Michael J. Kovacs, Division of Hematology, London Health Sciences Centre, 800 Commissioners Rd E, London, Ontario N6A 5W9, Canada. Tel.: +1 519 685 8500 ext. 52254; fax: +1 519 685 8477. E-mail: email@example.com
Summary. Background: Guidelines for perioperative warfarin management in patients with venous thromboembolic disease (VTE) are largely based on expert opinion.
Objectives: To assess the effectiveness and safety of a conservative perioperative anticoagulation strategy in patients with VTE on chronic warfarin therapy. Our center uses a conservative bridging approach for chronic VTE patients consisting of withholding warfarin for 5 days preoperatively, with prophylactic low-molecular-weight heparin (LMWH) post-procedure only if patients are admitted to hospital.
Patients/Methods: We performed a single-center retrospective cohort study. During the study period (1997–2011) there were 634 procedures in 416 patients that were reviewed for postoperative outcomes at 30 and 90 days.
Results: Of the 634 procedures, 156 procedures (24.6%) were completed as inpatients. Pre- and post-procedure LMWH bridging was used in 15 (2.4%) and 152 (24.0%) of all procedures, respectively. The 30-day VTE incidence was 0.32% (95% confidence interval [CI] 0.087–1.14), all non-fatal DVTs. The 30-day incidence of major and total bleeding events was 1.26% (95% CI 0.64–2.47) and 3.00% (95% CI 1.93–4.63), respectively. The all-cause mortality rate was 0.32% (95% CI 0.087–1.14) at 30 days; two patients died from arterial thrombosis events.
Conclusions: A randomized controlled trial is needed to provide definitive conclusions but a conservative bridging approach appears promising.
Despite a considerable number of patients on chronic warfarin therapy for venous thromboembolic disease (VTE), there is little evidence to guide perioperative anticoagulation management. There is significant inter-hospital variation in Vitamin K antagonist (VKA) management perioperatively, independent of surgical or patient-specific risk factors . The American College of Chest Physicians (ACCP) 2012 guidelines recommend that patients at low risk of perioperative thromboembolism (defined as previous VTE > 12 months ago) should not receive bridging with heparin products during interruption of VKA therapy . For moderate risk patients (defined as either VTE within the past 3–12 months, malignancy, non-severe thrombophilia or recurrent VTE), the bridging decision is based on their perceived risk profile; however, the authors suggest low-dose heparin products to decrease bleeding risk when appropriate . Unfortunately, there is a paucity of evidence to support these guidelines. Most studies have evaluated therapeutic-dose unfractionated heparin (UH) or low-molecular-weight heparin (LMWH) bridging regimes [3–5]. Of the studies examining no or prophylactic-dose LMWH use, the main study population was atrial fibrillation with only a small sub-population of VTE patients included [6,7].
The decision to use full-dose LMWH bridging to decrease perioperative thrombotic risk must be weighed against bleeding risk. There are significantly higher bleeding rates postoperatively in patients receiving LMWH bridging therapy compared with those without . If bleeding occurs with LMWH bridging, the subsequent temporary cessation of all anticoagulation can lead to a paradoxical increased thrombotic risk due to the postoperative state (prolonged immobilization, endothelial tissue injury and hypercoagulability) . In patients with chronic VTE, with or without other risk factors, simply withholding warfarin or using prophylactic LMWH may be preferable. Moreover, the need for preoperative LMWH bridging is not clear. In a recent anticoagulation study, there was no difference in recurrent thromboembolic events in a clinic group with short-term subtherapeutic INR values, compared with a matched cohort with therapeutic INRs at 90 days .
Our centre has used a conservative bridging protocol in patients with long-term (> 3 months) VKA use consisting of withholding warfarin for 5 days preoperatively and then resuming warfarin the same day as surgery. Postoperatively LMWH is given at prophylactic doses only if the patient is admitted to hospital. We conducted a single-center retrospective cohort study to examine the effectiveness of this conservative perioperative anticoagulation strategy in patients on chronic (> 3 months) warfarin therapy for previous VTE.
Materials and methods
We included VTE patients from our institution’s thrombosis unit between 1997 and 2011 who required temporary cessation of VKA therapy for planned procedures. Hematology clinic charts were reviewed for demographic, procedure and perioperative anticoagulant (VKA or LMWH) details. Procedures were included when there was preoperative review with the thrombosis clinic. Patients were excluded if they had other indications for warfarin, had an acute VTE less than 90 days before the procedure, had an atypical site for VTE, or had no follow-up planned after the procedure. Multiple procedures for a patient were included unless a second procedure was within 90 days of the first, in which case the second procedure was excluded (Fig. 1). Procedure details were verified with operative or procedure reports through an online database when available. This study was approved by the Lawson Health Research Institute and the University of Western Ontario’s Research Ethics Board.
