Optimal strategies for mitigating gastrointestinal bleeding in patients receiving antiplatelet therapy: Real-world study

Abstract

BACKGROUND

Aspirin and clopidogrel are cornerstone antiplatelet agents for cardiovascular prevention but substantially increase the risk of gastrointestinal (GI) bleeding. Although proton pump inhibitors (PPIs) are recommended for gastroprotection in high-risk patients, the comparative efficacy of alternative protective agents (PAs), including H2 receptor antagonists (H2RAs) and rebamipide, has not yet been fully elucidated in real-world clinical settings. Rebamipide is a mucoprotective agent by the pleiotropic mechanism such as enhancing PG synthesis, anti-oxidation, antiinflammation, promoting mucosal repair, and increasing mucus production. Evidence is particularly limited regarding combination strategies, risk stratification, and agent-specific benefits of antiplatelet regimens.

AIM

To identify the optimal gastroprotective strategy for preventing significant hemoglobin drop (SHD) in patients exposed to mucosal-damaging antiplatelet agents.

METHODS

We conducted a retrospective cohort study consisting of 98404 patients who received aspirin and/or clopidogrel at Samsung Medical Center between 2002 and 2019. The primary endpoint was SHD, defined as a hemoglobin decline > 2 g/dL. We performed a self-controlled case series (SCCS) analysis of 3649 patients and matched landmark analysis of 14008 patients. Patients were stratified into high- and low-risk groups for GI bleeding based on established clinical criteria, including age, prior GI events, concomitant medications, and comorbidity burden.

RESULTS

In the SCCS analysis, most PA strategies use reduced SHD risk compared with mucosal damaging agent alone. For single PA use, PPI showed larger protective effect [adjusted incidence rate ratio (IRR) 0.42; 95% confidence interval (CI): 0.34-0.53], compared to rebamipide (IRR 0.68; 95%CI: 0.50-0.53) and H2RA (IRR 0.75; 95%CI: 0.63-0.90). Concomitant PAs strategies resulted in a greater reduction in SHD risk than single PA strategies. Consistent results were observed in the high-risk subgroups, whereas no significant benefit was observed in low-risk patients. Additional matched landmark analyses confirmed protective effects of PA monotherapies in aspirin-alone users, while no agent significantly reduced SHD in clopidogrel-alone users.

CONCLUSION

Combination therapy with rebamipide and PPIs most effectively prevented hemoglobin decline in high-risk antiplatelet users, while PPI monotherapy sufficed for aspirin alone. Findings support individualized, risk-based gastroprotection strategies.

Key Words: Gastrointestinal bleeding; Aspirin; Clopidogrel; Proton pump inhibitor; Rebamipide; Antiplatelet therapy; High risk; Low risk; Protective agents; Real-world study

Core Tip: Aspirin and clopidogrel increase the risk of gastrointestinal bleeding; however, the comparative effectiveness of these mucosal protective agents remains unclear. In a large real-world cohort, we found that proton pump inhibitor monotherapy provided substantial protection, whereas combination therapy with rebamipide offered the greatest risk reduction, particularly in high-risk patients. No significant benefit was observed in clopidogrel-alone users or low-risk populations. These findings support a personalized, risk-stratified approach to gastroprotection, and suggest that combination mucosal protective strategies may provide incremental benefits beyond standard regimens.

INTRODUCTION

Aspirin and clopidogrel are antiplatelet agents that play pivotal roles in the prevention and management of atherothrombotic events, particularly in patients with coronary artery disease, peripheral artery disease and cerebrovascular diseases[1-5]. Aspirin exerts its antiplatelet effect by irreversibly inhibiting cyclooxygenase-1, thereby suppressing thromboxane A₂ production and platelet aggregation. Clopidogrel, a thienopyridine prodrug, selectively inhibits the P2Y₁₂ subtype of adenosine diphosphate receptor on the platelet surface, resulting in reduced platelet activation and aggregation. Combination therapy with aspirin and clopidogrel, often referred to as dual antiplatelet therapy (DAPT), is one of the standards of care for patients undergoing percutaneous coronary intervention with stent placement and for those with acute coronary syndromes. Although DAPT significantly reduces the risk of ischemic complications, it is associated with an increased risk of gastrointestinal (GI) bleeding and other hemorrhagic events. According to previous studies, the incidence of GI bleeding has been reported to be 0.7%-1.3% with aspirin use, approximately 1.3% with clopidogrel use, and 1.2%-2% among patients receiving DAPT[6-10]. Randomized trials comparing different antiplatelet agents have reported varying mortality rates[8,11,12].

