Risks and prevention of bleeding with oral anticoagulants

INTRODUCTION — There is no anticoagulant that reduces thrombotic risk without simultaneously increasing the risk of bleeding. The decision to administer an anticoagulant is based on the assessment that the risk of thrombosis and its complications is a greater clinical concern than the risk of bleeding and its complications for the specific patient at a specific point in time.

This topic reviews the risks of bleeding with oral anticoagulants, comparison of bleeding rates, and strategies to reduce the risk of clinically serious bleeding.

The management of bleeding with specific anticoagulants is discussed in separate topic reviews:

●Direct oral anticoagulants (apixaban, dabigatran, edoxaban, and rivaroxaban) – (See "Management of bleeding in patients receiving direct oral anticoagulants".)

●Warfarin-associated intracerebral hemorrhage – (See "Reversal of anticoagulation in intracranial hemorrhage".)

●Other warfarin-associated bleeding or supratherapeutic INR – (See "Biology of warfarin and modulators of INR control", section on 'Overview of INR control'.)

●Heparins – (See "Heparin and LMW heparin: Dosing and adverse effects", section on 'Bleeding'.)

●Fondaparinux – (See "Fondaparinux: Dosing and adverse effects", section on 'Bleeding'.)

PATHOGENESIS OF ANTICOAGULANT-ASSOCIATED BLEEDING

Loss of vascular integrity — Technically, anticoagulants do not cause bleeding; bleeding is caused by a breach in the wall of a blood vessel. However, anticoagulants interfere with the normal hemostatic process that resolves microscopic bleeding events that would otherwise never become clinically apparent; as a result, anticoagulation may contribute to hematoma expansion and may convert clinically insignificant bleeding to clinically significant bleeding.

Breaches of vascular integrity may be mechanical (eg, from trauma, tumor invasion, thrombosis, or hypertension) or due to altered endothelial cell barrier function (eg, with sepsis, ischemia, or certain chemotherapeutic drugs or biologic agents).

Microbleeds and other subclinical bleeding events — Subclinical bleeding (cerebral microbleeds and occult bleeding in other sites such as the gastrointestinal tract) may present as clinically significant bleeding when the individual is receiving an anticoagulant.

Microbleeds are clinically silent minor bleeding events that are apparent on imaging studies; the term is generally restricted to bleeding in the brain and generally refers to bleeding in the setting of hypertensive vasculopathy and/or cerebral amyloid angiopathy. (See "Cerebral amyloid angiopathy", section on 'Microbleeds'.)

Neuroimaging data suggest that microscopic pseudoaneurysm formation with subclinical leakage of blood is a potential mechanism for developing cerebral microbleeds. The regional distribution of these bleeds is consistent with the usual location of intracerebral hemorrhage (ICH) in these conditions.

In addition to hypertension and cerebral amyloid angiopathy, cerebral microbleeds have been associated with other angiopathies, dementia, trauma, diabetes, cigarette smoking, chronic obstructive pulmonary disease (independent of smoking), and normal aging [1]. They are often observed in individuals with ICH and ischemic stroke, as well as in healthy individuals without clinical evidence of stroke.

We remain uncertain whether cerebral microbleeds should be used, in addition to other factors, to identify patients who may experience net harm from anticoagulant therapy.

Subclinical bleeding in other tissues or organs (eg, gastrointestinal) may be responsible for early bleeding seen with anticoagulant initiation. (See 'Anticoagulant initiation' below.)

Screening for cerebral microbleeds or occult gastrointestinal bleeding is not routinely used in individuals on anticoagulation, as discussed below. (See 'Risk factors for bleeding in specific sites' below.)

RISK FACTORS FOR BLEEDING — Bleeding risk depends on a number of anticoagulant-related factors and individual patient characteristics (table 1). Some of these are modifiable, such as the choice of anticoagulant, dosing, and use or avoidance of other medications that increase bleeding risk. Others are fixed, including patient age and ethnicity. Some underlying medical conditions and comorbidities can be modified and others cannot. Patients should be informed that anticoagulation may increase their baseline bleeding risk. Information should be provided about strategies they can use to minimize this risk and symptoms of bleeding for which they should seek medical attention. Given the high fatality rate related to these bleeding events, patients should be advised to promptly seek medical attention should they develop signs or symptoms of major bleeding.

Risk factors related to the anticoagulant

Drug class — In general, the risk of life-threatening and fatal bleeding is lower with direct oral anticoagulants (DOACs; dabigatran and direct factor Xa inhibitors apixaban, edoxaban, rivaroxaban) than with vitamin K antagonists, including warfarin.

Evidence for this lower risk comes from randomized trials in patients with atrial fibrillation (AF) or venous thromboembolism (VTE). The absolute risks in patients with AF and VTE may differ because these conditions may be distributed across the population differently (eg, individuals with AF tend to be older and take more medications that may increase the risk of bleeding; those with VTE are generally younger and may have other factors, such as menstrual blood loss, that increase their risk of bleeding).

However, individuals ≥75 years of age who are taking warfarin may not derive significant benefit from switching to a DOAC [2].

Summaries of available data include:

●AF – An increase in bleeding risk (compared with the general population) is seen with all anticoagulants, but the absolute risk difference associated with anticoagulant therapy is generally small.

•Warfarin – Rates of major bleeding in patients in early clinical trials who were receiving warfarin versus aspirin are presented in the table (table 2). In an observational study involving 125,195 adults receiving warfarin for AF who were age ≥66 years (median, 77 years), the overall bleeding rate was approximately 4 percent per person-year [3]. Bleeding that resulted in a visit to the hospital was seen at a rate 3.8 events per 100 person-years; during the first month of therapy, this risk was 11.8 per 100 person-years. Additional discussions of bleeding risks in individuals with AF are presented separately.

•DOACs – In each of three major randomized trials comparing warfarin with a DOAC for primary stroke prevention in AF (eg, ARISTOTLE [apixaban], RELY [dabigatran], ROCKET AF [rivaroxaban]), the overall bleeding rate was lower with the DOAC than with warfarin. Compared with warfarin, the relative risks (RR) for major bleeding and intracranial bleeding with DOACs were on the order of 0.8 and 0.5, respectively [4,5]. These findings have been confirmed in numerous studies and systematic reviews [6].

