for fatal bleeding were 2.15 and 0.2 (p = 0.32). This analysis concluded that treating 1,000 patients with atrial fibrillation for 1 year with oral anticoagulants rather than aspirin would prevent 23 ischemic strokes while causing nine additional major bleeds.142

Because ≥ 50% of patients with atrial fibrillation are > 75 years old, the risk-benefit of oral anticoagulant therapy in this clinical subgroup is of particular interest. One study (SPAF II)145 raised concern that the risk for warfarin-related bleeding, especially intracranial hemorrhage, may be increased substantially in patients ≥ 75 years old. The rate of major bleeding while receiving warfarin was 2.3%/yr, compared with 1.1%/yr for patients receiving aspirin, 325 mg/d. However, the rate of major warfarin-related bleeding was 4.2%/yr in patients ≥ 75 years old, compared with 1.7%/yr in younger patients; the corresponding rates for intracranial bleeding were 1.8%/yr and 0.6%/yr, respectively. The reason why these rates are substantially higher than those observed in the other clinical trials of warfarin in patients with atrial fibrillation is likely related to the intensity of anticoagulant therapy: virtually all intracranial hemorrhages in SPAF II, as in the other clinical trials, were associated with an INR > 3.0.145 In contrast, in the SPAF III trial (targeted INR, 2.0 to 3.0), the mean age was 71 years and the rate of intracranial hemorrhage was 0.5%/yr.15

There have been two trials16,114,135 evaluating a fixed low dose of warfarin (1.25 mg/d) in atrial fibrillation. These trials were stopped early because of the SPAF III trial15 results that demonstrated that low-intensity warfarin (ie, INR < 1.5) was insufficient for stroke prevention. The rates of major bleeding were low in these studies (Table 3).

The relative risk-benefit of warfarin therapy at a targeted INR of 1.5 to 2.1 compared with warfarin therapy at a targeted INR of 2.2 to 3.5 has been evaluated in a randomized trial in patients with atrial fibrillation.17 Major bleeding occurred in 6 of 55 patients in the conventional-intensity group (rate, 6.6%/yr), compared with none of the 60 patients (0%/yr) in the low-intensity group (p = 0.01). The six patients with major bleeding were all elderly (mean, 74 years) and older than the other 109 patients without major bleeding (mean, 66 years) [p < 0.01]. The INR before bleeding was < 3.0 in four patients, and was 3.1 and 3.6 in the remaining two patients, respectively. The annual rate of ischemic stroke was low in both the conventional-intensity group (1.1%/yr) and in the low-intensity group (1.7%/yr); however, the 95% CIs for these rates overlap widely, and the study is too small to make definitive conclusions about the relative effectiveness of these two different intensities of anticoagulation.

A systematic review144 compared the rates of stroke, intracranial bleeding, and major bleeding from studies of patients treated in actual clinical practice with the pooled data from RCTs. Patients in clinical practice were older and had more comorbid conditions than the patients in clinical trials. Nevertheless, the rates of ischemic stroke were similar between clinical practice and the clinical trials (1.8 and 1.4 per 100 patient-years, respectively), as were the corresponding rates of intracranial bleeding (0.1 and 0.3 per 100 patient-years, respectively), and major bleeding (1.1 and 1.3 per 100 patient-years, respectively).144 There was a higher rate of minor bleeding in clinical practice (12.0 per 100 patient-years) than in clinical trials (7.9 per 100 patient-years) [p = 0.002].

1.2.4 Ischemic heart disease

Anand and Yusuf146 conducted a meta-analysis of trials evaluating vitamin K antagonists in patients with coronary artery disease. Trials were stratified based on the intensity of the vitamin K antagonist and on the use of aspirin. In 16 trials (10,056 patients) of high-intensity therapy (INR, 2.8 to 4.8), the reduction in mortality and thromboembolic complications was offset by a sixfold increase (95% CI, 4.4- to 8.2-fold) in major bleeding. For moderate-intensity therapy (INR, 2 to 3) vs control (four trials; 1,365 patients), the relative risk for major bleeding was 7.7 (95% CI, 3.3 to 18). For moderate-to-high-intensity therapy (INR ≥ 2) vs aspirin (seven trials; 3,457 patients), there was a relative risk of 2.4 (95% CI, 1.6 to 3.6) for increase in major bleeding. For low-intensity therapy (INR < 2.0) and aspirin vs aspirin alone (three trials; 8,435 patients), the relative risk for major bleeding was 1.3 (95% CI, 1.0 to 1.8), with no significant reductions in mortality or cardiovascular events.

