Incidence and predictors of bleeding or thrombosis after polypectomy in patients receiving and not receiving anticoagulation therapy

Authors


Daniel M. Witt, Kaiser Permanente Colorado, 16601 E. Centretech Pkwy, Aurora, CO 80011, USA.
Tel.: +1 303 739 3571; fax: +1 303 739 3574.
E-mail: dan.m.witt@kp.org

Abstract

Summary. Background and aims: To assess the effect of warfarin anticoagulation therapy (AC) on the incidence of colon bleeding after elective colonoscopy with polypectomy and to identify independent predictors of post-polypectomy colon bleeding. Methods: This was a retrospective cohort analysis. Patients interrupting warfarin AC therapy for polypectomy (AC group) were matched on age (± 3 years) with up to two patients who underwent polypectomy but were not receiving AC (non-AC group). Data were extracted from electronic medical, pharmacy and laboratory claims and records and manual medical chart review. Incidence rates of colon bleeding requiring hospitalization, other gastrointestinal bleeding, thrombosis and death in the 30 days post-polypectomy were compared between groups. Multivariate regression techniques were used to identify independent predictors of post-polypectomy colon bleeding. Results: A total of 425 AC group patients were matched to 800 non-AC group patients. Post-polypectomy colon bleeding occurred more often in AC group patients (2.6% vs. 0.2%, P = 0.005). There were no differences in the rates of other outcomes (P > 0.05). Independent predictors of colon bleeding included AC group status [adjusted odds ratio (AOR) = 11.6; 95% confidence interval (CI) = 2.3–57.3], number of polyps removed (AOR = 1.2; 95% CI = 1.1–1.4) and male gender (AOR = 9.2, 95% CI = 1.1–74.9). Conclusions: The incidence of post-polypectomy colon bleeding was higher in patients receiving AC even although warfarin was interrupted for the procedure. Independent predictors of colon bleeding were identified as: receiving AC, removal of multiple polyps and male gender. Our findings suggest that additional methods to reduce the likelihood of post-polypectomy colon bleeding in AC patients should be investigated.

Introduction

Colonoscopy is the most commonly performed gastrointestinal (GI) endoscopic procedure in the United States [1]. Complications of colonoscopy are uncommon but can be life threatening [2]. For example, polypectomy increases the likelihood of complications nearly 9-fold compared with colonoscopy without polypectomy [3], of which colon bleeding is the most common serious complication.

Patients receiving anticoagulant therapy (AC) with warfarin routinely undergo colonoscopy with polypectomy. Current guidelines suggest warfarin therapy should be interrupted for several days prior to polypectomy in an attempt to minimize the risk of post-polypectomy bleeding [4]. For patients at high risk of thrombosis in the absence of warfarin, cross-coverage with low-molecular-weight heparin (so-called bridging therapy) is frequently prescribed [4]. Anticoagulation therapy has been reported to increase the likelihood of post-polypectomy colon bleeding by up to thirteenfold [5,6]. These estimates, however, may be unreliable as they are derived from investigations with only a few AC patients (e.g. 23 out of 5152 patients) [6]; limiting their generalizability. In contrast, another study reported an equivocal relationship between AC and bleeding risk post-polypectomy [7]. Additionally, the potential influence of important factors such as indication for AC, number of warfarin doses held in preparation for, and anticoagulation intensity immediately prior to the procedure on the rate of bleeding in AC patients undergoing polypectomy has not been reported.

The purpose of this investigation was to assess the effect of warfarin AC on the incidence of serious colon bleeding after elective colonoscopy with polypectomy and to identify independent predictors of post-polypectomy bleeding in two large cohorts of patients who underwent polypectomy: one requiring interruption of warfarin AC for the procedure and the other not prescribed AC.