Given the clinical significance of GI bleeding, the current guidelines emphasize risk stratification and preventive strategies, including the co-administration of proton pump inhibitors (PPIs) in high-risk patients[13]. However, the efficacy and safety of alternative gastroprotective strategies, such as H2 receptor antagonists (H2RAs) and cytoprotective agents, remain to be investigated. In addition, there are several adverse events associated with the long-term administration of PPIs, H2RAs, antacids, and mucosal damage-preventive agents (PAs). For example, PPI has been associated with several adverse outcomes, including impaired absorption of essential nutrients such as calcium, iron, magnesium, and vitamin B₁₂. Additionally, its use may result in side effects such as nausea, constipation, and nasopharyngitis and has been linked to serious complications, including pneumonia, Clostridium difficile infection, cardiovascular events, and chronic kidney disease[14]. Rebamipide is a mucosal-damage PA used in the management of peptic ulcer disease and gastroesophageal reflux disease[15]. Rebamipide is generally better tolerated and is associated with a lower incidence of adverse effects than conventional anti-ulcer therapies. Its mechanism involves the stimulation of prostaglandin synthesis within the gastric mucosa, thereby promoting accelerated and qualitatively improved ulcer healing[16,17]. However, the common side effects of rebamipide are GI complications such as constipation, bloating, diarrhea, nausea, and vomiting[18]. Hypersensitivity and rash were observed in less than 1% of patients. Considering the potential adverse effects of various agents, it would be preferable to administer an agent that provides the greatest protective effect against GI bleeding.

This study aimed to identify protective agents that could effectively reduce GI injury induced by aspirin, clopidogrel, or other mucosal-damaging agents (DAs). To achieve this, we categorized patients into high- and low-risk groups according to their GI bleeding risk and performed stratified analyses.

MATERIALS AND METHODS

This retrospective cohort study included 136095 patients who received antiplatelet agents (aspirin or clopidogrel) at the Samsung Medical Center (SMC) between January 1, 2002, and December 31, 2019 (Figure 1). We excluded patients with cancer (International Classification of Diseases, 10^th Revision; C00-C97), intracerebral/intraventricular hemorrhage (I60-I62), external injuries to the head, neck, thorax, abdomen/pelvis, hip, thigh or multiple body regions (S00-S39, S70-S79, and T00-T09), advanced chronic kidney disease stage 3 or higher (N183-N185), decompensated liver cirrhosis with varices and ascites (K74, I85, and R18), bleeding disorders including hereditary factor VIII/IX deficiency and other coagulation defects (D66-D68) and those younger than 18 years of age. Analyses were conducted using a cohort of 98404 antiplatelet users.

Figure 1 Study population flowchart. ¹Significant hemoglobin drop (SHD): The occurrence of SHD event is declared, if hemoglobin (Hb) decreases by > 2 g/dL compared with the baseline Hb. SHD: Significant hemoglobin drop; ICH: Intracerebral hemorrhage; IVH: Intraventricular hemorrhage; CKD: Chronic kidney disease; PPI: Proton pump inhibitor; H2RA: H2 receptor antagonist; GI: Gastrointestinal; DA: Damaging agent; PA: Protective agent.

Data were extracted from DARWIN-C, a clinical data warehouse at SMC. The study protocol was reviewed and approved by the Institutional Review Board of SMC (SMC 2025-04-120-002). The study was conducted in accordance with the Declaration of Helsinki, and only data from patients who consented to access electronic medical records were collected.

Variables of interest

Mucosal-DAs: Mucosal-DAs are a class of drugs recognized for their potential to harm the GI mucosa. We classified DAs into three classes: (1) Antiplatelet agents, including aspirin and clopidogrel; (2) Anticoagulant agents, including Warfarin and direct oral anticoagulants (DOACs); and (3) Anti-inflammatory agents, including non-steroidal anti-inflammatory drugs (NSAIDs). In this study, DA users were defined as patients who were prescribed these medications for more than 30 days[19,20].

Mucosal damage PA: Mucosal damage-PAs include anti-ulcer medications, acid-suppressing agents, and mucosal protectants. PAs were categorized into the following classes: (1) PPIs, including lansoprazole, esomeprazole, pantoprazole, and rabeprazole; (2) Rebamipide; (3) H2RAs, including ranitidine, famotidine, cimetidine, and lafutidine; and (4) Others, such as sodium alginate, ecabet sodium hydrate, polaprezinc, and stillen (an ethanol extract of Artemisiae Argyi herba). Rebamipide was included separately in the analysis because of its frequent use at the study center, well-established mucosal protective effects, and extensively characterized mechanism of action. PA users were defined as patients who had taken PAs for at least seven days[21].

Risk classification for GI bleeding

In accordance with the 2008 and American College of Cardiology Foundation (ACCF)/American College of Gastroenterology (ACG)/American Heart Association (AHA) 2010 guidelines, patients were stratified into high- and low-risk groups for GI bleeding (Supplementary Figure 1). Individuals with a prior history of GI bleeding, or with established risk factors such as history of ulcer complication, history of ulcer disease (non-bleeding), old age (age 60 years or more); DAPT, concurrent use of anticoagulants (warfarin, DOACs, or low molecular weight heparin, corticosteroids, or NSAIDs (including aspirin); and Helicobacter pylori infection were categorized as high-risk. Patients without these factors were classified as low risk.

Outcome of interest

The primary outcome of this study was the significant hemoglobin drop (SHD), which was used to evaluate the protective efficacy of GI medications against mucosal injury. SHD was defined as a hemoglobin (Hb) reduction of > 2 g/dL from baseline, with the baseline determined as the average of the three most recent consecutive measurements[22-24].