There are 13 published randomized trials71,74,76,147158 of long-term oral anticoagulant therapy in patients with acute myocardial infarction (Table 4 ). These 13 trials compared the following regimens: anticoagulant therapy was compared with placebo or control (n = 7),71,147149,152156; anticoagulant therapy with aspirin (n = 1)150; anticoagulant therapy with aspirin or placebo (n = 1)151; fixed low doses of warfarin (1 mg or 3 mg) combined with aspirin were compared with aspirin alone (n = 1)74; anticoagulant therapy alone or combined with aspirin compared to aspirin alone (n = 2)157158; and anticoagulant therapy with aspirin compared to aspirin (n = 1).76 The frequency of major bleeding ranged from 0 to 10%, and fatal bleeding ranged from 0 to 2.9%.

Smith et al147 reported the results of a randomized trial that renewed interest in the long-term use of oral anticoagulants after myocardial infarction. The targeted INR was 2.8 to 4.8. Five patients in the warfarin group (0.8%) had intracranial hemorrhages, and three of these were fatal. Eight warfarin-treated patients (1.3%) experienced major extracranial bleeds. There were no major bleeds in the placebo group.

In a trial conducted by the Anticoagulants in the Secondary Prevention of Events in Coronary Thrombosis (ASPECT)-1 investigators, patients who had sustained a myocardial infarction were randomized to either oral anticoagulant therapy at a targeted INR of 2.8 to 4.8 or placebo.156 The mean follow-up was 37 months. Seventy-three patients (4.3%) in the anticoagulant group experienced major bleeding, compared with 19 placebo-treated patients (1.1%). Three extracranial bleeds in the anticoagulant group were fatal; all were GI in origin. Cerebral hemorrhage was more common in patients who had been treated with anticoagulants (n = 17; 1%), 8 of which were fatal, compared with 2 cerebral hemorrhages in placebo-treated patients, none of which were fatal. The rate of major bleeding in the anticoagulant-treated group was 1.5%/yr, compared with 0.2%/yr in the placebo-treated group. This difference was statistically significant.

The Coumadin-Aspirin Reinfarction Study (CARS)74 compared long-term treatment using fixed low doses of warfarin (1 mg or 3 mg) combined with aspirin, 80 mg, to treatment with aspirin alone (160 mg) using a randomized, blinded study design. The median follow-up of the CARS was 14 months. The median INR values 4 weeks and 6 months after beginning treatment were 1.27 and 1.19, respectively, for patients receiving 3 mg of warfarin with 80 mg of aspirin. Major hemorrhage, including those related to invasive procedures, occurred in 75 patients (2.0%/yr) receiving 3 mg of warfarin with 80 mg of aspirin, 42 patients (1.7%/yr) receiving 1 mg of warfarin with 80 mg of aspirin, and 57 patients (1.5%/yr) receiving aspirin alone. For spontaneous major hemorrhage (not procedure related), annual rates were 1.4% in the group receiving 3 mg of warfarin plus 80 mg of aspirin, 1.0% in the group receiving 1 mg of warfarin plus 80 mg of aspirin, and 0.74% in the group receiving 160 mg of aspirin alone.

The Combination Hemotherapy and Mortality Prevention (CHAMP) study76 compared the efficacy of warfarin (target INR, 1.5 to 2.5) with 81 mg of aspirin to 162 mg of aspirin alone in patients after myocardial infarction who were followed up for a median of 2.7 years. In the CHAMP study,76 major bleeding occurred more frequently in the combination therapy group than in the aspirin-alone group (1.28 events vs 0.72 events/100 patient-years; p < 0.001).

