bs_bs_banner
doi:10.1111/jgh.13830
GASTROENTEROLOGY
New predictive model for acute gastrointestinal bleeding in patients taking oral anticoagulants: A cohort study Akira Shimomura,* Naoyoshi Nagata,* Takuro Shimbo,† Toshiyuki Sakurai,* Shiori Moriyasu,* Hidetaka Okubo,* Kazuhiro Watanabe,* Chizu Yokoi,* Junichi Akiyama* and Naomi Uemura‡ *Department of Gastroenterology and Hepatology, National Center for Global Health and Medicine, Tokyo, †Clinical Research and Informatics, Ohta Nishinouchi Hospital, Koriyama, and ‡Department of Gastroenterology and Hepatology, National Center for Global Health and Medicine, Kohnodai Hospital, Chiba, Japan
Key words acute gastrointestinal hemorrhage, direct oral anticoagulants (DOACs), proton-pump inhibitors (PPIs), warfarin. Accepted for publication 16 May 2017. Correspondence Dr Naoyoshi Nagata, Department of Gastroenterology and Hepatology, National Center for Global Health and Medicine (NCGM), 1-21-1 Toyama, Shinjuku-ku, Tokyo 162-8655, Japan. Email:
[email protected] Declaration of conflict of interest: The authors declare no conflicts of interest. The funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. Financial support: This study was supported in part by a grant from the National Center for Global Health and Medicine (26A-201 and 29-2001). Author contribution: Shimomura A. and Nagata N. designed the study and main authors; Nagata N. equally first author, Sakurai T. and Moriyasu S. collected clinical information and were the main authors of the manuscript; Okubo H., Watanabe K., Yokoi C., and Akiyama J. performed endoscopy. Shimbo T. helped with statistical analysis; Uemura N. edited the manuscript.
Abstract Background and Aim: The study developed a predictive model of long-term gastrointestinal (GI) bleeding risk in patients receiving oral anticoagulants and compared it with the HAS-BLED (Hypertension, Abnormal renal/liver function, Stroke, Bleeding history or predisposition, Labile international normalized ratios, Elderly, Drugs/alcohol concomitantly) score. Methods: The study periodically followed a cohort of 508 patients taking oral anticoagulants (66 direct oral anticoagulants users and 442 warfarin users). Absence of GI bleeding at an initial examination and any subsequent GI bleeding were confirmed endoscopically. The bleeding model was developed by multivariate survival analysis and evaluated by Harrell’s c-index. Results: During a median follow-up of 31.4 months, 42 GI bleeds (8.3%) occurred: 42.8% in the upper GI tract, 50.0% in the lower GI tract, and 7.1% in the middle GI tract. The cumulative 5 and 10-year probability of GI bleeding was 12.6% and 18.5%, respectively. Patients who bled had a significantly higher cumulative incidence of all-cause mortality (hazard ratio 2.9, P < 0.001). Multivariate analysis revealed that absence of proton pump inhibitor therapy, chronic kidney disease, chronic obstructive pulmonary disease, history of peptic ulcer disease, and liver cirrhosis predicted GI bleeding. The c-statistic for the new predictive model using these five factors was 0.65 (P < 0.001), higher than the HAS-BLED score of 0.57 (P = 0.145). Conclusions: Gastrointestinal bleeding increased the risk of subsequent mortality during follow-up of anticoagulated patients, highlighting the importance of prevention. The study developed a new scoring model for acute GI bleeding risk based on five factors (no-proton pump inhibitor use, chronic kidney disease, chronic obstructive pulmonary disease, history of peptic ulcer disease, and liver cirrhosis), which was superior to the HAS-BLED score.
Introduction Oral anticoagulants are used to prevent stroke and systemic embolism and are increasingly prescribed worldwide, mostly because of the increasing age of the population.1,2 Nevertheless, anticoagulation can cause severe side effects, of which bleeding is the most common.3 The incidence of major hemorrhage (including intracranial, gastrointestinal [GI], genito-urinary, and respiratory bleeding) is reportedly 1–3% per person-years in patients taking warfarin4; the proportion of patients diagnosed
164
with acute GI bleeding who take warfarin is 8–15% and 7% for upper-GI and lower-GI bleeding, respectively.5,6 Warfarin-related GI bleeding events are associated with long duration of hospitalization, substantial resource utilization, and a 30-day mortality of up to 15%.7,8 It is therefore important to understand the influence of oral anticoagulation on GI bleeding episodes and identify the associated risk factors; however, the risk factors for GI bleeding in patients taking oral anticoagulation have not been explored in long-term cohort study.
Journal of Gastroenterology and Hepatology 33 (2018) 164–171 © 2017 Journal of Gastroenterology and Hepatology Foundation and John Wiley & Sons Australia, Ltd
A Shimomura et al.
There are several bleeding risk scoring systems, including the HASBLED (Hypertension, Abnormal renal/liver function, Stroke, Bleeding history or predisposition, Labile international normalized ratios, Elderly, Drugs/alcohol concomitantly) score,9 the anticoagulation and risk factors in atrial fibrillation (ATRIA) score,10 and the outcomes registry for better informed treatment of atrial fibrillation (ORBIT) score.11 The main outcomes of these scoring systems are composite bleeding events, including intracranial, intraspinal, intraocular, retroperitoneal, intra-articular, pericardial, and intramuscular (with compartment syndrome) hemorrhage, as well as GI bleeding. The severity of a bleeding event may vary depending on the organ or organs affected, but GI bleeding often requires blood transfusion, long hospitalization, and intensive care, and there is a high risk of recurrence.12,13 There is no risk scoring system that focuses specifically on GI bleeding in patients taking oral anticoagulants. We undertook long-term follow-up of a cohort of patients taking oral anticoagulants who had no endoscopic evidence of GI bleeding at an index endoscopic examination and recorded details of any subsequent GI bleeding events. Our aims were to determine the cumulative incidence of GI bleeding and mortality in patients taking oral anticoagulants, to develop a risk model for GI bleeding, and to compare our scoring system with the HAS-BLED score.
Methods Study design, setting, and participants. We performed a retrospective cohort study of Japanese adults taking systemic anticoagulation treated at the National Center for Global Health and Medicine, Tokyo, Japan, from January 2001 to October 2015. Our hospital is one of the largest emergency hospitals in the Tokyo metropolitan area. All patients were confirmed not to have GI bleeding or a GI-bleeding-related lesion by endoscopy at an initial index examination. Data were collected from the prospectively recorded electronic endoscopy database (Solemio Endo; Olympus Medical Systems, Tokyo, Japan),
Gut bleeding model on anticoagulants
supplemented by review of our institution’s electronic medical records (MegaOak online imaging system; NEC, Tokyo, Japan).14 A study flow chart is shown in Figure 1. First, we searched the electronic database using the keywords “heparin,” “warfarin,” “anticoagulant,” “dabigatran,” “rivaroxaban,” “edoxaban,” and “apixaban.” The search identified 564 consecutive patients taking a systemic anticoagulant who had undergone endoscopy. We reviewed all endoscopic findings, clinical findings, and laboratory data for each patient. We systematically excluded patients with (i) GI bleeding or endoscopically proven GI bleeding at the initial endoscopy examination, (ii) < 30 days follow-up, or (iii) insufficient data for subsequent analysis. Finally, a cohort of 508 patients treated with an oral anticoagulant but with no evidence of GI bleeding was identified. The progress of this cohort was examined periodically. Conduct of this study was approved by the institutional review board; the requirement for patient consent was waived because of its retrospective nature.
Drugs and comorbidities. We reviewed and collected data for each patient from the electronic database (MegaOak), which is a searchable collection of records into which physicians or nurses immediately input clinical information, which included the following: endoscopic findings, clinical findings, radiologic findings, comorbidities, drugs, the results of laboratory investigations, GI bleeding events, and death. Drugs recorded included anticoagulants, low-dose aspirin, other antiplatelet drugs, non-steroidal anti-inflammatory drugs, and proton pump inhibitors (PPIs). Anticoagulants included warfarin, dabigatran, rivaroxaban, edoxaban, and apixaban. We defined the use of drugs as oral administration starting at least 6 months before the date of index endoscopy. The following comorbidities were recorded: diabetes mellitus (defined as the use of a diabetes drug), hypertension (defined as the use of any antihypertensive drug), dyslipidemia (defined as the use of any anti-hyperlipidemic drug), cerebro-cardiovascular
Figure 1 Study flow chart. (*) Endoscopically proven gastrointestinal bleeding includes erosion, ulcer, malignancy, or other lesion with blood. (**) We included one patient who died 16 days after initial endoscopy. GI, gastrointestinal.
Journal of Gastroenterology and Hepatology 33 (2018) 164–171 © 2017 Journal of Gastroenterology and Hepatology Foundation and John Wiley & Sons Australia, Ltd
165
A Shimomura et al.
Gut bleeding model on anticoagulants
disease history (including a history of cerebral infarction, myocardial infarction, or angina pectoris), cerebral hemorrhage history, peripheral arterial disease history, atrial fibrillation, cardiac valve replacement or valvuloplasty history due to valvular disease, hemiplegia, chronic kidney disease (defined as a serum creatinine concentration > 2.3 mg/dl), chronic obstructive pulmonary disease (COPD), connective tissue disease, peptic ulcer disease history, and liver cirrhosis. The HAS-BLED9 and CHA2DS2-VASc (Congestive heart failure, Hypertension, Age ≥ 75, Diabetes mellitus, Stroke, Vascular disease, Sex female) scores15 were calculated for the entire cohort as previously described. Labile prothrombin time international normalized ratio was defined as 3.5 or greater,16 and abnormal liver function was defined as patients with Child–Pugh grade B or C disease.
Clinical outcomes and follow-up. The main outcomes were overt acute GI bleeding (hematochezia, passage of fresh blood per rectum, melena, tarry stool, or hematemesis) and death after the index examination (initial endoscopy date). To identify the bleeding source, we first tried to perform endoscopy. When a source of bleeding was not identified on high-resolution esophagogastroduodenoscopy or colonoscopy (model GIFH260H or CF-H260; Olympus Optical), capsule endoscopy or double-balloon endoscopy was performed as necessary.17 A water-jet device (Olympus Flushing Pump; Olympus Optical) was used to obtain optimal visualization during endoscopy.18 In patients with negative routine endoscopic studies, or patients who could not undergo endoscopy (e.g., because of advanced age, inability to tolerate bowel preparation for endoscopy, or risk of aspiration), multiple detector-row computed tomography (MDCT) was performed to evaluate the source of bleeding as necessary.19 Evidence of GI bleeding on MDCT was defined as visualization of the extravasation of contrast medium into the lumen.20 If a patient could not tolerate either endoscopy or MDCT, an overt GI bleeding event was defined as a clinically significant decrease in hematocrit of ≥ 10% and/or a decrease in hemoglobin concentration of ≥ 2 g/dl from baseline levels according to previous criteria.21 The GI bleeding source was categorized into three groups (upper, middle, or lower).3 Middle GI bleeding was defined as a source between the ampulla of Vater and the terminal ileum.22 Date and cause of death were ascertained from death certificates and electronic medical record reviews. Cause of death was additionally determined from laboratory tests, multiple imaging modalities, pathology, or autopsy.5,23 During the follow-up period, most patients visited our hospital every 1–3 months. In Japan, the prescription period under the healthcare system is limited to 3 months. Consequently, patients must attend the hospital at least every 3 months for repeat prescriptions, when they are assessed by internal physicians or surgeons. For example, patients with hypertension, diabetes mellitus, dyslipidemia, or arrhythmia, or those recovering from thromboembolism or surgery, were periodically followed up every 3 months for prescriptions, monitoring of symptoms and clinical signs, laboratory testing and imaging. The attending physician would ask about potential episodes of overt GI bleeding as part of routine clinical practice, and laboratory tests were performed to detect anemia if appropriate. As our institution is one of the
166
largest emergency referral hospitals in Japan, almost all patients with suspected acute GI bleeding presenting to nearby hospitals or clinics that lacked experienced staff or endoscopy equipment were referred to our hospital. Statistical analysis. We followed up patients from the index examination date to the diagnosis of an overt GI bleeding event, and censored patients at the time of the last visit, end of follow-up (October 31, 2015), or death. The Kaplan–Meier method was used to estimate the cumulative incidence of GI bleeding at 1, 5, and 10 years. Predictors of GI bleeding were evaluated using the log–rank test in a univariate analysis. The Cox proportional hazards model was used to estimate crude hazard ratios (HRs), adjusted HRs, and confidence intervals (CI). In a multivariate analysis, we adjusted for factors with a cut-off P value of 0.2 in a univariate analysis, and a final model was then developed by backward selection of factors showing values of P < 0.1. The weight of each predictor was determined based on the coefficients in the model. Each points value in the scoring system was assigned to each rounded coefficient value. The accuracy of the predictive model for GI bleeding was evaluated by the c-statistic using Harrell’s method.24 The endpoint in the survival analysis was death; data were censored at the time of the last visit, or the end of follow-up (October 30, 2015). The Kaplan–Meier method was used to estimate cumulative mortality at 1, 5, and 10 years. We compared the cumulative mortality rate with the occurrence of a GI bleeding event using the log–rank test. A P value of < 0.05 was considered to indicate statistical significance. All statistical analyses were conducted using STATA version 13 software (StataCorp, College Station, TX, USA).
Results Baseline characteristics. A total of 508 patients were selected for analysis (Fig. 1); their characteristics are shown in Table 1. The mean age of the cohort was 69.4 years, and the majority was male (329 patients, 64.8%). The proportion who drank alcohol was 52.6%, and the proportion who smoked tobacco was 16.7%. Most of those taking an anticoagulant took warfarin (87.0%). Just under half the cohort took an antiplatelet drug (47.1%), while 16.5% took a non-steroidal anti-inflammatory drug and 48.2% took a PPI. The most common comorbidities were diabetes (16.7%), hypertension (55.5%), dyslipidemia (32.9%), cerebro-cardiovascular disease history (38.4%), atrial fibrillation (53.4%), valve replacement or valvuloplasty history (14.8%), and peptic ulcer disease history (17.1%). The mean HAS-BLED score was 2.3, and the mean CHA2DS2-VASc score was 2.9. Occurrence of gastrointestinal bleeding episodes. During a median follow-up of 31.4 months, 42 acute GI bleeding (8.3%) episodes were recorded (Table 2). The overall incidence of acute GI bleeding was 20.5 per 1000 person-years. The cumulative probability of GI bleeding at 1, 5, and 10 years was 3.1%, 12.6%, and 18.5%, respectively (Fig. 2a). The bleeding source was identified in the upper GI tract in 42.8%, the lower GI tract in 50.0%, and the middle GI tract in 7.1% (Table 2).
Journal of Gastroenterology and Hepatology 33 (2018) 164–171 © 2017 Journal of Gastroenterology and Hepatology Foundation and John Wiley & Sons Australia, Ltd
A Shimomura et al.
Gut bleeding model on anticoagulants
Table 1 Baseline demographic and clinical characteristics of the patient cohort (n = 508)
Table 2
Clinical outcomes in the patient cohort (n = 508)
Gastrointestinal bleeding episode Characteristic Age, years Male sex Alcohol consumer Tobacco smoker Drugs Anticoagulant Warfarin DOAC† Dabigatran Rivaroxaban Apixaban Antiplatelet drug Low-dose aspirin‡ Other antiplatelet drug§ NSAIDs¶ Proton pump inhibitor Histamine H2-receptor antagonists †† Labile PT-INR Comorbidities Diabetes mellitus Hypertension Dyslipidemia Cerebro-cardiovascular disease history Cerebral hemorrhage history Peripheral arterial disease history Atrial fibrillation Valve replacement or valvuloplasty history Hemiplegia ‡‡ Chronic kidney disease COPD Connective tissue disease Peptic ulcer disease history Liver cirrhosis Mean HAS-BLED score (±SD) Mean CHA2DS2-VASc score (±SD) §§ Grade of gastric atrophy on endoscopy None or mild Moderate Severe
69.4 329 269 85
(±9.9) (64.8%) (52.6%) (16.7%)
442 66 24 22 20 239 166 73 84 245 41 2
(87.0%) (13.0%) (4.7%) (4.3%) (3.9%) (47.1%) (32.7%) (14.4%) (16.5%) (48.2%) (8.1%) (0.4%)
85 282 167 195 26 24 271 75 16 13 16 53 87 9 2.3 2.9
(16.7%) (55.5%) (32.9%) (38.4%) (5.1%) (4.7%) (53.4%) (14.8%) (3.2%) (2.6%) (3.2%) (10.4%) (17.1%) (1.8%) (±1.2) (±1.7)
219 (45.9%) 140 (29.4%) 118 (24.7%)
Notes: DOAC included dabigatran, rivaroxaban, and apixaban. ‡ Low-dose aspirin comprised enteric-coated aspirin (100 mg) or buffered aspirin (81 mg). § Other antiplatelet drugs were ticlopidine, clopidogrel, cilostazol, dipyridamole, sarpogrelate hydrochloride, ethylicosapentate, dilazep, limaprost, and beraprost. ¶ NSAIDs included non-selective non-steroidal anti-inflammatory drugs and celecoxib. †† Labile PT-INR was defined as PT-INR > 3.5. ‡‡ Chronic kidney disease was defined as a serum creatinine concentration > 2.3 mg/dl. §§ Grades of atrophic gastritis at baseline were categorized into three (1, mild or no atrophy; 2, moderate atrophy; and 3, severe atrophy) according to the Kimura–Takemoto classification. Data are presented as the mean (±standard deviation) or number (proportion, %) as appropriate. Abbreviations: CHA2DS2-VASc, Congestive heart failure, Hypertension, Age ≥ 75, Diabetes mellitus, Stroke, Vascular disease, Sex female; COPD, chronic obstructive pulmonary disease; DOAC, direct oral anticoagulant; HAS-BLED, Hypertension, Abnormal renal/liver function, Stroke, Bleeding history or predisposition, Labile INR, Elderly, Drugs/ alcohol; NSAID, non-steroidal anti-inflammatory drug; PT-INR, prothrombin time-international normalized ratio. †
42 (8.3%)
Value Upper/lower/middle gastrointestinal tract Follow-up period, months All-cause deaths Follow-up period, months
18 (42.8%)/21 (50.0%)/3 (7.1%) 31.4 (19.8–57.7) 59 (11.6%) 32.4 (21.9–59.8)
Notes: Data are presented as the median (interquartile range) or the number (proportion, %) as appropriate.
All-cause mortality. During a median follow-up of 32.4 months, 59 patients (11.6%) died (Table 2). The overall incidence of mortality was 21.3 per 1000 person-years. The cumulative probability of mortality at 1, 5, and 10 years was 1.9%, 13.8%, and 30.1%, respectively (Fig. 2b). Patients with GI bleeding had significantly higher cumulative incidence of all-cause mortality than those without (log–rank test, P < 0.001) (Fig. 2c). The HR for GI bleeding was 2.89 (95% CI 1.55–5.39, P < 0.001), and the age-adjusted and sex-adjusted HR was 2.73 (95% CI 1.45–5.14, P = 0.020). Risk factors for gastrointestinal bleeding and development of predictive scoring system. The risk factors for GI bleeding are shown in Table 3. In the univariate analysis, factors associated with GI bleeding (P < 0.2) were PPI use (P = 0.110), peripheral arterial disease history (P = 0.044), atrial fibrillation (P = 0.086), chronic kidney disease (P = 0.001), COPD (P = 0.023), peptic ulcer disease history (P = 0.041), and liver cirrhosis (P = 0.005). On multivariate analysis, a final GI bleeding model was developed that included PPI use, chronic kidney disease, COPD, peptic ulcer disease history, and liver cirrhosis (Table 3). A predictive model for GI bleeding was developed using these five factors. Based on the final model’s regression coefficients, PPI use was assigned 1 point, chronic kidney disease and cirrhosis 2 points each, and COPD and peptic ulcer disease history 1 point each. Evaluation of the new scoring system. The c-statistic for the new predictive model was 0.65 (95% CI 0.58–0.72, P < 0.001) (Table 4). The HRs for GI bleeding with 1, 0, 1, 2, and 3 predictors were 1 (reference), 3.84, 5.98, 12.9, and 59.0, respectively (P < 0.001 for trend). The c-statistic for the HAS-BLED score was 0.57 (95% CI 0.48–0.65, P = 0.145). The HRs for GI bleeding using the HAS-BLED score with 0, 1, 2, 3, 4, and ≥ 5 predictors were 1, 0.35, 0.75, 1.10, 1.19, and 1.02, respectively (P = 0.50 for trend). The c-statistic was higher in the new predictive model compared with that of the HAS-BLED score, but this was not significant (P = 0.139).
