Acute Severe Gastrointestinal Tract Bleeding Is Associated With an Increased Risk of Thromboembolism and Death

Accepted Manuscript Acute Severe Gastrointestinal Tract Bleeding is Associated with an Increased Risk of Thromboembolism and Death Naoyoshi Nagata, Toshiyuki Sakurai, Takuro Shimbo, Shiori Moriyasu, Hidetaka Okubo, Kazuhiro Watanabe, Chizu Yokoi, Mikio Yanase, Junichi Akiyama, Naomi Uemura PII: DOI: Reference:

S1542-3565(17)30735-8 10.1016/j.cgh.2017.06.028 YJCGH 55310

To appear in: Clinical Gastroenterology and Hepatology Accepted Date: 14 June 2017 Please cite this article as: Nagata N, Sakurai T, Shimbo T, Moriyasu S, Okubo H, Watanabe K, Yokoi C, Yanase M, Akiyama J, Uemura N, Acute Severe Gastrointestinal Tract Bleeding is Associated with an Increased Risk of Thromboembolism and Death, Clinical Gastroenterology and Hepatology (2017), doi: 10.1016/j.cgh.2017.06.028. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT

Acute Severe Gastrointestinal Tract Bleeding is Associated with an Increased Risk

RI PT

of Thromboembolism and Death

Authors: Naoyoshi Nagata1, Toshiyuki Sakurai1, Takuro Shimbo2, Shiori Moriyasu1,

SC

Hidetaka Okubo1, Kazuhiro Watanabe1, Chizu Yokoi1, Mikio Yanase1, Junichi Akiyama1,

1

M AN U

and Naomi Uemura3.

Department of Gastroenterology and Hepatology, National Center for Global Health

and Medicine, 1-21-1 Toyama, Shinjuku 162-8655, Tokyo, Japan Ohta Nishinouchi Hospital, 2-5-20 Nishinouchi, Koriyama, Fukushima 963-8022,

Japan

Department of Gastroenterology and Hepatology, Kohnodai Hospital, National Center

EP

3

TE D

2

AC C

for Global Health and Medicine, 1-7-1 Kohnodai, Ichikawa 272-8516, Chiba, Japan

Running Head: GIB as a risk for thromboembolism and mortality

Correspondence: Naoyoshi Nagata, Department of Gastroenterology and Hepatology National Center for Global Health and Medicine, 1-21-1 Toyama, Shinjuku-ku, Tokyo

1

ACCEPTED MANUSCRIPT

162-8655, Japan. Tel.: +81-03-3202-7181; Fax: +81-03-3207-1038; E-mail:

RI PT

[email protected]

SC

Word count: 3663 words

M AN U

Grant Support: This work was supported in part by Grants-in-Aid for Research from the National Center for Global Health and Medicine (26A-201 and 29-2001). The funding agency played no role in the study design, data collection and analysis, decision

TE D

to publish, or preparation of the manuscript.

Abbreviations: ACS, acute coronary syndrome; BMI, body mass index; CI, confidence

EP

interval; EGD, esophagogastroduodenoscopy; GIB, gastrointestinal bleeding; HR,

AC C

hazard ratio; LDA, low-dose aspirin; MDCT, multidetector-row computed tomography; NCGM, National Center for Global Health and Medicine; NSAIDs, non-steroidal anti-inflammatory drugs; OR, odds ratio; SD, standard deviation; SHR, subdistribution hazard ratio.

Disclosures: The authors declare no conflict of interest.

2

ACCEPTED MANUSCRIPT

Author Contributions: NN designed the study; TS helped to perform statistical

RI PT

analysis; NN, TS, and SM mainly performed data collection and are the main authors of the manuscript; HO, KW, CY, and JA assisted with treatment; and MY and NU edited

AC C

EP

TE D

M AN U

SC

the manuscript. All authors read and approved the submitted version of the manuscript.

3

ACCEPTED MANUSCRIPT

Abstract: Background & Aims: We performed a retrospective cohort study of patients with and

RI PT

without gastrointestinal bleeding (GIB) to determine whether GIB increases the risks of thromboembolism and death.

SC

Methods: We collected data from 522 patients with acute severe GIB and 1044 patients

M AN U

without GIB (controls, matched for age, sex, year of diagnosis, history of thromboembolism, and use of antithrombotic drugs) who underwent endoscopy at the National Center for Global Health and Medicine in Japan from January 2009 through December 2014. Hazard ratios (HRs) of GIB for thromboembolism and mortality risk

TE D

were estimated, adjusting for confounders. We also compared standardized mortality ratios between the GIB cohort and the age- and sex-matched general population in

EP

Japan.

