Predictors of major bleeding in peri-procedural anticoagulation management


Robert D. McBane II, Gonda Vascular Center, Mayo Clinic, 200 First Street SW, Rochester, MN, USA.
Tel.: +1 507 266 3964; fax: +1 507 266 1617.


Summary.  Background: Appropriate periprocedural management for chronically anticoagulated patients requires assessment of patient-specific thrombosis and bleeding risks. However, predictors of post-procedure bleeding are unknown. Objectives: To determine the 3-month cumulative incidence and independent predictors of peri-procedural bleeding in chronically anticoagulated patients requiring temporary warfarin interruption for an invasive procedure. Methods: In a protocol driven, cohort study design, all patients referred to the Mayo Clinic Thrombophilia Center for peri-procedural anticoagulation management (1997–2007; n = 2182), were followed forward in time to determine the 3-month cumulative incidence of peri-procedural bleeding (Kaplan–Meier product limit) and potential predictors of bleeding (Cox proportional hazards). Decisions to ‘bridge’ with low-molecular-weight heparin were based on estimated thromboembolism and bleeding risk. Results: Indications for chronic anticoagulation included venous thromboembolism (38%), atrial fibrillation (30%) and mechanical heart valves (27%). Of these, 1496 (69%) patients received bridging therapy. The 3-month cumulative incidence rates of major and overall bleeding were 2.1% and 5.1%, respectively. Major bleeding occurred more frequently in patients receiving bridging therapy (3% vs. 1%; P = 0.017). Independent predictors (hazard ratio; 95% confidence interval) of major bleeding included mitral mechanical heart valve (2.2; 1.1–4.3), active cancer (1.8; 1.0–3.1), prior bleeding history (2.6; 1.5–4.5) and re-initiation of heparin therapy within 24 h after the procedure (1.9; 1.1–3.4). Conclusion: Factors predisposing to peri-procedural bleeding are primarily patient-specific. Premature heparin re-initiation is an avoidable provider-specific variable to consider.


Each year, nearly 10% of the more than 2.5 million chronically anticoagulated Americans require temporary warfarin discontinuation for an invasive procedure [1]. Prescribing the most appropriate peri-procedural anticoagulation management strategy for these patients requires balancing the risk of peri-procedural thrombosis with the risk of major hemorrhage. For patients at moderate to high risk of thromboembolism, ‘bridging’ therapy with either low-molecular-weight heparin (LMWH) or unfractionated heparin is provided during both the warfarin interruption period prior to the procedure and the re-initiation period post-procedurally once hemostasis is assured. For patients deemed to be at low risk of thromboembolism, warfarin is simply stopped and then restarted post-procedurally without interval heparin therapy. This general approach is similar for patients with a variety of warfarin indications, including mechanical heart valves [2,3], atrial fibrillation [4,5] or venous thromboembolism [6,7]. In general, bleeding outcomes exceed thromboembolism rates for patients receiving chronic anticoagulation therapy. [8] Guidelines for peri-procedural anticoagulation management place a relatively high value on preventing thromboembolism and a relatively lower value on preventing bleeding, particularly in patients with moderate to high thromboembolism risk [9]. In low-risk patients, the recommendations place a higher value on bleeding prevention. These guidelines are primarily based on expert opinion. There is a paucity of data to assist the healthcare provider in identifying those individuals at increased risk of major hemorrhage in the post-procedural setting. Predictors of peri-procedural bleeding remain undefined. Compared with a post-procedural thromboembolism, major bleeding can result in equally devastating morbidity and mortality. To address these limitations, consecutive patients referred to the Mayo Clinic Thrombophilia Center for peri-procedural anticoagulation management over the 11-year period, 1997–2007, were followed forward in time to estimate the 3-month cumulative incidence of major and overall bleeding. We further sought to identify independent predictors of major hemorrhage for this patient population.