Discussion We focused on the risk of long-term GI bleeding episodes and mortality in patients taking oral anticoagulants who had no GI bleeding on endoscopy at an initial index examination. First, we found that the cumulative probability of GI bleeding at 1, 5, and 10 years was 3.1%, 12.6%, and 18.5%, respectively. Second,
Journal of Gastroenterology and Hepatology 33 (2018) 164–171 © 2017 Journal of Gastroenterology and Hepatology Foundation and John Wiley & Sons Australia, Ltd
167
A Shimomura et al.
Gut bleeding model on anticoagulants
Figure 2 Cumulative probability of gastrointestinal bleeding and all-cause mortality using the Kaplan–Meier method. (a) The cumulative probability of gastrointestinal bleeding (95% confidence interval) at 1, 5, and 10 years was 3.1% (1.9–5.1%), 12.6% (8.9–17.6%), and 18.5% (12.8–26.3%), respectively. (b) The cumulative probability of mortality (95% confidence interval) at 1, 5, and 10 years was 1.9% (1.0–3.5%), 13.8% (10.0–18.7%), and 30.1% (22.2–40.4%), respectively. (c) The cumulative probability of mortality compared with the presence of gastrointestinal bleeding. Patients with gastrointestinal bleeding had significantly higher cumulative incidence of mortality than those without (log–rank test, P < 0.001). , GI bleeding , GI bleeding ( ). GI, gastrointestinal. [Color figure can be viewed at wileyonlinelibrary.com] (+);
patients who experienced an episode of acute GI bleeding had a significantly elevated risk of subsequent mortality. Third, we found that chronic renal disease, COPD, peptic ulcer disease history, and liver cirrhosis increased the risk of GI bleeding, whereas PPI use reduced the risk. We then used these five factors to develop a new predictive model for GI bleeding that had an accuracy superior to the HAS-BLED score. Our new risk assessment model can distinguish patients at high and low risk of acute GI bleeding and may be a useful means of reducing the risk of mortality in patients taking oral anticoagulants. Long-term rates of GI bleeding and mortality in patients taking anticoagulants are broadly comparable with those observed in our cohort. Ruff et al. undertook a meta-analysis of randomized controlled trials of patients with atrial fibrillation taking oral anticoagulation and reported that the incidence of GI bleeding was 2.0% and mortality was 7.7% in a median follow-up period of 26.4 months.25 Chen et al. reported that the incidence of GI bleeding was 10% in a retrospective cohort study of patients taking warfarin followed-up for a median of 40 months.26 The risk of death from GI bleeding in patients taking oral anticoagulants has not been examined before. We found that patients who experienced an acute GI bleeding episode had a significantly higher cumulative incidence of all-cause mortality than those who did not. Staerk et al. have reported that during a 24-month follow-up period, the incidence of death in patients taking an anticoagulant who had a history of GI bleeding was 40%.27 These findings underline the importance of preventing GI bleeding and developing a precise and accurate means of predicting which patients are at greatest risk. The five risk factors for GI bleeding including not taking a PPI, chronic renal disease, COPD, peptic ulcer disease history, and liver cirrhosis that we identified were incorporated into a scoring system that had superior accuracy than the HAS-BLED score. Our scoring system would be straightforward to incorporate into clinical practice, as a screening method for those patients taking long-term anticoagulants. In our study, PPI use decreased the risk of GI bleeding by half (HR 0.5), which closely reflects the findings of Ray et al., who reported that PPI use was associated with a substantial reduction in the risk of warfarin-related upper GI bleeding (HR 0.76).28 Although PPI therapy was not included in
168
either the HAS-BLED score, ATRIA score,10 or ORBIT score11, PPI use is an important means of preventing GI bleeding in the long term. We also found that a diagnosis of COPD increased the GI bleeding risk (HR 4.0). Sherwood et al. reported that COPD increased the GI bleeding risk (HR 1.3) in their randomized controlled trial.29 The exact mechanism by which COPD might increase GI bleeding risk remains unclear, but we propose that patients with COPD might have an extensive smoking history, which is recognized as risk factor for acute GI bleeding.3,30 We also found that patients with a peptic ulcer history were at an elevated risk of GI bleeding (HR 1.8), which chimes with the findings of Chan et al., who reported that history of peptic ulcer increased the risk of GI bleeding in their population-based study (incidence rate ratio 2.31).31,32 In our study, liver cirrhosis substantially elevated the GI bleeding risk (HR 5.6), which concurs with the findings of Chen et al..26 Patients with cirrhosis are more likely to have gastric or esophageal varices as a complication of portal hypertension, along with other abnormalities of coagulation function, all of which would increase the risk of GI bleeding. We also found that chronic renal disease substantially increased the risk of GI bleeding (HR 6.7); Matthew et al. have reported that chronic renal disease increased the incidence of GI bleeding in patients on oral anticoagulation, but to a lesser extent (HR 1.06).29 Chronic renal disease causes abnormalities in blood coagulation and platelet function and causes a bleeding predisposition that may result in GI hemorrhage.33 A study carried out in western countries has shown that coadministration of aspirin use is a risk factor for GI bleeding in patients receiving anticoagulant therapy,29 but we could not show a significant association in our study. In agreement with our findings, however, a study from Taiwan demonstrated that aspirin use was not associated with GI bleeding risk in anticoagulant users (relative risk 0.9 [0.3–2.7], P = 1.00)26, and thus, we speculate that bleeding risk may differ between Asians and Westerners. One of the strengths of our study was that a relatively large sample of patients (n = 508) was followed for a substantial period. Second, all patients underwent endoscopy at an initial index examination and were confirmed to have no evidence of GI bleeding. Nonetheless, our study had some limitations. First,
Journal of Gastroenterology and Hepatology 33 (2018) 164–171 © 2017 Journal of Gastroenterology and Hepatology Foundation and John Wiley & Sons Australia, Ltd
A Shimomura et al.
Table 3
Gut bleeding model on anticoagulants
Gastrointestinal bleeding risk on univariate and multivariate analysis (n = 508)
Factor Age > 70 years Sex Alcohol consumer Tobacco smoker DOAC† Low-dose aspirin‡ Other antiplatelet drug§ NSAID therapy¶ Proton pump inhibitor therapy Histamine H2-receptor antagonists Labile PT-INR†† Diabetes mellitus Hypertension Dyslipidemia Cerebro-cardiovascular disease history Cerebral hemorrhage history Peripheral arterial disease history Atrial fibrillation Valve replacement or valvuloplasty Hemiplegia Chronic kidney disease§§ COPD Collagen disease History of peptic ulcer disease Liver cirrhosis Grade of gastric atrophy on endoscopy¶¶ None or mild Moderate Severe
Unadjusted hazard ratio (95% CI) 1.13 1.19 1.04 0.77 1.22 1.15 1.09 0.94 0.59 0.69 1.10 1.2 0.69 1.08 0.83 2.6 1.8 1.0 1.8 5.4 3.3 1.3 2.0 5.4
(0.61–2.09) (0.62–2.28) (0.56–1.90) (0.30–1.97) (0.48–3.13) (0.62–2.15) (0.50–2.37) (0.37–2.42) (0.31–1.12) (0.21–2.26) N/A‡‡ (0.60–2.4) (0.63–2.2) (0.35–1.3) (0.58–2.0) (0.20–3.4) (1.0–6.6) (0.92–3.4) (0.44–2.2) (0.56–6.0) (1.9–15.3) (1.8–9.2) (0.57–3.2) (1.0–3.8) (1.7–17.4)
1 (reference) 0.73 (0.34–1.6) 0.78 (0.36–1.7)
P
Adjusted hazard ratio (95% CI) Coefficient (95% CI)
0.71 0.61 0.91 0.59 0.68 0.66 0.82 0.90 0.11 0.55 1.00 0.81 0.62 0.28 0.81 0.80 0.044 0.086 1.0 0.31 0.001 0.023 0.50 0.041 0.005
— — — — — — — — 0.52(0.27–1.0) — — — — — — — — — — — 6.7(2.3–19.6) 4.0(1.4–11.2) — 1.8(0.95–3.6) 5.6(1.7–18.6)
— 0.41 0.53
— — —
0.65
1.9 1.4 0.61 1.7
— — — — — — — — ( 1.3–0.096) — — — — — — — — — — — (0.84–3.0) (0.31–2.4) — ( 0.54–1.3) (0.54–2.9) — — —
P — — — — — — — — 0.053 — — — — — — — — — — — 3.5. ‡‡ There were no cases of labile PT-INR in patients with gastrointestinal bleeding. §§ Chronic kidney disease was defined as a serum creatinine concentration > 2.3 mg/dl. ¶¶ Grades of atrophic gastritis at baseline were categorized into three (1, mild or no atrophy; 2, moderate atrophy; and 3, severe atrophy) according to the Kimura–Takemoto classification. Abbreviations: CI, confidence interval; COPD, chronic obstructive pulmonary disease; DOAC, direct oral anticoagulant; NA, not applicable; NSAID, non-steroidal anti-inflammatory drug; PT-INR, prothrombin time-international normalized ratio. †
it was a retrospective study conducted in a single institution. Second, we have not yet undertaken validation of our GI bleeding prediction model, so the generalizability of this scoring system must be evaluated further. Given the relatively low GI bleeding event rate in those taking oral anticoagulants, a large prospective multicenter study will be needed. Third, in Japan, indications for endoscopy are varied, including evaluation for GI symptoms or GI cancer screening in the absence of symptoms, because Japan has a high incidence of colorectal and gastric cancer, and thus, upper and lower GI endoscopy is readily available at moderate cost due under a national insurance system in Japan.34,35 The situation in Japan may be different from that in Western countries. Fourth, because this was a retrospective study, we could not access sufficient
information on the types, frequency, or detailed doses of medications used. In conclusion, an episode of GI bleeding increased the risk of subsequent mortality during follow-up of patients taking oral anticoagulants, highlighting the importance of preventing GI bleeding. We have developed a new scoring system to assess acute GI bleeding risk based on five factors (no-PPI use, chronic kidney disease, COPD, peptic ulcer disease history, and liver cirrhosis), which was superior to the HAS-BLED score.
Acknowledgments We would like to thank clinical research coordinators Ms Hisae Kawashiro, Ms Eiko Izawa, Ms Kenko Yoshida, Ms Yoko
Journal of Gastroenterology and Hepatology 33 (2018) 164–171 © 2017 Journal of Gastroenterology and Hepatology Foundation and John Wiley & Sons Australia, Ltd
169
A Shimomura et al.
Gut bleeding model on anticoagulants
Table 4
Predictive ability and accuracy of the new score compared with the HAS-BLED score (n = 508)
Score New score 1 0 1 2 3 HAS-BLED score 0 1 2 3 4 5
Number of bleeding/number with no bleeding episode
c-statistic (95% CI)
P
Hazard ratio (95% CI)
P for trend
— 4/176 25/239 7/41 4/10 2/0 — 2/22 3/92 11/140 16/134 8/63 2/15
0.65 (0.58–0.72) — — — — — 0.57 (0.48–0.65) — — — — — —
< 0.001 — — — — — 0.145 — — — — — —
— 1 (reference) 3.84 (1.34–11.1) 5.98 (1.75–20.5) 12.9 (3.23–51.7) 59.0 (10.7–325.8) — 1 (reference) 0.35 (0.58–2.09) 0.75 (0.17–3.40) 1.10 (0.25–4.81) 1.19 (0.25–5.62) 1.02 (0.14–7.27)
— — — — — < 0.001 — — — — — — 0.50
Abbreviations: CI, confidence interval; HAS-BLED, Hypertension, Abnormal renal/liver function, Stroke, Bleeding history or predisposition, Labile international normalized ratios [INR], Elderly, Drugs/alcohol.
Tanigawa, Ms Aiko Gotannda, and Ms Kuniko Miki for their help with data collection.
References 1 Gomes T, Mamdani MM, Holbrook AM, Paterson JM, Hellings C, Juurlink DN. Rates of hemorrhage during warfarin therapy for atrial fibrillation. CMAJ 2013; 185: E121–E127. 2 Wysowski DK, Nourjah P, Swartz L. Bleeding complications with warfarin use: a prevalent adverse effect resulting in regulatory action. Arch. Intern. Med. 2007; 167: 1414–9. 3 Rondonotti E, Marmo R, Petracchini M, de Franchis R, Pennazio M. The American society for gastrointestinal endoscopy (ASGE) diagnostic algorithm for obscure gastrointestinal bleeding: eight burning questions from everyday clinical practice. Dig. Liver Dis. 2013; 45: 179–85. 4 Granger CB, Alexander JH, McMurray JJ et al. Apixaban versus warfarin in patients with atrial fibrillation. N. Engl. J. Med. 2011; 365: 981–92. 5 Aoki T, Nagata N, Niikura R et al. Recurrence and mortality among patients hospitalized for acute lower gastrointestinal bleeding. Clin. Gastroenterol. Hepatol. 2015; 13: 488 494.e1. 6 Marmo R, Koch M, Cipolletta L et al. Predicting mortality in nonvariceal upper gastrointestinal bleeders: validation of the italian PNED score and prospective comparison with the Rockall score. Am. J. Gastroenterol. 2010; 105: 1284–91. 7 Irwin J, Ferguson R, Weilert F, Smith A. Supratherapeutic anticoagulation at presentation is associated with reduced mortality in nonvariceal upper gastrointestinal hemorrhage. Endosc. Int. Open. 2014; 2: E148–E152. 8 Halbritter K, Beyer-Westendorf J, Nowotny J, Pannach S, Kuhlisch E, Schellong SM. Hospitalization for vitamin-K-antagonist-related bleeding: treatment patterns and outcome. J. Thromb. Haemost. 2013; 11: 651–9. 9 Pisters R, Lane DA, Nieuwlaat R, de Vos CB, Crijns HJ, Lip GY. A novel user-friendly score (HAS-BLED) to assess 1-year risk of major bleeding in patients with atrial fibrillation: the euro heart survey. Chest 2010; 138: 1093–100. 10 Fang MC, Go AS, Chang Y et al. A new risk scheme to predict warfarin-associated hemorrhage: the ATRIA (anticoagulation and risk factors in atrial fibrillation) study. J. Am. Coll. Cardiol. 2011; 58: 395–401.
170
11 O’Brien EC, Simon DN, Thomas LE et al. The ORBIT bleeding score: a simple bedside score to assess bleeding risk in atrial fibrillation. Eur. Heart J. 2015; 36: 3258–64. 12 Strate LL, Gralnek IM. ACG clinical guideline: management of patients with acute lower gastrointestinal bleeding. Am. J. Gastroenterol. 2016; 111: 459–74. 13 Laine L, Jensen DM. Management of patients with ulcer bleeding. Am. J. Gastroenterol. 2012; 107: 345,60 quiz 361. 14 Aoki T, Nagata N, Shimbo T et al. Development and validation of a risk scoring system for severe acute lower gastrointestinal bleeding. Clin. Gastroenterol. Hepatol. 2016; 14: 1562 1570.e2. 15 Lip GY, Nieuwlaat R, Pisters R, Lane DA, Crijns HJ. Refining clinical risk stratification for predicting stroke and thromboembolism in atrial fibrillation using a novel risk factor-based approach: the euro heart survey on atrial fibrillation. Chest 2010; 137: 263–72. 16 Peloquin JM, Seraj SM, King LY, Campbell EJ, Ananthakrishnan AN, Richter JM. Diagnostic and therapeutic yield of endoscopy in patients with elevated INR and gastrointestinal bleeding. Am. J. Med. 2016; 129: 628–34. 17 Niikura R, Nagata N, Aoki T et al. Predictors for identification of stigmata of recent hemorrhage on colonic diverticula in lower gastrointestinal bleeding. J. Clin. Gastroenterol. 2015; 49: e24–e30. 18 Niikura R, Yamada A, Nagata N et al. New predictive model of rebleeding during follow-up of patents with obscure gastrointestinal bleeding: a multicenter cohort study. J. Gastroenterol. Hepatol. 2016; 31: 752–60. 19 Kulkarni C, Moorthy S, Sreekumar K et al. In the workup of patients with obscure gastrointestinal bleed, does 64-slice MDCT have a role? Indian. J. Radiol. Imaging. 2012; 22: 47–53. 20 Nagata N, Niikura R, Aoki T et al. Role of urgent contrast-enhanced multidetector computed tomography for acute lower gastrointestinal bleeding in patients undergoing early colonoscopy. J. Gastroenterol. 2015; 50: 1162–72. 21 Chan FK, Cryer B, Goldstein JL et al. A novel composite endpoint to evaluate the gastrointestinal (GI) effects of nonsteroidal antiinflammatory drugs through the entire GI tract. J. Rheumatol. 2010; 37: 167–74. 22 Nagata N, Niikura R, Yamada A et al. Acute middle gastrointestinal bleeding risk associated with NSAIDs, antithrombotic drugs, and PPIs: a multicenter case-control study. PLoS One 2016; 11 e0151332. 23 Nagata N, Niikura R, Sakurai T et al. Safety and effectiveness of early colonoscopy in management of acute lower gastrointestinal bleeding on the basis of propensity score matching analysis. Clin. Gastroenterol. Hepatol. 2016; 14: 558–64.
Journal of Gastroenterology and Hepatology 33 (2018) 164–171 © 2017 Journal of Gastroenterology and Hepatology Foundation and John Wiley & Sons Australia, Ltd
A Shimomura et al.
24 Harrell FE Jr, Lee KL, Mark DB. Multivariable prognostic models: issues in developing models, evaluating assumptions and adequacy, and measuring and reducing errors. Stat. Med. 1996; 15: 361–87. 25 Ruff CT, Giugliano RP, Braunwald E et al. Comparison of the efficacy and safety of new oral anticoagulants with warfarin in patients with atrial fibrillation: a meta-analysis of randomised trials. Lancet 2014; 383: 955–62. 26 Chen WC, Chen YH, Hsu PI et al. Gastrointestinal hemorrhage in warfarin anticoagulated patients: incidence, risk factor, management, and outcome. Biomed. Res. Int. 2014; 2014: 463767. 27 Staerk L, Lip GY, Olesen JB et al. Stroke and recurrent haemorrhage associated with antithrombotic treatment after gastrointestinal bleeding in patients with atrial fibrillation: nationwide cohort study. BMJ 2015; 351: h5876. 28 Ray WA, Chung CP, Murray KT et al. Association of proton pump inhibitors with reduced risk of warfarin-related serious upper gastrointestinal bleeding. Gastroenterology 2016; 151: 1105 1112.e10. 29 Sherwood MW, Nessel CC, Hellkamp AS et al. Gastrointestinal bleeding in patients with atrial fibrillation treated with rivaroxaban or warfarin: ROCKET AF trial. J. Am. Coll. Cardiol. 2015; 66: 2271–81.
Gut bleeding model on anticoagulants
30 Laursen SB, Hansen JM, Hallas J, Schaffalitzky de Muckadell OB. The excess long-term mortality in peptic ulcer bleeding is explained by nonspecific comorbidity. Scand. J. Gastroenterol. 2015; 50: 145–52. 31 Chan EW, Lau WC, Leung WK et al. Prevention of dabigatran-related gastrointestinal bleeding with gastroprotective agents: a populationbased study. Gastroenterology 2015; 149: 586 95.e3. 32 Goodman SG, Wojdyla DM, Piccini JP et al. Factors associated with major bleeding events: insights from the ROCKET AF trial (rivaroxaban once-daily oral direct factor xa inhibition compared with vitamin K antagonism for prevention of stroke and embolism trial in atrial fibrillation). J. Am. Coll. Cardiol. 2014; 63: 891–900. 33 Gerson LB. Causes of gastrointestinal hemorrhage in patients with chronic renal failure. Gastroenterology 2013; 145: 895 7; discussion 897. 34 Uemura N, Okamoto S, Yamamoto S et al. Helicobacter pylori infection and the development of gastric cancer. N. Engl. J. Med. 2001; 345: 784–9. 35 Nagata N, Sakamoto K, Arai T et al. Visceral abdominal fat measured by computed tomography is associated with an increased risk of colorectal adenoma. Int. J. Cancer 2014; 135: 2273–81.
Journal of Gastroenterology and Hepatology 33 (2018) 164–171 © 2017 Journal of Gastroenterology and Hepatology Foundation and John Wiley & Sons Australia, Ltd
171