Acute gastrointestinal bleeding (GIB) is linked to increased morbidity, mortality, and healthcare expenditures.¹ The incidence of GIB has risen with the increased use of anticoagulant and nonsteroidal anti-inflammatory drugs (NSAIDs). In randomized trials of anticoagulant, GIB was the most commonly observed complication of anticoagulant users.² Among anticoagulant users, the main cause of upper GIB is gastroduodenal ulcers, whereas the main causes of lower GIB are mainly diverticular bleeding, followed by bleeding from angiodysplasia and hemorrhoids.² The risk of GIB has more increased with the simultaneous use of anticoagulant and NSAIDs.³ In a post hoc analysis of 2,279 patients, major GIB was significantly increased with the combotherapy of anticoagulant and NSAIDs than anticoagulant monotherapy.⁴ In patients with atrial fibrillation taking anticoagulant, however, the risk of GIB was not significantly different between anticoagulant and NSAIDs combotherapy and anticoagulant monotherapy.⁵ Therefore, further researches are needed to explore the risk of GIB for the combined use of NSAIDs in patients taking anticoagulant. Moreover, only a few Asian patients were included in previous clinical trials to evaluate the risk of GIB among anticoagulant users.⁶ The risk of GIB may differ between Western and Asian patients because there are substantial variations⁷ in the metabolism of warfarin and aspirin as well as the prevalence of Helicobacter pylori infection.⁸ However, few data have been reported from Asian populations regarding the risk of GIB associated with anticoagulant and NSAIDs. Given the increasing use of anticoagulant and NSAIDs, it is crucial to evaluate the risk of GIB associated with the combined use of these medications.

This study aimed to compare the risk of GIB in anticoagulant and NSAIDs combousers versus anticoagulant monousers in real-world practice using a “common data model (CDM).”

1. Data source and ethics

We used CDM data converted from electronic health record data of three hospitals: Kyung Hee University Hospital at Gangdong (n=887,370), Kangdong Sacred Heart Hospital (n=1,194,685), and Kyung Hee University Medical Center (n=1,168,640). Table 1 shows the baseline data for the number of patients, study period, and use of anticoagulant and NSAIDs before and after propensity score matching (PSM). To minimize mapping errors and information loss, consensus on concept mapping was achieved through review by all authors. This study was approved by the Institutional Review Board of Kyung Hee University Hospital at Gangdong (IRB number: KHNMC 2023–10–039). The need for written informed consent was waived for this study as it was based on the CDM data without data security issues.

Table 1 . Summary of Data Source.

Database	Total number of participants	Study period	Before propensity score matching			After propensity score matching
			Anticoagulant + NSAIDs users	Anticoagulant users		Anticoagulant + NSAIDs users	Anticoagulant users
Kyung Hee University Hospital at Gangdong	887,370	2006/06–2023/03	2,670	1,255		800	800
Kangdong Sacred Heart Hospital	1,194,685	1986/11–2022/09	3,337	957		735	735
Kyung Hee University Medical Center	1,168,640	2008/01–2022/02	3,914	1,831		1,416	1,416
Total			9,921	4,043		2,951	2,951

NSAIDs, nonsteroidal anti-inflammatory drugs..

As described in prior CDM-based studies,^9-12 CDM comprises standardized data with the uniform structure and uses the same analysis program across multiple electronic health records. In CDM research, raw data from each institution is not extracted, but only the statistical results analyzed within the institution are extracted without patient data. As source variable terms are typically expressed in nonstandard terminology, these terms undergo standardization into concepts with mapping, a crucial process in facilitating patient data standardization.^9,10 In CDM, the meaning of each content is represented using standard concepts, and concepts associated with content are stored with their concept-ids as foreign keys to the concept table in a standardized vocabulary. CDM-based research has advantages to maintain patient confidentiality, privacy, and security. CDM data quality from electronic health records has been verified in many previous studies.¹³