AC C

Results: During a mean follow up of 23.7 months, thromboembolism was identified in 11.5% of patients with GIB and 2.4% of controls (HR, 5.3; 95% CI, 3.3–8.5; P<.001). Multivariate analysis revealed GIB as a risk factor for all-thromboembolic events, cerebrovascular events, and cardiovascular events. During a mean follow up of 24.6 months, 15.9% of patients with GIB and 8.6% of controls died (HR, 2.1; 95% CI, 1.6–2.9; P<.001). Multivariate analysis revealed GIB as a risk factor for all-cause

4

ACCEPTED MANUSCRIPT

mortality. Compared with the general population, patients with GIB were at increased risk of death (standardized mortality ratio, 12.0).

RI PT

Conclusion: In a retrospective analysis of patients undergoing endoscopy in Japan, we identified acute GIB was a significant risk factor for late thromboembolism and death,

SC

compared to patients without GIB. GIB also increased risk of death compared with the

M AN U

general population.

KEY WORDS upper gastrointestinal hemorrhage; lower gastrointestinal hemorrhage;

AC C

EP

TE D

stroke; myocardial infarction

5

ACCEPTED MANUSCRIPT

INTRODUCTION Previous studies have indicated that acute gastrointestinal bleeding (GIB) is associated

RI PT

with increased mortality and major thromboembolic events, particularly in selected

patients such as those with acute coronary syndrome (ACS) or stroke mainly treated

SC

with antithrombotic drugs1-5. However, in real-world clinical settings, more than half of acute GIB events occur in the absence of antithrombotic drug use6-8. Because patients

M AN U

with acute GIB have potentially hypercoagulable blood9, thromboembolic events may occur even in the absence of antithrombotic therapy. However, the associations between acute GIB and risks of thromboembolism and mortality in non-selected populations

TE D

have never been studied. If acute GIB episodes are associated with increased risks of subsequent thromboembolism and mortality, gastroenterologists and specialists for

AC C

episodes.

EP

cerebro-cardiovascular disease may have to perform a careful follow-up after GIB

We therefore conducted a long-term cohort study of patients with acute severe GIB and matched non-GIB diagnosed by endoscopy, and aimed to determine whether acute GIB episodes increase the risk of subsequent thromboembolism and mortality.

METHODS

6

ACCEPTED MANUSCRIPT

Study Design, Setting, and Participants We conducted a retrospective cohort study of severe acute GIB and non-GIB patients to

RI PT

estimate the risks of thromboembolism and mortality. All patients were adults who

underwent endoscopy at the National Center for Global Health and Medicine, Japan

SC

between January 2009 and December 2014. Our hospital is one of the largest emergency

M AN U

hospitals (900 beds) in the Tokyo metropolitan area. Data were collected from the recorded electronic endoscopic database (Solemio Endo; Olympus Medical Systems, Tokyo, Japan)7, 8, supplemented by review of the electronic medical records (MegaOak online imaging system; NEC, Tokyo, Japan). This electronic database is a searchable

TE D

collection of records into which physicians or nurses prospectively input endoscopic or clinical findings, diagnosis, and therapy based on certain forms7, 8. This study was

EP

approved by the institutional review board.

AC C

First, we selected overt GIB patients and non-GIB patients as defined below using the endoscopic database (Fig. 1). The search identified 1,267 consecutive patients who presented with acute, continuous, or frequent overt GIB (“hematochezia”, “red blood pre rectum”, “melena”, “tarry stool”, or “hematemesis”) and 2,657 patients with no apparent episode of GIB who underwent both upper and lower endoscopy at the same time. Endoscopic and clinical data for each patient were then reviewed by expert

7

ACCEPTED MANUSCRIPT

gastroenterologists (n=3,924). To define the acute severe GIB cohort, we excluded patients: i) with no overt bleeding within 3 days of endoscopy (n=211); ii) for whom

RI PT

clinical information could not be accurately collected (n=89); iii) with less than 1 month of follow-up of the GIB episode (n=173); or iv) who did not fit the following definition

SC

of severe acute GIB (n=308): ongoing bleeding with transfusion of more than 2 units of

M AN U

packed red blood cells in the emergency room or during hospital stay and/or with signs of shock (decrease in systolic blood pressure to <90 mmHg, pallor, cold sweats, dizziness, syncope, or loss of consciousness)10, 11. A cohort of 522 acute severe outpatient-onset GIB patients who underwent endoscopy was selected. To define the

TE D

non-GIB cohort, we excluded patients: i) with a prior history of acute overt GIB (n=178); ii) with endoscopically verified ulcers or bleeding lesions (n=102); iii) clinical

EP

information could not be accurately collected (n=211); or iii) with less than 1 month

AC C

follow-up of endoscopy (n=263). After application of the exclusion criteria, 2,095 non-GIB patients remained. To minimize confounding effects, non-GIB subjects were randomly selected from among individuals matched for decennial age, sex, year of GIB diagnosis, history of thromboembolism, and use of antithrombotic drugs, with a ratio of 1:2. Ultimately, cohorts of 522 severe acute GIB patients and 1,044 non-GIB patients who underwent endoscopy were selected for analysis.