Study design

To evaluate the protective efficacy of PAs against SHD in patients receiving DAs, we employed a self-controlled case series (SCCS) design supplemented by a retrospective cohort design for matched landmark analysis.

SCCS design

The SCCS method is a within-individual case-only design that controls for all time-invariant confounders by comparing the exposure and non-exposure periods within the same individual[25,26]. This approach eliminates the need for an independent control group while providing robust estimates of the relative risk.

In our cohort, we identified patients who meet the following conditions: (1) Received PAs as well as DAs; (2) Experienced at least one episode of SHD; and (3) Had PA prescriptions within one year before and/or after SHD occurrence (i.e., index date) (Figure 1).

Details of the SCCS design using a representative patient are illustrated in Figure 2. For each agent, we defined a continuous “prescription period” when gaps between consecutive prescriptions were less than 7 days. A 7-day “latent period” was appended after each prescription period to account for residual drug effects. The “observation period” was defined as the entire duration of continuous DA prescription periods plus the latent period (Figure 2A). Within the observation period, the PA exposure periods included all continuous durations of exposure to PAs (either a single agent or a combination of agents), plus latent periods. The remaining observation period (i.e., non-PA or non-exposure periods) served as a reference period for within-individual comparisons.

Figure 2 Schematic illustration of self-controlled case series design for a representative patient. A: Observation period is defined as the entire duration of continuous mucosal damaging agent (DA) prescription periods plus the latent period. Within the observational period, protective agent (PA)-exposure period includes all continuous durations exposed to PAs (either a single agent or combination of agents) plus latent periods. The remaining observation period (i.e., non-PA periods or non-exposure periods) is the duration of DA alone; B: Outcome period is defined as the 30-day period following any significant hemoglobin drop (SHD) events. Overlapping outcome periods due to multiple SHDs were merged into a single outcome. The start of outcome period is considered the index date; C: For each SHD event, all observation periods are restricted to one year before and after the index date. The outcome periods are removed from the analysis; D: The incidence of SHD was assessed and compared between PA-exposure and non-PA periods. Using conditional Poisson regression models, incidence rate ratios for PA-exposure period were estimated compared to non-PA period. SDH: Significant hemoglobin drop; DA: Damaging agent; PA: Protective agent; LP: Latent period.

"Outcome periods" were defined as the 30-days periods following any SHD event (Figure 2B). These outcome periods were excluded from the analysis to prevent confounding, as PAs prescribed during these periods were likely aimed at preventing further acute bleeding. Overlapping outcome periods owing to multiple SHDs were merged into a single outcome.

We excluded the periods when iron supplementation was administered because of its direct influence on Hb levels. We also excluded patients with concurrent use of PPI, rebamipide, and H2RA (n = 6); concurrent use of PPI, rebamipide, and others (n = 4); concurrent use of rebamipide and others (n = 23); concurrent use of PPI, H2RA, and others (n = 4); and concurrent use of FRBRBC and/or other PAs (n = 1852) because of possible instability in the estimation due to the small number of patients. All observation periods were restricted to one year before and after the index date to minimize potential bias from time-varying confounders (Figure 2C).

Retrospective cohort design for matched landmark analysis

To confirm the robustness of the SCCS design findings, we conducted additional matched landmark analysis. This method addresses the immortal time bias, which can occur when the exposure status is determined during follow-up rather than at baseline[27-29].

For landmark analysis, we constructed annual SHD-free subcohorts from 2003 to 2019 (Figure 3). For each sub-cohort, a landmark point was set as January 1st of each year, and the baseline period was defined as the one-year duration preceding the landmark point.

Figure 3  Schematic illustration of retrospective cohort design for the matched landmark analysis.

For each sub-cohort, the eligibility and exposure status of patients were determined during the baseline period. Eligible patients met the following conditions: (1) Alive during the baseline period; (2) Used DAs for at least 180 days; (3) Remained free of SHD within 1 month before and during the baseline period; and (4) Did not receive iron therapy. The PA-exposure group comprised patients who used single PAs (rebamipide, PPI, and H2RA) for at least 180 days without other PAs or with other PAs used for less than 120 days. The reference group used only DAs (i.e., non-exposure group).

For each subcohort, 1:1:1:1 exact matching was performed to overcome selection bias between the three exposure and one non-exposure groups. We matched patients based on the following variables: Sex, history of ulcer disease, GI bleeding history, DAPT, concomitant anticoagulant therapy, age 60 years or more, corticosteroid history, Helicobacter pylori infection history, and number of previous GI bleeding events. Each matched sub-cohort was followed up for one year after the landmark point (i.e., the follow-up period). Finally, the matched sub-cohorts were pooled into a single SHD-free matched cohort for analysis.

Statistical analysis

For SCCS analyses, conditional Poisson regression models were used to estimate the incidence rate ratios (IRRs) with 95% confidence intervals (CIs) for PA exposure periods compared with non-exposure periods. By conditioning the total number of SHD events per individual, a comparison between these periods can be performed with an inherent adjustment for time-invariant confounding factors. In the main analysis, we assumed that the multiple SHD events in a patient were independent. However, we also performed a sensitivity analysis restricted to the first SHD event to reflect possible non-independence between multiple events. The models were further adjusted for the lengths of the PA and DA prescription periods to reduce residual confounding factors.