2. Study design and cohort definition

This study was a retrospective, observational, and comparative cohort study of new anticoagulant users with or without NSAIDs aged >18 years from the three hospitals (Fig. 1). NSAID users were defined as any individuals using NSAIDs or aspirin. The target cohort was defined as anticoagulant and NSAIDs combousers for >90 consecutive days. A comparative cohort was defined as anticoagulant monousers for >90 consecutive days. Participants who fulfilled any of the following criteria were excluded from the cohort: (1) history of GIB before cohort entry; (2) NSAIDs use in the comparative cohort; and (3) age <18 years. We also excluded patients with chronic liver disease (concept_id 4212540), chronic kidney disease (concept_id 46271022), renal dialysis (concept_id 4146536), or kidney transplantation (concept_id 4322471). All anticoagulant, NSAIDs, and aspirin available in the South Korean market were included in the analysis. Eligible patients were defined as those who were continuously exposed to the study medication for the duration of their prescription from the index date until a gap of 30 days, the occurrence of GIB, hospital discharge, or treatment termination. Both cohorts were censored or terminated at the time of the GIB occurrence. In this study, the “time-at-risk” was defined from 1 day after the index date to the end of the observation, during which the patients were followed up and the longest observation period for any patient was 455 days. The end of patient observation included the following events: no follow-up prescription of the cohort medications that were first prescribed at the index date 180 days after the last prescription; cohort medication use that differed from the medication that was initially prescribed at the index date; death of the patient; or end of data availability. For validation purposes, the “time-at-risk” endpoint was also assessed with 455 days after cohort start and with 30 days after cohort end.

Figure 1. Study flowchart of the cohort data from the three hospitals. Patients included in both cohorts with a history of gastrointestinal bleeding who did not have at least 1 day at risk and were not matched on propensity scoring were excluded. Finally, 2,951 propensity-matched pairs of patients treated with anticoagulant with (target cohort) or without NSAIDs (comparative cohort) were included. NSAIDs, nonsteroidal anti-inflammatory drugs; First cohort, new user of study medication; No prior outcome, no gastrointestinal bleeding before the cohort entry..

Outcome cohort (GIB) was defined as diagnosis of gastrointestinal hemorrhage (concept_id 192671, 4100660), upper GIB (concept_id 4291649, 193250, 4318535) or lower GIB (concept_id 4338544, 4318536, 4318829), and/or treatment with endoscopic control of bleeding (concept_id 2109184, 2108900, 44784306, 2109100) identified by SNOMED-CT codes for relevant primary and secondary discharge diagnoses. Coding for GIB is much more accurate than other disease coding, and widely used in recent studies without an operational definition.¹⁴ To compare the risk of GIB, we matched the target and comparative cohorts using 1:1 PSM generated using logistic regression estimation with age group, sex, and comorbidity score as fixed independent variables. Each patient in both cohorts were matched with a similar propensity score based on nearest-neighbor matching without replacement using a caliper width equal to 0.2 of the standard deviation of the logit of the propensity score. The standardized mean difference was used to assess the imbalance between the baseline characteristics after matching, and values of <10% were considered acceptable.¹⁵

3. Covariates

The CDM provides regularized logistic regression models for extensive PSM.^10-12,15 A comprehensive set of 237 covariates were used for PSM between the target and comparative cohorts, including age, sex, index year, Charlson Comorbidity Index, comorbidities, and prescribed drugs 365 days before the index date (Table 2).^10-12,15 In this study, we showed covariates in >5% of the total patients before PSM. General medical history included hypertensive disorder, hyperlipidemia, diabetes mellitus, gastroesophageal reflux disease, pneumonia, chronic obstructive lung disease, and neoplasms. Cardiovascular disease history included heart disease, atrial fibrillation, ischemic heart disease, venous thrombosis, pulmonary embolism, cerebrovascular disease, and coronary arteriosclerosis. The prescribed drugs analyzed included antithrombotic agents, drugs for acid-related disorders, beta-blocking agents, diuretics, lipid-modifying agents, agents acting on the renin-angiotensin system, opioids, psycholeptics, antidepressants, drugs for obstructive airway diseases, and drugs used for diabetes. The overall comorbidity burden was assessed using the Charlson Comorbidity Index and CHADS₂ score.