The ASPECT-2 trial157 compared 80 mg of aspirin, high-intensity (INR, 3.0 to 4.0) vitamin K antagonist, or combination of 80 mg of aspirin and moderate-intensity (INR, 2.0 to 2.5) vitamin K antagonist in patients who had survived acute coronary events; the median follow-up was 12 months. Major bleeding rates were 1%, 1%, and 2% per patient-year in the aspirin, warfarin, and combination groups, respectively. The frequency of minor bleeding was 5%, 8%, and 15% per patient-years in the aspirin, warfarin, and combination groups, respectively.

The Warfarin-Aspirin Reinfarction Study (WARIS) II158 compared the efficacy and safety of warfarin (target INR, 2.8 to 4.2), aspirin (160 mg), and the two combined (75 mg of aspirin with warfarin with target INR of 2.0 to 2.5) in a long-term, randomized, unblinded multicenter study involving 3,606 patients after acute myocardial infarction (1,202 patients in each treatment group). The mean duration of follow-up in WARIS II was 4 years. Major nonfatal bleeding occurred in 0.62% of patients per treatment-year in both warfarin groups, and in 0.17% of patients receiving aspirin (p < 0.001).

1.2.5 Venous thromboembolism Heparins vs vitamin K antagonists.

Linkins and colleagues159 performed a meta-analysis of 33 prospective studies to determine how often major episodes of bleeding that occurred during vitamin K antagonist therapy for venous thromboembolism (VTE) were fatal (target INR, 2.0 to 3.0). Most of the 10,757 patients included in the analysis were initially treated with unfractionated heparin (UFH) or low molecular weight heparin (LMWH) and received vitamin K antagonists for 3 months.159 Of 275 major bleeds, 37 bleeds were fatal, for an overall case fatality of 13%. Case fatality with intracranial bleeds was 46%, and for major extracranial bleeds was 10%. Extracranial bleeds, which accounted for 81% of major bleeding episodes, caused approximately two thirds of the deaths from bleeding. Although the risk of bleeding was higher during the first 3 months of anticoagulant therapy than subsequently, case fatality with major bleeding was similar during the acute and long-term phases of treatment.159

There have been 16 randomized trials160176 in patients with VTE in which vitamin K antagonists were compared with various subcutaneous heparin regimens, usually over a 3-month period (Table 5 ). In one study,,7 two intensities or vitamin K antagonist therapy were compared following initial heparin therapy (Table 5). Higher intensities of anticoagulation (ie, INR, 2.6 to 4.4) were evaluated in earlier studies,160162 than in more recent trials (ie, INR, 2.0 to 3.0).,7,163173,175176 The higher-intensity regimens were consistently associated with more total bleeding than the comparison study arms, with a similar trend for major bleeding (Table 5). In the study,7 that compared two intensities of oral anticoagulation, the frequency of total bleeding was also substantially lower with the less-intense regimen (4% vs 22%), without being associated with a loss of antithrombotic efficacy.

The 12 most recent of these studies163173,175 compared oral anticoagulant therapy at a targeted INR of 2.0 to 3.0 anticoagulation with widely differing regimens of three LMWH preparations (Table 5). The daily LMWH dose was as low as 4,000 IU,,163,167 to as high as 200 IU/kg,169,173approximately a 3.5-fold difference. Two meta-analyses174,177 of studies that compared LMWH with vitamin K antagonist, each administered for 3 months after initial heparin therapy, were performed. In the analysis by Iorio and colleagues,177 which included seven studies163165,167169,171 and a total of 1,379 patients, there was a trend toward less bleeding with long-term LMWH therapy (OR, 0.45; 95% CI, 0.2 to 1.1).177 Importantly, the relative frequency of major bleeding with LMWH therapy was dose dependent, varying from an OR of approximately 0.2 at a dose of approximately 4,000 IU/d, to an OR of approximately 0.7 at a dose of approximately 12,000 IU/d (p = 0.03).

The results of randomized trials in which patients with VTE were treated with less-intense oral anticoagulation (not part of primary comparison) following initial treatment with either UFH or LMWH generally have found an overall frequency of major bleeding of approximately ≤ 3% during 3 months of therapy,178 with higher rates among patients with cancer.170,173 Different durations of anticoagulation.