Methods

Setting

This analysis was conducted at Kaiser Permanente Colorado (KPCO), an integrated healthcare delivery system providing care to approximately 450 000 members in the Denver–Boulder metropolitan area, more than 7000 of which are patients treated with warfarin. All anticoagulated patients receive comprehensive services, including plans for peri-procedural warfarin interruption, from a centralized Clinical Pharmacy Anticoagulation Service (CPAS) [8]. Details of these plans are recorded in the patient’s electronic medical record (EMR) as well as an anticoagulant database (Dawn-AC; 4S Systems, Cumbria, UK).

Study design

All study activities were approved by the KPCO Institutional Review Board. This was a matched retrospective analysis. All KPCO members 18 years or older who had an elective, outpatient colonoscopy with polypectomy performed between January 2000 and December 2006 were eligible for study inclusion. Patients interrupting chronic warfarin therapy for polypectomy (AC group) were randomly matched using the GREEDY Matching Algorithm (Mayo Clinic College of Medicine 2003) on age (± 3 years) with up to two patients who underwent polypectomy but were not receiving warfarin at that time (Non-AC group). If a patient had more than one polypectomy during the study period, only data from the first episode were analyzed.

Study population

All included patients were continuously enrolled in the Kaiser Foundation Health Plan during the 180 days prior to and 30 days after polypectomy (unless death occurred during the 30 days following). Patients who received colonoscopy to evaluate emergent GI bleeding, had a history of bleeding from intestinal arteriovenous malformations (ICD-9 code 569.85), had insufficient EMR information to assess study-related variables, had their warfarin monitored outside of CPAS, or were receiving prophylaxis with warfarin to prevent venous thromboembolism after orthopedic surgery (because the duration of warfarin therapy in this setting is typically ≤ 6 weeks) were excluded.

Study outcomes

The primary outcome was the incidence rate of colon bleeding requiring hospitalization [including emergency department (ED) visits] within 30 days of polypectomy. The 30-day follow-up period was chosen as the majority of post-polypectomy bleeding complications have been reported to occur within this time frame [5,9]. Descriptions of each case were summarized. Secondary outcomes included the incidence rates of hospitalization for non-colon GI bleeding, hospitalization for thromboembolism and death attributable to bleeding or thromboembolism during the 30-day post-polypectomy follow-up. In addition, independent predictors of post-polypectomy colon bleeding were identifed.

Data collection

Patients who had undergone elective colonoscopy with polypectomy were identified by querying integrated KPCO electronic administrative databases using International Classification of Diseases, Ninth Revision (ICD-9) codes 45.42 and 48.36 and Current Procedural Terminology (CPT) codes 44392, 44394, 45384 and 45385. Information on concurrent AC use and primary indication for AC at the time of the colonoscopy were identified via the Dawn AC management system. Electronic membership databases were queried to identify patient age (on date of polypectomy), gender and KPCO membership eligility (including date of death when applicable). Integrated electronic medical office databases were queried to identify co-morbid diagnoses including intestinal arteriovenous malformations, previous history of hypertension, diabetes, stroke/transient ischemic attack, recent heart failure, active malignancy, bleeding and arterial or venous thrombosis.

Manual medical chart reviews were perfomed on all included patients to document information on indication for colonoscopy (anemia, colon cancer screening, constipation, diarrhea, irritable bowel, follow up from previous polypectomy and non-emergent rectal bleeding); number, location (ascending, transverse, descending, sigmoid and rectum) and size of polyps removed; description of polyp morphology (diminutive, flat, hyperplastic, laterally spreading, mixed, pedunculated and sessile); method of polyp removal (forceps, cold snare, snare cautery and combination of methods); and the use of aspirin and/or other antiplatelet therapy at the time of colonoscopy. A sum of all polyps removed was calculated for each patient. A global variable was created to indicate if the largest polyp removed was > 10 mm (as per the endoscopist’s dictated report).

Additional information collected for AC group patients only included indication for AC therapy, number of warfarin doses omitted prior to polypectomy, international normalized ratio (INR) value prior to polypectomy, use of unfractionated heparin or the low-molecular-weight heparin enoxaprin during the 7 days prior to and 30 days after polypectomy (bridging therapy), elapsed time (days) between polypectomy and resumption of warfarin (and enoxaparin if applicable) and INR value recorded at the time of anticoagulation therapy-related complications (when available and applicable).

