Hemorrhagic Complications of Anticoagulation

Sam Schulman

Mark A. Crowther

Bleeding is the most common complication of anticoagulant therapy. The spectrum is from seemingly harmless subcutaneous ecchymoses, albeit cosmetically disturbing for some individuals, to severely disabling intracerebral hemorrhages and fatal bleeding. Parenteral anticoagulants, for example, unfractionated heparin (UFH) or low molecular weight heparin (LMWH), may cause serious hemorrhage despite careful monitoring or predictable plasma levels, respectively, because of the underlying illness and invasive procedures performed in the hospital. Vitamin K antagonist (VKA) therapy is in many patients difficult to control, and excessive anticoagulation increases dramatically the risk of bleeding. New anticoagulants with more predictable pharmacokinetics and a minimum of drug-drug and drug-food interactions may have a lower risk of bleeding. However, a substantial number of bleeding events occur without provoking risk factors and while patients are within the therapeutic range—these events are impossible to avoid during anticoagulant therapy. It is therefore also crucial to be familiar with the reversal of anticoagulants and management of hemorrhage.

DEFINITIONS

Comparison of incidences of bleeding between clinical studies is hampered by the plethora of definitions for serious or major bleeding. Some of the most used or well-reasoned classifications and their criteria are provided in Table 107.1.¹, ², ³, ⁴, ⁵ Many studies have, however, used their own definition or a “modified” version of any of those in Table 107.1.¹

The International Society on Thrombosis and Haemostasis (ISTH) definition of major bleeding for nonsurgical patients² is supported by the European Medicines Agency,³ and many contemporary clinical trials have chosen this definition. A criterion for major bleeding that has been used in several clinical trials on prophylaxis against venous thromboembolism (VTE) after major orthopedic surgery is the bleeding index of at least 2. This index is the sum of the number of red cell and whole blood units transfused plus the difference between pre- and postbleeding hemoglobin (in g/dL).

The number of major bleedings is often compared with the number of thromboembolic events to evaluate the net benefit of a treatment. It is then necessary to analyze the proportion of different types of major bleeding events, since an intracranial hemorrhage has about the same clinical burden as an ischemic stroke, whereas a gastrointestinal bleeding usually has far less severe consequences. In many large trials, the greatest proportion of major hemorrhages were indeed gastrointestinal.⁸

INCIDENCE OF ANTICOAGULATIONASSOCIATED BLEEDING

Vitamin K Antagonists

The risk of bleeding complications in randomized control trials (RCTs) is lower than that seen in routine clinical practice since the study population is selected and patients with risk of bleeding are usually excluded. Paradoxically, the risk of bleeding reported in observational studies is often lower than in RCTs. This could be due to recall bias in retrospective studies, missed events in registry studies, or a result of studying a VKA-experienced cohort. Conversely, RCTs may demonstrate a high incidence of bleeding if they are pivotal registration trials, where all events are carefully captured, and if a large proportion of patients are VKA naïve.⁴

The largest RCTs on VKA therapy have been performed in patients with atrial fibrillation. In Table 107.2, the annualized incidence of fatal, intracranial, and major bleeding on VKA is shown for RCTs versus placebo, aspirin + clopidogrel, and dabigatran as well as for observational studies of an inception cohort and of VKA-experienced patients. The annual incidences of intracranial bleeding and of major hemorrhage range from 0.3% to 1.7% and from 1.3% to 4.7%, respectively.⁴, ⁵, ⁶, ⁷, ⁸, ⁹. Detailed information on incidence of bleeding for each RCT can be found in the guidelines of the American College of Chest Physician.¹⁰

In a pooled analysis of five studies on VKA versus placebo the attributable risk of major bleeding or of intracranial hemorrhage for VKA was 0.3% and 0.2% per year, respectively.⁶