Four randomized trials have compared 4 weeks179180 or 6 weeks181182 with 3 months179180 or 6 months,181182 and three studies have compared 3 months182184 with 6 months182,184 or 12 months183184 of oral anticoagulation (INR, approximately 2.0 to 3.0) for the treatment of VTE. Following the initial phase of treatment during which all patients were treated with anticoagulants, major bleeding occurred infrequently without convincing evidence of more bleeding with the longer durations of therapy.179183 Two randomized trials185186 evaluated long-term oral anticoagulation for the prevention of recurrent VTE following an acute episode. Schulman et al185randomized 227 patients to regimens of either 6 months or 4 years of anticoagulation (INR, 2.0 to 2.85) after a second episode of VTE. Major bleeding occurred more frequently in patients who were treated with long-term anticoagulation (2.4%/yr vs 0.7%/yr). Kearon et al186 randomized 162 patients with a first episode of idiopathic VTE to remain on a regimen of warfarin (INR, 2.0 to 3.0), or to receive placebo, for a further 2 years after an initial 3 months of anticoagulation. Major bleeding occurred more frequently in patients who continued to receive anticoagulants (3.8%/yr vs 0%/yr).

A reasonable interpretation of the findings of these studies is that, for patients without a high risk of bleeding (ie, such as those entered in clinical trials), differences in duration of anticoagulation do not translate into clinically important differences in frequencies of major bleeding until duration of therapy differs by a year or longer. Different intensities of anticoagulation for extended treatment.

The Extended Low-intensity Anticoagulation trial,61 which compared warfarin therapy with INR of 1.5 to 1.9 and INR of 2.0 to 3.0 for long-term prevention of recurrent unprovoked VTE in 738 patients who had completed at least 3 months of initial treatment with an INR of 2.0 to 3.0, found no difference in the frequency of major bleeding between the two groups during an average of 2.4 years of follow-up (INR, 1.5 to 1.9 vs INR, 2.0 to 3.0: 1.1%/yr vs 0.9%/yr; hazard ratio, 1.2; 95% CI, 0.4 to 3.0). The Prevention of Recurrent Venous Thromboembolism trial,187 which compared warfarin therapy with INR of 1.5 to 2.0 and placebo in 508 patients who had completed at least 3 months of warfarin treatment with an INR of 2.0 to 3.0, found no significant difference in the rate of major bleeding between the two groups during an average of 2.1 years of follow-up (INR, 1.5 to 2.0 vs placebo: 0.9%/yr vs 0.4%/yr; hazard ratio, 2.5; 95% CI, 0.5 to 13.0).

2.1 Risk of hemorrhage and clinical disorders

The direct thrombin inhibitor ximelagatran has been compared to warfarin in patients with nonvalvular atrial fibrillation in the Stroke Prevention Using an Oral Thrombin Inhibitor in Atrial Fibrillation (SPORTIF) II trial.188In SPORTIF II, 257 patients were randomized to receive one of three twice-daily doses of ximelagatran (20 mg, 40 mg, or 60 mg) or warfarin (INR, 2.0 to 3.0) for 3 months. There were no major bleeds in the ximelagatran group and one in the warfarin group. In the SPORTIF III trial,189 3,407 patients with nonvalvular atrial fibrillation received ximelagatran, 36 mg bid, or warfarin (INR, 2.0 to 3.0). The rates of major bleeding were 1.3% and 1.8%, respectively. This difference was not statistically significant.

Oral direct thrombin inhibition with ximelagatran at doses of 24 mg, 36 mg, 48 mg, or 60 mg bid plus 160 mg/d of aspirin was compared to 160 mg/d of aspirin alone in a recent multicenter blinded trial190 for secondary prevention of myocardial infarction. There were 1,883 patients followed up for a 6-month treatment period. The rates of major bleeding did not differ between treatment groups (1% for aspirin alone vs 2% for combined ximelagatran doses), but patients in the combined ximelagatran groups were three times more likely to stop therapy due to bleeding (hazard ratio, 3.35; 95% CI, 1.87 to 6.01). In addition, any bleeding (major and minor) was more frequent in the combined ximelagatran group (22%) compared to the aspirin-alone group (13%) [hazard ratio, 1.76; 95% CI, 1.38 to 2.25].