Table 2 . Distribution of Baseline Characteristics in the Overall Population from the Three Hospitals before and after Propensity Score Matching.

	Before propensity score matching				After propensity score matching
	Anticoagulant + NSAIDs users (n=9,921)	Anticoagulant users (n=4,043)	SMD		Anticoagulant + NSAIDs users (n=2,951)	Anticoagulant users (n=2,951)	SMD
Age group, No. (%)
<40 yr	306 (3.1)	103 (2.5)	0.032		94 (3.2)	90 (3.0)	0.008
40–49 yr	583 (5.9)	182 (4.5)	0.062		169 (5.7)	140 (4.7)	0.044
50–59 yr	1,324 (13.3)	486 (12.0)	0.040		388 (13.1)	375 (12.7)	0.013
60–69 yr	2,453 (24.7)	1,037 (25.6)	0.021		783 (26.5)	757 (25.7)	0.020
70–79 yr	3,477 (35.0)	1,415 (35.0)	0.001		997 (33.8)	1,020 (34.6)	0.016
≥80 yr	1,778 (17.9)	820 (20.3)	0.060		520 (17.6)	569 (19.3)	0.043
Female sex, No. (%)	4,768 (48.1)	1,766 (43.7)	0.088		1,297 (44.0)	1,333 (45.2)	0.025
General medical history, No. (%)
Hypertensive disorder	3,398 (34.3)	1,432 (35.4)	0.025		1,050 (35.6)	1,073 (36.4)	0.016
Hyperlipidemia	1,843 (18.6)	837 (20.7)	0.054		611 (20.7)	631 (21.4)	0.017
Diabetes mellitus	1,227 (12.4)	437 (10.8)	0.049		334 (11.3)	348 (11.8)	0.015
Gastroesophageal reflux disease	599 (6.0)	190 (4.7)	0.059		116 (3.9)	147 (5.0)	0.051
Pneumonia	474 (4.8)	145 (3.6)	0.060		97 (3.3)	121 (4.1)	0.043
Chronic obstructive lung disease	287 (2.9)	125 (3.1)	0.012		70 (2.4)	90 (3.0)	0.042
Neoplasm history, No. (%)	1,514 (15.3)	347 (8.6)	0.207		305 (10.3)	297 (10.1)	0.009
Cardiovascular disease history, No. (%)
Heart disease	6,234 (62.8)	2,972 (73.5)	0.231		2,095 (71.0)	2,058 (69.7)	0.027
Atrial fibrillation	2,385 (24.0)	1,555 (38.5)	0.315		1,002 (34.0)	965 (32.7)	0.027
Ischemic heart disease	1,627 (16.4)	403 (10.0)	0.191		325 (11.0)	321 (10.9)	0.004
Venous thrombosis	1,017 (10.3)	314 (7.8)	0.087		267 (9.0)	272 (9.2)	0.006
Pulmonary embolism	841 (8.5)	256 (6.3)	0.082		215 (7.3)	211 (7.2)	0.005
Cerebrovascular disease	1,145 (11.5)	281 (7.0)	0.159		227 (7.7)	239 (8.1)	0.015
Coronary arteriosclerosis	131 (1.3)	31 (0.8)	0.055		29 (1.0)	21 (0.7)	0.030
Medication use, No. (%)
Antithrombotic agents	6,343 (63.9)	1,529 (37.8)	0.541		1,154 (39.1)	1,319 (44.7)	0.114
Drugs for acid-related disorders	5,795 (58.4)	1,386 (34.3)	0.499		1,019 (34.5)	1,165 (39.5)	0.103
Beta blocking agents	4,464 (45.0)	1,659 (41.0)	0.080		1,196 (40.5)	1,177 (39.9)	0.013
Diuretics	5,080 (51.2)	1,480 (36.6)	0.297		1,084 (36.7)	1,168 (39.6)	0.059
Lipid modifying agents	4,640 (46.8)	1,323 (32.7)	0.290		997 (33.8)	1,030 (34.9)	0.024
Agents acting on the RAS	4,576 (46.1)	1,578 (39.0)	0.144		1,184 (40.1)	1,187 (40.2)	0.002
Opioids	4,299 (43.3)	893 (22.1)	0.465		669 (22.7)	786 (26.6)	0.092
Antibacterials for systemic use	4,135 (41.7)	934 (23.1)	0.405		711 (24.1)	798 (27.0)	0.068
Calcium channel blockers	4,698 (47.4)	1,502 (37.2)	0.208		1,067 (36.2)	1,144 (38.8)	0.054
Psycholeptics	3,727 (37.6)	869 (21.5)	0.358		620 (21.0)	719 (24.4)	0.080
Antidepressants	2,491 (25.1)	500 (12.4)	0.331		350 (11.9)	437 (14.8)	0.087
Drugs for obstructive airway diseases	2,997 (30.2)	616 (15.2)	0.363		460 (15.6)	503 (17.0)	0.039
Drugs used in diabetes	2,629 (26.5)	684 (16.9)	0.234		493 (16.7)	557 (18.9)	0.057