Our centre uses a conservative bridging strategy in patients with chronic (> 3 months) VTE consisting of holding warfarin by missing five doses preoperatively with no preoperative LMWH given. Postoperatively, patients resume warfarin as soon as they can swallow. For ambulatory (i.e. day) procedures no LMWH is given postoperatively, whereas patients undergoing inpatient procedures are given prophylactic LMWH starting the morning after surgery until discharge or until the INR is > 1.9, whichever occurs first. Patient and surgical risk factors contribute to LMWH bridging decisions. For example, patients with protein S or C deficiency usually receive LMWH postoperatively to reduce the risk of warfarin skin necrosis.
Primary and secondary outcomes
The primary study outcome was the incidence of new symptomatic VTE within 30 days of the procedure. The Scientific and Standardization Committee (SSC) of the International Society on Thrombosis and Haemostasis (ISTH) recommends standardized reporting using a 30-day post-procedural follow-up period, to best capture related events . Data up to postoperative day 90 were also collected. Recurrent deep vein thrombosis was diagnosed using compression ultrasound of the extremities and was defined by a new non-compressible deep vein segment or if the diameter of thrombus had increased by at least 4 mm, compared with previous imaging . Recurrent pulmonary embolism was diagnosed when a new intraluminal filling defect on CT pulmonary angiogram (segmental or more proximal) or a new high probability perfusion defect on ventilation-perfusion studies was seen. It is our clinic’s practice to have new baseline imaging done at 3–6 months from the original diagnosis . Secondary outcomes were 30 and 90-day incidences of major and total bleeding, and all-cause mortality . Major bleeding was defined as bleeding associated with death, a critical site (intracranial, intraspinal, intraocular, pericardial or retroperitoneal), need for transfusion of at least 2 units of blood, or a drop in hemoglobin of at least 20 g L−1 (2.0 g dL−1) with overt bleeding . Total bleeding included all major and non-major bleeding. All deaths were reviewed in detail using hospital charts and online databases. Postoperative arterial events were recorded in all patients and included myocardial infarction, peripheral arterial thromboembolism, ischemic stroke and transient ischemic attack. Peripheral arterial thromboembolism was confirmed by angiography or surgery, and stroke was documented by computed tomography (CT) or magnetic resonance imaging (MRI). All outcomes were measured from time of warfarin cessation.
Outcomes were reviewed by each author in a non-blinded fashion. Charts were evaluated by a second reviewer when retrospective data were unclear, and consensus achieved by all authors.
Outcomes were analyzed per procedure. Due to the retrospective nature of the study we conducted a power analysis to determine the appropriateness of our sample. We estimated that a sample of 634 procedures achieved 95% power to detect a difference of 3% assuming that the underlying proportion of bleeding events is 6% using a two-sided binomial hypothesis test at the 0.05 level of significance. Confidence intervals for proportions were calculated using the Wilson’s score method. Groups were compared using chi-squared or Fisher’s exact tests. Spearman’s correlation coefficient was used to analyze correlations. Survival analysis was carried out using the Kaplan–Meier method. Two-sided P values ≤ 0.05 were considered statistically significant. All calculations were done using open epi Version 2.3, pasw 18.0 or spss 20.0 software (IBM Corporation, Armonk, NY, USA) .
There were 416 patients included (Fig. 1, Table 1); the mean patient age at time of first procedure was 64.9 years (SD 15.8) and 47.4% were male. There were 141 (33.9%) patients who had recurrent (≥ 1) VTEs. A patient’s most recent VTE was categorized as idiopathic (79.1%), malignancy-associated (12.7%), provoked with ongoing risk factors (5.8%) or line-associated VTE (2.4%). Malignancy-associated VTE was defined as patients with active malignancy treated within 6 months or palliative care . Of the venous thromboembolic events, 238 (57.2%) were only lower extremity DVTs, 15 (3.6%) upper extremity DVTs and 163 (39.2%) PEs, with (46) or without (117) DVTs. Thrombophilia testing was completed in 298 patients (71.6%), and 54.4% of those tested had a thrombophilia. There were 114 (38.3%) patients tested who had a significant thrombophilia (defined as homozygous or compound heterozygous mutation of factor V Leiden and prothrombin G20210A mutation, lupus anticoagulant, antiphospholipid antibodies, or antithrombin, protein S or C deficiency). There were 13 (3.1%) patients with chronic IVC filters in place. There were 15.9% of patients who had their first procedure within 3–12 months of their VTE diagnosis (Table 1).