Ximelagatran has been evaluated for both short-term and long-term treatment of VTE (Thrombin Inhibitors in Venous Thromboembolism studies).191192 In the short-term treatment study,192 2,491 patients with acute DVT were treated for 6 months with ximelagatran, 36 mg bid, or LMWH followed by vitamin K antagonist therapy (INR, 2.0 to 3.0), using a blinded design. An “on-treatment” analysis suggested less major bleeding with ximelagatran (1.3% vs 2.2%; 95% CI for difference, −2.0 to + 0.2%); intention-to-treat analyses have not been reported for bleeding.192

In a long-term treatment study,191 18 months of ximelagatran, 24 mg bid, was compared with placebo in 1,224 patients with DVT or pulmonary embolism who had completed 6 months of initial treatment with vitamin K antagonists. There was no apparent increase of major bleeding with ximelagatran (0.7%/yr; hazard ratio, 1.2; 95% CI, 0.4 to 3.8).

Heparin is usually administered in low doses by subcutaneous injection to prevent venous thrombosis (prophylactic heparin), in higher doses to treat patients with acute VTE or with acute coronary syndromes (therapeutic heparin), and in very high doses in patients during open-heart surgery. In this chapter, we will discuss only bleeding associated with therapeutic heparin (see the chapter by Geerts et al for a discussion of bleeding associated with prophylactic heparin). Heparin has the potential to induce bleeding by inhibiting blood coagulation, by impairing platelet function,193and by increasing capillary permeability.194 Heparin can also produce thrombocytopenia, but this is rarely an important cause of bleeding.

3.1 Determinants of bleeding
3.1.1 Relationship between risk of bleeding and heparin dose/response

Since the anticoagulant response to heparin (measured by a test of blood coagulation, eg, the activated partial thromboplastin time [APTT]) is influenced by the heparin dose, it was not possible from reported studies to separate the effects of these two variables (dose and laboratory response) on hemorrhagic rates. To our knowledge, there have been no randomized trials in patients with established VTE directly comparing different doses of heparin. In a study195 evaluating prophylaxis in patients with recent-onset traumatic spinal cord injuries, the incidence of bleeding was significantly greater in patients randomized to receive heparin adjusted to maintain the APTT at 1.5 times control than compared with heparin, 5,000 U bid. The mean dose of heparin for the adjusted-dose regimen was 13,200 U bid. Bleeding occurred in seven adjusted-dose patients compared with none in the fixed-dose group.

Subgroup analysis of randomized trials and prospective cohort studies provide suggestive evidence for an association between the incidence of bleeding and the anticoagulant response. In the Urokinase Pulmonary Embolism Study,196bleeding occurred in 20% of patients assigned to heparin in whom whole-blood clotting time was > 60 min, compared to 5% of those whose whole-blood clotting time was < 60 min (relative risk, 4.0). Norman and Provan197reported five major bleeds in 10 patients whose APTT was prolonged to more than twice the upper limit of their therapeutic range for at least 50% of their assays, but in only 1 of 40 patients whose APTT remained therapeutic (relative risk, 20.0). Wilson et al198described 80 nonsurgical patients receiving heparin monitored by the whole-blood clotting time. Fifty-six percent who received “excessive heparin” bled, whereas only 16% who did not receive excessive heparin bled (relative risk, 3.5). Anand et al199 examined the relationship between the APTT and bleeding in 5,058 patients with acute coronary syndrome who received IV heparin in the Organization to Assess Strategies for Ischemic Syndromes-2 trial. For every 10-s increase in the APTT, the major bleeding was increased by 7% (p = 0.0004).