During the 30 days after polypectomy, hospitalizations (including ED visits) were identified through queries of integrated electronic KPCO inpatient and ambulatory claims databases using predefined colon bleeding and thromboembolic event ICD-9 codes (available upon request). Thromboembolic complications included an embolic or thrombotic cerebrovascular accident, pulmonary embolism, deep vein thrombosis, or other systemic thromboembolic events. All events were verified and assessed for the association with AC therapy (where applicable) via manual medical chart review using a standardized abstraction form and a modified Naranjo scale. The Naranjo scale is a scoring system used to quantify the relationship between an adverse event and a given drug therapy [10]. All outcomes were reviewed by two adjudicators blinded to AC therapy status, disagreements regarding event categorization were resolved by a third adjudicator. Cause of death was verified through medical record and/or death certificate review.

Data analysis

Patient characteristics were reported as means, medians and standard deviations for interval-level variables and proportions for categorical variables. Colon bleeding rate was calculated by dividing the number of patients experiencing at least one bleeding episode by the total number of patients undergoing polypectomy. Rates of secondary outcomes were similarly calculated.

Conditional logistic regression with adjustment for the matching variable (age) and clustering of matched observations was used to assess differences in baseline characteristics and outcomes between groups. Wilcoxon rank-sum tests and Fisher’s exact tests were used to assess differences between patients with and without a primary bleeding outcome for interval-level and categorical variables, respectively.

Patient AC status, age, gender, colonoscopy indications, co-morbidities and polyp variables (size, total count, retrieval method, and morphology) were entered into an unconditional, stepwise multivariate logistic regression model to identify independent predictors of colon bleeding after polypectomy among all patients. To preserve power, only variables with a P < 0.30 were entered into the model and kept in the final model with a P < 0.10. Two-way interactions were not included in the model because the primary outcome occurred infrequently. Unconditional regression was used because the incidence of colon bleeding among the clustered, matched patients was low. Similar predictive modeling was performed among AC group patients only; but with the addition of a variable indicating the use of enoxaparin.

Results

A total of 614 potential AC group patients were identified. Of these, 83 with insufficient medical record information, 61 with colonoscopy to evaluate emergent GI bleeding, 39 with warfarin monitored outside of CPAS, 4 with non-continuous health plan enrollment, 1 with intestinal arteriovenous malformation and 1 for warfarin prophylaxis after orthopedic surgery were excluded, resulting in 425 eligible-for-matching AC patients. A potential pool of 9219 non-AC group patients was identified; however, 125 and 91 were excluded for non-continuous health plan enrollment and colonoscopy to evaluate emergent GI bleeding, respectively. Of the remaining 9003 eligible-for-matching non-AC patients, 800 were randomly matched on age (± 3 years) to 425 AC patients.

Baseline characteristics for the AC and Non-AC groups are summarized in Table 1. The groups were well-matched with regards to age. Patients in the AC group were more likely to be male, to have had their colonoscopy for non-emergent rectal bleeding follow up, to have co-morbid heart failure and prior venous thrombosis (all P < 0.05). Patients in the Non-AC group were more likely to have had colonoscopy for cancer screening (P = 0.001).

Table 1.   Baseline characteristics by anticoagulant use status at time of colonoscopy
Characteristic AC group (n = 425)Non-AC group (= 800)P-value*
  1. AC, anticoagulant therapy; SD, standard deviation. *Adjusted for clustering of matched observations and matching age variable. As of date of colonoscopy. During the 180 days prior to the colonoscopy.