Table 1. Baseline characteristics
n (%) (n = 416)
PE, pulmonary embolism; DVT, deep vein thrombosis; VTE, venous thromboembolism.
Age, mean ± SD
64.9 ± 15.8
Weight, mean ± SD (kg)
88.7 ± 22.6
PE ± DVT
Etiology of VTE
> 12 months
Number of procedures
Four or more
All patients were seen or contacted postoperatively. There were 27 patients who were seen postoperatively before the 90-day follow-up period was complete.
There was a total of 634 procedures, 478 (75.4%) were outpatient procedures and 156 (24.6%) were completed as inpatients. The most common outpatient procedures were colonoscopy (n = 133), dental procedures or surgeries (n = 118), and skin excisions/biopsies (n = 45). The most common inpatient procedures were total knee (n = 26) and hip replacements (n = 18). A variety of procedures, ranging from simple (minor plastic or dental surgery) to complex (major orthopedic, vascular and neurosurgery) were recorded (Table 2).
Table 2. Type of procedure
n (%) (n = 634)
AAA, abdominal aortic aneurysm; ENT, ear nose throat.
Pre- and postoperative LMWH bridging was used in 15 (2.4%) and 152 (24.0%) of all procedures, respectively. Specifically, 97.6% of procedures received no bridging anticoagulation preoperatively. There were seven (1.1%) and eight (1.3%) procedures with preoperative prophylactic-dose and full-dose LMWH, respectively. Postoperatively, 95% of outpatient procedures used no LMWH. There were 20 (4.2%) and four (0.8%) outpatient procedures that used postoperative prophylactic-dose and full-dose LMWH, respectively. In contrast, 17.9% of inpatient procedures used no bridging anticoagulation. The most common inpatient bridging strategy was postoperative prophylactic-dose LMWH (79.5%). Only four (2.6%) inpatient procedures used full-dose LMWH bridging. All patients who received preoperative full-dose LMWH also received postoperative full-dose LMWH. There were no IVC filters placed for perioperative anticoagulation management; those with IVC filters had them already in situ.
There were two episodes of recurrent VTE, giving a 30-day cumulative incidence of symptomatic VTE of 0.32% (95% confidence interval [CI] 0.087–1.14). At 90 days, there were four episodes of recurrent VTE, giving a 90-day cumulative incidence of 0.63% (95% CI 0.25–1.61). This consisted of four DVTs with no associated mortality. One patient developed a left leg DVT despite prophylactic-dose LMWH after undergoing a left total knee replacement, 43 days postoperatively. A second patient had no LMWH bridging and developed a left leg DVT on postoperative day (POD) 3 after her warfarin was withheld for an esophageal dilatation. The remaining two postoperative DVTs developed after warfarin was withheld for an extended period of time. One patient developed a right leg DVT on POD 29 from a dacryocystorhinostomy after warfarin had been discontinued due to a major gastrointestinal bleed. The other patient developed a right leg DVT 41 days after a complicated postoperative course from a craniopharyngioma resection and had been off and on prophylactic LMWH after an unrelated complication of a cerebral spinal fluid (CSF) hygroma. Three of the four patients who had VTE recurrence also had additional risk factors (Table 3).
TKR, total knee replacement; Proph, prophylactic dose; LMWH, low-molecular-weight heparin; LDVT, left lower extremity deep vein thrombosis; RDVT, right lower extremity deep vein thrombosis; POD, postoperative day; PE, pulmonary embolism; A/C, anticoagulation; GI, gastrointestinal.