Despite this favorable comparison with warfarin, DOACs probably confer a small increased risk of intracerebral hemorrhage (ICH) when compared with patients who are not receiving an oral anticoagulant. This was illustrated in a 2018 systematic review and meta-analysis of the risk of ICH with a DOAC versus aspirin (five randomized trials and nearly 40,000 patients), which found a trend towards increased likelihood of ICH with DOACs that reached statistical significance at higher doses [7]:

-Rivaroxaban 15 to 20 mg once daily: Odds ratio (OR) 3.31, 95% CI 1.42-7.72

-Rivaroxaban 10 mg once daily or 5 mg twice daily: OR 1.43, 95% CI 0.93-2.21

-Apixaban 5 mg twice daily: OR 0.84, 95% CI 0.38-1.88

While the annual risk for ICH in DOAC-treated patients with AF was approximately 3 per 1000 (0.3 percent), the risk of ICH with low-dose aspirin would be expected to be smaller (perhaps 1 to 2 per 1000 per year). The primary data from these trials are discussed in detail separately, along with other considerations in the choice of anticoagulant in AF. (See "Atrial fibrillation in adults: Use of oral anticoagulants".)

●VTE – The overall annualized major bleeding risk in large randomized trials was in the range of 1.2 to 2.2 percent during the acute/initial phase of treatment [8]. Additional discussions of bleeding risk in individuals with VTE are presented separately. (See "Overview of the treatment of proximal and distal lower extremity deep vein thrombosis (DVT)", section on 'Assessing bleeding risk' and "Venous thromboembolism: Initiation of anticoagulation".)

In many cases, real-world bleeding risk is likely to be higher than the risk reported from clinical trials in which both patients and centers have typically been carefully selected.

Comparisons of bleeding risks between the various available DOACs are limited by the lack of direct comparisons in randomized, controlled head-to-head studies. Indirect comparisons are limited by bias associated with the rationale for prescribing one DOAC over another in an individual patient. Thus, while we are confident there is a class effect favoring DOACs over warfarin with respect to bleeding, there is insufficient high-quality evidence to recommend any one DOAC over any other, particularly when considering the risk of life-threatening or fatal bleeding.

Some individuals believe that the wider availability of a reversal agent for warfarin (prothrombin complex concentrates [PCCs] and vitamin K) counterbalances the higher risk of bleeding with warfarin, but preventing bleeding in the first place is always preferable to treating bleeding after it has occurred. Even in warfarin-treated patients in clinical trials, the time to reversal is often hours and the morbidities associated with bleeding are great. Further, mortality was higher in warfarin-treated patients who bled than it was in DOAC-treated patients in the era when there were two effective reversal agents for warfarin (vitamin K and PCC) and no effective reversal agents for DOAC bleeding [4,9]. Subsequently, reversal agents for DOACs have become more widely available. (See "Management of warfarin-associated bleeding or supratherapeutic INR", section on 'PCC products'.)

Anticoagulant initiation — Numerous studies, including those mentioned above, have demonstrated that the risk of bleeding is highest during the initial period (typically defined as the first three months) of anticoagulation, regardless of the indication for the anticoagulant. Examples of the absolute differences at different bleeding risks are shown in the table (table 3). The mechanism is not completely understood. One hypothesis is that subclinical bleeding, if present, becomes apparent early during anticoagulation. A related concept is that individuals who have not bled during the first three months of anticoagulation are a select group of individuals who have an inherently lower risk of future bleeding [10].

Dose level — The intensity of anticoagulation for warfarin is based on the INR; for the DOACs, it is based on the standard drug dose (table 4).

To maximize therapeutic benefit, clinicians should use the DOAC dosing and appropriate target international normalized ratio (INR) for warfarin approved by US Food and Drug Administration (FDA) or other comparable agency in other regions of the world. Systematic underdosing of DOACS, particularly in patients with atrial fibrillation, has emerged as a major concern; this practice is driven by the mistaken impression that lower drug doses are associated with lower risks of complications, including bleeding. This underdosing has been associated with an increase in death and avoidable thrombosis, including stroke, without a reduced risk of bleeding [11,12].

Although dose intensity tends to correlate with bleeding risk, studies using "low-intensity" warfarin did not demonstrate a substantial reduction of bleeding risk by targeting a lower INR [13]. Thus, clinicians should be aware of the appropriate dose level for the patient's indication and clinical characteristics, rather than trying to reduce bleeding risk by lowering the dose.

Risk factors related to the patient — These risk factors are expected to apply to all anticoagulants, despite most of the relevant evidence coming from patients treated with warfarin, which has been in use for decades.

Bleeding risk scores that incorporate these risk factors are discussed below. (See 'Bleeding risk scores' below.)

Age, race, and sex — Age and race (as well as other undefined genetic variables) may impact the risk of bleeding. However, the clinical significance is mainly in an awareness that some individuals may have a slightly greater risk. These factors are not used in bleeding risk scores and are not a reason to withhold anticoagulation. Age-related changes in anticoagulant dosing generally are not made, with the exception of dose reductions in individuals who are older than 75 or 80 years and are receiving a DOAC for AF, depending on the specific drug and other bleeding risk factors. These adjustments are discussed separately. (See "Direct oral anticoagulants (DOACs) and parenteral direct-acting anticoagulants: Dosing and adverse effects".)

●Older age – Older age is generally cited as a risk factor for bleeding, even after adjustment for comorbidities associated with aging. The age cutoff is variously defined as >60, >65, >75, or >80 years; the risk increase is approximately linear. However, the risk of bleeding attributable to older age is often overestimated, and anticoagulants are underused in older individuals who may derive more benefit than younger individuals [14,15]. Dose adjustment for age-related decline in kidney function based on the package insert and close attention to fall risk are prudent in older adults. (See 'Comorbidities' below.)

●Race/ethnicity – In a cohort of nearly 20,000 hospitalized patients with nonvalvular AF who had not had a prior stroke, after adjustment for comorbidities, the risk of ICH in the ensuing three years was higher in non-White compared with White individuals [16]. The reasons for these differences have not been determined.

●Sex – Unlike thrombosis risk, which appears to differ in males and females, bleeding risk does not appear to differ significantly according to sex, with the exception of menstrual blood loss, which may be more frequent with some anticoagulants than with others.

