Risk of postpolypectomy bleeding in patients taking direct oral anticoagulants or clopidogrel

The usage of direct oral anticoagulants (DOACs) to prevent and treat thromboembolic events is gradually increasing. We aimed to evaluate the outcomes of patients taking DOACs after polypectomy. We retrospectively reviewed 131 patients taking DOACs and 270 taking clopidogrel who underwent polypectomy between November 2010 and December 2017. The risk of delayed postpolypectomy bleeding (PPB) was evaluated and compared. A total of 989 polyps were removed (320 polyps in the DOAC and 669 polyps in the clopidogrel group). DOACs and clopidogrel were discontinued for 2.8 ± 1.7 days and 5.8 ± 2.5 days before polypectomy, respectively. DOACs and clopidogrel were restarted on 1.6 ± 2.9 days and 1.7 ± 1.1 days after polypectomy, respectively. According to per polyp analysis, delayed PPB rate was 1.6% in both groups (p = 0.924). Logistic regression analysis was performed after propensity score matching and revealed that DOACs did not increase the delayed PPB risk compared to clopidogrel (OR 0.929, 95% CI 0.436–1.975, p = 0.847). With the majority following the antithrombotic discontinuation guidelines, the incidence of delayed PPB was 3.1% in the patients taking DOACs. The delayed PPB risk was not greater in those taking DOACs than in those taking clopidogrel.

Colonoscopic polypectomy is effective in preventing colorectal cancer (CRC) and reducing CRC-related mortality 1 but carries a risk of procedure-related adverse events. Hemorrhage is one of the significant complications related to polypectomy. Patients taking antithrombotic medication are exposed to a greater risk of delayed postpolypectomy bleeding (PPB) than those not taking antithrombotic agents 2,3 . Current guidelines suggest discontinuation of antiplatelets and anticoagulants according to the compromise between the bleeding and thromboembolic risk considering endoscopic procedures, category of antithrombotic agents, and the medical condition requiring anticoagulation 4,5 . For example, clopidogrel increases the risk of PPB 3,6 , but aspirin does not 3 . In addition, discontinuing both clopidogrel and aspirin in patients on dual antiplatelet therapy for drug-eluting coronary stents increases the risk of early-onset stent thrombosis compared with stopping clopidogrel only 7 . The current guidelines recommend maintaining aspirin and stop clopidogrel 5-7 days before polypectomy 8 . Therefore, understanding the risk of bleeding with each drug is essential to determine the antithrombotic agent cessation strategy in relation to polypectomies.
Compared with warfarin, a representative vitamin K antagonist, direct oral anticoagulants (DOACs) take advantage of a rapid onset, little drug interaction, and no need for regular monitoring of anticoagulation effect. Thus, DOACs are increasingly selected for patients with nonvalvular atrial fibrillation (AF), venous thromboembolism, and other hypercoagulable conditions [9][10][11] . According to meta-analyses, the major gastrointestinal (GI) bleeding risk is lower in patients with DOACs than in those with vitamin K antagonists [12][13][14] . However, there is still a lack of evidence regarding the risk of PPB in those taking DOACs. Interestingly, a recent large-scale retrospective study about postpolypectomy adverse events suggested that not only warfarin but also clopidogrel carried a higher risk of PPB than DOACs 15 . Because clopidogrel has been used for a long time, it is familiar to the clinician and the associated risk of PPB has been well-investigated through previous studies. Therefore, the risk of PPB in those taking clopidogrel would be a good reference to assess the risk of PPB related to DOACs. Here, we aimed to investigate the adverse events related to polypectomy in patients taking DOACs or clopidogrel and to assess the risk factors of delayed PPB. Case-matched analysis. We conducted a propensity score matching to lower any effects of selection bias.
Two hundred and seventy-six of the 320 polyps in the DOAC group were matched with the same number of polyps in the clopidogrel group as a control, based on age, sex, comorbidities (respiratory disease, diabetes mellitus, hypertension, and cardiovascular disease), polypectomy techniques (cold snare polypectomy vs. others), polyp size, location (right versus left colon), histopathology (adenoma vs. others), compliance with the antithrombotic discontinuation guideline, and prophylactic hemostasis (Table 3). Then, a logistic regression analysis was conducted, which revealed that DOAC did not increase the risk of delayed PPB in comparison with clopidogrel (OR 0.929, 95% CI 0.436 -1.975, p = 0.847).