Study population, design and setting

Consecutive, chronically-anticoagulated patients who were referred to the Mayo Clinic Thrombophilia Center for peri-procedural anticoagulation management over the 11-year period 1 January 1997 to 31 December 2007 were eligible for inclusion; 97% consented to participate. All patients were followed forward in time for 3 months from the date of the Thrombophilia Center consultation. Patients (or family members for deceased patients) who did not return for a clinic visit were mailed a questionnaire and/or contacted by telephone for any symptoms or signs of thromboembolism or bleeding in the 3 months after the Thrombophilia Center consultation, and for vital status. The local medical records of patients reporting thromboembolism or bleeding, and death certificates and autopsy reports for deceased patients, were obtained and reviewed by the study endpoint adjudication committee. Two experienced study nurse abstractors reviewed the complete inpatient and outpatient medical records for each patient. The study was approved by the Mayo Clinic Institutional Review Board.

Peri-procedural anticoagulation management

Patients were referred to the Thrombophilia Center for peri-procedural anticoagulation management recommendations 4–7 days before the anticipated procedure. Each patient was evaluated for the patient-specific risk of thromboembolism and the procedure-specific risk of major bleeding. For patients requiring outpatient procedures associated with either a low risk of bleeding or easy access for physical hemostasis, the intensity of warfarin anticoagulation was reduced to the lower limit of the therapeutic range. Procedures in this low-risk category include dental extractions, cataract surgery, dermatologic surgery and bone marrow biopsy. For patients undergoing procedures with moderate to high bleeding risk, warfarin was stopped 5 days before surgery and resumed as soon as possible after surgery, starting with the patient’s usual daily warfarin dose.

For patients deemed to be at moderate to high thromboembolism (TE) risk, warfarin was stopped 5 days before surgery and the patient was ‘bridged’ with LMWH as an outpatient when the international normalized ratio (INR) was anticipated to be below the lower limit of the therapeutic range. The last LMWH injection was given on the morning prior to the procedure day and at 50% of the daily dose. For these patients, post-procedure warfarin and LMWH therapy were overlapped for at least 5 days and until the INR exceeded the lower limit of the therapeutic range for at least 24 h. When indicated, aspirin was typically stopped 1 week before the procedure and restarted after the procedure when hemostasis was assured. Over the 10-year time period, three successive LMWHs (ardeparin, enoxaparin and dalteparin) were on the Mayo Clinic formulary and were used for peri-procedural anticoagulation therapy.

The ‘preoperative’ anticoagulation management for these patients was under direct supervision of personnel in the Thrombophilia Clinic, as was the post-hospital discharge management. Specific post-surgical recommendations were provided at the time of consultation. Surgical teams overseeing the postoperative management of these patients directed the anticoagulation re-initiation during the hospital stay, often with input from Thrombophilia Center consultants.

Definition of outcomes

Major bleeding was defined as overt bleeding and a hemoglobin decrease of ≥ 20 g L−1 after the procedure or transfusion of ≥ 2 units of packed red blood cells, or intracranial, intraspinal, intraocular, retroperitoneal, pericardial or fatal bleeding [10]. Minor bleeding was defined as overt bleeding that did not meet criteria for major bleeding. All events were adjudicated using a priori study criteria by a committee comprised of four Thrombophilia Center physicians blinded to patient name and healthcare provider.

In order to determine impact of surgical complexity on bleeding outcomes, we divided procedures by bleeding risk. Biopsies, endoscopy, endovascular interventional procedures, dental extractions, cataract surgery, pacemaker placement, arthroscopic surgery and dermatologic surgery were classified as ‘low bleeding risk procedures’. Some procedures within this low-risk category required warfarin discontinuation. Procedures not included in this category were thereby categorized as ‘moderate-high bleeding risk’. All procedures within this category required warfarin interruption.

Thromboembolic complications, including symptomatic arterial (ischemic stroke, transient ischemic attack, amaurosis fugax, unstable angina, myocardial infarction or other peripheral artery thromboembolism) and venous thromboembolic (deep vein thrombosis or pulmonary embolism) events, were defined as previously described [2,4,6].

Statistical analysis

Continuous numeric variables were reported as means with SD. Frequencies were reported when appropriate. The main group combinations included major bleeding, overall bleeding, and bridged and non-bridged patients. From the univariate results, variables that met a 0.10 level of significance were included in the multivariate Cox proportional hazards models for each patient group/outcome. A final Cox model was constructed for each of the patient group/outcome combinations.