8

ACCEPTED MANUSCRIPT

Comorbidity and Medication

RI PT

We evaluated comorbidity using the Charlson comorbidity index12, a validated and

commonly used predictive index of mortality in GIB research13. Among patients on

SC

these drugs before and after the index date (defined as the day of discharge due to acute

M AN U

GIB for the GIB cohort, or day of endoscopy for the matched non-GIB cohort), we categorized patients into 3 groups: i) no drug, no drug use within 1 month before and after the index date; ii) continue, drug use within 1 month before and after the index date; and iii) discontinue, drug use only within 1 month before the index date. Because

TE D

specific medications (NSAIDs, LDA, and non-aspirin antiplatelet drugs) were occasionally interrupted in continued group or resumed in discontinued group, we

EP

defined continued group as 50% or more use of the observational period, and

AC C

discontinued group as less than 20% use of the follow-up period according to their cumulative duration of medication use, as previously reported14.

Outcomes and Follow-up The main outcomes of interest were thromboembolism, and death after the index date (defined as the day of discharge due to acute GIB for the GIB cohort, or day of

9

ACCEPTED MANUSCRIPT

endoscopy for the matched non-GIB cohort). In Japan, because the prescription period under the healthcare system is limited to 3 months, GIB and non-GIB patients need to

RI PT

visit at least every 3 months for disease-related prescriptions. Non-GIB patients were outpatient subjects who were under long-term follow-up for metabolic disorder or

SC

chronic disease. These patients were also periodically followed-up every 3 months for

M AN U

monitoring of symptoms, laboratory testing, and imaging, and also had occasion to undergo endoscopy for cancer screening, even as part of the examination of patients without symptoms of the disease. We defined a thromboembolic event as the presence of acute coronary syndrome, stroke, transient ischemic attack, venous

TE D

thromboembolism (pulmonary embolism or deep vein thrombosis), or arterial thromboembolism. The diagnosis of thromboembolism was based on typical symptoms

EP

with imaging modalities including computed tomography, magnetic resonance imaging,

AC C

coronary angiography, ventilation-perfusion scan, ultrasonography, or electrocardiography. 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, or autopsy.

Statistics

10

ACCEPTED MANUSCRIPT

The main clinical outcomes were thromboembolism and mortality, and exposures were the GIB episode and other clinical factors. In the thromboembolic risk analysis, we

RI PT

followed up patients from the index date to the diagnosis of any thromboembolic event, and censored patients at the time of the last visit, end of follow-up (December 31, 2014),

SC

or death. In the survival risk analysis, the endpoint was death and data were censored at

M AN U

the time of the last visit, or the end of follow-up (December 31, 2014). The Kaplan-Meier method was used to estimate cumulative incidences of outcomes, and differences were compared with log-rank test. Unadjusted and adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) of GIB for the risk of outcomes were

TE D

estimated by Cox proportional hazards modeling. In multivariate analysis, we adjusted for matching factors, propensity score for GIB, plus 7 factors (hypertension, diabetes,

EP

dyslipidemia, cerebrovascular disease history, acute coronary syndrome history, chronic

AC C

kidney disease, and liver cirrhosis. To estimate the propensity score for GIB, we have employed a logistic regression model for GIB including 17 factors that are potentially clinically important variables (BMI ≥25 kg/m2, hypertension, diabetes mellitus, dyslipidemia, chronic kidney disease, liver cirrhosis, treatment history of malignancy, leukemia, history of peptic ulcer disease, congestive heart failure, dementia, hemiplegia, and use of NSAIDs, LDAs, non-LDA antiplatelets, and anticoagulants). Some of these

11

ACCEPTED MANUSCRIPT

were shown to differ (p<0.10) between GIB and non-GIB. The area under the receiver operating characteristic curve for propensity scores for GIB was 0.78 (95%CI,

RI PT

0.76–0.81).

Next, using Gray’s test15 in the competing risk analysis, we calculated subdistribution

SC

hazard ratios (SHRs) with 95%CI, treating death without thromboembolism as a

M AN U

competing risk.

To determine whether GIB is a risk factor for mortality in another non-endoscopy cohort, we selected a comparative cohort comprising the general population in Japan. The expected number of all-cause deaths was determined using age-stratified and

TE D

sex-specific total mortality rate data for Japan, provided by the Statistics Information Department, Minister's Secretariat, Ministry of Health and Welfare, 2014 in Japan16.

EP

Standardized mortality ratios (SMRs) were calculated as the ratio of observed deaths to

AC C

the expected number of patient deaths. The 95%CI of the SMR was estimated assuming a Poisson distribution.

Values of p<0.05 were considered significant. All statistical analyses were performed using STATA version 13 software (StataCorp, College Station, TX).