NSAIDs, nonsteroidal anti-inflammatory drugs; SMD, standardized mean difference; RAS, renin-angiotensin system..

4. Statistical analysis

Distributed research platform, a health big-data platform based on CDM supported by the Korean National Project, was used in this study.¹¹ Categorical variables are presented as number (percentage), while normally distributed continuous variables are expressed as mean (standard deviation). Initial analysis utilized the tools for the CDM analysis embedded in the ATLAS platform (version 2.12.0). The Cox model analysis and Kaplan-Meier estimation were supported by R package version 4.3.1 (R Foundation for Statistical Computing, Vienna, Austria). The Cox proportional hazards models were used to compare GIB in the matched cohorts, calculating hazard ratio (HR) and 95% confidence interval (CI). The Kaplan-Meier curves depicted the percentage of event-free patients over time using log-rank tests of the statistical significance of the differences between two cohorts. Only the first event was included in the time-to-event analyses. A significance level of p<0.05 (two-sided) was applied to all tests. We used 0.2 of the pooled standard deviation of the logit of the propensity score as the caliper width for PSM. Statistical heterogeneity was assessed using chi-square and I² statistics. A fixed-effects model was used in the absence of heterogeneity (p<0.05, I²>50%); otherwise, a random-effects model was used.

Before the PSM, a total of 9,921 patients in the target cohort (anticoagulant and NSAIDs combotherapy) and 4,043 in the comparative cohort (anticoagulant monotherapy) met the eligibility criteria. Patients included in both cohorts who had a history of GIB, were not at least 1 day at risk, and did not match the propensity score were excluded (Fig. 1). Finally, 2,951 propensity-matched pairs of the target and comparative cohorts were included. After PSM, each standardized mean difference was lower than 0.1 PSM regarding age group, sex, medical history, cardiovascular disease history, and medication use.

1. Baseline characteristics of study subjects

Table 2 shows the baseline characteristics of the study population. After PSM, the proportion of anticoagulant users with or without NSAIDs consistently increased up to 70–79 years; however, decreased after 80 years. After PSM, common general medical histories were hypertensive disorders, followed by hyperlipidemia and diabetes mellitus. The prevalent cardiovascular disease was heart disease, followed by atrial fibrillation and ischemic heart disease. In our study, medications that potentially impact the risk of GIB, such as antithrombotic agents, anti-inflammatory drugs, antirheumatic drugs, and drugs for acid-related disorders, were well-matched. The most frequently prescribed medications before cohort entry were “beta-blockers” and “antithrombotic agents,” followed by “drugs for acid-related disorders” (34.5% vs 39.5% of the target and comparative cohorts, respectively).

2. Risk of GIB between anticoagulant/NSAIDs combotherapy versus anticoagulant monotherapy

Cox proportional hazards analyses compared the risk of GIB between anticoagulant and NSAIDs combotherapy versus anticoagulant monotherapy after PSM (Fig. 2). The risk of GIB was significantly higher in the anticoagulant and NSAIDs combotherapy group than the anticoagulant monotherapy group (HR, 1.66; 95% CI, 1.30 to 2.12; p<0.001). When the “time-at-risk” end was reevaluated within 455 days after the cohort start, the risk of GIB was still significantly higher in the anticoagulant and NSAIDs combotherapy group than anticoagulant monotherapy group (HR, 2.27; 95% CI, 1.37 to 3.77; p=0.001) (Supplementary Fig. 1A). When the “time-at-risk” end was modified as 30 days after the cohort end, a similar result was achieved (HR, 1.39; 95% CI, 1.01 to 1.92; p=0.041) (Supplementary Fig. 1B).

Figure 2. Kaplan-Meier plots of gastrointestinal bleeding in patients treated with anticoagulant with and without nonsteroidal anti-inflammatory drugs (NSAIDs) use. The risk of gastrointestinal bleeding was significantly higher in anticoagulant and NSAIDs combination therapy recipients than in anticoagulant monousers (hazard ratio, 1.66; 95% confidence interval, 1.30 to 2.12; p<0.001).

3. Risk of GIB between anticoagulant/NSAIDs combotherapy versus anticoagulant monotherapy in older patients

Fig. 3 shows the Kaplan-Meier plots for GIB between anticoagulant and NSAIDs combotherapy versus anticoagulant monotherapy in elderly patients. Anticoagulant and NSAIDs combotherapy were associated with significantly higher risk of GIB compared to anticoagulant monotherapy in elderly patients >65 years (HR, 1.53; 95% CI, 1.15 to 2.03; p=0.003) (Fig. 3A) as well as in elderly patients >75 years (HR, 1.89; 95% CI, 1.23 to 2.90; p=0.003) (Fig. 3B).

Figure 3. Kaplan-Meier plots for gastrointestinal bleeding in older patients treated with anticoagulant with and without nonsteroidal anti-inflammatory drugs (NSAIDs). (A) Older adults >65 years and (B) older adults >75 years. Combotherapy was significantly associated with gastrointestinal bleeding compared to anticoagulant monotherapy in patients aged >65 years (A: hazard ratio [HR], 1.53; 95% confidence interval [CI], 1.15 to 2.03; p=0.003) as well as in patients aged >75 years (B: HR, 1.89; 95% CI, 1.23 to 2.90; p=0.003).

4. Sex difference of GIB risk for anticoagulant/NSAIDs combotherapy versus anticoagulant monotherapy

Fig. 4 shows the Kaplan-Meier plots for GIB between anticoagulant and NSAIDs combotherapy versus anticoagulant monotherapy according to sex. Anticoagulant and NSAIDs combotherapy was significantly associated with higher risk of GIB compared to anticoagulant monotherapy in both male cohort (HR, 1.48; 95% CI, 1.07 to 2.05; p=0.016) (Fig. 4A) and female cohort (HR, 1.64; 95% CI, 1.12 to 2.40; p=0.010) (Fig. 4B).

Figure 4. Kaplan-Meier plots of gastrointestinal bleeding in patients treated with anticoagulant with and without nonsteroidal anti-inflammatory drugs (NSAIDs) stratified by sex. (A) Male cohort and (B) female cohort. Combotherapy was significantly associated with gastrointestinal bleeding compared to anticoagulant monotherapy in the male cohort (A: hazard ratio [HR], 1.48; 95% confidence interval [CI], 1.07 to 2.05; p=0.016) and female cohort (B: HR, 1.64; 95% CI, 1.12 to 2.40; p=0.010).

5. Risk of GIB between anticoagulant/NSAIDs combotherapy versus anticoagulant monotherapy according to upper and lower GIB

New independent upper and lower GIB cohorts were made with the same methods and variable definitions with the original GIB cohort (Supplementary Fig. 2). For the upper and lower GIB, duodenal bleeding was classified as upper GIB, while jejunal and ileal bleeding was classified as lower GIB. The risk of upper GIB (Supplementary Fig. 2A) was significantly higher in the anticoagulant/NSAIDs combotherapy group than the anticoagulant monotherapy group (HR, 1.46; 95% CI, 1.01 to 2.11; p=0.043), but the risk of lower GIB (Supplementary Fig. 2B) did not reach a statistical significance for the difference between two groups (HR, 1.36; 95% CI, 0.99 to 1.85; p=0.054).

To the best of our knowledge, this is the first distributed network analysis using CDM data to compare the risk of GIB in anticoagulant and NSAIDs combotherapy versus anticoagulant monotherapy. In this analysis of real-world database, the risk of GIB was 1.7 times higher for anticoagulant and NSAIDs combotherapy than anticoagulant monotherapy (HR, 1.66; 95% CI, 1.30 to 2.12; p<0.001). Our finding was duplicated when the “time-at-risk” end was reevaluated within 455 days after the cohort start and modified as 30 days after the cohort end. We also found that anticoagulant and NSAIDs combotherapy was associated with significantly higher risk of GIB compared to anticoagulant monotherapy in older patients >65 years as well as in older patients >75 years (HR, 1.89; 95% CI, 1.23 to 2.90; p=0.003). Moreover, higher risks of GIB in anticoagulant and NSAIDs combotherapy than anticoagulant monotherapy were duplicated in both sex cohorts. Our CDM-based analysis supports the Asian evidence that the risk of GIB is significantly increased with anticoagulant and NSAIDs combotherapy versus anticoagulant monotherapy.^3,4

It is well-known that anticoagulant use increases the risk of GIB, especially in the elderly population,¹⁵ and anticoagulant and NSAIDs combotherapy may further increase this risk. As NSAIDs use increases the risk of GIB by more than fourfold,¹⁶ the concurrent use of anticoagulant and NSAIDs may naturally escalate the risk of GIB. In a meta-analysis of 11 studies,¹⁷ the concurrent use of warfarin and NSAIDs significantly increased the risk of GIB by 2.0 folds compared with warfarin use alone, but limited by the high degree of heterogeneity among the included studies. In a systematic review, meta-analysis, and meta-regression,¹⁸ the odds ratio for GIB in high-risk patients of arterial thromboembolism was 1.27 (95% CI, 0.84 to 1.92), showing no statistically significant difference. In a cohort of 27,395 patients with coronary or peripheral artery disease, anticoagulant (rivaroxaban) and aspirin combotherapy increased the major GIB risk by 2.5 folds (HR, 2.15; 95% CI, 1.60 to 2.89; p<0.001) compared with aspirin use alone.¹⁹ However, a recent trial investigating the risk of GIB in patients with atrial fibrillation taking anticoagulant (apixaban or warfarin) and NSAIDs showed different results.⁵ Incident NSAIDs use combined with anticoagulant use was not associated with a higher risk of GIB (HR, 1.08; 95% CI, 0.64 to 1.82), and the number of GIB events was low.⁵ In this study, statistical power for the detection of a difference in GIB may have lacked due to low GIB events in both groups.

The risk of GIB was reduced by the combined use of acid suppressants in anticoagulant users.²⁰ In a meta-analysis of five observational studies, the use of proton pump inhibitors also reduced the risk of GIB in anticoagulant users, with a common relative risk of 0.67 (95% CI, 0.61 to 0.74).²¹ In a real-world practice, therefore, anticoagulant or NSAIDs are often prescribed along with acid suppressants to reduce the risk of GIB. In our study, for example, 34.5% of the anticoagulant and NSAIDs combotherapy group and 39.5% of the anticoagulant monotherapy group were prescribed “drugs for acid-related disorders” (i.e., acid suppressants in CDM) before cohort entry. Our study showed that many anticoagulant users concurrently used acid suppressants for the prevention of GIB. However, the combined use of acid suppressants may not decrease the risk of GIB in patients taking anticoagulant and NSAIDs combotherapy.