Although none of the studies were designed to compare the effects on bleeding of either different doses of heparin or different levels of heparin response, there is a suggestion that bleeding is more likely to occur when an in vitro test of coagulation is prolonged excessively, but this evidence is by no means definitive. In addition, there is good evidence that serious bleeding during heparin treatment can occur when the anticoagulant response is in the therapeutic range. Finally, the results of Global Utilization of Streptokinase and t-PA for Occluded Coronary Arteries IIa and the Thrombolysis in Myocardial Infarction (TIMI) 9A studies in patients with ischemic coronary syndromes indicated that a 20% increase in the IV heparin dose above the 1,000 U/h that was used in the Global Utilization of Streptokinase and t-PA for Occluded Coronary Arteries I study increased the risk of intracranial bleeding when combined with thrombolytic therapy.200201

3.1.2 Relationship between risk of bleeding and method of administering heparin

The evidence for a relationship between the risk of bleeding and the method of administering heparin comes from randomized trials in which UFH was either administered by continuous IV infusion with intermittent IV injection,198,202206 continuous IV heparin with subcutaneous heparin,207210 continuous IV heparin for approximately 10 days with a shorter course (4 to 5 days),211212 continuous IV heparin and oral anticoagulants compared with oral anticoagulants alone,213continuous IV heparin administered on a weight-adjusted basis with a standard clinical approach (5,000-U bolus, 1,000 U/h),214 and continuous IV heparin monitored using either the APTT or monitored using a heparin assay.

In summary, there was an increased rate of major bleeding with intermittent IV heparin compared with continuous IV infusion. Continuous IV heparin caused less bleeding than intermittent IV heparin; continuous IV heparin and subcutaneous heparin were associated with a similar amount of bleeding; and continuous IV heparin for approximately 10 days and 5 days caused a similar amount of bleeding.

3.1.3 Relationship between the risk of bleeding and patient risk factors

There is good evidence that comorbid conditions, particularly recent surgery or trauma, are very important risk factors for heparin-induced bleeding.52,205,215 This association was demonstrated in the study by Hull and associates215 in patients with proximal vein thrombosis. Patients without clinical risk factors for bleeding were treated with a starting dose of 40,000 U of heparin by continuous infusion, while those with well-recognized risk factors for bleeding (recent surgery, trauma) received a starting dose of 30,000 U. Bleeding occurred in 1 of 88 low-risk patients (1%) who received 40,000 U initially and 12 of 111 high-risk patients (11%) who received 30,000 U.

The concomitant use of aspirin was identified as a risk factor in early retrospective studies216and corroborated by Sethi and associates.217 In their study in patients undergoing aortocoronary bypass surgery, the preoperative use of aspirin caused excessive operative bleeding in patients who receive very high doses of heparin as part of the routine for bypass procedures. Although the concomitant use of aspirin is associated with heparin-induced bleeding, this combination is used frequently in the initial treatment of acute coronary artery syndromes without serious bleeding. The risk of heparin-associated bleeding increases with concomitant thrombolytic therapy4 or GP IIb/IIIa antagonists.218219 In a retrospective analysis of 5,216 patients undergoing percutaneous coronary intervention using UFH, bleeding complications incrementally increased at all activated clotting times > 200 s.220

Renal failure and patient gender have also been implicated as risk factors for heparin-induced bleeding.221222 The reported association with female gender has not been consistent among studies and remains in question.

Other studies221,223have reported that older patients had a higher risk of heparin-induced bleeding. In an analysis of a randomized trial,224 age > 70 years was associated with a clinically important increased risk of major bleeding.

3.2 Risk of hemorrhage and clinical disorder
3.2.1 VTE Initial therapy.

Trials that have evaluated acute treatment of VTE with different heparin regimens have generally compared fixed-dose LMWH with adjusted-dose UFH administered IV225237 (Table 6 ) or subcutaneously,229,238 (Table 7 ). The results of these studies have been combined in a number of meta-analysis.,178,239240 In relationship to major bleeding, the findings of these studies and meta-analysis can be summarized as follows. Overall, LMWH has generally been associated with less bleeding than UFH (eg, OR, 0.60; 95% CI, 0.39 to 0.93),240; however, this finding has been changing over time. LMWH was associated with less bleeding than UFH in studies published before 1997 (OR, 0.53; 95% CI, 0.28 to 0.98), but has been associated with a similar frequency of bleeding in more recent studies (OR, 0.97; 95% CI, 0.52 to 1.81).240 It is not clear if between study differences in the relative frequency of bleeding with LMWH and UFH reflect differences in LMWH regimens, differences in UFH regimens, or occurred by chance. Direct comparisons among LMWH regimens.