Discussion
In the current study, we evaluated the risk of delayed PPB and clinical outcome in patients taking DOACs or clopidogrel. Compliance rates to the antithrombotic discontinuation guidelines for polypectomy and delayed PPB rates were 85.5% and 3.1%, respectively, in the DOAC group and 80.0% and 3.0%, respectively, in the clopidogrel group. The propensity score-matched case-control analysis revealed no difference in the delayed PPB rate per polypectomy site between the two groups (1.6% in both groups). Meanwhile, a recent large retrospective study analyzed the risk of GI bleeding within 30 days after polypectomy in patients taking DOAC, clopidogrel, warfarin, or no antithrombotic agent 15 16 . According to their predominantly colonic procedure data, the delayed bleeding rate was not significantly different between patients taking warfarin or DOAC (13.8% vs. 9.5%, respectively), but endoscopic hemostasis was more frequently required in those taking warfarin than DOAC (12.0 vs. 6.9%, p < 0.05). Therefore, taken together with the results of our study, the risk of delayed PPB between patients taking DOACs and those taking clopidogrel is deemed at least equivalent. Unlike DOACs, the incidence of delayed PPB in patients taking clopidogrel is relatively well-known and ranges from 0.8 to 4% in previous reports 3,17,18 . In a meta-analysis including five observational studies, the pooled relative risk of delayed PPB was 4.66 (95% CI 2.37-9.17) if clopidogrel was not discontinued before polypectomy 6 . A randomized controlled trial (RCT) compared the PPB rate of those who continued thienopyridine (mostly clopidogrel) with that of controls (thienopyridine non-users). In that study, delayed PPB occurred in 2.4% of those who continued thienopyridine but none of the controls. In contrast, a recent double-blind RCT evaluated the effect of 7 days of clopidogrel interruption before polypectomy, which showed no significant difference in the delayed PPB rate between the uninterrupted clopidogrel (3.8%) and the placebo (3.6%) groups 19 . In the current study, delayed PPB occurred in 3.0% of patients in the clopidogrel group, comparable with the studies mentioned above.
We evaluated the potential risk factors of delayed PPB per polypectomy site and only immediate PPB was positively associated with the risk of delayed PPB. Therefore, if immediate PPB occurs in patients taking either DOAC or clopidogrel, these patients should be carefully monitored. A previous study suggested that EMR was an independent risk factor of delayed PPB 15 . However, no association of the polypectomy methods with the risk of delayed PPB was observed in our study, which may be due to the relatively small average polyp size in this study. In our study population, the DOACs included four major medicines: apixaban, dabigatran, edoxaban, and rivaroxaban. The pharmacological properties are different between each drug, and as such, the delayed PPB risk may be different among the various DOACs and may be influenced by the clinical setting. In a large populationbased retrospective study regarding any kind of GI bleeding in patients taking DOACs, apixaban showed a relatively lower risk of GI bleeding than did dabigatran or rivaroxaban 20 . On the other hand, apixaban showed a higher risk of bleeding than rivaroxaban according to a Japanese multicenter retrospective study focused on GI bleeding after therapeutic endoscopy for patients taking DOACs 16 . Our data did not show any difference in delayed PPB according to the type of DOACs, which may be due to our limited number of cases.
This study has several limitations. Because of its retrospective design, the DOAC and clopidogrel groups differed with respect to the mean age of patients and the proportion of prophylactic hemostasis. One of the Korean reimbursement criteria for DOAC is nonvalvular AF in patients ≥ 75 years or with a history of a thromboembolic event, which may explain why the DOAC group was older than the clopidogrel group. Prophylactic hemostasis was also more often performed in the DOAC group than in the clopidogrel group, which may be because the endoscopists were less familiar with DOAC compared with clopidogrel. We used propensity score matching to perform a comparison between the DOAC and clopidogrel groups to overcome such differences. Meanwhile, our database includes a relatively small number of patients taking DOACs. Therefore, risk analysis was possible after propensity score matching for polyps in each group; however, risk analysis would not be appropriate after propensity score matching for patients in each group. Moreover, we could not examine the differences between the various DOACs in this study. Last, we did not compare the risk of delayed PPB in the antithrombotic-naive group with the DOAC or clopidogrel groups because a significant proportion of the retrospective data were not as qualified as the data available for the DOACs and clopidogrel groups. Therefore, prospective cohort studies should be followed to compare the risk of delayed PPB in patients taking antiplatelets, warfarin, and DOACs with those not taking such medicines. In order to provide critical evidence on the risk of PPB in those who taking DOACs further studies should be performed in the future, such as a RCT comparison between 2-day versus 1-day cessation of DOACs before polypectomy.
In conclusion, according to our retrospective, single-center study, the incidence of delayed PPB was 3.1% in patients taking DOACs, with the majority of patients following the antithrombotic discontinuation guidelines. The delayed PPB risk was not increased in patients taking DOACs in comparison with patients taking clopidogrel. All patients provided informed consent for the colonoscopic procedures. Patients taking DOAC or clopidogrel that was prescribed and/or monitored at the outpatient clinic of the same institution were included in this study. Patients who did not take either DOAC or clopidogrel in the 2 weeks before and after polypectomy were also excluded (Fig. 1). The clopidogrel group was included as a reference group. At 2-4 weeks after colonoscopic polypectomy, patients visited the outpatient clinic to check for adverse events and receive their pathologic results. During the consultation, a physician checked whether delayed PPB had occurred. We reviewed medical records for age, sex, medication before and after endoscopy, and comorbidities.

Methods
Ethical approval. The study protocol followed the ethical guidelines of the Declaration of Helsinki Principles. The institutional review board of the Asan Medical Center approved this study (No. 2018-0184).

Medication.
We reviewed the data on medication for the use of clopidogrel and DOACs within 1 month preceding the colonoscopy. Data regarding the time intervals between drug cessation and the endoscopic procedure and drug resumption were retrieved. The DOAC group included patients taking apixaban, dabigatran, edoxaban, or rivaroxaban. The antithrombotic discontinuation guideline for polypectomy was considered to have been complied with if clopidogrel was interrupted 5 days before the procedure and DOAC 2 days before the procedure 4,5 .
Endoscopic procedures. All procedures were performed using a single-channel colonoscope (CF-Q260AL/I, CF-H260AL/I, CF-HQ290AL/I; Olympus Medical, Tokyo, Japan). The endoscopic resection technique was determined by the endoscopists depending on the size and shape. Cold snare polypectomy was performed for lesions < 10 mm for which histology was estimated as benign. EMR was performed for nonpedunculated lesions between 10 to 20 mm in size or pedunculated lesions ≥ 20 mm. A 1:100,000 mixture of epinephrine saline solution was injected submucosally. After snaring the lesion, Endocut Q current (effect 2 or 3, cut duration 2, and cut interval 6) was applied via an electrosurgical unit (VIO300D; ERBE, Tubingen, Germany). ESD was performed according to our institutional protocol as described in previous studies [21][22][23] .
Outcome measures and variables. We defined delayed PPB as the hematochezia that appeared after completion of the polypectomy, which required therapeutic interventions such as endoscopic hemostasis, angiography, and/or surgery. Immediate PPB was defined as intraprocedural hemorrhage that needed endoscopic hemostasis. In addition to the patients' age, sex, and comorbidities, data regarding the time intervals between drug cessation, the endoscopic procedure, and drug resumption were retrieved. Characteristics of polyps and details of procedural outcomes were prospectively recorded in the structured histology reports and colonoscopy reports. Factors related to polyps and procedures were retrieved from the histology and colonoscopy reports. www.nature.com/scientificreports/ The location, size, and the final histology of the polypectomy specimen were investigated. The location of the polyp was classified to those on the right (from transverse colon to cecum) or the left colon (from rectum to splenic flexure). The procedure type, the occurrence of immediate PPB, and prophylactic hemostasis were also included as factors related to procedures. The methods and outcomes of hemostatic procedures were assessed in cases that had delayed PPB.
Statistical analysis. The Student's t-test and Pearson's chi-square test were used to assess the differences between the two groups. Logistic regression analysis was performed to assess the risk factors for delayed PPB, including DOAC or clopidogrel use. Propensity scores were matched to minimize the potential bias due to unbalanced characteristics between the two groups and to improve causality between treatment and "delayed PPB" and "medication. " In calculating the propensity score, we considered factors that could influence outcomes including age, sex, drug discontinuation interval before polypectomy, execution of prophylactic hemostasis, location and size of the polyp, resection methods, and pathology of the resected specimen. The propensity score was calculated by the logistic model and matched between the two groups within a caliper of 0.07. For propensity score matching, AF and coronary artery disease were combined as cardiovascular disease. Resection technique was classified as cold snare polypectomy, and other techniques and histopathology findings were classified as adenoma and other findings, respectively. We also evaluated the balance between two groups using the standardized mean difference and Hosmer-Lemeshow test, and C-statistic was conducted to check the model fit. A p-value of < 0.05 was interpreted as statistically significant. All statistical analyses were performed using SAS software (version 9.4 m6 Cary, NC, USA).

Data availability
All data generated or analyzed during this study are included in this published article.