Mean age (SD)69.5 (9.6)69.6 (9.4)0.484
Male (%)61.953.50.003
Primary indication for colonoscopy (%)
 Anemia7.85.40.064
 Cancer screening40.149.70.001
 Constipation0.00.60.998
 Diarrhea2.62.40.898
 Irritable bowel0.20.40.607
 Previous polypectomy follow-up27.126.00.624
 Non-emergent rectal bleeding follow-up18.213.60.043
 Other4.22.00.025
Antiplatelet use (%)
 Yes19.16.1< 0.001
 Unknown49.992.1< 0.001
Co-morbidity (%)
 Diabetes mellitus3.13.50.775
 Hypertension37.432.10.053
 Heart failure11.81.4< 0.001
 Prior venous thrombosis10.43.0< 0.001
 Any prior hemorrhage12.99.60.060
 Prior stroke0.00.01.000
 Cancer0.90.30.180

Overall, 13 out of 1225 patients (1.06%) experienced post-polypectomy colon bleeding. Colon bleeding was more common in the AC group (2.6% vs. 0.2%, P = 0.005) (Table 2). Case summaries of all primary outcome events are presented in Table 3. An average of 9 days (median = 7, range = 3–18) elapsed between polypectomy and the primary outcome. No bleeding complications occurred during the colonoscopy. Secondary outcomes (thrombosis, death and non-colon GI bleeding) were uncommon and differences between groups were not statistically significant (P > 0.05 for all) (Table 2). Details pertaining to the quantity and morphology of polyps removed in the two groups as well as the polyp retrieval method are summarized in Table 4. Patients in the AC group were more likely to have had polyps retrieved using snare cautery whereas patients in the non-AC group were more likely to have had cold snare retrieval (both P < 0.05).

Table 2.   Outcomes during the 30-day follow-up period by anticoagulant use status at time of colonoscopy
Outcome AC group (n = 425)Non-AC group (n = 800)P-value*
  1. AC, anticoagulant therapy; GI, gastrointestinal. *Adjusted for clustering of matched observations and age matching variable.

Colon bleeding (n, %)11, 2.62, 0.20.005
Death (n, %)1, 0.21, 0.10.624
Thrombosis (n, %)1, 0.21, 0.10.624
Non-colon GI bleed (n, %)1, 0.20, 0.00.347
Table 3.   Summary of all colon bleeding events
Pre-polypectomy INREvent INREnoxaparin bridge therapyASA doseAC indicationLOS (days)Event descriptionTransfusion (# of units)
  1. ASA, aspirin; AC, anticoagulation; LOS, length of stay; na, not applicable; UNK, unknown; Hgb, hemoglobin; BRBPR, bright red blood per rectum; ED, emergency department.