The risk of GIB in anticoagulant users in Asian countries may differ from that in Western countries,^1,4,5,15 as the prevalence of H. pylori infection in South Korea was approximately 78%,^22,23 and H. pylori infection increases the risk of GIB in NSAIDs users.²⁴ Unfortunately, H. pylori infection rate and the GIB risk according to H. pylori infection could not be assessed in this study. No study has compared the risk of GIB in patients receiving anticoagulant and NSAIDs combotherapy in Asian versus Western countries. In a multinational study from Europe, Asia, and the United States, the HR of GIB was 1.8 (95% CI, 1.35 to 2.43), which was similar to our HR of GIB of 1.6 (95% CI, 1.30 to 2.12).⁴ Therefore, further research is needed to determine any difference in GIB risk between Asian and Western countries.

Our study has several strengths. First, previous studies compared the risk of GIB between anticoagulant and NSAIDs combotherapy and anticoagulant monotherapy for specific indications, such as atrial fibrillation or venous thrombosis.⁴ Our, study showed the risk of GIB in anticoagulant users with all real-world indications. However, the risk of GIB in the real world should be carefully interpreted as we excluded some patients with exclusion criteria. Second, our target and comparative cohorts were well-matched after extensive PSM for any medications that may alter the risk of GIB, such as acid suppressants. Although this study was a retrospective analysis, we attempted to overcome the limitations of confounding bias with PSM. Clinical trials of GIB usually exclude potential medications that may affect GIB, such as antithrombotic agents, anti-inflammatory drugs, antirheumatic drugs, and acid suppressants. Consequently, they do not reflect the risk of GIB in real-world medical practice. To address this, we adjusted for these drugs without excluding them. Third, we applied a new user design and limited the included subjects to those who were first exposed to either the target or comparative medications.

Despite its strengths, this study had several limitations. First, the baseline characteristics of the three hospitals may be different because of different prescription patterns in each hospital. However, these effects may not have been significant because this study was based on standardized data with the same structure from the CDM database. Second, we designated the comparative group as anticoagulant users to enhance the comparability between the target and comparative cohorts. Therefore, the comparative group may not reflect the general healthy population. However, healthy controls may be inappropriate for analyzing GIB risk because anticoagulant users usually have comorbidities. Third, the risk of GIB may have been affected by previous anticoagulant use or dosage. However, patients who had used anticoagulant before cohort entry were excluded, and anticoagulant are used at a fixed (direct oral anticoagulant) or an optimized dose (warfarin) titrating bleeding tendency. Therefore, the risk of GIB may be less affected by previous anticoagulant use or dose. Finally, our study may be limited by the participation of only three hospitals, which may not adequately represent the general population. It would be beneficial to analyze a more diverse range of healthcare institutions and ethnicities in future studies.

In conclusion, the concurrent use of anticoagulant and NSAIDs is associated with higher risk of GIB compared to anticoagulant monotherapy. In addition, the risk of GIB was much higher in individuals aged >75 years. Therefore, the combined use of anticoagulant and NSAIDs should be carefully monitored for the risk of GIB, especially in elderly population.

We thank Myoungsuk Kim (Department of Big Data Center, Kyung Hee University Hospital at Gangdong, Seoul, Korea) for assistance with the data and statistical analyses.

No potential conflict of interest relevant to this article was reported.

Study concept and design: M.L., J.M.C. Data acquisition: M.L. Data analysis and interpretation: M.L. Manuscript drafting: M.L., J.M.C. Critical revision of the manuscript for important intellectual content: J.M.C. Statistical analysis: M.L. Study supervision: J.M.C. Approval of final manuscript: all authors.

Supplementary materials can be accessed at https://doi.org/10.5009/gnl230541.