Prior bleeding — Prior bleeding may be a risk factor for future bleeding. However, the magnitude of risk is variable, and details of the prior bleed should be elicited. Most individuals with prior bleeding can re-initiate anticoagulation after recovery, especially if their baseline risk of thrombosis remains high and the reversible factors that contributed to bleeding are addressed. As discussed in a 2018 guideline from the American Society of Hematology (ASH), several studies have found that resuming anticoagulation following a major bleed (ICH or gastrointestinal [GI] bleeding) was associated with a reduced risk of all-cause mortality, mostly due to reduced risk of recurrent VTE (relative risk [RR] 0.62, 95% CI 0.43-0.89) [17]. Details of the timing of resuming the anticoagulant are discussed in separate topic reviews related to specific sites of bleeding.

●ICH – Prior intracerebral bleeding confers an increased risk of recurrent ICH (approximately 2 to 3 percent per year, which is 10-fold higher than the general population risk). Patients with ICH can restart anticoagulants if the risk of re-bleeding is less than the risk of ischemic stroke and its consequences. (See "Spontaneous intracerebral hemorrhage: Secondary prevention and long-term prognosis", section on 'Anticoagulation'.)

●GI – Gastrointestinal lesions can re-bleed, with the risk of rebleeding somewhat predicted by endoscopic findings (table 5). The approach to restarting anticoagulation is discussed separately. (See "Management of anticoagulants in patients undergoing endoscopic procedures", section on 'Resuming anticoagulants after hemostasis'.)

●Postsurgical – Sites of surgical bleeding are generally considered to be transient risk factors, and anticoagulation following surgery is often initiated within one to three days, as long as there were no unexpected surgical issues that would increase bleeding risk. Specific intervals for each anticoagulant are discussed in detail separately. (See "Perioperative management of patients receiving anticoagulants", section on 'Timing of anticoagulant interruption'.)

The interval between bleeding and resumption of anticoagulation depends on the site of bleeding and the cause and is discussed in the linked topic reviews. In general, clinicians err on the side of over-estimating the risk of recurrent bleeding, and underestimating the consequences of thrombosis, resulting in otherwise avoidable thrombotic events.

Comorbidities — Bleeding risk is increased with several chronic conditions that can interfere with normal hemostasis by a variety of mechanisms. These are codified in several bleeding risk scores (see 'Bleeding risk scores' below) and include:

●Liver disease – Liver disease can affect circulating levels of several endogenous procoagulant and anticoagulant factors; this effect is probably greatest in individuals with severe liver disease. In a systematic review and meta-analysis that included data on nearly 20,000 patients with AF and cirrhosis, anticoagulation did not result in a major increase in bleeding risk (pooled HR 1.45, 95% CI 0.96-2.17) [18]; the expected reduction in stroke risk with anticoagulation was observed. Management of anticoagulation and treatment of thrombosis in patients with liver disease are discussed in more detail separately. (See "Hemostatic abnormalities in patients with liver disease".)

Judicious use of anticoagulants in patients with portal vein thrombosis may actually reduce the risk of bleeding, presumably due to reductions in portal venous pressure with clot resolution; in general, anticoagulants can be used safely in such patients [19]. Low molecular weight (LMW) heparin is often preferred in individuals with cirrhosis due to challenges in adjusting warfarin with elevated baseline prothrombin time and international normalized ratio (PT/INR). (See "Hemostatic abnormalities in patients with liver disease", section on 'Portal vein thrombosis (PVT)'.)

●Kidney disease – Kidney disease can cause uremic platelet dysfunction and anemia, both of which may increase bleeding risk. (See "Uremic platelet dysfunction".)

Chronic kidney disease affects metabolism of the DOACs, which are all excreted by the kidney to some degree. Of the DOACs, dabigatran is the most dependent on kidney clearance (approximately 80 to 85 percent) and apixaban is the least dependent (approximately 25 percent). Kidney failure also increases the risk of thrombosis; thus, in many patients, anticoagulation remains reasonable despite degrees of chronic kidney disease that might seem to contraindicate anticoagulation. Close attention to anticoagulant dosing based on the glomerular filtration rate is important in individuals with kidney disease, as discussed separately. Compared to patients without kidney disease, patients with chronic kidney disease are more likely to experience major bleeding on anticoagulants, irrespective of which anticoagulant they receive. For some patients with advanced kidney disease, the risk of anticoagulant related bleeding will outweigh any potential benefit. (See "Venous thromboembolism: Anticoagulation after initial management" and "Direct oral anticoagulants (DOACs) and parenteral direct-acting anticoagulants: Dosing and adverse effects", section on 'Settings in which a heparin or vitamin K antagonist may be preferable'.)

●Diabetes – Diabetes may increase bleeding risk by effects on the vasculature as well as other complications related to chronic inflammation. In a regression analysis for fatal ICH with warfarin, diabetes increased the likelihood of ICH slightly (OR 1.56, 95% CI 1.06-2.28) [20]. Other studies have also found a modest association between diabetes and spontaneous ICH [21]. Diabetes can also increase risk of bleeding through other mechanisms. For example, diabetic retinopathy may place patients at risk of anticoagulant-associated intraocular bleeding.

●Cancer – Cancer and cancer therapy can increase bleeding risk by causing thrombocytopenia, increasing inflammatory cytokines, and disrupting vascular integrity at the primary tumor site or metastases. Tumor blood vessels are more likely to have structural and functional immaturity, which may also make them more inherently prone to bleeding [22]. In a network meta-analysis involving nearly 5000 patients with cancer-associated VTE, major bleeding occurred in 4 to 5 percent of patients treated with an anticoagulant (4.9 percent with a DOAC; 4.1 percent with a vitamin K antagonist) [23]. Among cancer patients, the risk of major bleeding with DOACs appears to be highest in those with gastrointestinal or genitourinary cancers [24,25]. However, the concomitant increased risk of thrombosis related to cancer and cancer therapies may pose a more serious risk for some individuals than the increased risk of bleeding, as discussed separately. (See "Anticoagulation in individuals with thrombocytopenia", section on 'Risk factors for bleeding and thrombosis in cancer'.)