Post-op proph LMWH
Recurrent idiopathic VTE
Initial PE 6 months prior
A/C held for GI bleed No thrombophilia testing
Post-op proph LMWH
A/C held for complication Antiphospholipid syndrome
There were eight episodes of major bleeding within 30 days postoperatively, giving a 30-day incidence of 1.26% (95% CI 0.64–2.47). There were 10 episodes of major bleeding, giving a 90-day cumulative incidence of 1.58% (95% CI 0.86–2.88). At 30 and 90 days, there were 19 and 22 episodes of any postoperative bleeding, respectively (3.00%, 95% CI 1.93–4.63; 3.47%, 95% CI 2.30–5.20). Of the major bleeds, there were five gastrointestinal (GI), two subdural, one vaginal, one urological and one localized to the surgical site. The procedures prior to major bleeds were left hip aspiration (GI bleed POD 13), dacryocystorhinostomy (see ‘Primary outcomes’, GI bleed POD 8), esophageal dilatation (see ‘Primary outcomes’, GI bleed POD 73), vaginal hysterectomy and bladder suspension surgery (GI bleed POD 2), open abdominal aortic aneurysm (AAA) repair (GI bleed POD 8), craniopharyngioma resection (see ‘Primary outcomes’, subdural hematoma POD 75), colonoscopy (subdural hematoma POD 6), cervical cone biopsy (vaginal bleed POD 14), trans-urethral bladder tumor resection and urethrotomy (hematuria POD 26), and lipoma excision (surgical site bleed POD 4). No patient with major bleeding received full-dose LMWH. Only two of the 10 patients with major bleeding received postoperative prophylactic LMWH (hysterectomy, craniopharyngioma resection); the remaining eight patients received no LMWH.
There were four deaths giving an all-cause 90-day mortality rate of 0.63% (95% CI 0.25–1.61), two of which were within 30 days (0.32%, 95% CI 0.087–1.14). There were no deaths due to pulmonary embolism. One patient had an endovascular abdominal aortic aneurysm repair and received postoperative prophylactic-dose LMWH, and on POD 3 suffered a fatal myocardial infarction. A second patient had a large ischemic stroke several hours before her colonoscopy after withholding warfarin for 5 days. At presentation to the emergency department this patient was found to be in new atrial fibrillation. She died 18 days later from her ischemic stroke. There were two patients with non-VTE related deaths; one patient died 85 days after she had a dental extraction, from malnutrition/electrolyte abnormalities. The second patient’s cause of death was unknown because he died at home, 60 days after a dental extraction. He presented to the emergency department the day before with dyspnea. Basic investigations were completed, including an INR which was 2.9. While VTE is unlikely, it cannot be excluded. All arterial events were fatal.
There was no difference in outcomes between patients who received LMWH vs. no LMWH therapy (Table 4; Fig. 2A–B). When comparing post-procedure LMWH use with no LMWH use, there was no statistical difference in VTE events (P = 0.244, Table 2), or VTE-free survival at 90 days using a Kaplan–Meier analysis (log-rank P = 0.220, Fig. 2A). There was also no significant difference in major bleeding events (P = 1.0, Table 2) or major bleeding-free survival (log-rank P = 0.755, Fig. 2B) between the two groups at 90 days. There were significantly more total bleeding events with inpatient procedures compared with outpatient procedures at 90 days (log-rank P = 0.020, Fig. 2C). However, there was no difference in major bleeding or VTE events when comparing inpatient and outpatient procedures (P = 0.678 and 0.237, respectively).
Table 4. Effect of type of procedure and use of LMWH bridging on outcome
There was a significant correlation between bleeding and VTE events at 90 days: 13.6% of patients with any bleeding and 30% with major bleeding developed a VTE complication, compared with 0.2% of those without bleeding. Spearman’s correlation coefficients were 0.311 and 0.470, respectively (P < 0.001, Table 5).
Table 5. Bleeding and VTE correlation
No (n = 630)
Yes (n = 4)
VTE, venous thromboembolism.
Any bleeding, n (%)
Major bleeding, n (%)
This study supports a conservative perioperative anticoagulation strategy in patients on chronic (> 3 months) warfarin therapy for VTE, evidence further strengthening the 2012 ACCP guidelines. Our strategy consists of withholding warfarin 5 days preoperatively with no LMWH given. Postoperatively warfarin is resumed on the same day of surgery, and prophylactic-dose LMWH is given postoperatively only if patients are admitted to hospital. With this strategy the postoperative recurrent VTE rate was only 0.32%, while the major bleeding rate was 1.26% at 30 days.