Non-AC group patients
 nanaNoneUNKNone6Developed venous thrombosis following polypectomy, had just started enoxaparin and warfarin (INR = 1.0) when presented 11 days after polypectomy with hemodynamically stable BRBPR, and ↓ Hgb ≥ 2 g dL−1, intact polypectomy sites without bleeding noted on exam3
 nanaNoneUNKNone1Presented 7 days after polypectomy with bloody diarrhea and ↓ Hgb ≥ 2 g dL−1, observed in ED overnight, two ulcers at polypectomy sites noted on exam and cauterized0
AC group patients
 1.73.0Therapeutic enoxaparin pre- and post-polypectomyUNKAtrial fibrillation2Daily rectal bleeding reported following polypectomy, presented 6 days after polypectomy with ↓ Hgb ≥ 2 g dL−1, bleeding resolved without intervention during hospital admission1
 1.5UNKNoneUNKAtrial fibrillation10Presented 3 days after polypectomy with hypotensive BRBPR and ↓ Hgb ≥ g dL−1, required hemicolectomy complicated by two episodes of recurrent bleeding18
 1.32.4None81 mgDialysis access thrombosis2Presented 18 days after polypectomy with maroon stools, lightheadedness and ↓ Hgb ≥ 2 g dL−1, small ulcer right colon without active bleeding noted on exam2
 1.13.1Therapeutic enoxaparin pre- and post-polypectomyUNKAtrial fibrillation2Presented 7 days after polypectomy with hypotension, exam revealed bleeding from a previously cauterized lesion, bleeding resolved following repeat cauterization2
 1.32.2NoneUNKAtrial flutter0Presented 6 days after polypectomy with BRBPR and ↓ Hgb ≥ 2 g dL−1; bleeding resolved without intervention during ED observation0
 1.42.54Therapeutic enoxparin pre- and post-polypectomyNoneMechanical mitral valve2Presented 10 days after polypectomy with bloody diarrhea, a bleeding ulceration at polypectomy site was noted on exam and clipped0
 1.92.47None325 mgCoronary artery disease9Presented 15 days after polypectomy with acute worsening of chronic blood in stools, lightheadedness and ↓ Hgb ≥ 2 g dL−1, bleeding polypectomy ulcers and an antral ulcer were noted on exam and cauterized3
 1.41.9None81 mgAtrial fibrillation6Presented 10 days after polypectomy with rectal bleeding and ↓ Hgb ≥ 2 g dL−1, a cecal ulcer was noted on exam and injected and clipped, subsequently required repeat colonoscopy (× 3) due to persistent bleeding> 10
 UNK1.4None325 mgCoronary artery disease1Presented 4 days after polypectomy with rectal bleeding and ↓ Hgb ≥ 2 g dL−1, bleeding at polypectomy site noted on exam and injected, resection of cecum required 2 weeks later due to large persistent pneumoperitoneum5
 UNK1.73NoneUNKPulmonary embolism1Presented 13 days after polypectomy with abdominal pain and BRBPR and ↓ Hgb ≥ 2 g dL−1, bleeding resolved following discontinuation of warfarin0
 1.32.51Therapeutic enoxaparin pre- and post-polypectomyUNKTransient ischemic attack4Presented 7 days after polypectomy with hypotension, dizziness and syncope and ↓ Hgb ≥ 2, became non-responsive with respiratory arrest requiring resuscitation, no active bleeding from polypectomy sites noted on exam9
Table 4.   Polyp characteristics by anticoagulant use status at time of colonoscopy
CharacteristicAC group (n = 425)Non-AC group (n = 800) P-value*
  1. AC, anticoagulant therapy. *Adjusted for clustering of matched observations and age matching variable.

Largest polyp > 10 mm (% yes)21.717.90.100
Mean total count of polyps removed (median, SD)2.6 (2.0, 2.3)2.4 (2.0, 2.4)0.103
Gross polyp morphology (%)
 Diminutive5.56.60.297
 Flat0.20.60.403
 Hyperplastic1.01.40.472
 Laterally spreading0.00.10.999
 Mixed polyp types12.213.30.543
 Pedunculated9.85.60.008
 Sessile48.548.00.973
 Not stated22.924.30.535
Polyp-retrieval method (%)
 Cold biopsy forceps8.16.90.483
 Cold snare2.421.2< 0.001
 Combination of methods17.113.00.045
 Snare cautery57.843.4< 0.001
 Other2.83.30.609
 Not stated11.912.30.854

Patients hospitalized with post-polypectomy colon bleeding were more likely to have been receiving AC (P < 0.001), to be male (P = 0.005) and to have had a higher number of polyps removed (P < 0.001) (Table 5). Predictive modeling indicated that independent predictors of post-polypectomy colon bleeding included being in the AC group [adjusted odds ratio (AOR) = 11.6; 95% confidence interval (CI) = 2.3–57.3], number of polyps removed [AOR = 1.2 (i.e. 20% increased likelihood of bleeding for each additional polyp removed); 95% CI = 1.1–1.4] and male gender (AOR = 9.2, 95% CI = 1.1–74.9) (model fit c-statistic = 0.89). Age was included in the final model iteration but was not a significant predictor (P = 0.13). Among patients who received colonoscopy for colon cancer screening, the rates of colon bleeding were 1.76% and 0.50% in the AC and non-AC groups, respectively.

Table 5.   Potential colon bleeding predictors by colon bleeding status
CharacteristicBleeding (n = 13)No bleeding (n = 1212) P-value
  1. AC, anticoagulant therapy. *As of date of colonoscopy. During the 180 days prior to the colonoscopy.