RESULTS

12

ACCEPTED MANUSCRIPT

Baseline characteristics of the GIB cohort and non-GIB cohort are shown in Table 1. Scores for the Charlson Comorbidity Index and CHA2DS2-VASc were higher in the

RI PT

GIB cohort than in the non-GIB cohort. Rates of drug discontinuation were higher in the GIB cohort than in the non-GIB cohort. Blood transfusions were administered to 88%

SC

of GIB patients (mean±standard deviation, 7.0±8.3 units; median, 4.0 units; interquartile

shock.

M AN U

range [IQR], 4-8 units) during the hospital stay, and 36% of patients presented with Among comorbidities, there was a higher rate of chronic kidney disease in the

GIB cohort than in the non-GIB cohort (Supplementary Table 1). Endoscopy revealed that 71.5% of all GIB were from upper sources and 28.5% were from lower sources

TE D

(supplementary Table 1). Peptic ulcer disease was the major cause of upper GIB,

EP

whereas colonic diverticular bleeding was the major cause of lower GIB.

AC C

Thromboembolism Risk

During a mean follow up of 23.7 months (IQR, 10.1-34.0 months), thromboembolic event was identified in 60 of the 522 patients (11.5%) in the GIB cohort and 25 of the 1,044 patients (2.4%) in the non-GIB cohort (Table 2). Cumulative thromboembolism rates at 5 years were 21.5% in the GIB cohort and 6.0% in the non-GIB cohort (log-rank test, p<0.001; Fig. 2A). Overall incidences of thromboembolism were 5.0 per 1,000

13

ACCEPTED MANUSCRIPT

person-years in the GIB cohort and 0.9 per 1,000 person-years in the non-GIB cohort. Various risk factors for thromboembolism other than GIB were considered. Log-rank

RI PT

testing revealed that risk factors for thromboembolism other than GIB included age ≥65 years (p=0.231), male sex (p=0.088), past history of thromboembolism (Fig. 3A;

SC

p=0.007), BMI ≤25 kg/m2 (p=0.139), CHAD2DS2-VASc ≥2 (Fig. 3B; p=0.269),

M AN U

comorbidity index ≥2 (Fig. 3C; p<0.001), NSAID use (p=0.098), LDA use (Fig. 3D; p<0.001), non-aspirin antiplatelet drug use (Fig. 3E; p=0.004), and anticoagulant use (Fig. 3F; p<0.001). Multivariate analysis revealed GIB as an independent risk factor for subsequent all-thromboembolic events, cerebrovascular events, cardiovascular events,

TE D

and other thromboembolic events (Table 2). In subgroup analysis for antithrombotic drug user and non-users, a risk of

EP

thromboembolism was evident, irrespective of antithrombotic drug use in univariate

AC C

analysis (Table 2). Multivariate analysis revealed that GIB as an independent risk factor for subsequent all-thromboembolic events, cerebrovascular events, and cardiovascular events, but not other thromboembolic events irrespective of antithrombotic drug use (Table 2). In competing risk analysis, cumulative incidence of thromboembolism was higher in the GIB cohort than that in the non-GIB cohort (Fig. 2B; SHR, 5.0; 95%CI, 3.1-8.0;

14

ACCEPTED MANUSCRIPT

p<0.001). Multivariate analysis also revealed that GIB episodes increased the risk of subsequent all-thromboembolic events, cerebrovascular events, and cardiovascular

RI PT

events (Table 2).

SC

Mortality Risk

M AN U

During a mean follow up of 24.6 months (IQR, 11.2-35.6), 83 patients (15.9%) died in the GIB cohort and 86 patients (8.6%) died in the non-GIB cohort (Table 2). Cumulative all-cause mortality rates at 5 years were 22.1% in the GIB cohort and 11.6% in the non-GIB cohort (log-rank test, p<0.001; Fig. 2C). Overall incidences of

in the non-GIB cohort.

TE D

death were 6.6 per 1,000 person-years in the GIB cohort and 3.1 per 1,000 person-years

EP

Various risk factors for mortality other than GIB were considered. Log-rank testing

AC C

revealed that risk factors for mortality other than GIB were age ≥65 years (Fig. 4A; p=0.004), male sex (p=0.707), history of thromboembolism (p=0.882), BMI ≤25 kg/m2 (Fig. 4B; p<0.001), CHAD2DS2-VASc ≥2 (p=0.479), comorbidity index ≥2 (Fig. 4C; p<0.001), NSAID use (Fig. 4D; p=0.004), LDA use (Fig. 4E; p=0.047), non-aspirin antiplatelet drug use (Fig. 4F; p=0.082), and anticoagulant use (p=0.333). Multivariate analysis revealed GIB as an independent risk factor for all-cause mortality (Table 2). In

15

ACCEPTED MANUSCRIPT

subgroup analysis for antithrombotic drug user and non-users, mortality risk was evident, irrespective of antithrombotic drug use in univariate analysis (Table 2).

antithrombotic drug users, but not in non-users.