During the 11-year study period, 2182 unique patients (2484 invasive procedures) receiving chronic warfarin therapy (42% women; mean age 66 ± 14 years) were referred to the Thrombophilia Center for peri-procedural anticoagulation management (Table 1). The main indications for chronic anticoagulation were venous thromboembolism (38%), atrial fibrillation (30%) and mechanical heart valves (27%). Of these, 1496 patients (43% women; mean age 65 ± 14 years) were provided with bridging therapy with LMWH. Patients receiving bridging therapy with LMW heparin were on average younger and more frequently female. Bridging therapy was more frequently provided for patients with mechanical heart valves, prior cardioembolic events including stroke, heart failure and those receiving chemotherapy for active malignancy. In contrast, patients with non-valvular atrial fibrillation and those with severe renal insufficiency were less frequently bridged with LMWH. Procedural classification is provided in Table 2. Post-procedural heparin use did not differ for ‘low bleeding risk’ vs. ‘moderate-high bleeding risk’ surgeries (P = 0.120). Ninety-day follow-up was available on all patients.

Table 1.   Baseline demographics
VariableNo heparin (n = 686)Heparin (n = 1496)P-value
Age67.6± 13.464.9± 14.0< 0.001
Gender (female), No. (%)263 (38)643 (43)0.041
Warfarin indication
 Mechanical heart valves, No. (%)115 (17)465 (31)< 0.001
  Aortic, No. (%)95 (14)330 (22)< 0.001
  Mitral, No. (%)23 (3)171 (11)< 0.001
  Tricuspid, No. (%)3 (0)7 (0)0.92
  Pulmonary, No. (%)1 (0)1 (0)0.57
 VTE, No. (%)281 (41)552 (37)0.07
 Atrial fibrillation, No. (%)236 (34)414 (28)0.001
 Prior stroke, No. (%)67 (10)207 (14)0.008
 Dilated cardiomyopathy, No. (%)15 (2)33 (2)0.98
 Prior cardioembolic event, No. (%)12 (2)57 (4)0.01
 Intracardiac thrombus, No. (%)5 (1)21 (1)0.18
Malignant neoplasm, No. (%)225 (33)490 (33)0.98
 Chemotherapy, No. (%)24 (19)87 (28)0.03
Past medical history
 Anemia, No. (%)38 (6)105 (7)0.19
 Bleeding, No. (%)134 (20)299 (20)0.81
 Chronic liver disease, No. (%)4 (1)8 (1)0.89
 Chronic renal disease, No. (%)41 (6)61 (4)0.05
  On dialysis, No. (%)7 (1)9 (1)0.29
 CHF, No. (%)72 (10)207 (14)0.03
 CAD, No. (%)194 (28)471 (31)0.13
 Diabetes mellitus, No. (%)109 (16)262 (18)0.35
 ASA, No. (%)112 (16)258 (17)0.60
 Clopidogrel, No. (%)5 (1)18 (1)0.31
 Statin therapy, No. (%)194 (28)415 (28)0.79
 Hemoglobin < 10, No. (%)32 (4)90 (5)0.28
 Platelet < 100, No. (%)14 (2)28 (2)0.74
 Platelet < 150, No. (%)69 (9)167 (10)0.54
 Platelet > 150, No. (%)652 (85)1469 (86)0.50
 White blood cell count  < 10, No. (%)654 (89)1498 (90)0.49
  < 60, No. (%)358 (46)794 (46)0.94
  < 30 or renal dialysis,  No. (%)40 (5)39 (2)< 0.001
  > 60, No. (%)323 (42)789 (46)0.06
Table 2.   Procedural types
Type of procedure/surgeryAll procedures
n (% of total)
‘Moderate-high bleeding risk’
n (% of procedure type)
Gastrointestinal631 (24)195 (31)
Orthopedic491 (19)466 (95)
Cardiothoracic381 (15)70 (18)
Urologic344 (13)168 (49)
Dental138 (5) 
Gynecologic112 (4)102 (91)
Neurosurgical111 (4)55 (50)
Vascular87 (3)79 (91)
Plastic75 (3)21 (28)
Interventional radiology68 (3) 
Ophthalmologic67 (3)1 (1)
ENT66 (3)57 (86)
Other9 (0.4) 