AC group status (%)84.634.2< 0.001
Mean age* (SD)72.8 (6.4)69.5 (9.5)0.208
Male (%)92.356.00.009
Primary indication for colonoscopy (%)
 Non-emergent rectal bleeding15.415.21.000
 Colon cancer screening38.546.40.781
Mean total count of polyps removed (median, SD)5.2 (5.0, 2.4)2.4 (2.0, 2.4)< 0.001
Risk factors (%)
 Hypertension38.533.90.772
 Heart failure15.44.90.134
 Prior venous thrombosis7.75.50.552
 Prior hemorrhage7.710.80.719
Pedunculated polyp (%)0.07.10.319
Largest polyp > 10 mm (% yes)38.518.90.085
Polyp-retrieval method (%)
 Cold snare15.414.70.943
 Combination30.814.20.103
 Snare cautery38.548.30.582

For patients in the AC group only, those experiencing post-polypectomy colon bleeding were more likely to have had a primary indication for AC of coronary artery disease compared with AC patients without bleeding (16.7% vs. 3.4%, P = 0.034). No other differences in AC group baseline characteristics were detected (all P > 0.05; data not shown). Patients in the AC group who experienced colon bleeding also had a higher number of polyps removed (P < 0.001) (Table 6).

Table 6.   Anticoagulant patient characteristics overall and by primary bleeding status
CharacteristicOverall (n = 425)Bleeding (n = 11)No bleeding (n = 414)
  1. SD, standard deviation; INR, international normalized ratio. *Therapeutic dose enoxaparin (n = 1), other enoxaparin dose (n = 3). Prophylactic dose enoxaparin (n = 14), therapeutic dose enoxaparin (n = 10), other enoxaparin dose (n = 4). Therapeutic dose enoxaparin before and after (n = 56), prophylactic dose enoxaparin before and after (n = 6), therapeutic dose enoxaparin before prophylactic dose enoxaparin after (n = 2); intravenous unfractionated heparin before and after, other enoxaparin dose before and after, and prophylactic dose enoxaparin before therapeutic dose enoxaparin after (n = 1 each). §< 0.001.

Mean warfarin doses held prior to colonoscopy (n, median, SD)3.6 (404, 4.0, 1.0)3.6 (9, 4.0, 1.4)3.6 (395, 4.0, 1.0)
Bridging therapy (%)
 Prior to but not after colonoscopy*0.90.01.0
 After but not prior to colonoscopy6.40.06.5
 Prior to and after colonoscopy15.836.415.2
 None76.963.677.3
Mean INR prior to colonoscopy (n, median, SD)1.3 (388, 1.2, 0.2)1.4 (9, 1.4, 0.2)1.3 (379, 1.2, 0.2)
Resumed warfarin with a loading dose after colonoscopy (%)72.354.672.8
Mean days to re-starting warfarin after colonoscopy (n, median, SD)0.5 (412, 0, 1.5)0.5 (11, 0, 1.0)0.5 (400, 0, 1.5)
Mean total count of polyps removed (median, SD)2.6 (2.0, 2.3)5.4 (5.0, 2.4)§2.5 (2.0, 2.3)
Largest polyp > 10 mm (% yes)21.727.321.5
Cold snare polyp-retrieval (%)2.49.12.2
Mean days to enoxaparin after colonoscopy (n, median, SD)1.1 (89, 1.0, 0.8)1.0 (4, 1.0, 0.0)1.1 (85, 1.0, 0.8)

For most AC group patients, warfarin was stopped 4 days prior to and resumed on the evening of colonoscopy. Differences in the use of bridging therapy between AC group patients who did and did not experience colon bleeding were not statistically significant. Among AC group patients, 5.6% (n = 4) of those who received bridging therapy after polypectomy experienced colon bleeding (all received therapeutic enoxaparin before and after the procedure) compared with 2.0% (n = 7) of those who did not receive bridging (Table 3). The most common bridging strategy was therapeutic enoxaparin before and after the procedure (56.6% of all bridging therapies). Most patients did not receive bridging therapy (Table 6). The mean INR on the day prior to colonoscopy was 1.4 ± 1.4 and 1.3 ± 1.2 in AC group patients with and without colon bleeding, respectively (P > 0.05).

For patients in the AC group, the number of polyps removed was the sole independent predictor of colon bleeding (AOR = 1.3; 95% CI = 1.1–1.6; model fit c-statistic = 0.86). The cold snare retrieval method (P = 0.070) and male gender (P = 0.088) were included in the final multivariate analysis but were not independently associated with bleeding. The mean INR at the time of post-polypectomy bleeding for AC patients was 2.3 (median = 2.4, range = 1.4–3.1).

Discussion

In our cohort of 1225 patients undergoing polypectomy during routine colonoscopy (425 of which interrupted AC therapy for the procedure), we detected an overall incidence of subsequent colon bleeding episodes related to the procedure of 11 per 1000 polypectomies. However, the majority of these complications (85%) were observed in patients receiving warfarin (26 per 1000 AC patients undergoing polypectomy). Our observations suggest that patients interrupting warfarin are at a substantially increased risk (elevenfold higher) of colon bleeding after polypectomy compared with patients not prescribed AC. This confirms observations from the aforementioned studies involving limited AC populations that also reported high rates of post-polypectomy colon bleeding in AC patients [5,6].

In addition to AC use, we identified other factors that independently predict colon bleeding in patients undergoing polypectomy including male gender and the number of polyps retrieved. Greater numbers of polyps may be more common in men and, therefore, compared with women, more polyps are likely removed during colonoscopy [2]. We found the count of polyps was a predictor independent of patient gender. We were unable to address if there is an interaction between gender and count of polyps in our model as a result of the low incidence of bleeding in our cohorts. Nevertheless, we did find that men had a higher mean count of polyps removed than women (2.6 vs. 2.2, P < 0.001). Although data regarding polyp location were collected, they were not included in regression modeling because many patients had polyps removed from multiple locations and we were unable to link the removal site to the bleeding site. Among AC group patients, only the number of polyps removed independently predicted subsequent colon-bleeding, suggesting that the likelihood of post-polypectomy colon bleeding increases by 30% for each polyp removed.

To date, recommendations on the management of patients receiving AC during the period surrounding colonoscopy are limited [4]. Recommendations regarding the need for bridging therapy during warfarin interruption vary for polypectomy (a high-bleeding risk procedure) based on the underlying risk of thromboembolism [11]. The standardized approach employed by KPCO CPAS pharmacists reserves bridging therapy for only those patients at highest thromboembolic risk, such as those with mechanical heart valves or atrial fibrillation with previous thromboembolic events [4,11]. For AC patients with post-polypectomy colon bleeding in our study, a median of 4 (SD = 1.4) doses of warfarin were withheld prior to colonoscopy resulting in a median pre-procdedure INR of 1.4 (SD = 0.2). Warfarin was resumed on the evening of colonoscopy for the vast majority of AC patients; loading doses of warfarin were used commonly when therapy was resumed. However, AC patients who did and did not experience post-polypectomy colon bleeding were managed similarly (Table 6). For patients who received bridging therapy, the last dose of enoxaparin was administered approximately 24 h prior to colonoscopy and resumed, unless otherwise directed by the endoscopist, the morning after the procedure. Both once and twice daily doses of enoxaparin were used. The small number of bleeding events precludes any conclusion about what (if any) increase in the risk of colon bleeding might be attributed to enoxaparin use for bridging therapy.

Although bleeding was the primary focus of this study, preventing thromboembolic complications during AC therapy interruption was also assessed. Thrombotic events were uncommon and not significantly different between AC and non-AC groups. Althought this observation suggests the strategies used to interrupt AC therapy in this study minimized the likelihood of thromboembolic complications, the very small number of events precludes definitive conclusions regarding this outcome.