For all these comorbidities, the relative magnitude of increased risk is not well established and is likely to vary with the severity and nature of the condition. In an analysis of major bleeding in individuals enrolled in the RIETE registry (Registry of Patients with Venous Thromboembolism), increases in bleeding were as follows:

●Age >75 years – OR 2.16, 95% CI 1.49-3.16

●Recent major bleeding – OR 2.64, 95% CI 1.44-4.83

●Metastatic cancer – OR 3.80, 95% CI 2.56-5.64

●Creatinine clearance <30 mL/minute – OR 2.27, 95% CI 1.49-3.44

Most of this information is derived from studies of individuals receiving warfarin and extrapolated to other anticoagulants. (See 'Risk factors related to the patient' above.)

Thrombocytopenia and bleeding disorders

●Thrombocytopenia – Thrombocytopenia is not necessarily a contraindication to anticoagulation, and thrombocytopenia generally is not thought to be protective against thrombosis. In most cases, individuals with platelet counts ≥50,000/microL can be treated with anticoagulation at therapeutic doses if indicated. The approach to anticoagulation in individuals with platelet counts <50,000/microL depends on the thrombotic risk. This subject is discussed in detail separately. (See "Anticoagulation in individuals with thrombocytopenia".)

●Coagulation factor deficiencies including hemophilia – Individuals with bleeding disorders such as hemophilia or other coagulation factor deficiencies can also develop thromboses and in some cases may require anticoagulation. This is generally managed by a hemostasis and thrombosis expert who can help the patient balance and mitigate the bleeding and thrombotic risks. (See "Chronic complications and age-related comorbidities in people with hemophilia", section on 'Cardiovascular disease' and "Factor XI (eleven) deficiency", section on 'Anticoagulation or antiplatelet therapy'.)

Concomitant antiplatelet medications — Antiplatelet agents include several medications such as aspirin, nonsteroidal antiinflammatory drugs (NSAIDs), P2Y₁₂ receptor blockers such as clopidogrel, prasugrel, or ticagrelor, and GPIIb/IIIa receptor blockers such as tirofiban or eptifibatide.

Combined use of an anticoagulant and an antiplatelet medication increases bleeding risk, as illustrated by the comparisons below. These findings support the practice of avoiding routine use of nonselective NSAIDs for treatment of pain or fever when other agents such as acetaminophen are available. If an NSAID is used, we advise patients to limit its use to the shortest possible duration. A selective COX-2 inhibitor may be safer than other agents. Specific agents and details of use are presented separately. (See "Overview of COX-2 selective NSAIDs", section on 'Lack of platelet inhibition and use during anticoagulation'.)

For some individuals, such as those with prosthetic heart valves or recent coronary artery stent placement, combined anticoagulation and antiplatelet therapy may be appropriate, as discussed separately. These individuals should be aware of the increased risk of bleeding, and the relative risks and benefit of the antiplatelet (or anticoagulant) medication(s) should be frequently reassessed, with a plan to stop any antithrombotic agent that does not clearly justify the added risk. (See "Antithrombotic therapy for mechanical heart valves", section on 'Approach to antithrombotic therapy' and "Coronary artery disease patients requiring combined anticoagulant and antiplatelet therapy".)

Risk factors for bleeding in specific sites

Intracranial — Intracranial bleeding, including intracerebral hemorrhage (ICH), subarachnoid hemorrhage (SAH), and hemorrhagic transformation of ischemic stroke, is usually the greatest concern because the risks of significant long-term deficits and mortality are high.

In a prospective series of 419 consecutive patients with atrial fibrillation who had an ICH while receiving a DOAC, multivariable analysis identified the following as risk factors: increasing age, use of an antiplatelet agent, white matter changes, hyperlipidemia, active cancer, high risk of falls, and low creatinine clearance [26].

In addition to the factors above (see 'Risk factors related to the anticoagulant' above and 'Risk factors related to the patient' above), risk factors for intracranial bleeding include the following:

●Prior stroke – Prior stroke is included in many risk models and clinical prediction scores for major bleeding; in most cases, these models (and the primary data from which they were derived) do not distinguish between prior ischemic and hemorrhagic stroke (see 'Bleeding risk scores' below). Nevertheless, it is almost certain that the risk of recurrent ICH is higher for individuals with a prior spontaneous ICH and underlying hypertensive vasculopathy or cerebral amyloid angiopathy than it is for individuals with a prior ischemic stroke and no history of ICH.

●Hypertension – Hypertensive vasculopathy tends to affect small penetrating arteries that branch off major cerebral arteries. Microscopic pseudoaneurysm formation with microhemorrhage can result. Hypertensive vasculopathy typically causes ICH in deep brain locations (eg, putamen, internal capsule, caudate nucleus, thalamus, pons, or cerebellum). This subject is discussed in more detail separately. (See "Spontaneous intracerebral hemorrhage: Pathogenesis, clinical features, and diagnosis".)

●Cerebral amyloid angiopathy – Cerebral amyloid angiopathy (CAA), although usually asymptomatic, is an important cause of primary lobar ICH in older adults. CAA is characterized by the deposition of amyloid beta-peptide in the vascular wall of small- to medium-sized blood vessels in the brain and leptomeninges. This weakens the structure of the vessel walls and makes them prone to bleeding. (See "Cerebral amyloid angiopathy".)

The use of anticoagulants in patients with CAA is discussed separately. (See "Cerebral amyloid angiopathy", section on 'Managing anticoagulant and antiplatelet medications'.)

●Cerebral microbleeds – Cerebral microbleeds detected by brain magnetic resonance imaging (MRI) are a risk factor for both ischemic stroke and ICH. However, screening for microbleeds is not routinely performed, as the MRI criteria for their assessment and the implications for clinical management are unclear. (See 'Microbleeds and other subclinical bleeding events' above.)

Management decisions regarding anticoagulation are based on a comprehensive, individualized risk assessment for clinically significant thrombosis and bleeding. (See 'Anticoagulant selection, dosing, and monitoring' below.)

●Intracranial vascular abnormalities – Arteriovenous malformations, cavernous malformations, and intracranial aneurysms are associated with a variable risk of intracerebral hemorrhage and/or subarachnoid hemorrhage, depending upon individual factors. The available data are limited and insufficient to determine whether anticoagulant therapy increases the risk of bleeding from intracranial vascular malformations or unruptured intracranial aneurysms. However, there is suspicion that anticoagulant therapy worsens the severity of bleeding from one of these lesions, should it occur. (See "Brain arteriovenous malformations" and "Vascular malformations of the central nervous system", section on 'Cavernous malformations' and "Anticoagulant and antiplatelet therapy in patients with an unruptured intracranial aneurysm", section on 'Effect of anticoagulation'.)