At 3 months follow-up, the overall bleeding rate was 5% (n = 127) and 2% (n = 51) were major bleeding events. For those patients receiving peri-procedural heparin (n = 1496 patients, 1714 procedures), the rate of any bleeding was 6% (n = 104) and the rate of major hemorrhage was 3% (n = 43; Fig. 1). For those patients not receiving peri-procedural heparin (n = 686 patients, 770 procedures), the overall bleeding rate was 3% (n = 23) with a major bleeding rate of 1% (n = 8). Two deaths were attributed to major bleeding.

Figure 1.

 Three-month cumulative incidence of peri-procedural bleeding stratified by bridging strategy. Survival free of any bleeding (Panel A; P = 0.001) or major bleeding (Panel B; P = 0.018) was significantly lower for patients receiving peri-procedural heparin bridging therapy (dashed line) compared with those not receiving heparin therapy (solid line).

Of the 1714 patients who used heparin, 967 resumed heparin at 24 h and 210 resumed heparin at 48 h post-procedure. The remainder (n = 537) resumed heparin beyond 48 h. Of the 1177 patients who resumed heparin, 27 patients experienced a major bleeding event. All of these major bleeding events occurred in patients who resumed heparin within 24 h of the procedure. There were no events in the patients who resumed heparin ≥ 48 h following the procedure.

In the multivariate analysis, independent predictors of major bleeding for all patients included mitral mechanical heart valve (hazard ratio [HR], 2.2; 95% confidence interval [CI], 1.1–4.3), active cancer (HR, 1.8; 95% CI, 1.0–3.1), prior bleeding history (HR, 2.6; 95% CI, 1.5–4.5) and re-initiation of heparin therapy within 24 h of the procedure (HR, 1.9; 95% CI, 1.1–3.4; Table 3). Aspirin use was not associated with major bleeding.

Table 3.   Independent predictors of major bleeding
HR95% CIP-valueHR95% CIP-value
Mitral mechanical heart valve2.21.1––4.80.02
Malignant neoplasm1.81.0–3.10.04 NS 
Bleeding history2.61.5–4.5< 0.0012.21.2–4.20.01
Re-started low-molecular-weight heparin within 24 h (n = 926)1.91.1–3.40.02 NS 
PLT < 150.000 NS 2.31.1–4.70.02
‘Moderate-high risk’ procedure NS 2.21.2–4.10.01

In a separate multivariate analysis, independent predictors of major bleeding specific to those patients receiving heparin bridging therapy included mitral mechanical heart valve (HR, 2.4; 95% CI, 1.1–4.9), ‘moderate-high bleeding risk’ procedure (HR, 1.9; 95% CI, 1.0–3.6), active cancer (HR, 1.9; 95% CI, 1.0–3.5), thrombocytopenia < 150 000 (HR, 2.3; 95% CI, 1.1–4.6) and a history of bleeding (HR, 2.1; 95% CI, 1.1–4.1).

Using the independent predictors of major bleeding, a bleeding risk score was devised for ease of clinical application. ‘BleedMAP’ assigns one point for each risk factor: history of prior bleeding (Bleed), mechanical mitral heart valve (M), active cancer (A), and low platelets (P). For all patients and for those receiving heparin therapy, the risk of major bleeding according to this bleeding risk score is provided in Table 4. Using this risk stratification tool, the bleeding rates varied more than 10-fold from highest to lowest risk scores (Fig. 2A). The thromboembolic event rates are stratified by BleedMAP score (Fig. 2B). For risk score of 0, the rate of thromboembolic complications was 0.71% for all patients and 0.60% for those receiving bridging therapy. For risk scores of 1–2, the rate was 0.89% for all patients and also for bridged patients. For those patients with risk scores ≥ 3, there were no events irrespective of bridging status.

Table 4.   BleedMAP score and rate of major hemorrhage
BleedMAP scoreMajor bleed rate (%)95% confidence intervals
All patients
 Score 00.810.35–1.60
 Score 1–22.671.90–3.63
 Score ≥ 310.002.79–23.66
Patients receiving LMWH
 Score 00.750.24–1.74
 Score 1–23.362.23–4.66
 Score ≥ 312.123.40–28.20
Figure 2.

 Major event rates according to BleedMAP score. A gradual progressive rate of major hemorrhage was observed with increasing BleedMAP scores for patients receiving bridging therapy with low-molecular-weight heparin (light shade) and all patients (dark shade) (Panel A). Thromboembolism rates were low and non-existent at high scores (Panel B).


In a protocol driven study, we have identified independent predictors of major bleeding from a large patient cohort (n = 2182 patients; 2484 procedures) referred for peri-procedural bridging management. These include mitral mechanical heart valve, active cancer, prior bleeding, thrombocytopenia and post-procedural heparin re-initiation within 24 h. We have proposed a novel score system for risk assessment of peri-procedural bleeding for which we assigned the acronym, BleedMAP. This novel score system with an easy to remember acronym assigns patients to low, intermediate and high-risk groups, providing an estimate of major hemorrhage stratified by bridging strategy (Fig. 2A). BleedMAP also appears to provide an assessment of thromboembolic risk, which conveniently trends in the opposite direction (Fig. 2B). The 3-month cumulative incidence of major bleeding was 2%. The rate of any bleeding during this period was 5%. Previously reported peri-procedural bleeding rates for patients receiving chronic warfarin therapy are variable and depend on the population studied, the bridging protocol utilized and the duration of follow-up. Factors associated with increased bleeding rates in these patients have been proposed, including patient age [11], active cancer [6], an elevated Charlson score [5], and the dosing of LMWH.[12,13] Despite adequate thromboembolic risk stratification guidelines [9], until now there has been a paucity of data validating predictors of peri-procedural bleeding in these patients.

Predictors can be divided into variables intrinsic to the patient, including mitral mechanical valve prosthesis, history of prior bleeding, thrombocytopenia and active cancer. These factors must be accounted for in the bridging recommendations but are not generally alterable. Premature re-initiation of therapeutic heparin, defined as restarted heparin within 24 h of the procedure, however, is an independent risk factor that could be altered. In our cohort, all major bleeding events occurred in those patients who resumed heparin use within 24 h. There were no events in patients who resumed heparin at 48 h or beyond. In our current practise, we have restructured our post-procedural heparin strategy such that therapeutic heparin (either unfractionated heparin or LMWH) is held for 48 h to allow adequate hemostasis following the invasive procedure. During this time interval, prophylactic dose heparin can be used in keeping with published venous thromboembolism prevention guidelines. Warfarin is re-initiated immediately post-procedure once the patient is able to take oral medications as the anticoagulant effect of this medication is delayed.

A recent prospective, multicenter, observational study was performed to characterize rates of bridging therapy as related to patient characteristics and 30-day outcomes at nine US medical centers with affiliated anticoagulation clinics [13]. Among 492 patients participating in this study, the rate of major bleeding was 3.2% at 30 days of follow-up. As in our study, aggressive post-procedural anticoagulation was the strongest predictor of bleeding complications. For medical patients, risk factors included age over 65 years, creatinine exceeding 2 mg dL−1, prior gastrointestinal hemorrhage, prior stroke, history of liver disease or active cancer, each equally weighted. For surgical patients, risk factors for major hemorrhage were assigned relative to the extent and complexity of the surgical procedure. Considerable heterogeneity of anticoagulation management existed across the nine medical centers. Several factors tested by these investigators (history of chronic kidney disease, calculated glomerular filtration rate, age, prior stroke and liver disease) were not independent predictors in our study. Prospective validation of any risk score is therefore warranted.

Active cancer has been suggested as a strong predictor of bleeding complications [6,8,14,15]. Cancer-specific therapy is associated with bleeding complications among patients who are on chronic anticoagulation [14]. Chemotherapy did not have an impact on our model for bleeding outcome association. Prandoni et al. [8] found a high risk of bleeding among patients with active cancer and venous thromboembolism receiving anticoagulant therapy. In a prospective cohort, the 12-month cumulative incidence of major bleeding was 12.4% (95% CI, 6.5–18.2) in patients with cancer compared with 4.9% (95% CI, 2.5–7.4) in patients without cancer, for a hazard ratio of 2.2 (95% CI, 1.2–4.1) [8].

While the incidence of peri-procedural thromboembolism may be as high as 3% during warfarin interruption in high-risk patients [16], others have reported rates more than double these percentages [4–9]. The current guidelines for peri-procedural anticoagulation management for patients on chronic warfarin reflect the preference of the guideline authors for preventing thromboembolism and a relatively low value is placed on preventing bleeding, particularly for patients at greater than low risk for thromboembolism [15]. And yet the morbidity and mortality associated with a major bleeding event can be substantial. Furthermore, these guidelines are based on limited data regarding bleeding risk prediction. We now provide a list of independent predictors for major bleeding in this setting. These data will be helpful in developing future guidelines for peri-procedural anticoagulant management for these patients.

Several limitations of this study should be noted. First, the delivery of low-molecular-weight heparin was not assigned randomly. Careful patient stratification based on perceived risks and benefits of LMWH by the attending physicians may have contributed to the low rate of bleeding complications observed. Patient preferences may have also contributed to these outcomes. Moreover, we did not control for total dose of LMWH. Others have noted a correlation between higher rates of bleeding and residual anti-Xa activity, which may reflect higher LMW heparin dosing [3,11,12,17]. Secondly, although referral to the Mayo Clinic Thrombophilia Center was open to all patients, we cannot exclude the possibility of referral bias. Referral bias may have had an impact on bleeding event rates in either a positive or negative manner. It is possible that healthcare providers referred the more complicated patients while managing the more straightforward cases themselves. Bleeding outcomes were adjudicated by a committee blinded to patient provider identifiers, thus increasing the internal validity of our study. Finally, by extending observations to 90 days, event capture was maximized to include the entire transition to outpatient anticoagulation management. A factor not included in our analysis was the INR time in range, which has been associated with bleeding outcomes in other scenarios [18,19]. If this variable was demonstrated to be predictive of bleeding events, evaluation of CYP2C9 polymorphisms may aid in determining the patients at highest risk of bleeding [20].

Our current practise has evolved to error on reducing the risk of peri-procedural bleeding given the low rates of thromboembolism [2,4,6]. Warfarin is restarted as soon as feasible once the patient is able to take oral medications. The re-initiation of post-procedural heparin is postponed for 48 h and limited to prophylaxis doses where feasible, depending on the chronic anticoagulation indication. This approach is especially warranted for patients on warfarin therapy for remote venous thromboembolism in keeping with the guidelines [21]. For these patients, full-dose heparin therapy postoperatively while restarting warfarin is unnecessary and likely to increase bleeding complications. This general approach is also warranted for atrial fibrillation patients without a prior history of thromboembolism, known intra-cardiac thrombus, or high CHADS 2 score [4]. Patients with bileaflet aortic valve prostheses and no additional risk factors (atrial fibrillation, prior thromboembolism, known intracardiac thrombus or heart failure) are treated similarly. Full-dose heparin initiated 48 h postoperatively is reserved for those patients deemed to be at high risk of thromboembolism: venous thromboembolism within 3 months; atrial fibrillation with prior stroke, thromboembolism, or known intracardiac thrombus; prosthetic heart valves other than aortic bileaflet prostheses or any mechanical prosthetic heart valve where there is a history of coexistent atrial fibrillation, prior stroke, thromboembolism or known intracardiac thrombus. Even under these high-risk conditions, heparin therapy is used carefully and conservatively wherever feasible and warranted.


Funded, in part, by grants from the Centers for Disease Control and Prevention (DD00235), U.S. Public Health Service, and by the Mayo Foundation.

Disclosure of Conflict of Interests

The authors state that they have no conflict of interest.