In a meta-analysis of 33 RCTs and prospective cohort studies with 10,757 patients treated with VKAs for VTE for an average of 5 months, the annualized risk of fatal, intracranial, and major bleeding was 1.3%, 1.15%, and 7.2%, respectively.¹¹ It should be noted that these patients are all VKA naïve and the risk of bleeding is highest shortly after start of therapy. For the subpopulation with duration of anticoagulation longer than 3 months, the annualized incidence of fatal, intracranial, and major bleeding for the period after 3 months was 0.6%, 0.6%, and 2.7%, respectively.¹¹

In nine RCTs on VKA versus combined VKA-antiplatelet therapy for patients with mechanical prosthetic heart valves, there were 59 major bleeding events in the VKA-alone arm during a total of 1,926 patient-years, corresponding to an annualized risk of 3.1%.¹² Three of these studies aimed for a therapeutic international normalized ratio (INR) of 3.0 to 4.5, and the risk includes nonfatal intra- and extracranial major bleeding.

Whenever VKA is combined with other antithrombotic agents, the risk of bleeding increases. The annual risk of bleeding requiring hospitalization was in a large registry study 4.3% on VKA, 5.7% with aspirin added, and 12% with aspirin and clopidogrel added¹³ (Table 104.2, Chapter 104).

Table 107.1 Commonly used and/or well-reasoned classifications for major or severe bleeding in nonsurgical and surgical patients

*Nonsurgical Patients*
TIMI Major Bleeding¹⁴⁹	GUSTO Severe Bleeding¹⁵⁰	ISTH Major Bleeding²
Intracranial bleeding ≥5 g/dL decrease in Hgb ≥15% decrease in Hct	At least one of: Intracranial bleeding Bleeding causing hemodynamic compromise	At least one of: Fatal bleeding Bleeding in critical organ Bleeding with either of (a) ≥2 g/dL decrease in Hgb or (b) transfusion of ≥2 units of whole blood or red cells
*Surgical Patients*
ISTF—Adjudicated Serious Bleeding¹⁵¹	ISTH Major Bleeding (Surgical)¹⁵²
At least one of: Fatal bleeding Intracranial bleeding Retroperitoneal bleeding Bleeding causing treatment cessation Bleeding causing reoperation	At least one of: Fatal bleeding Bleeding in critical organ Extrasurgical site bleeding with either of: (a) ≥2 g/dL decrease in Hgb or (b) transfusion of ≥2 U of whole blood or red cells Complicated hemarthrosis Surgical site bleeding requiring second intervention Surgical site bleeding with (a) ≥2 g/dL decrease in Hgb or (b) transfusion of ≥2 U of whole blood or red cells AND that, as assessed by the surgeon, is either (c) unexpected and prolonged or (d) causing hemodynamic instability
TIMI, thrombolysis in myocardial infarction; GUSTO, global utilization of streptokinase and tPA for occluded coronary arteries; ISTH, International Society on Thrombosis and Haemostasis; ISTF, International Surgical Thrombosis Forum.

Unfractionated Heparin

The administration of UFH should be by continuous infusion or subcutaneous injections, since the risk of bleeding is increased with intermittent intravenous injections.¹⁰ The risk of bleeding depends very much on the indication for treatment. In patients with acute myocardial infarction, there is a high risk due to invasive procedures, frontloaded thrombolysis, and concomitant (frequently multiagent) antiplatelet therapy. The frequent use of revascularization procedures in non-ST-elevation acute coronary syndromes probably explains the high risk of major bleeding seen in this setting (Table 107.3).¹⁴ Patients with ischemic stroke have a risk of hemorrhagic stroke transformation,¹⁵ whereas those treated for VTE are less susceptible unless they have underlying cancer.