Once-daily and twice-daily administration of the same LMWH have been directly compared in six studies241246 (the same total daily dose of LMWH has not always been compared within studies). A meta-analysis247 of five of these studies241244,246 that had unconfounded comparisons found no increase in major bleeding with once-daily treatment (OR, 1.2; 95% CI, 0.4 to 3.2). Outpatient and inpatient administration of LMWH (three preparations were used) was compared in a single study of 201 patients; two major bleeds occurred in each group.248Tinzaparin and dalteparin, each administered once daily, have been compared for outpatient treatment of VTE in a study of 497 patients; there was no difference in major bleeding (five bleeds with tinzaparin, and three bleeds with dalteparin).249 Long-term treatment with UFH and LMWH.

Trials of ≥ 3 months of treatment with UFH or LMWH compared with oral anticoagulants have been described earlier in this chapter (Table 5). In another study of 80 patients that compared 10,000 U of UFH with 5,000 IU of dalteparin, with each administered subcutaneously twice daily for 3 months after acute VTE, there was no difference in the frequency of total bleeding and no episodes of major bleeding in either group.250 Fondaparinux vs UFH or LMWH.

The synthetic pentasaccharide fondaparinux has been evaluated for acute treatment of pulmonary embolism and DVT (Matisse studies).251252 In the Matisse-PE trial,251 2,213 patients were treated with a single subcutaneous dose of fondaparinux (7.5 mg if 50 to 100 kg) or IV UFH (APTT ratio, 1.5 to 2.5) for at least 5 days using an open-label design. Major bleeding occurred in 1.3% of patients receiving fondaparinux, and in 1.1% of patients receiving UFH during the initial treatment period.251

In the Matisse-DVT trial,252 2,205 patients were treated with a single subcutaneous dose of fondaparinux (7.5 mg if 50 to 100 kg) or twice-daily subcutaneous LMWH (enoxaparin, 1 mg/kg) for at least 5 days using a blinded design. Major bleeding occurred in 1.1% of patients receiving fondaparinux, and in 1.2% of patients receiving LMWH during the initial treatment period.252

3.3 Ischemic cerebral vascular disease
3.3.1 Anticoagulants vs control

Approximately 20, mostly small, studies have compared early anticoagulant therapy with control in patients with acute ischemic stroke. The finding of studies that were published by 1999 (approximately 23,000 patients) were combined in a Cochrane systematic review by Gubitz and colleagues.253The trials in this overview evaluated UFH, LMWH, heparinoid, oral anticoagulants, and direct thrombin inhibitors administered in differing doses (eg, prophylactic doses for VTE; therapeutic doses for stroke) and by different routes (eg, subcutaneously or IV). In relationship to bleeding, the review found that acute anticoagulation had the following effects: (1) increased symptomatic intracranial bleeding approximately 2.5-fold (an excess of 9 per 1,000 patients), and (2) increased major extracranial bleeding approximately threefold (an excess of 9 per 1,000 patients). As the International Stroke Trial (IST),254 with > 19,000 patients, accounted for > 75% of patients in the analysis, it will be considered further.

In the IST, patients with acute ischemic stroke were treated with aspirin (300 mg/d), subcutaneous heparin (5,000 U bid or 12,500 U bid), both, or neither (Table 8 ).254 Heparin was associated with a dose-dependent increase of both intracranial and extracranial bleeding (all major bleeds: control, 0.6%; heparin 5,000 U bid, 1.1%; heparin 12,500 U bid, 3.2%), which, at the higher dose, more than offset antithrombotic benefit. Patients who had the highest risk of recurrent ischemic stroke also had the highest risk of intracerebral bleeding. For example, in patients with acute ischemic stroke and atrial fibrillation, although the frequency of hemorrhagic stroke after 14 days was 2.1% (32 of 1,557 patients) with heparin therapy (either dose) compared with 0.4% (7 of 1,612 patients) without heparin therapy, there was no difference in the combined end point of recurrent ischemic or hemorrhagic stroke.