Our conclusions are potentially confounded by the lack of information regarding aspirin use, particularly in the non-AC group. We performed two post-hoc regression analyses with aspirin use added to the model under the assumption that all patients with unknown aspirin use status were either (i) not taking aspirin or (ii) were all taking aspirin. Neither of these assumptions altered the results of our origninal analysis to an appreciable extent. At KPCO, discontinuation of aspirin or non-steroidal anti-inflammatory drugs (NSAIDs) prior to colonoscopy is managed on a case-by-case basis. It is possible that antiplatelet agents were used more commonly by patients in the AC group and that some of the excess bleeding in that cohort could be attributable to antiplatelet therapy (rather than warfarin). That possibility notwithstanding, several studies have reported that the concurrent use of aspirin or NSAIDs does not increase the risk of post-polypectomy bleeding [5,6,12,13]. Current American Society for Gastrointestinal Endoscopy guidelines recommend that, in the absence of a pre-existing bleeding disorder, endoscopic procedures may be performed on patients taking aspirin and NSAIDs in standard doses [4]. However, no information on the concommitant use of aspirin, NSAIDs and/or AC has been reported; nevertheless, it is plausible that interactions between these agents may increase bleeding.

The small number of primary outcome events limits our ability to discern predictors of post-polypectomy colon bleeding. Nevertheless, to our knowledge, this is by far the largest evaluation of post-polypectomy bleeding in AC patients. In addition, we included a comprehensive set of predictive factors to evaluate. Finally, as KPCO employs a centralized anticoagulation management program, processes may have been in place that decreased the risk of bleeding. Other healthcare systems without anticoagulation management service may experience different rates of post-polypectomy colon bleeding among their AC patients.

Our results are likely to be valid. We examined real-world patients with a variety of indications for warfarin and colonoscopy. The dataset used to complete this study is robust and has been used previously in health records and data extraction research [8,14]. The large number of patients included in our analysis increases the generalizability of our results and reduces the likelihood that unmeasured bias may have influenced them. Clinical events were comprehensively collected and described and all clinical events were independently assessed for causality by two expert reviewers.

Conclusions

Exposure to AC therapy around the time of colonoscopy with polypectomy increased the likelihood of post-polypectomy colon bleeding by approximately elevenfold. As no bleeding complications occurred during colonoscopy, it appears that a strategy of witholding warfarin to achieve an INR below 1.5 and giving any pre-procedure bridging doses of enoxaparin at least 24 h prior to colonoscopy is sufficient to allow safe polyp retrieval. However, prompt resumption of warfarin and/or enoxaparin therapy after colonoscopy confers an increased likelihood of colon bleeding in the 30 days post procedure. As each polyp retrieved increased bleeding likelihood by about 20%, resumption of AC therapy could be delayed selectively when removal of a large number of polyps is required, especially in men. We are unable to determine if delaying the resumption of AC therapy in this manner would attenuate bleeding risk or would unacceptably increase the risk of thromboembolism. Local strategies to control bleeding such as endoscopic clip application immediately after polypectomy might also reduce bleeding risk in patients requiring resumption of AC therapy. These strategies should be explored in future research.

Addendum

Contribution: D.M. Witt, T. Delate, K.H. McCool, M.B. Dowd, and N.P. Clark designed the research and extracted information from medical records; T. Delate performed the statistical analysis; D.M. Witt and F. Dentali wrote the initial draft of the manuscript; D.M. Witt, T. Delate, K.H. McCool, M.B. Dowd, N.P. Clark, M.A. Crowther, D.A. Garcia, W. Ageno, F. Dentali, E.M. Hylek and W.G. Rector interpreted the analysis and revised the manuscript.

Acknowledgements

The authors would like to acknowledge M. L. Smith, JD, PharmD, BCPS, M. Kershner and D. Winn for their contributions to the data collection. D.M. Witt had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. This study was funded in its entirety by Kaiser Permanente Colorado.

Disclosure of Conflict of Interests

The authors state that they have no conflict of interest.

Ancillary