RI PT

Multivariate analysis revealed GIB as an independent risk factor for mortality in

SC

Compared with the expected mortality rate based on vital statistics data for Japan, the

SMR of 12.0 (95%CI, 9.4-14.6).

DISCUSSION

M AN U

expected number of all-cause deaths in patients with acute GIB was 6.92, yielding an

TE D

We found that patients with a severe GIB episode were at increased risk of subsequent all-thromboembolic events, cerebrovascular events, cardiovascular events, other

EP

thromboembolic events, and all-cause mortality compared with non-GIB. Mortality risk

AC C

in patients with severe GIB was significantly higher compared to the age-sex adjusted general population in Japan. An increased risk of all-thromboembolism and mortality was evident, irrespective of antithrombotic drug use. In the ACUITY trial of ACS patients2, GIB was significantly associated with 1-year mortality (HR, 4.0), myocardial infarction (HR, 1.7), and composite ischemia (HR, 1.9) as well as 30-day outcomes. In results from the CHARISMA trial, bleeding (GIB was

16

ACCEPTED MANUSCRIPT

most common) and mortality were significant (HR, 2.6), along with myocardial infarction (HR, 2.9) and stroke (HR, 4.2)3. In 6853 patients with ischemic stroke, GIB

RI PT

was associated with mortality (odds ratio [OR], 3.3), recurrent stroke (OR, 3.7), and

myocardial infarction (OR 2.8) at discharge4. Although previous studies have focused on

SC

patients with ACS, cerebrovascular disease, or peripheral artery disease, as these

M AN U

patients usually take antithrombotic drugs1-5, the risk ratios for thromboembolism or mortality due to GIB were similar to our results.

The exact mechanisms by which episodes of GIB increase the risk of subsequent thromboembolism or mortality remains unknown, but we suggest four possibilities.

TE D

First, acute episodes of GIB potentially cause a hypercoagulable state. GIB may result in abnormal platelet activation or coagulation cascade at many different levels17. When a

EP

blood vessel is injured, a cascade of biochemical reactions begins as a necessary

AC C

pathway to achieving hemostasis by developing a clot17, 18. One study indicated that plasma thrombin-antithrombin III complex concentrations were raised during acute GIB episodes (p<0.001), compared with an age-matched reference group9. Second, in terms of cardiovascular risk, bleeding may reduce oxygen delivery to the myocardium downstream of coronary stenosis19. Third, GIB is a well-known cause of cessation of antithrombotic drug, particularly in patients with ACS or atrial fibrillation20,21. Two

17

ACCEPTED MANUSCRIPT

observational studies of GIB patients on anticoagulants demonstrated that discontinuing anticoagulant use at discharge increased the risks of thromboembolism within 90 days

RI PT

(HR, 14-20) and death (HR, 3.3)20,21.

A major limitation of this study is that patients with severe GIB had more comorbidities

SC

at baseline and therefore were more likely to develop thromboembolism or die during

M AN U

follow-up. Although we tried to adjust for matching factors, propensity score for GIB, and 7 comorbidities in the current analysis, the multivariate model did not account for all potential adjusted confounders associated with GIB. To determine whether GIB is an independent predictor of thromboembolism or death, a large prospective cohort study

TE D

should enroll patients in the GIB and non-GIB cohorts that have the same background, including past medical history, comorbidities, and drugs used.

EP

We also recognize other limitations to this study. First, the number of thromboembolic

AC C

events was relatively low, which might have resulted in insufficient statistical power in multivariate analysis. Second, information was not collected on other drugs known to affect the risk of GIB, including proton-pump inhibitors, corticosteroids, selective serotonin reuptake inhibitors, and spironolactone. Third, we could not analyze thromboembolic risk between the GIB cohort and the general population, because no data are available on thromboembolic incidence in Japan. We tried to make the baseline

18

ACCEPTED MANUSCRIPT

characteristics in the non-GIB cohort as close as possible to those in the GIB cohort, but our non-GIB cohort might not be representative of general population. Despite these

RI PT

limitations, the strengths of this study are that all patients underwent endoscopy at the index date and we could assess whether drugs with known risks were continued or

SC

discontinued after the index date.

M AN U

In conclusion, acute GIB was a significant risk factor for late thromboembolism and death, irrespective of antithrombotic drug use compared to non-GIB evaluated by endoscopy. GIB was also a risk factor for mortality compared with the general population. Prevention of thromboembolism and death seems likely to become

TE D

increasingly important in the long-term management of acute GIB.