Despite a common belief that this topic is well studied, there are few studies in this area. Pengo et al.  created an individualized bridging protocol using a prophylactic dose or intermediate (70%) dose of LMWH based on risk of thromboembolic disease. While the main study population was atrial fibrillation, there was a subgroup of 210 VTE patients. Unlike our study, the low-moderate-risk VTE patients (VTE > 3 months) received prophylactic LMWH pre- and post-procedure, and had no thromboembolic events. All thrombotic outcomes were in high-risk patients (30-day incidence 0.4%), 40% of whom were in the VTE patient population. Major bleeding was reported by Pengo et al. as a 30-day incidence of 1.2%. A study completed by Garcia et al. examined perioperative bridging practices at 101 practice sites using a prospective registry. Of the 1293 episodes of warfarin interruption, perioperative heparin products were used in 8.3% of cases, giving a 30-day thrombosis incidence of 0.7% . However, only a small subgroup (n = 201) of these interruptions were in patients with previous VTE, limiting conclusions in the chronic VTE population. Unlike our study, Garcia et al. did not include vascular or major surgical procedures. McBane et al.  assessed a therapeutic-dose LMWH bridging strategy in patients with venous thromboembolic disease. Decisions to use the LMWH bridging strategy were based on patient and procedure-specific risk; however, risk stratification details were not described. In contrast to our study, there were 59% of patients with chronic VTE who received bridging with full-dose LMWH. The 90-day thrombosis incidence was 1.8% in all patients, with a bleeding risk of 1.8%. Of note, their total patient population was different to ours; they included acute and subacute VTE groups, and had more patients with chronic IVC filters (12.4% vs. 3.1%) and provoked VTEs (70% vs. 6%). Like McBane et al., we used a 90-day outcome to ensure all events were included. Thus our study is the largest to date and has the most conservative approach compared with previous studies but has similar favourable outcomes with respect to recurrent VTE.
In our study we also found a significant correlation between bleeding and recurrent venous thromboembolism. There were 30% of patients with major bleeding and 13.6% of patients with any bleeding that had recurrent VTE compared with only 0.2% for those with no bleeding. This correlation has been previously described in a prospective cohort study evaluating perioperative anticoagulation in patients with mechanical heart valves or atrial fibrillation. Kovacs et al.  found that six out of the eight thromboembolic outcomes occurred when warfarin was withheld because of bleeding. Clinicians should be cautious when considering therapeutic-dose LMWH; prolonged cessation of anticoagulation because of bleeding may subsequently place a patient at increased thrombotic risk.
Chronic VTE patients may require prophylactic LMWH post-procedure, a decision based on patient and surgical risk factors. Risk stratification tools exist in the general surgical population to guide prophylaxis use for primary prevention of VTE . Although retrospective, the Caprini scoring method is a tool validated in general, vascular and urological surgery inpatients that takes into account surgery type, medical co-morbidities and VTE risk factors . Based on guidelines using the Caprini scoring method, our inpatients would have been considered for prophylactic LMWH post-procedure . The Caprini scoring system is comprehensive but not always practical to perform. Unfortunately, a similar perioperative risk stratification model does not exist for chronic VTE patients. In our study the event rate was not large enough to detect differences in patients with thrombophilias, malignancy or recurrent VTE. Further studies are still needed to validate these risk factors perioperatively in patients with previous VTE.
There are limitations inherent in our study design. Reviewing clinic charts retrospectively has the potential for incomplete data or lack of adequate follow-up. We utilized an online hospital database to confirm procedures and test results. Our clinic charts also included supplemental notes from a nurse who contacted patients regularly for INR management. We are not likely to have missed any significant outcomes, as patients with serious symptoms would have contacted a physician, and all patients have at least an annual clinic assessment at which they are asked specifically about bleeding or VTE complications. There is potential for treatment or selection bias because this study was non-blinded. Our study observed the effectiveness of a real-life clinical practice, where anticoagulation decisions were individualized and recommended to another treating physician.
Because of the retrospective cohort design and low proportion of VTE events, a larger prospective study is needed to verify our preliminary results, including higher risk surgical procedures such as arthroplasty. Two randomized controlled trials are underway to examine optimal perioperative anticoagulation management for patients with atrial fibrillation (the BRIDGE study) and with atrial fibrillation or mechanical heart valves (PERIOP 2) (ClinicalTrials.gov identifiers: NCT00786474 and NCT00432796). A similar trial for patients with venous thromboembolic disease would be ideal.
In patients with venous thromboembolic disease (DVT or PE) on warfarin therapy for at least 3 months duration, a conservative perioperative anticoagulation strategy appears promising. A prospective study or carefully designed registry is needed to confirm our findings.
Disclosure of Conflict of Interests
A. Lazo-Langner has received honoraria from Pfizer Inc. and Leo Pharma. The other authors state that they have no conflict of interest.