Low Molecular Weight Heparin

Whereas UFH is almost exclusively given for short durations, LMWH is, due to the possibility of once-daily administration without monitoring, sometimes used for many months, particularly during pregnancy or in patients with VTE and cancer. The risk of bleeding for different indications and durations of treatment is shown in Table 107.4.¹⁴, ¹⁵, ¹⁶, ¹⁷, ¹⁸ In comparison with UFH, LMWH resulted in more bleeding in acute coronary syndromes, and the difference was statistically significant in ST-elevation myocardial infarction, although this was outweighed by better efficacy for LMWH.¹⁴ In acute stroke and in pulmonary embolism, there was a nonsignificant decrease of risk of major hemorrhage with LMWH compared to UFH,¹⁵ whereas in short-term treatment of deep vein thrombosis LMWH resulted in significantly fewer major bleedings.¹⁷

Fondaparinux

The pentasaccharide fondaparinux is almost exclusively used for durations up to approximately 10 days, both in prevention and treatment of thromboembolism. The risks of major bleeding during treatment for cardiac or for venous indications are presented in Table 107.5.¹⁹, ²⁰, ²¹, ²²

Dabigatran

The orally available direct thrombin inhibitor, dabigatran etexilate, was evaluated in a study with 18,000 patients versus warfarin and was given as either 110 mg b.i.d. or 150 mg b.i.d. for a mean of 2 years.⁷ The annualized risk for major bleeding was 2.7% and 3.1%, respectively, and for intracranial bleeding it was 0.2% and 0.3%, respectively. The risk of intracranial
bleeding was for both dose levels significantly lower than for patients randomized to warfarin (0.7%).

Table 107.2 Annual risk (in %) for different types of hemorrhages with VKA therapy in studies with various designs

Study/Year	Design	Fatal	Intracranial	Major Bleed
*Atrial Fibrillation*
Atrial fibrillation Investigators/1994⁶	Pooled analysis of RCTs; warfarin vs. control	NA	0.3	1.3
SPAF III/1996⁵	RCT; vs. low, fixed dose VKA plus aspirin	NA	0.5	2.1
ACTIVE W/2006⁴	RCT; vs. aspirin plus clopidogrel; 23% of patients VKA naïve	0.26	0.36	2.21
RE-LY/2009⁷	RCT; vs. dabigatran; 50% of patients VKA naïve	NA	0.74	3.36
Hylek/2007⁹	Inception cohort; elderly	0.8	1.7	4.7
Go/2003⁸	ATRIA cohort, retro and prospective	NA	0.5	1.5
VTE
Linkins/2003¹¹	Meta-analysis, RCTs, and prospective cohorts	1.3	1.15	7.2
Linkins/2003¹¹	Same, but 1st 3 mo excluded	0.6	0.6	2.7
Akl/2008¹⁶	Meta-analysis, four RCTs, VTE and cancer, 3-6 mo	NA	NA	1.9
*Mechanical Heart Valves*
Little/2003¹²	Meta-analysis of RCTs	NA	NA	3.1

Table 107.3 Risk for major hemorrhage with UFH according to indication

Study/Year	Design; Duration	Intracranial	Major Bleed
*Acute Coronary Syndromes, ST-Elevation*
Murphy/2007¹⁴	Meta-analysis, six RCTs; 30 d	NA	1.8%^a
*Acute Coronary Syndromes, Non-ST-Elevation*
Murphy/2007¹⁴	Meta-analysis, six RCTs; 30 d	NA	5.4%^a
*Acute Ischemic Stroke*
Sandercock/2008¹⁵	Meta-analysis, eight RCTs; 10 d	1.1%^b	0.1%^c
*Acute VTE*
Quinlan/2004¹⁸	Meta-analysis, 12 RCTs, pulmonary embolism; 7 d	NA	2.3%
Erkens/2010¹⁷	Meta-analysis, 8 RCTs, deep vein thrombosis; 7 d	NA	2.1%
^aIncludes intracranial bleeding but not fatal bleeding. ^bSymptomatic bleeding; for any—including asymptomatic hemorrhagic transformation—the risk was 2.6%. ^cExtracranial, nonfatal bleeding.

Parenteral Direct Thrombin Inhibitors

Several parenteral, direct thrombin inhibitors (argatroban, bivalirudin, and lepirudin) are used for limited indications and often with small populations studied.¹⁰

Table 107.4 Risk for major hemorrhage with LMWH according to indication

Study/Year	Design; Duration	Intracranial	Major Bleed
*Acute Coronary Syndromes, ST-Elevation*
Murphy/2007¹⁴	Meta-analysis, six RCTs; 30 d	NA	2.6%^a
*Acute Coronary Syndromes, Non-ST-Elevation*
Murphy/2007¹⁴	Meta-analysis, six RCTs; 30 d	NA	6.3%^a
*Acute Ischemic Stroke*
Sandercock/2008¹⁵	Meta-analysis, four RCTs; 10 d	0.7%^b	0.6%^c
*Acute VTE, Short-Term*
Quinlan/2004¹⁸	Meta-analysis, 12 RCTs, pulmonary embolism; 7 d	NA	1.4%
Erkens/2010¹⁷	Meta-analysis, eight RCTs, deep vein thrombosis; 7 d	NA	1.0%
*VTE, Long-Term*
Akl/2008¹⁶	Meta-analysis, four RCTs, patients with cancer, 3-6 mo	NA	6.4%^d
^aIncludes intracranial bleeding but not fatal bleeding. ^bSymptomatic bleeding; for any—including asymptomatic hemorrhagic transformation—the risk was 1.7%. ^cExtracranial, nonfatal bleeding. ^dAnnualized risk was 2.0%.

Argatroban has been studied in treatment after stroke, and the incidence of intracranial bleeding was 3.4% or 5.1% with a bolus dose of 100 µg/kg followed by continuous infusion at 1 or 3 mg/kg/min, respectively, compared to 0% with placebo.²³ In three trials including patients with myocardial infarction after initial thrombolytic therapy, the risk of bleeding was similar²⁴ or lower²⁵, ²⁶ with argatroban compared to UFH. In a pooled analysis of two cohort studies with argatroban for patients with heparin-induced thrombocytopenia (HIT), the incidence of major bleeding was 6% compared to 7% in historical controls for whom heparin was stopped and sometimes replaced with VKA.²⁷ Argatroban is eliminated by hepatic metabolism, and the dose has to be reduced in case of liver impairment or a drug with alternative route of elimination chosen.

Table 107.5 Risk for major hemorrhage with fondaparinux according to indication

Study/Year	Design	Fatal	Intracranial	Major Bleed
*Acute Coronary Syndromes, ST-Elevation*
OASIS 6/2006²¹	RCT, 9 d	0.6%	0.2%	1.8%^a
*Acute Coronary Syndromes, Non-ST-Elevation*
OASIS 5/2006²²	RCT, 9 d	NA	NA	2.2%
	Same, 30 d	NA	NA	3.1%
*Acute VTE, Short-Term*
MATISSE PE/2003²⁰	RCTs, pulmonary embolism, 7 d	0.1%	NA	1.3%
MATISSE DVT/2004¹⁹	RCTs, deep vein thrombosis, 7 d	0.2%	NA	1.1%
^aFatal and intracranial hemorrhages are included in this result. Severe bleeding according to modified TIMI definition occurred in 1.0%.

Bivalirudin was compared with UFH in patients with ST-elevation myocardial infarction in the HERO-2 trial and was associated with a trend to more severe bleeding and intracranial bleeding events.²⁸ Conversely, after percutaneous coronary interventions, bivalirudin has consistently resulted in only about half the number of major bleeding events as with UFH (˜3% vs. 6%).²⁹, ³⁰

Lepirudin as anticoagulant after thrombolysis in myocardial infarction resulted in a similar risk of major or intracranial bleeding as UFH.³¹ In patients with acute coronary syndrome
without thrombolysis, lepirudin was, however, associated with a higher incidence of major bleeding than UFH.³² In patients with HIT treatment, lepirudin appears in indirect or historical comparisons to give more major bleeding (10% to 14%)³³, ³⁴ than with argatroban (6%),²⁷ danaparoid (2.5%),³³ or controls (8.5%).³⁴ Both bivalirudin and lepirudin are eliminated renally and are contraindicated in severe kidney disease.