EP

ACKNOWLEDGMENT: We wish to thank Ms. Hisae Kawashiro, Ms. Eiko Izawa,

AC C

Ms. Kenko Yoshida, Ms. Akiko Shimizu, Ms. Tomoe Watanabe, Ms. Haruko Sawamoto, and Ms. Kuniko Miki, and for assistance with data collection. None of these contributors received any financial compensation.

19

ACCEPTED MANUSCRIPT

REFERENCES 1. Staerk L, Lip GY, Olesen JB, et al. Stroke and recurrent haemorrhage associated with

RI PT

antithrombotic treatment after gastrointestinal bleeding in patients with atrial

fibrillation: nationwide cohort study. BMJ (Clinical research ed.) 2015;351:h5876.

SC

2. Nikolsky E, Stone GW, Kirtane AJ, et al. Gastrointestinal bleeding in patients with

M AN U

acute coronary syndromes: incidence, predictors, and clinical implications: analysis from the ACUITY (Acute Catheterization and Urgent Intervention Triage Strategy) trial. Journal of the American College of Cardiology 2009;54:1293-1302. 3. Berger PB, Bhatt DL, Fuster V, et al. Bleeding complications with dual antiplatelet

TE D

therapy among patients with stable vascular disease or risk factors for vascular disease: results from the Clopidogrel for High Atherothrombotic Risk and Ischemic Stabilization,

EP

Management, and Avoidance (CHARISMA) trial. Circulation 2010;121:2575-2583.

AC C

4. O'Donnell MJ, Kapral MK, Fang J, et al. Gastrointestinal bleeding after acute ischemic stroke. Neurology 2008;71:650-655. 5. van Hattum ES, Algra A, Lawson JA, et al. Bleeding increases the risk of ischemic events in patients with peripheral arterial disease. Circulation 2009;120:1569-1576. 6. Lanas A, Carrera-Lasfuentes P, Arguedas Y, et al. Risk of upper and lower gastrointestinal bleeding in patients taking nonsteroidal anti-inflammatory drugs,

20

ACCEPTED MANUSCRIPT

antiplatelet agents, or anticoagulants. Clinical gastroenterology and hepatology : the official clinical practice journal of the American Gastroenterological Association

RI PT

2015;13:906-12.e2.

7. Nagata N, Niikura R, Sekine K, et al. Risk of peptic ulcer bleeding associated with

SC

Helicobacter pylori infection, nonsteroidal anti-inflammatory drugs, low-dose aspirin,

hepatology 2015;30:292-298.

M AN U

and antihypertensive drugs: a case-control study. Journal of gastroenterology and

8. Nagata N, Niikura R, Aoki T, et al. Lower GI bleeding risk of nonsteroidal anti-inflammatory drugs and antiplatelet drug use alone and the effect of combined

TE D

therapy. Gastrointestinal endoscopy 2014.

9. Henriksson AE, Nilsson TK, Svensson JO. Hypercoagulability in acute bleeding

EP

peptic ulcer disease assessed by thrombin-antithrombin III concentrations. The

AC C

European journal of surgery = Acta chirurgica 1993;159:167-169. 10. Velayos FS, Williamson A, Sousa KH, et al. Early predictors of severe lower gastrointestinal bleeding and adverse outcomes: a prospective study. Clinical gastroenterology and hepatology : the official clinical practice journal of the American Gastroenterological Association 2004;2:485-490.

21

ACCEPTED MANUSCRIPT

11. Fujino Y, Inoue Y, Onodera M, et al. Risk factors for early re-bleeding and associated hospitalisation in patients with colonic diverticular bleeding. Colorectal

RI PT

disease : the official journal of the Association of Coloproctology of Great Britain and Ireland 2013.

SC

12. Charlson ME, Pompei P, Ales KL, et al. A new method of classifying prognostic

diseases 1987;40:373-383.

M AN U

comorbidity in longitudinal studies: development and validation. Journal of chronic

13. Aoki T, Nagata N, Niikura R, et al. Recurrence and mortality among patients hospitalized for acute lower gastrointestinal bleeding. Clinical gastroenterology and

TE D

hepatology : the official clinical practice journal of the American Gastroenterological Association 2015;13:488-494.e1.

EP

14. Chan FK, Leung Ki EL, Wong GL, et al. Risks of Bleeding Recurrence and

AC C

Cardiovascular Events With Continued Aspirin Use After Lower Gastrointestinal Hemorrhage. Gastroenterology 2016;151:271-277. 15. Fine J, Gray R. A proportional hazards model for the subdistribution of a competing risk. Journal of the American Statistical Association 1999;94:496-509. 16. Vital Statistics of Japan. Tokyo, Japan: Statistics and Information Department, Minister's Secretariat, Ministry of Health and Welfare, 2014.

22

ACCEPTED MANUSCRIPT

17. Adams RL, Bird RJ. Review article: Coagulation cascade and therapeutics update:

anticoagulants. Nephrology (Carlton, Vic.) 2009;14:462-470.