More recently, the Therapy Of Patients With Acute Stroke study255 compared four doses of a LMWH (certoparin) in 400 patients with ischemic stroke and found a trend to more major bleeding with the highest dose compared to the three lower doses combined (9.0% vs 2.0%).

3.3.2 Anticoagulants vs antiplatelet agents

Four trials254,256258 have compared anticoagulants with antiplatelet agents in patients with acute ischemic stroke. UFH254,258 and LMWH256257 in low doses254,258 or high doses254,256257 were compared with aspirin254,256257 or aspirin and dipyridamole.258 Three trials254,257258 included cardioembolic and noncardioembolic strokes, whereas one study256 was confined to cardioembolic strokes. The findings of these four studies, with data from > 16,000 patients (88% from the IST),254 have been combined in a Cochrane systematic review by Berge and Sandercock.259 The review found that, compared to antiplatelet therapy, acute anticoagulation increased symptomatic intracranial hemorrhage 2.3-fold (an excess of 10 per 1,000 patients) and increased major extracranial hemorrhage 1.9-fold (an excess of 5 per 1,000 patients treated). The increase in major bleeding with anticoagulants was mostly confined to high-dose regimens.259

3.4 Ischemic coronary syndromes

There have been two trials260261 in which patients with ischemic coronary artery disease were randomized to treatment with heparin or no heparin, one trial260 in which heparin was compared with aspirin, and one trial262 in which high-dose heparin therapy was compared with a lower dose of heparin. The results of these trials have shown that heparin administered alone in patients with coronary artery disease (without concurrent thrombolytic therapy) is not associated with an increased risk of major bleeding.4

LMWH has been compared with a no-treatment control or IV UFH in several trials in patients with unstable coronary artery disease.263266 For IV UFH, the rates of major bleeding range from 0 to 6.3% during the initial 8 days of treatment, and from 0.3 to 3.2% during the long-term treatment phase between approximately 1 week and 3 months. For several of the trials, explicit data for the incidence of fatal bleeding were not reported. The data in Table 8 support the inference that LMWH does not result in an increased risk of major bleeding compared with IV UFH. The absolute rates of major bleeding were higher in more recent trials269 than were observed in the initial large trials of LMWH.264265 This is probably due to inclusion in the more recent studies of patients who undergo cardiac catheterization or coronary bypass surgery; much of the major bleeding in these trials267269 was associated with invasive vascular procedures or coronary bypass surgery. In contrast, the earlier studies264265 excluded patients for whom catheterization, angioplasty, or coronary bypass surgery was planned. A meta-analysis270of randomized trials comparing UFH or LMWH with placebo or untreated control, or comparing UFH with LMWH, for the short-term and long-term management of patients with acute coronary syndrome without ST-segment elevation identified 12 trials, involving a total of 17,157 patients. Long-term LMWH (> 7 days) was associated with a significantly increased risk of major bleeding (OR, 2.26; 95% CI, 1.63 to 3.14], p < 0.0001), which is equivalent to 12 major bleeds per 1,000 patients treated. In a multicenter, blinded, placebo-controlled trial271 to examine the efficacy and safety of twice-daily injections of weight-adjusted enoxaparin or placebo for 14 days after stenting in patients at high risk for stent thrombosis, the groups had comparable rates of major bleeding (3.3% for enoxaparin, and 1.6% for placebo; p = 0.08), but minor bleeding was increased with enoxaparin (25% vs 5.1%; p ≤ 0.001).

Bleeding is the major complication of anticoagulant therapy. The criteria for defining the severity of bleeding varied considerably between studies, accounting in part for the variation in the rates of bleeding reported. Since the last review, there have been several meta-analyses published on the rates of major bleeding in trials of anticoagulants for atrial fibrillation, VTE, and ischemic heart disease.