Multivariate conditional Cox hazard models were used to estimate the hazard ratios (HRs) with 95%CIs, adjusting for PA and DA prescription periods during the baseline period. We added cluster effects to the models to reflect the dependency within the matched set. Because patients could experience multiple SHD events during the follow-up period, we conducted analyses for both the first and recurrent events (i.e., all events). Recurrent event analysis included a shared frailty term to address within-subject correlations. The proportional hazards assumption and multicollinearity were verified using Schoenfeld residuals and variance inflation factor, respectively. No severe violations were observed.

For subgroup analyses, antiplatelet users were divided into the following subgroups: (1) Group A: Aspirin alone; (2) Group B: Clopidogrel alone; (3) Group C-E: Concomitant users of aspirin and clopidogrel; and (4) Group E: Concurrent users of aspirin and clopidogrel, as illustrated in Figure 4. Aspirin-alone users and clopidogrel-alone users were defined as patients receiving either agent without exposure to the other. Concomitant users were defined as patients exposed to both agents, regardless of whether they took both agents simultaneously. Concurrent users, as a subset of concomitant users, are those who use two agents simultaneously without using a single agent.

Figure 4 Conceptual diagram of antiplatelet user subgroups. Subgroups of antiplatelet users based on aspirin and clopidogrel use are illustrated. Aspirin-alone users (A) and clopidogrel-alone users (B) are defined as patients receiving either agent without exposure to the other. Concomitant users are defined as patients exposed to both agents, regardless of taking two agents at the same time (overlapping). Concomitant users include sequential users (C) of two agents without overlapping, partially concurrent users (D) of two agents with partial overlap, and concurrent users (E), taking two agents at the same time without using single agent alone.

All analyses were conducted using R software (version 4.0.5, Vienna, Austria), with a two-sided P value < 0.05 considered statistically significant.

RESULTS

Baseline characteristics

For the 98404 patients in our cohort, including all antiplatelet users, the median age was 62 years [interquartile range (IQR), 53-70], and 55.7% were males (Table 1). In the cohort, 90766 (92.2%), 34921 (35.5%), 6869 (7.0%), 3994 (4.0%), and 3003 patients (3.1%) used aspirin, clopidogrel, warfarin, NSAIDs, and DOAC, respectively. In addition, 38209 patients (38.8%) were co-administered PAs. Among the PA users, 13337 (12.3%), 5890 (54.4%), and 19919 patients (18.4%) used PPI, rebamipide, and H2RA, respectively. The SCCS dataset consisted of 1137 patients with a median age of 64 years (63.4% of male. The proportion of aspirin users was 85.9% and that of clopidogrel users was 40.9%. The matched landmark dataset included 6029 patients with a median age of 61. Patients taking PPI, rebamipide, and H2RA comprised 22.7%, 17.7%, and 22.4% of the dataset, respectively.

Table 1 Baseline characteristics for all antiplatelet ± anticoagulant/non-steroidal anti-inflammatory drug patients across different designs, n (%).

Variables	Cohort dataset (n = 98404)	SCCS dataset (n = 1137)	Matched landmark dataset (n = 6029)
Age, median (IQR)	62 (53-70)	64 (56-71)	61 (54-69)
Sex
Male	54789 (55.7)	703 (61.8)	2934 (48.7)
Female	43615 (44.3)	434 (38.2)	3095 (51.3)
Mucosal DAs
Aspirin	90766 (92.2)	977 (85.9)	5510 (91.4)
Clopidogrel	34921 (35.5)	465 (40.9)	3316 (55)
Warfarin	6869 (7.0)	151 (13.3)	487 (8.1)
NSAIDs	3994 (4.0)	66 (5.8)	469 (7.8)
DOAC	3003 (3.1)	70 (6.2)	274 (4.5)
PAs co-administered with DAs
PPI	13337 (12.3)	476 (41.9)	1366 (22.7)
Rebamipide	5890 (54.4)	196 (17.2)	1068 (17.7)
H2RA	19919 (18.4)	594 (52.2)	1352 (22.4)
Others	3647 (34.7)	103 (9.1)

SCCS: Self-controlled case series; IQR: Interquartile range; DA: Damaging agent; NSAID: Non-steroidal anti-inflammatory drug; DOAC: Direct oral anticoagulant; PA: Protective agent; H2RA: H2-receptor antagonist; PPI: Proton pump inhibitor; Others: Agents such as sodium alginate, ecabet sodium hydrate, polaprezinc, and stillen (an ethanol extract of Artemisiae Argyi herba).

Table 2 describes the baseline characteristics of patients who received antiplatelet therapy alone. As described in Figure 4, concurrent aspirin and clopidogrel users referred specifically to patients who were prescribed both agents during exactly the same period. In contrast, concomitant use indicated that both aspirin and clopidogrel were prescribed during the study period, although not necessarily on the same date. As illustrated in Figure 4, partially concurrent use represents cases in which the treatment periods overlap for a certain duration. There were more male patients in aspirin/clopidogrel-alone users, concurrent and concomitant users of aspirin and clopidogrel than females. For those who used PAs, the majority used H2RA, followed by PPI and rebamipide in all subgroups.

Table 2 Baseline characteristics for aspirin or clopidogrel alone users, n (%).

Variables	Aspirin-alone users (n = 55295)	Clopidogrel-alone users (n = 6882)	Concurrent users of aspirin & clopidogrel (n = 7080)	Concomitant users of aspirin & clopidogrel (n = 22452)
Age, median (IQR)	61 (52-69)	65 (56-73)	65 (57-73)	63 (54-70)
Sex
Male	28344 (51.3)	3465 (50.3)	5081 (71.8)	15176 (67.6)
Female	26951 (48.7)	3417 (49.7)	1999 (28.2)	7276 (32.4)
Mucosal DAs
Aspirin	55295 (100)	0 (0)	7080 (100)	22452 (100)
Clopidogrel	0 (0)	6882 (100)	7080 (100)	22452 (100)
PAs co-administered with DAs
PPI	5737 (31.8)	715 (31.9)	742 (23.1)	3745 (30.3)
Rebamipide	2143 (11.9)	450 (20.1)	224 (7.0)	1484 (12.0)
H2RA	8065 (44.7)	927 (41.4)	2187 (68.0)	6503 (52.5)
Others	2085 (11.6)	147 (6.6)	63 (1.9)	643 (5.2)

IQR: Interquartile range; DA: Damaging agent; PA: Protective agent; H2RA: H2-receptor antagonist; PPI: Proton pump inhibitor; Others: Agents such as sodium alginate, ecabet sodium hydrate, polaprezinc, and stillen (an ethanol extract of Artemisiae Argyi herba).

The high- and low-risk groups comprised 73691 patients (74.9%) and 24713 patients (25.1%), respectively, of all antiplatelet users, with median ages of 66 (IQR 60-72) and 51 (IQR 45-56), respectively (Supplementary Table 1). Both groups included more males (54.7% and 58.5%, respectively). Aspirin was the most commonly used antiplatelet agent (92.5% and 91.5%), followed by clopidogrel (43.9% and 10.3%) in both groups. Among the PAs co-administered with DAs, H2RA was the most commonly used in both groups (23.1% and 11.7%). The second most commonly used PA was PPI (15.5%) in the high-risk group and rebamipide (5.3%) in the low-risk group.

Comparative effect of PAs on SHD for all antiplatelet users in SCCS analyses

Compared with DA alone, all types of PA significantly decreased the risk of SHD, except for the concomitant use of H2RA and others (Table 3). All single agents showed significant protective effects on SHD: PPI (IRR 0.42, 95%CI: 0.34-0.53) with greatest reduction in the risk of SHD, followed by rebamipide (IRR 0.68, 95%CI: 0.50-0.53) and H2RA (IRR 0.75, 95%CI: 0.63-0.90). Among concomitant PA users, PPI and rebamipide (IRR 0.07, 95%CI: 0.02-0.30) was significantly associated with the lowest risk of SHD, followed by PPI & others (IRR 0.09, 95%CI: 0.02-0.52) and rebamipide and H2RA (IRR 0.23, 95%CI: 0.09-0.59). For those with high-risk, PPI & rebamipide lowered the risk of SHD the most (IRR 0.07, 95%CI: 0.02-0.29), followed by PPI & others (IRR 0.09, 95%CI: 0.01-0.51) and rebamipide & H2RA (IRR 0.25, 95%CI: 0.10-0.64). In the low-risk group, no PA type showed a significant protective effect against SHD.

Table 3 Effect of protective agent uses on significant hemoglobin drop incidences by risk groups for all antiplatelet ± anticoagulant/non-steroidal anti-inflammatory drug patients in self-controlled case series analyses.

Exposure period	Total (n = 1137)			High-risk group (n = 1028)			Low-risk group (n = 109)
	Number of SHD incidence	IRR (95%CI)	P value	Number of SHD incidence	IRR (95%CI)	P value	Number of SHD incidence	IRR (95%CI)	P value
All SHD events
Mucosal DA alone	939	Reference		845	Reference		94	Reference
PPI	233	0.42 (0.34-0.53)	< 0.001	222	0.42 (0.33-0.53)	< 0.001	11	0.46 (0.18-1.20)	0.114
PPI & rebamipide	2	0.07 (0.02-0.30)	< 0.001	2	0.07 (0.02-0.29)	< 0.001	0
PPI & H2RA	8	0.27 (0.11-0.65)	0.003	8	0.28 (0.12-0.68)	0.005	0
PPI & others	4	0.09 (0.02-0.52)	0.007	4	0.09 (0.01-0.51)	0.007	0
Rebamipide	122	0.68 (0.50-0.93)	0.017	110	0.66 (0.47-0.92)	0.014	12	0.92 (0.39-2.15)	0.840
Rebamipide & H2RA	6	0.23 (0.09-0.59)	0.002	6	0.25 (0.10-0.64)	0.004	0
H2RA	356	0.75 (0.63-0.90)	0.002	333	0.77 (0.64-0.93)	0.006	23	0.59 (0.31-1.13)	0.113
H2RA & others	2	0.24 (0.05-1.18)	0.078	2	0.34 (0.07-1.74)	0.196	0
Others	29	0.37 (0.22-0.60)	< 0.001	23	0.31 (0.17-0.53)	< 0.001	6	1.01 (0.27-3.79)	0.985
First SHD event only
Mucosal DA alone	938	Reference		844	Reference		94	Reference
PPI	219	0.43 (0.34-0.53)	< 0.001	208	0.42 (0.34-0.54)	< 0.001	11	0.43 (0.16-1.13)	0.087
PPI & rebamipide	2	0.07 (0.02-0.29)	< 0.001	2	0.07 (0.02-0.28)	< 0.001	0
PPI & H2RA	8	0.32 (0.13-0.78)	0.0120	8	0.34 (0.14-0.83)	0.017	0
PPI & others	4	0.09 (0.02-0.53)	0.0080	4	0.09 (0.02-0.53)	0.008	0
Rebamipide	118	0.68 (0.49-0.94)	0.0200	106	0.65 (0.46-0.92)	0.015	12	0.96 (0.40-2.28)	0.923
Rebamipide & H2RA	4	0.19 (0.06-0.55)	0.0020	4	0.20 (0.07-0.61)	0.005	0
H2RA	331	0.76 (0.63-0.91)	0.0030	309	0.78 (0.64-0.94)	0.010	22	0.60 (0.32-1.15)	0.122
H2RA & others	2	0.29 (0.06-1.45)	0.1300	2	0.43 (0.08-2.31)	0.328	0
Others	29	0.39 (0.24-0.65)	< 0.001	23	0.33 (0.19-0.58)	< 0.001	6	0.92 (0.24-3.49)	0.898

Adjusted incidence rate ratios were estimated after adjusting prescription period for protective agents and damaging agents. SHD: Significant hemoglobin drop; CI: Confidence interval; DA: Damaging agent; PPI: Proton pump inhibitor; H2RA: H2-receptor antagonist; IRR: Incidence rate ratio; Others: Agents such as sodium alginate, ecabet sodium hydrate, polaprezinc, and stillen (an ethanol extract of Artemisiae Argyi herba).

In the first sensitivity analysis, in which only the first SHD events were included, these results were consistently observed (Table 3). In the second sensitivity analysis for antiplatelet-alone users, single use of rebamipide did not show significant effect (IRR 0.92, 95%CI: 0.62-1.38), while all other results were consistent (Supplementary Table 2).

Comparative effect of PAs on SHD for all antiplatelet users in matched landmark analyses

All PAs (PPI, rebamipide, and H2RA) showed protective effects in the high- and low-risk groups; however, the difference was not statistically significant (Supplementary Table 3). For all antiplatelet users, PPI users HR was 0.77 (95%CI: 0.28-2.13), rebamipide’s was 0.36 (95%CI: 0.12-1.08), and H2RA’s was 0.62 (95%CI: 0.20-1.92). Similar results were observed when only the first event was considered (Supplementary Table 3).

Subgroup analyses according to DAs in SCCS analyses

For aspirin-alone users, PPI (IRR 0.33, 95%CI: 0.21-0.52) and H2RA (IRR 0.60, 95%CI: 0.43-0.83) showed significant protective effects on SHD (Table 4). All other types of PA use tend to show protective effects on SHD, but with no statistical significance (e.g., IRR 0.25 with 95%CI: 0.02-2.87 for rebamipide-alone users). For clopidogrel-alone users and concurrent DA users, all types of PA use were not statistically associated with SHD risk. For concomitant DA users, the single use of PPI showed significantly preventive effect (IRR 0.43, 95%CI: 0.30-0.62), while other PA types did not show significant effect. Similar results were obtained when only the first event was considered (Table 4).

Table 4 Effect of protective agent uses on significant hemoglobin drop incidences for aspirin or clopidogrel alone users in self-controlled case series analyses.

Exposure period	Aspirin-alone users (n = 351)		Clopidogrel-alone users (n = 35)		Concurrent users of aspirin & clopidogrel (n = 30)		Concomitant users of aspirin & clopidogrel (n = 348)
	IRR (95%CI)	P value	IRR (95%CI)	P value	IRR (95%CI)	P value	IRR (95%CI)	P value
All SHD events
Mucosal DA alone	Reference		Reference
PPI	0.33 (0.21-0.52)	< 0.001	0.43 (0.09-2.13)	0.301	0.39 (0.10-1.60)	0.191	0.43 (0.30-0.62)	< 0.001
PPI & rebamipide	0.35 (0.06-1.90)	0.223
PPI & H2RA							0.54 (0.18-1.64)	0.281
PPI & others							0.14 (0.02-1.02)	0.052
Rebamipide	0.69 (0.36-1.34)	0.276	0.25 (0.02-2.87)	0.263			1.49 (0.87-2.57)	0.148
Rebamipide & H2RA	0.61 (0.12-3.17)	0.553					0.27 (0.05-1.35)	0.111
H2RA	0.60 (0.43-0.83)	0.002	1.85 (0.55-6.23)	0.322	1.02 (0.42-2.53)	0.958	0.93 (0.67-1.29)	0.661
H2RA & others	0.46 (0.07-2.95)	0.411
Others	0.47 (0.20-1.10)	0.081	2.34 (0.32-17.19)	0.403			0.40 (0.15-1.07)	0.068
First SHD event only
Mucosal DA alone	Reference		Reference
PPI	0.33 (0.21-0.51)	< 0.001	0.42 (0.08-2.08)	0.286	0.39 (0.10-1.60)	0.191	0.45 (0.31-0.65)	< 0.001
PPI & rebamipide	0.35 (0.06-1.91)	0.224
PPI & H2RA							0.81 (0.26-2.50)	0.708
PPI & others							0.15 (0.02-1.04)	0.055
Rebamipide	0.74 (0.38-1.44)	0.380	0.26 (0.02-3.07)	0.285			1.46 (0.85-2.52)	0.173
Rebamipide & H2RA	0.60 (0.12-3.15)	0.549
H2RA	0.62 (0.44-0.86)	< 0.001	2.22 (0.59-8.33)	0.239	1.02 (0.42-2.53)	0.958	0.96 (0.69-1.34)	0.814
H2RA & others	0.45 (0.07-2.89)	0.401
Others	0.48 (0.20-1.12)	0.090	2.30 (0.31-16.93)	0.413

SHD: Significant hemoglobin drop; IRR: Incidence rate ratio; CI: Confidence interval; DA: Damaging agent; PPI: Proton pump inhibitor; H2RA: H2-receptor antagonist; Others: Agents such as sodium alginate, ecabet sodium hydrate, polaprezinc, and stillen (an ethanol extract of Artemisiae Argyi herba).

DISCUSSION

This retrospective study assessed the efficacy of various PAs in patients treated with aspirin, clopidogrel, or other mucosal-DAs. We further stratified patients into high- and low-risk groups to explore risk-specific outcomes. In addition, we examined patients prescribed aspirin or clopidogrel alone, without concurrent mucosal-DAs, to determine the preventive effects of different PAs and evaluate the role of risk stratification in these subgroups. Importantly, we also differentiated patients who received aspirin and clopidogrel concurrently from those who were exposed to these drugs during overlapping periods, thereby enabling a more granular analysis. The strength of this study lies in its comprehensive subgroup analyses and detailed comparisons across a wide range of mucosal-damaging and mucosal-damage-PAs.

In designing the cohort, our objective was to capture Hb decline attributable primarily to GI bleeding related to antiplatelet exposure. Therefore, we excluded patients with conditions known to cause chronic progressive anemia independent of GI pathology - such as malignancy, advanced chronic kidney disease, decompensated cirrhosis with portal hypertensive bleeding risk, coagulation disorders, or major trauma. Stable chronic anemia itself was not an exclusion criterion, as the SCCS design relies on within-individual comparisons and is robust against time-invariant hematologic differences. These exclusions helped ensure that the observed Hb reductions more accurately reflected GI bleeding-related events rather than non-GI-driven progressive anemia.

The ACCF, ACG, and AHA revised the expert consensus document of the 2010 guidelines, which states that the use of a PPI or H2RA reduces the risk of upper GI (UGI) bleeding compared to no therapy, and PPIs reduce UGI bleeding to a greater degree than H2RAs. Previous studies have consistently shown that PPIs outperform H2RAs in preventing low-dose aspirin-associated UGI bleeding (odds ratio, 0.28; 95%CI: 0.16-0.50)[19,30-32]. Our findings add to the growing body of evidence that PPIs provide superior GI protection compared to H2RA. The degree to which PPIs exert a protective effect against H2RAs varies depending on the type of DA used. However, as cardiovascular pharmacotherapy has become increasingly refined, as illustrated in a recent study[33], there is a need to refine strategies for preventing bleeding complications. A key advantage of the present study is its real-world design, which allowed us to analyze multiple PA combinations in patients receiving aspirin or clopidogrel and to further stratify outcomes by risk category.

When analyzed according to the antiplatelet regimen, aspirin monotherapy was most effectively protected by PPI alone, with H2RA also showing a significant benefit. However, no PAs have demonstrated efficacy in patients receiving clopidogrel monotherapy. Previous studies comparing the GI effects of aspirin and clopidogrel reported conflicting results[34-37]. In our study, the combination of aspirin and PPI provided preventive benefits, whereas clopidogrel did not show a meaningful response to PAs, which is consistent with earlier findings[35,36]. Given the well-documented interaction between clopidogrel and PPIs, aspirin may be the preferred antiplatelet agent in certain cases. Clopidogrel is a prodrug requiring hepatic activation via CYP2C19, which is inhibited to varying degrees by different PPIs (omeprazole > pantoprazole > lansoprazole > rabeprazole)[38,39]. Furthermore, the more rapid recovery of platelet aggregation after discontinuation of aspirin (4 days) compared with clopidogrel (10 days) may also contribute to the differences in bleeding risk[40,41].

PPI monotherapy was associated with the greatest protective effect in patients who received dual therapy with aspirin and clopidogrel. Notably, in the subgroup analyses of aspirin or clopidogrel monotherapy stratified by risk, rebamipide plus PPI remained the most effective strategy in high-risk patients, followed by PPI combined with other agents, and H2RA plus PPI. Importantly, bleeding in patients receiving antiplatelet therapy was not confined to the UGI tract. In a 10-year prospective study by Ray[42], lower GI (LGI) bleeding accounted for 45% of all antiplatelet drug-related GI bleeding events, with the small intestine being the predominant source (84%). This underscores the need for preventive strategies targeting both upper and LGI bleeding. Rebamipide has demonstrated efficacy in reducing aspirin- and NSAID-induced small bowel injuries in randomized controlled trials[43]. In contrast, PPIs, although highly effective in preventing UGI bleeding, have limited protective effects against LGI bleeding and may even increase the risk, possibly through alterations in the microbiota of the small bowel[44,45]. Taken together, these findings suggest that, particularly in high-risk patients, the combination of rebamipide with PPI may provide broader protection against both upper and LGI bleeding than PPI monotherapy.

The superior efficacy of the combination of PPI and rebamipide observed in high-risk patients is biologically plausible given these complementary mechanisms: PPIs effectively reduce gastric acid-mediated UGI damage, while rebamipide enhances mucosal defense and promotes healing throughout the small and large bowel. This synergistic effect has also been proposed in prior clinical and experimental studies evaluating mucosal protectants[15,37]. Taken together, these mechanistic insights align closely with our clinical findings and support the rationale for combination therapy, particularly in patients at elevated risk for both upper and LGI bleeding.

In this study, we found that the combination of rebamipide and PPI provided the strongest protective effect in patients taking aspirin or clopidogrel along with other mucosal-DAs, followed by PPI with other agents and PPI alone. This protective hierarchy persisted after risk stratification, with rebamipide plus PPI conferring particular benefits to the high-risk patients. The incidence of GI bleeding events was higher in the high-risk group than in the low-risk group. In high-risk patients, our analysis confirmed that PPI monotherapy offers benefits but yields the greatest protection when combined with rebamipide. The protective role of PPIs in reducing UGI bleeding through the mucosal protection of the UGI tract has been well documented[46]. However, emerging evidence indicates that PPIs may have limited efficacy in preventing LGI bleeding, and may contribute to its occurrence in some cases[47-49]. Despite guideline recommendations focusing solely on PPIs for GI protection, our findings, corroborated by previous studies, demonstrate that adding rebamipide may enhance safety and GI outcomes, especially where a higher GI risk is a concern. This finding supports a tailored approach of PPI plus rebamipide in clinical scenarios with compounded GI bleeding risk. In low-risk patients, mucosal protective agents did not demonstrate a significant benefit, which is consistent with the current AHA guidelines stating that the routine use of either a PPI or H2RA is not recommended for patients at a lower risk of UGI bleeding. The lack of statistical significance in low-risk patients may be explained by the lower incidence of SHD in this population, the limited sample size, or the possibility that preventive therapy may not be necessary in patients with inherently low bleeding risk.

This study also possesses several methodological strengths that enhance the robustness and reliability of its findings. First, the integration of two complementary analytical approaches - a SCCS design and matched landmark analysis - provides a rigorous framework that minimizes both time-invariant confounding and immortal-time bias. The dual-design strategy allows cross-validation of results across fundamentally different methodological structures, thereby strengthening causal inference. Second, this work addresses an important clinical gap by evaluating the comparative effectiveness of alternative gastroprotective strategies in antiplatelet users within real-world practice, where polypharmacy and heterogeneous risk profiles are common. Third, the large cohort of 98404 antiplatelet users, along with substantial sample sizes for both the SCCS (n = 1137) and matched landmark cohorts (n = 6029), affords high statistical power and enables reliable subgroup analyses across diverse clinical scenarios. These strengths collectively reinforce the validity and clinical applicability of our results.

This study has some limitations. First, because this was a single-center study, the generalizability of our findings to a broader population may be limited. Second, the retrospective cohort study introduced a potential selection bias and residual confounding factors that could not be fully eliminated. Although we performed exact matching based on guideline-recommended risk factors, unmeasured variables such as the presence of esophageal varices or vascular ectasia may also influence bleeding risk. Concomitant medications, including certain antibiotics (e.g., amoxicillin, cefepime, and ceftazidime), may also increase the risk of bleeding through pharmacological interactions with antiplatelet agents, complicating the interpretation of the outcomes. Third, we did not estimate the effect of concomitant PA use in the matched landmark analyses, anticipating reduced statistical power due to the relatively small number of patients receiving each concomitant PA use. Therefore, we focused primarily on single-agent users (rebamipide, PPI, or H2RA alone) to maintain statistical validity.

CONCLUSION

Overall, our findings suggest that the choice of a PA should be tailored to the specific antiplatelet regimen and individual risk profile. While PPI monotherapy may be sufficient for patients receiving aspirin alone, high-risk patients or those receiving additional mucosal-DAs appear to benefit the most from the combination of rebamipide and PPI.