RI PT

relevance to nephrology. Part 1: Overview of coagulation, thrombophilias and history of

18. McMichael M. New models of hemostasis. Topics in companion animal medicine

SC

2012;27:40-45.

M AN U

19. Most AS, Ruocco NA,Jr, Gewirtz H. Effect of a reduction in blood viscosity on maximal myocardial oxygen delivery distal to a moderate coronary stenosis. Circulation 1986;74:1085-1092.

20. Sengupta N, Feuerstein JD, Patwardhan VR, et al. The risks of thromboembolism vs.

TE D

recurrent gastrointestinal bleeding after interruption of systemic anticoagulation in hospitalized inpatients with gastrointestinal bleeding: a prospective study. The American

EP

Journal of Gastroenterology 2015;110:328-335.

AC C

21. Witt DM, Delate T, Garcia DA, et al. Risk of thromboembolism, recurrent hemorrhage, and death after warfarin therapy interruption for gastrointestinal tract bleeding. Archives of Internal Medicine 2012;172:1484-1491.

23

ACCEPTED MANUSCRIPT

FIGURE LEGENDS Figure 1. Study flow

RI PT

Abbreviation. GIB, gastrointestinal bleeding

M AN U

Kaplan-Meier method and competing risk regression.

SC

Figure 2. Cumulative probability of thromboembolism and mortality rate using the

A) Cumulative thromboembolism rates (95%CI) in the GIB cohort and non-GIB cohort were 8.8% (6.5-11.8%) vs. 1.2% (0.6-2.1%) at 1 year, and 21.5% (14.1-32.1%) vs. 6.0% (3.6-16.1%) at 5 years, respectively (log-rank test, p<0.001).

TE D

B) Cumulative thromboembolism rates by competing risk regression. C) The cumulative all-cause mortality rate (95%CI) in the GIB and non-GIB cohorts

EP

were 12.9% (10.2-16.3%) vs. 5.6% (4.3-7.2%) at 1 year, and 22.1% (16.2-29.8%) vs.

AC C

11.6% (9.2-14.6%) at 5 years, respectively (log-rank test, p<0.001).

Figure 3. Cumulative probability of thromboembolism according to risk factors other than GIB analyzed using the Kaplan-Meier method. A) History of thromboembolism; B) CHAD2DS2-VASc; C) comorbidity index; D) use of low-dose aspirin; E) use of non-aspirin antiplatelet drugs; and F) use of

24

ACCEPTED MANUSCRIPT

anticoagulants.

analyzed using the Kaplan-Meier method.

RI PT

Figure 4. Cumulative probability of mortality according to risk factors other than GIB

SC

A) Age; B) BMI; C) comorbidity index; D) use of non-steroidal anti-inflammatory

AC C

EP

TE D

M AN U

drugs ; E) use of low-dose aspirin; and F) use of non-aspirin antiplatelet drugs.

25

ACCEPTED MANUSCRIPT

Table 1. Baseline characteristics of patients with gastrointestinal bleeding (GIB) (n=522) and matched non-GIB cohort (n=1,044) Non-GIB cohort

(n=522)

(n=1,044)

Age ≥ 60 y

410 (78.5)

821 (78.6)

0.965

Sex, male

330 (63.2)

682 (65.3)

0.411

BMI >25

134 (25.7)

Past history of thromboembolism Charlson comorbidity index

CHA2DS2-VASc score

AC C

Drug use

Antithrombotic drug use NSAID use

NSAID use continued NSAID use discontinued LDA use

SC 299 (28.6)

0.968

2.0±2.2

1.7±1.9

0.002

259 (49.6)

421 (40.3)

< 0.001

2.5±1.8

2.2±1.7

0.006

344 (65.9)

723 (69.3)

0.180

157 (30.1)

294 (29.1)

0.702

108 (20.7)

71 (6.8)

< 0.001

13 (12.0)

64 (90.1)

95 (88.0)

7 (9.9)

< 0.001

114 (21.8)

193 (18.5)

0.115

EP

CHA2DS2-VASc score ≥ 2

0.152

149 (28.5)

TE D

Charlson comorbidity index ≥2

304 (29.1)

M AN U

Past-hisotry or comorbidities

p

RI PT

GIB cohort

26

ACCEPTED MANUSCRIPT

Non-LDA antiplatelet use Non-LDA antiplatelet continued Non-LDA antiplatelet discontinued

137 (71.0)

45 (39.5)

56 (29.0)

0.060

41 (7.9)

106 (10.2)

0.141

27 (65.9)

92 (86.8)

14 (34.2)

14 (13.2)

47 (9.0)

Anticoagulant use

Anticoagulant discontinued

94 (9.0)

9 (19.2)

0.004 1.000

93 (98.9)

M AN U

38 (80.9)

Anticoagulant continued

RI PT

LDA use discontinued

69 (60.5)

SC

LDA use continued

1 (1.1)

< 0.001

Abbreviations: BMI, body mass index; NSAIDs, non-steroidal anti-inflammatory drugs;

AC C

EP

TE D

LDA, low-dose aspirin.