●Brain tumor – Primary or metastatic brain tumors may develop intratumoral hemorrhage. Hemorrhagic brain metastases can occur with any underlying type of cancer but are most common in association with melanoma, renal cell cancer, thyroid cancer, choriocarcinoma, and non-small cell lung cancer. (See "Epidemiology, clinical manifestations, and diagnosis of brain metastases" and "Treatment and prevention of venous thromboembolism in patients with brain tumors", section on 'Pre-anticoagulation risk assessment'.)

●Drug abuse – Cocaine and other sympathomimetic drugs are associated with an increased risk of ICH. (See "Cocaine use disorder: Epidemiology, clinical features, and diagnosis", section on 'Central nervous system'.)

●Falls – Falls can lead to traumatic brain injury with associated intracranial hemorrhage, including epidural or subdural hematoma, SAH, ICH, and/or intraventricular hemorrhage. Patients with a history of multiple falls tend to be older and tend to have more comorbidities and an increased risk of stroke. Among patients with a history of falls or at high risk of falling, the risk of intracranial hemorrhage is increased among patients on warfarin, aspirin, or no antithrombotic therapy, but the absolute increased risk of intracranial hemorrhage related to anticoagulation is small [27-29]. Nonrandomized studies suggest that for patients with atrial fibrillation and high risk of falls, the benefit of anticoagulation (ie, a reduced risk of ischemic stroke and consequent disability) outweighs the risk of intracranial bleeding from a fall [28,30]. (See "Traumatic brain injury: Epidemiology, classification, and pathophysiology".)

Components of fall prevention include addressing environmental factors (electrical cords, slippery floors, loose carpets), vision, properly fitting shoes, and avoidance of medications that cause dizziness, drowsiness, problems with coordination, and overly tight control of blood sugar [31]. (See "Falls: Prevention in community-dwelling older persons".)

●Endocarditis – Endocarditis can cause emboli (septic or sterile) that are susceptible to bleeding [32]. (See "Antithrombotic therapy in patients with infective endocarditis".)

Of these risk factors, hypertension is the most modifiable. The risk for falls may include several modifiable factors as well; strategies to reduce fall risk are presented in detail separately. However, clinicians generally overestimate the risk of bleeding with falls, resulting in unneeded withdrawal of anticoagulants [33]. (See "Falls: Prevention in community-dwelling older persons".)

The absolute increase in the risk of ICH with antiplatelet therapy such as aspirin is likely to be very small, and the risk of ICH does not appear to be increased with nonsteroidal antiinflammatory drugs (NSAIDs), as discussed separately. (See "Spontaneous intracerebral hemorrhage: Pathogenesis, clinical features, and diagnosis", section on 'Antithrombotic medications'.)

The risk of subdural hematoma (SDH) is also increased with anticoagulants. Other risk factors specific to SDH, as well as SDH evaluation and management, are discussed separately. (See "Subdural hematoma in adults: Etiology, clinical features, and diagnosis" and "Subdural hematoma in adults: Management and prognosis".)

Gastrointestinal — Gastrointestinal (GI) bleeding is another common site for major bleeding. In an analysis from the RIETE registry (Registry of Patients with Venous Thromboembolism) that examined risk factors for fatal bleeding in over 24,000 patients treated with an oral anticoagulant for acute VTE, there were 135 fatal bleeds during three months of therapy, and of these, gastrointestinal bleeding was the most likely bleeding site to result in fatality, representing 40 percent of fatal bleeds [34].

Additional risk factors specific for GI bleeding include:

●GI tumors

●Gastric or esophageal varices

●Gastritis

●Excess alcohol

●Antiplatelet agents, especially aspirin

●Chemotherapy agents that affect the GI epithelium

●Prior GI bleeding

In some of the trials testing DOACs for prophylaxis against venous thromboembolism in patients with cancer, the increased risk of GI bleeding was at least in part attributed to the presence of a GI tumor [35].

Numerous reviews have examined the risk of GI bleeding comparing DOACs with warfarin. Although conclusions differ, a consensus has emerged that the overall rate of GI bleeding is not decreased in DOAC treated patients; however, whether one DOAC is superior to another cannot be determined with high certainty due to the lack of head-to-head comparative studies.

In a large retrospective cohort study that assessed rates of hospitalization for GI bleeding, the adjusted incidence of hospital admission for GI bleeding was 115 per 10,000 patient-years (1.15 per 100 patient-years) [36]. Bleeding risk was highest for rivaroxaban (144 per 10,000 patient-years), followed by dabigatran (120 per 10,000 patient-years), followed by warfarin (113 per 10,000 patient-years), followed by apixaban (73 per 10,000 patient-years).

The use of a proton pump inhibitor or other form of gastric protection may be reasonable in selected individuals with increased risk of GI bleeding. (See 'Gastric protection' below.)

Spinal epidural — The risk factors for spinal epidural hematoma in individuals receiving anticoagulants and approaches to reduce these risks in individuals for whom neuraxial anesthesia or analgesia are being considered are discussed in detail separately. (See "Neuraxial anesthesia/analgesia techniques in the patient receiving anticoagulant or antiplatelet medication".)

Bleeding risk scores — Several bleeding risk scores have been developed and validated, mostly in patients receiving warfarin for atrial fibrillation. However, these scores generally do not perform significantly better than clinician judgment based on close consideration of patient characteristics. As an example, in a prospective cohort of 515 patients, subjective estimates of bleeding risk made by the treating clinicians (mean clinical experience: three years) had similar accuracy in predicting bleeding as use of a risk score [37].

Thus, a major benefit of these scoring systems is the identification of potentially modifiable factors that can be remedied as a means of reducing bleeding risk in anticoagulated patients (table 1). However, not every association implies causation. As an example, anemia is likely to be a marker of increased risk because anemia is caused by bleeding [38]. Anemia has also been proposed to contribute to bleeding risk by reducing platelet interactions with the vessel wall.

While most of the scores include prior stroke as an independent risk factor for major bleeding, they do not appear to distinguish between hemorrhagic versus ischemic stroke.

●HAS-BLED – The HAS-BLED score (calculator 1) was derived from a cohort of 3978 patients with atrial fibrillation in the Euro Heart Study [39]. It has been validated in additional cohorts of patients with atrial fibrillation, and its use has been recommended in European and Canadian guidelines [40-45]. Variables include (table 6):

•Hypertension – 1 point

•Abnormal renal and/or hepatic function – 1 point each

•Stroke – 1 point

•Bleeding tendency/predisposition – 1 point

•Labile INR on warfarin – 1 point

•Elderly (age >65 years) – 1 point

•Drugs (aspirin or NSAIDs) and/or alcohol – 1 point each

●ATRIA – This score was developed from the results of the Anticoagulation and Risk Factors in Atrial Fibrillation (ATRIA) Study, one of the largest studies that evaluated risk prediction [46]. Variables include:

•Anemia – 3 points

•Severe kidney disease (estimated glomerular filtration rate <30 mL/minute or dialysis-dependent) – 3 points

•Age ≥75 years – 2 points

•Any prior hemorrhage – 1 point

•Diagnosed hypertension – 1 point

Bleeding rates for low- (0 to 3 points), intermediate- (4 points), and high-risk patients (5 to 10 points) were 0.76, 2.62, and 5.76 events per 100 patient-years, respectively [46].

●VTE-BLEED – This score was developed from an evaluation of over 2500 individuals with venous thromboembolism (VTE) in the RE-COVER trials who were assigned to receive dabigatran and verified in over 2500 individuals from the same trials assigned to warfarin [47]. Variables that predicted bleeding include (table 7):

•Active cancer – 2 points

•Anemia – 1.5 points

•History of bleeding – 1.5 points

•Creatinine clearance 30 to 60 mL/min – 1.5 points

•Age ≥60 years – 1.5 points

•Uncontrolled hypertension in a male – 1 point

A score of ≥2 points was associated with a high bleed risk (average bleed incidence, 13 percent); 0 to 1.5 points with a low bleed risk (average bleed incidence, 2.8 percent) [47]. The C-statistic was 0.72 to 0.78, which was better than that for other scores listed herein (ranging from 0.60 to 0.66).

●HEMORR2HAGES – The HEMORR2HAGES score was created by combining risk factors from existing scoring systems [48]. Each factor is assigned 1 point, with the exception of a previous bleeding episode (2 points):

•Hepatic or renal disease

•Ethanol abuse

•Malignancy

•Older age (>75 years)

•Reduced platelet count or function, including aspirin therapy

•Re-bleeding risk (history of prior bleed)

•Hypertension

•Anemia

•Genetic factors

•Excessive fall risk

•Stroke

Risks of major bleeding per 100 patient-years were 1.9 (0 points), 2.5 (1 point), 5.3 (2 points), 8.4 (3 points), 10.4 (4 points), and 12.3 (≥5 points).

These risk calculations are imprecise, do not include all possible risk factors, and cannot account for variations in the severity of comorbidities. They cannot predict bleeding risk in an individual patient; rather, they only predict bleeding in an exposed patient population. Studies have found higher predictive value when other variables such as cerebral microbleeds were also included [49].

APPROACHES TO RISK REDUCTION

Anticoagulant selection, dosing, and monitoring — The initial indication for anticoagulation as well as any changes in thrombotic risk and bleeding risk should be reviewed periodically for each patient. In some cases, the risk-benefit calculation may shift, and changes may be warranted in the decision to use an anticoagulant, in the specific agent selected, and/or in the dose.

Examples include discontinuation of the anticoagulant after three months of therapy for selected individuals with a first provoked deep vein thrombosis (DVT) or reduction from therapeutic-dose DOAC to prophylactic dose of DOAC after six months of therapy in selected individuals with DVT for whom continued anticoagulation is indicated. (See "Overview of the treatment of proximal and distal lower extremity deep vein thrombosis (DVT)".)

It is also important to pay attention to drug adherence. A switch from warfarin to a DOAC or the addition of low-dose vitamin K may be appropriate in individuals taking warfarin who have significant INR variability despite good medication adherence. In contrast, however, switching to a DOAC is not advised in individuals with poor INR control that is due to poor drug adherence, as one or two missed doses of a DOAC could reduce the efficacy of anticoagulation more than one or two missed doses of warfarin, and use of a DOAC eliminates the ability to monitor adherence effectively. (See "Warfarin and other VKAs: Dosing and adverse effects", section on 'Poor INR control/vitamin K supplementation' and "Direct oral anticoagulants (DOACs) and parenteral direct-acting anticoagulants: Dosing and adverse effects", section on 'Advantages over heparin and warfarin'.)

However, patients should be treated with a dose of anticoagulant that has been proven effective for their indication and comorbidities rather than with the lowest possible dose.

The selection of anticoagulant takes into account a number of factors related to the patient's underlying condition, values and preferences, and burdens of therapy, as discussed in separate topic reviews listed below. All other factors being equal, the risk of serious bleeding appears to be lower with direct oral anticoagulants (DOACs; including the direct thrombin inhibitor dabigatran and the direct factor Xa inhibitors apixaban, edoxaban, and rivaroxaban) compared with warfarin. However, these agents are not appropriate for all indications. Choice of anticoagulant for specific indications is discussed in the linked topics:

●Atrial fibrillation (AF), primary prophylaxis (see "Atrial fibrillation in adults: Use of oral anticoagulants")

●Stroke, secondary prophylaxis (see "Early antithrombotic treatment of acute ischemic stroke and transient ischemic attack" and "Stroke in patients with atrial fibrillation", section on 'Long-term anticoagulation')

●Venous thromboembolism (VTE), primary prophylaxis in hospitalized medical patients (see "Prevention of venous thromboembolic disease in acutely ill hospitalized medical adults")

●VTE, primary prophylaxis in surgical patients (see "Prevention of venous thromboembolism in adults undergoing hip fracture repair or hip or knee replacement" and "Prevention of venous thromboembolic disease in adult nonorthopedic surgical patients")

●VTE, primary and secondary prophylaxis in inherited thrombophilias (see "Factor V Leiden and activated protein C resistance" and "Prothrombin G20210A" and "Protein S deficiency" and "Antithrombin deficiency" and "Protein C deficiency")

●VTE, treatment and secondary prophylaxis (see "Venous thromboembolism: Initiation of anticoagulation" and "Venous thromboembolism: Anticoagulation after initial management")

●VTE in patients with cancer (see "Risk and prevention of venous thromboembolism in adults with cancer" and "Anticoagulation therapy for venous thromboembolism (lower extremity venous thrombosis and pulmonary embolism) in adult patients with malignancy")

●VTE in patients with antiphospholipid syndrome (see "Management of antiphospholipid syndrome")

DOACs are not used in pregnancy, during breastfeeding, or in patients with valvular AF with native valves, mechanical heart valves, or high-risk antiphospholipid syndrome. (See "Use of anticoagulants during pregnancy and postpartum" and "Venous thromboembolism in pregnancy and postpartum: Treatment" and "Management of antiphospholipid syndrome", section on 'Secondary thrombosis prevention'.)

Their use in other patients with a high risk of thrombosis (such as heparin-induced thrombocytopenia [HIT]) should only be undertaken in clinical studies or by clinicians experienced with these disorders. (See "Management of heparin-induced thrombocytopenia".)

Other medications — Combined anticoagulant and antiplatelet therapy is likely to increase bleeding risk (see 'Concomitant antiplatelet medications' above), and combined therapy should be restricted to settings in which the benefit is expected to outweigh the risk. (See "Antithrombotic therapy for mechanical heart valves", section on 'Approach to antithrombotic therapy' and "Coronary artery disease patients requiring combined anticoagulant and antiplatelet therapy".)

Unnecessary antiplatelet therapy should be discontinued. This may require specific conversations with the patient about over-the-counter antiplatelet medication such as aspirin or NSAIDs, which may not be listed on the medication list. A selective COX-2 inhibitor may be safer than other agents. (See "Overview of COX-2 selective NSAIDs", section on 'Lack of platelet inhibition and use during anticoagulation'.)

One series evaluated nearly 7000 patients who initiated warfarin and were enrolled in one of six anticoagulation clinics that participated in a quality improvement consortium in the state of Michigan [50]. Of these, over 3100 (45 percent) were receiving concomitant antiplatelet therapy, but the indication was unclear for over 1300 (nearly one-half of those on antiplatelet therapy, or 20 percent of the entire cohort). Of all the patients on concomitant antiplatelet therapy, over 2000 were not receiving a proton pump inhibitor (PPI). The authors concluded that there was a large opportunity to reduce the risk of gastrointestinal bleeding by discontinuing the antiplatelet agent when it was not indicated, or using a PPI when the antiplatelet agent was needed. These interventions are discussed below. (See 'Approaches to risk reduction' above.)

For individuals receiving anticoagulants, we generally advise patients to avoid routine use of NSAIDs for pain or fever when other agents such as acetaminophen are available. When an NSAID is indicated, we ask patients to limit the duration of use and/or to use a selective COX-2 inhibitor if appropriate.

Other medications may increase the risk of anticoagulant-associated bleeding by potentiating the effects of the anticoagulant, and these medications should be avoided when possible [51]. Details are presented in the tables and separate topic reviews:

●Medication interactions with warfarin – (table 8) (see "Biology of warfarin and modulators of INR control", section on 'Drug interactions')

●Medication interactions with DOACs – (table 9) (see "Direct oral anticoagulants (DOACs) and parenteral direct-acting anticoagulants: Dosing and adverse effects" and "COVID-19: Management of adults with acute illness in the outpatient setting", section on 'Drug interactions')

Modifiable risk factors — The following interventions may reduce the risk of bleeding in individuals treated with oral anticoagulants:

●Control of blood pressure to prevent hypertensive angiopathy, especially in the brain, and to prevent hypotension-induced loss of balance or falls. (See "Overview of hypertension in adults".)

●Optimize kidney and liver function. (See "Overview of the management of chronic kidney disease in adults" and "Approach to the patient with abnormal liver biochemical and function tests".)

●For those with risk factors for falling or a history of falls, a multidisciplinary risk factor screening/intervention program may be helpful. (See "Falls in older persons: Risk factors and patient evaluation" and "Falls: Prevention in community-dwelling older persons".)

●Limit the use of antiplatelet agents and other drugs that may influence DOAC levels or effect, including nonselective NSAIDs, and other interacting medications, as discussed above. (See 'Other medications' above.)

The mechanisms by which these factors increase bleeding risk is discussed above. (See 'Risk factors for bleeding' above.)

Gastric protection — Although not well studied, gastric protection, typically with a proton pump inhibitor (PPI), is given to many individuals treated with therapeutic-dose anticoagulation, especially those with additional risk factors for gastrointestinal bleeding.

In our practices, we would use a PPI in patients with a history of GI bleeding that had not been adequately treated or in individuals perceived to be at higher risk for gastrointestinal bleeding, such as those with concomitant use of a nonsteroidal antiinflammatory drug (NSAID). Other risk factors are listed above. (See 'Gastrointestinal' above.)

We would generally use a PPI at the approved doses, as there is better evidence for their use than a histamine 2 receptor blocker.

Efficacy of a PPI to reduce gastric bleeding risk includes [36,52,53]:

●A 2022 meta-analysis of six observational studies and one randomized trial that included patients receiving an oral anticoagulant found an association between PPI use (or histamine 2 receptor blocker use in one study) and reduced risk of upper gastrointestinal bleeding (relative risk [RR] 0.67, 95% CI 0.61-0.74) [52]. The risk reduction was greatest for individuals with higher baseline risk of gastrointestinal bleeding due to use of NSAIDs or other risk factors. The sole randomized trial, which included >11,000 patients, showed a trend towards reduced upper gastrointestinal bleeding that did not reach statistical significance, but bleeding risk was lower than expected due to stringent criteria for participation.

●A 2018 retrospective cohort study of >1.6 million patients receiving an anticoagulant (three-fourths for atrial fibrillation) found that PPI co-therapy was associated with a lower risk of hospitalization for gastrointestinal bleeding (incidence rate ratio [IRR] 0.66, 95% CI 0.62-0.69) [36]. Subgroup analysis revealed that the effect was seen independently for each anticoagulant. Bleeding requiring hospitalization was most likely with rivaroxaban.

Approaches to gastric protection in individuals taking NSAIDs (unrelated to anticoagulation) are presented separately. (See "NSAIDs (including aspirin): Primary prevention of gastroduodenal toxicity" and "NSAIDs (including aspirin): Treatment and secondary prevention of gastroduodenal toxicity".)

PROGNOSIS AND REINITIATION OF ANTICOAGULATION — As discussed in a 2016 American Society of Hematology (ASH) education program review, resuming anticoagulation after major bleeding (including intracerebral hemorrhage [ICH] and gastrointestinal [GI] bleeding) often has a favorable risk-benefit profile [54].

Often, the patient's anticoagulant can be resumed within a period of approximately two weeks following resolution of the bleed. Separate topic reviews discuss the features of decision-making and the optimal timing for restarting an anticoagulant after an invasive procedure or following bleeding in specific sites:

●Intracerebral bleeding – (See "Spontaneous intracerebral hemorrhage: Acute treatment and prognosis" and "Spontaneous intracerebral hemorrhage: Secondary prevention and long-term prognosis".)

●Gastrointestinal bleeding – (See "Management of anticoagulants in patients undergoing endoscopic procedures".)

●Neuraxial anesthesia – (See "Neuraxial anesthesia/analgesia techniques in the patient receiving anticoagulant or antiplatelet medication".)

●Surgery – (See "Perioperative management of patients receiving anticoagulants", section on 'Timing of anticoagulant interruption'.)

●Labor and delivery – (See "Use of anticoagulants during pregnancy and postpartum", section on 'Resuming or initiating anticoagulation postpartum'.)

CLINICAL CASE VIGNETTES

●Several published case reports have described individuals (some in their late 70s and 80s) being treated with anticoagulation for atrial fibrillation who subsequently developed gastrointestinal (GI) bleeding, with various etiologies; in some cases, a reversal agent for the anticoagulant was used in the acute setting [55-57]. These individuals have been appropriately evaluated and treated for the GI lesion and subsequently resumed anticoagulation without recurrent bleeding.

●A published case report described a 38-year-old woman who presented in labor with an early-term pregnancy and noted leg pain and swelling that was diagnosed as a new deep vein thrombosis (DVT) [58]. She was assessed by the obstetrics service and confirmed to be in active labor; an epidural was placed, and she delivered a healthy baby. Anticoagulation was initiated after delivery. This case illustrates a setting (active labor) in which the risks of anticoagulation were thought to outweigh the benefits. The timing of anticoagulation relative to neuraxial anesthesia is another important consideration that is discussed in detail separately. (See "Neuraxial anesthesia/analgesia techniques in the patient receiving anticoagulant or antiplatelet medication".)

SOCIETY GUIDELINE LINKS — Links to society and government-sponsored guidelines from selected countries and regions around the world are provided separately. (See "Society guideline links: Anticoagulation" and "Society guideline links: Stroke in adults".)

SUMMARY AND RECOMMENDATIONS

●Mechanism – Anticoagulants do not technically cause bleeding; they interfere with normal hemostasis that would otherwise prevent microbleeds from becoming clinically significant. Anticoagulation is used when the benefits of reducing thrombosis risk outweigh the increased risks of clinically significant bleeding. (See 'Pathogenesis of anticoagulant-associated bleeding' above.)

●Relative risks of different drugs – All else being equal, direct oral anticoagulants (DOACs; dabigatran and direct factor Xa inhibitors) carry a lower bleeding risk than warfarin, although the risk of gastrointestinal bleeding is likely similar between the DOACs and warfarin. The risk of anticoagulant-associated bleeding is greatest during the initial period of use and is often dose-dependent, but bleeding can occur at prophylactic and therapeutic doses. (See 'Risk factors related to the anticoagulant' above.)

●Comorbidities – Several patient factors contribute to bleeding risk, including the age of the patient; prior bleeding; comorbidities such as impaired kidney or liver function, diabetes, cancer, and obesity (table 1). The risk of bleeding is also increased with severe thrombocytopenia (eg, platelet count <50,000/microL) and with concomitant antiplatelet medications. However, none of these are absolute contraindications to anticoagulation, as discussed above. (See 'Risk factors related to the patient' above.)

●ICH – Intracranial bleeding is a feared complication of anticoagulation. Important risk factors include a history of stroke (especially intracerebral hemorrhage [ICH]), hypertension, and cerebral amyloid angiopathy. ICH risk is reduced with DOAC treatment compared with warfarin. Radiologically identified microbleeds strongly predict subsequent ICH. (See 'Intracranial' above.)

●Risk scores – Several bleeding risk scores have been developed and validated, mostly in patients receiving warfarin for atrial fibrillation. However, these scores generally do not perform significantly better than clinician judgment based on close consideration of patient characteristics. A major benefit of these scores is the identification of potentially modifiable factors that can be remedied. (See 'Bleeding risk scores' above.)

●Risk reduction – The risk of anticoagulant-associated bleeding can be minimized by periodically reviewing the indication for anticoagulation, risk-benefit ratio, dose, adherence, concomitant medications (nonsteroidal antiinflammatory drugs [NSAIDs], other antiplatelet agents, and other over-the-counter medications), and patient comorbidities. (See 'Anticoagulant selection, dosing, and monitoring' above.)

Combined use of an anticoagulant and antiplatelet medication should be used only when the benefit outweighs the risk. Routine use of NSAIDs for pain or fever should be avoided when other agents such as acetaminophen are available; when an NSAID is indicated, the duration should be limited and/or a selective COX-2 inhibitor used if appropriate. (See 'Other medications' above and "Overview of COX-2 selective NSAIDs", section on 'Lack of platelet inhibition and use during anticoagulation'.)

Good blood pressure control and attention to fall risk also may be helpful. (See 'Modifiable risk factors' above.)

For patients at increased risk of gastrointestinal bleeding, we suggest a proton pump inhibitor (PPI) (Grade 2C). Examples of increased risk include prior gastrointestinal bleeding, varices, excess alcohol, or need to take an NSAID more often than once or twice a month. (See 'Gastric protection' above and 'Risk factors for bleeding in specific sites' above.)

●Anticoagulant resumption – Resuming anticoagulation after major bleeding (including ICH and gastrointestinal bleeding) often has a favorable risk-benefit profile. Details are provided in topic reviews listed above. (See 'Prognosis and reinitiation of anticoagulation' above.)

●Examples – The case vignettes illustrate some of the aspects of decision-making. (See 'Clinical case vignettes' above.)