Rivaroxaban

Several direct and orally available factor Xa inhibitors are under development. Rivaroxaban was, in a phase III RCT for patients with deep vein thrombosis, associated with a similar risk of major bleeding and fatal bleeding as LMWH and VKA (0.8% vs. 1.2% and 0.1% vs. 0.3%, respectively).³⁵ The same was true in a study on 14,000 patients with atrial fibrillation with major hemorrhage in 3.6% per 100 patient-years with rivaroxaban versus 3.45% with warfarin.³⁶

RISK FACTORS FOR BLEEDING

The risk factors for bleeding can be classified as drug related and patient related. Drug-related risk factors may result in excessive drug levels—every known antithrombotic agent has a dose-bleeding relationship, although this curve may be more or less steep for different drugs.

Drug-Related Risk Factors

The risk of bleeding with overly intensive treatment has been best documented with VKAs. The risk of intracranial bleeding doubles with each increase of approximately 1 in the INR.³⁷ The risk for major bleeding on warfarin doubles when the treatment range is above INR 3.0 compared to when it is INR 2.0 to 3.0.³⁸, ³⁹, ⁴⁰, ⁴¹ Patients with myocardial infarction, treated with UFH in a large trial, had an increase of the risk of bleeding of 7% for every 10-second additional prolongation of the activated partial thromboplastin time (APTT).⁴² Increased bleeding with higher doses have also been reported for lepirudin,⁴³ argatroban,²⁵ and dabigatran.⁷

High variability in the drug level will inevitably result in frequent periods of excessive anticoagulation. The VKAs are most susceptible to such variations due to a multitude of interactions with other drugs and with food and a narrow therapeutic range. The variability is a risk factor that is independent of the mean INR.⁵, ⁴⁴, ⁴⁵ Furthermore, “time in therapeutic range” (TTR) is closely related to hemorrhagic as well as thromboembolic complications.⁴⁶ Nevertheless, a patient that has INR results consistently a few decimal points above the upper limit of the therapeutic range and accordingly a very low TTR has probably a lower risk of bleeding than another patient with alternating low and high INR results and a modest TTR. Therefore, a method that estimates both the variability of INR and the time spent in range probably provides the best estimate of risk for complications.⁴⁷

The management of anticoagulation with VKA needs to be meticulous and delivered by experienced staff to minimize the complication rate. A systematic review of 67 studies with more than 50,000 patients monitored by specialized anticoagulation clinics (68%), community practices (24%), or in clinical trials (7%) demonstrated that the study setting had the greatest influence on TTR, which was lowest in community practices.⁴⁸ Dosing decision support with manual algorithms⁴⁹, ⁵⁰ or computer software⁵¹ can effectively improve the quality of anticoagulant management.

Concomitant medications can result in increased risk of bleeding with VKAs either via direct interacting effects on the pharmacokinetics of VKA and resulting high INRs (see Chapter 106) or by exerting inhibition on other components in the hemostatic system, typically on the platelets (nonsteroid anti-inflammatory drugs, serotonin reuptake inhibitors, aspirin, glycoprotein IIb/IIIa, or P2Y₁₂ receptor inhibitors) (Table 104.2, Chapter 104; Table 106.1, Chapter 106). Multiple inhibitions of the hemostatic system or concomitant thrombolysis and increased bleeding have also been described for other drugs, for example, UFH.⁵², ⁵³, ⁵⁴, ⁵⁵

Only gold members can continue reading. Log In or Register to continue