27

ACCEPTED MANUSCRIPT

Table 2. Risk of thromboembolism and mortality in patients with GIB and matched non-GIB cohort (n=1,566) GIB /

Unadjusted

Adjusted HR*

Non-GIB

HR (95%CI)

P

<0.001 <0.001 6.7 (3.9-11.5)

<0.001

<0.001 <0.001 8.8 (3.6-21.5)

<0.001

<0.001 <0.001 7.3 (3.1-17.5)

<0.001

(n=522/

60/ 25

5.3 (3.3-8.5)

Cerebrovascular events

26/ 8

7.0 (3.1-15.5)

Cardiovascular events

21/ 11

4.5 (2.2-9.3)

Other thromboembolic events

13/ 6

Death

83/ 86

All subjects (competing risk

GIB /

analysis)

Non-GIB

M AN U

All thromboembolic events

4.5 (1.7-12.0)

0.003

2.1 (1.6-2.9)

<0.001 0.003

0.003

Unadjusted

SHR (95%CI) p

TE D

(95%CI)

p

SC

1,044)

p

RI PT

All subjects

3.5 (1.1-11.3)

0.003

1.9 (1.2-2.7)

0.001

Adjusted p

(n=522/

SHR*

P

(95%CI)

1,044) 60/ 25

5.0 (3.1-8.0)

<0.001 <0.001 6.1 (3.6-10.6)

<0.001

26/ 8

6.6 (3.0-14.5)

<0.001 <0.001 8.2 (3.2-20.6)

<0.001

Cardiovascular events

21/ 11

4.1 (1.9-8.8)

<0.001 <0.001 6.6 (2.8-15.2)

<0.001

Other thromboembolic events

13/ 6

4.3 (1.6-11.5)

0.004

Antithrombotic drug users

GIB /

Unadjusted

(n=451)

Non-GIB

HR (95%CI)

EP

All thromboembolic events

AC C

Cerebrovascular events

0.042

3.3 (1.0-11.0)

0.051

Adjusted HR* p

p

(95%CI)

P

(n=157/ 294)

28

ACCEPTED MANUSCRIPT

27/ 11

5.9 (2.9-11.9)

<0.001 <0.001 7.3 (3.2-16.4)

Cerebrovascular events

11/ 2

14.1 (3.1-64.4)

0.001

0.001

11.9 (2.3-60.5)

0.003

Cardiovascular event

9/ 7

3.2 (1.2-8.6)

0.022

0.002

9.3 (2.7-32.2)

<0.001

Other thromboembolic events

7/ 2

7.3 (1.5-35.2)

0.013

0.116

4.7 (0.7-32.2)

0.113

Death

24/ 14

3.9 (2.0-7.5)

<0.001 <0.001 4.2 (1.9-9.1)

Antithrombotic drug

GIB /

non-users

Non-GIB

Unadjusted

(n=1,080)

(n=365/

HR (95%CI)

33/ 13

Cerebrovascular events

15/ 6

Cardiovascular events

12/ 3

Other thromboembolic events

6/ 4 59/ 72

Death

p

SC

p

(95%CI)

p

<0.001 <0.001 6.5 (3.0-13.9)

<0.001

4.4 (1.7-11.7)

0.003

<0.001 8.2 (2.7-24.7)

<0.001

8.6 (2.4-30.5)

0.001

0.003

9.3 (2.1-41.5)

0.004

2.8 (0.7-10.5)

0.126

0.456

2.0 (0.4-11.1)

0.417

0.003

0.295

1.3 (0.8-2.1)

0.223

1.7 (1.2-2.4)

AC C

EP

Abbreviations: NA, not applicable; SHR, subdistributional hazard ratio; CI, confidence interval.

<0.001

4.9 (2.5-9.4)

TE D

All thromboembolic events

<0.001

Adjusted HR*

M AN U

715)

RI PT

All thromboembolic events

Note: Thromboembolic events include acute coronary syndrome (n=32), stroke (n=32), transient ischemic attack (n=2), pulmonary embolism (n=11), deep-vein thrombosis (n=6), and arterial thrombosis (n=2). *We adjusted for matching factors, propensity score for GIB, plus 7 factors (hypertension, diabetes, dyslipidemia, cerebrovascular disease history, acute coronary syndrome history, chronic kidney disease, and liver

29

ACCEPTED MANUSCRIPT

AC C

EP

TE D

M AN U

SC

RI PT

cirrhosis).

30

AC C

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

AC C

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

AC C

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

AC C

EP

TE D

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT