Fexuprazan vs esomeprazole for endoscopic submucosal dissection-induced gastric ulcer healing: A double-blind randomized controlled trial

Abstract

BACKGROUND

Endoscopic submucosal dissection (ESD) induces large artificial gastric ulcers requiring acid-suppressive therapy, but high-quality evidence comparing potassium-competitive acid blockers (P-CABs) with proton pump inhibitors is limited. We hypothesized that fexuprazan, a novel P-CAB, would achieve a higher complete healing rate of ESD-induced gastric ulcers than esomeprazole.

AIM

To evaluate the efficacy and safety of fexuprazan vs esomeprazole in the healing of ESD-induced gastric ulcers.

METHODS

In this multicenter, double-blind, double-dummy, parallel-group superiority randomized controlled trial, patients undergoing ESD for gastric neoplasms received fexuprazan 40 mg or esomeprazole 40 mg once daily for 8 weeks. The primary endpoint was the complete healing rate at 8 weeks, analyzed in both modified intention-to-treat (mITT) and per-protocol (PP) populations. Secondary endpoints included the ulcer shrinkage rate and delayed bleeding. The model-estimated time to a 99% reduction in ulcer area was compared as an exploratory endpoint.

RESULTS

Of 132 randomized patients, 128 received at least one dose of study medication and were included in the mITT analysis; 107 were included in the PP analysis. In the PP analysis, the 8-week complete healing rate was significantly higher with fexuprazan [90.2% vs 75.0%; risk difference (RD): 15.2%, 95% confidence interval (CI): 1.2-29.2; P = 0.040]; the mITT analysis showed a directionally consistent but non-significant difference (73.4% vs 65.6%; RD: 7.8%, 95%CI: -8.1 to 23.7; P = 0.337). Ulcer shrinkage rates at 4 weeks and 8 weeks were significantly greater with fexuprazan (both P = 0.049). In exploratory analysis, the model-estimated time to a 99% reduction in ulcer area was shorter with fexuprazan (40.31 days vs 47.22 days; P = 0.045), particularly among patients with large baseline ulcers (≥ 800 mm²). Safety was comparable between groups.

CONCLUSION

In the pre-specified PP analysis, fexuprazan achieved a significantly higher 8-week complete healing rate than esomeprazole, with comparable safety. Exploratory analysis suggested shorter model-estimated healing time in large ulcers, warranting larger trials.

Key Words: Fexuprazan; Potassium-competitive acid blocker; Esomeprazole; Endoscopic submucosal dissection; Gastric ulcer; Randomized controlled trial

Core Tip: This double-blind, double-dummy randomized controlled trial compared the potassium-competitive acid blocker fexuprazan with the proton pump inhibitor esomeprazole for the healing of artificial gastric ulcers after endoscopic submucosal dissection. Ulcer healing was assessed endoscopically at 4 weeks and 8 weeks. Fexuprazan achieved a significantly higher complete healing rate at 8 weeks in the pre-specified PP analysis (90.2% vs 75.0%) and greater ulcer shrinkage at both time points, although the difference was not significant in the modified intention-to-treat population. In an exploratory analysis, the model-estimated healing time was shorter with fexuprazan, particularly for large baseline ulcers (≥ 800 mm²). The safety profile was comparable between groups.

INTRODUCTION

Endoscopic submucosal dissection (ESD) is recommended as a standard treatment for early gastric neoplasms because it enables reliable en bloc resection and precise histologic assessment[1]. However, ESD inevitably induces large and deep artificial ulcers, which can lead to complications, including delayed bleeding or perforation[2]. Delayed bleeding occurs in approximately 5% of patients, and the risk increases to nearly 20% among those receiving antithrombotic therapy[2-4]. Accordingly, timely and complete healing of post-ESD ulcers is clinically important.

Proton pump inhibitors (PPIs) are widely used to manage ESD-induced ulcers, as acid suppression facilitates ulcer healing[5-7]. An 8-week treatment course is commonly used in clinical practice to promote ulcer healing[8]. However, the optimal duration of acid-suppressive therapy following ESD remains uncertain, as current guidelines do not specify a standardized treatment duration[1,9]. Moreover, the time required for complete ulcer healing after ESD varies across studies, and complete resolution is not consistently observed even after 8 weeks of therapy[8,10,11]. These findings indicate that the optimal duration of acid-suppressive therapy remains undefined and may require relatively prolonged PPI administration. In addition, PPIs have several inherent pharmacologic limitations, including the need for pre-meal administration, delayed onset of effectiveness, short plasma half-life, and CYP2C19-dependent metabolism, which can lead to variable therapeutic responses[5,12,13]. In contrast, potassium-competitive acid blockers (P-CABs) are a newer class of acid-suppressive agents that produce more rapid and sustained acid suppression than PPIs, thereby offering the potential to overcome many of these limitations[14]. Several studies have evaluated P-CABs, including vonoprazan and tegoprazan, for the treatment of ESD-induced ulcers[10,15,16]. Although some reports suggest that P-CABs may provide efficacy comparable to or greater than that of PPIs, the evidence remains inconsistent[16-19]. Furthermore, previous randomized controlled trials (RCTs) comparing P-CABs with PPIs for post-ESD ulcers have largely employed open-label designs, which may introduce performance and assessment bias in the evaluation of endoscopic ulcer healing. A well-designed, double-blind trial is therefore needed to provide more robust evidence.

Among available P-CABs, fexuprazan is a novel agent with minimal influence from CYP2C19 polymorphisms; however, no clinical trial has yet evaluated its efficacy in healing artificial ulcers after ESD[20]. This study aimed to compare the efficacy and safety of fexuprazan with esomeprazole in a double-blind RCT for the healing of ESD-induced gastric ulcers, and, as an exploratory model-based analysis, to estimate the time to ulcer healing with each agent rather than to compare treatment durations directly. We hypothesized that 8 weeks of fexuprazan 40 mg once daily is superior to esomeprazole 40 mg once daily with respect to the complete healing rate assessed endoscopically at 8 weeks.

MATERIALS AND METHODS

Study design and protocol

This prospective, multicenter, double-blind, double-dummy, parallel-group, active-controlled superiority randomized controlled trial enrolled patients undergoing ESD for gastric neoplasia between April 2024 and October 2025 at two referral hospitals in Korea. During the screening period, patient demographics, baseline characteristics, medical history, comorbidities, and concomitant medications were recorded. The study protocol is shown in Figure 1. All patients received intravenous esomeprazole 40 mg twice daily for the first 2 days after ESD. In the absence of post-procedural complications, patients subsequently received oral fexuprazan 40 mg once daily or esomeprazole 40 mg once daily, administered before breakfast, for an 8-week treatment period. Endoscopy was performed by the principal investigators at 4 weeks and 8 weeks following ESD.

Figure 1 Study protocol. All patients received intravenous esomeprazole 40 mg twice daily for the first 2 days after endoscopic submucosal dissection (ESD). Beginning on post-ESD day 3, patients were assigned to receive either oral fexuprazan 40 mg/day plus placebo esomeprazole or oral esomeprazole 40 mg/day plus placebo fexuprazan, administered once daily before breakfast for 8 weeks. Follow-up assessments included laboratory testing on days 1, 3, and 14 and upper gastrointestinal endoscopy at 4 weeks and 8 weeks after ESD to evaluate ulcer healing status. ESD: Endoscopic submucosal dissection.

All patients provided written informed consent. The study adhered to the Declaration of Helsinki, received Institutional Review Board approval at each site (Yongin Severance Hospital, No. 9-2023-0259; Severance Hospital, No. 4-2024-0167), and was registered at the Clinical Research Information Service, a primary registry of the World Health Organization International Clinical Trials Registry Platform (https://cris.nih.go.kr, No. KCT0009356). All authors had access to the study data and reviewed and approved the final manuscript.

Participants and randomization

Patients diagnosed with gastric adenoma or early gastric cancer who underwent ESD at the two participating centers were screened for eligibility. Eligible participants met the following inclusion criteria: (1) Aged 19-80 years; and (2) Histological and endoscopic confirmation of gastric adenoma or early gastric cancer, including: Differentiated-type adenocarcinoma without ulceration; differentiated-type adenocarcinoma with ulceration confined to the mucosa and measuring < 3.0 cm; differentiated-type adenocarcinoma infiltrating the submucosa to a depth < 500 μm without ulceration and measuring < 3.0 cm; or undifferentiated-type adenocarcinoma confined to the mucosa without ulceration and measuring < 2.0 cm. The exclusion criteria were as follows: (1) Known hypersensitivity to the investigational product or benzimidazole derivatives; (2) Use of gastro-esophageal reflux-related medications (H₂-receptor antagonists, prostaglandin analogues, mucosal protectants, or prokinetics) within the preceding 2 weeks or anticipated use during the study period; (3) Laboratory values exceeding predefined safety thresholds [total bilirubin or creatinine levels greater than 1.5 times the upper limit of normal (ULN)]; aspartate aminotransferase, alanine aminotransferase, alkaline phosphatase, or blood urea nitrogen levels greater than 2 times the ULN); (4) A history of other malignancies within the past 5 years; (5) Underlying gastrointestinal or systemic conditions that could alter gastric physiology or study drug absorption, including Zollinger-Ellison syndrome, Barrett’s esophagus, esophageal motility disorders, esophageal stricture, pancreatitis, malabsorption disorders, or severe cardiopulmonary disease; (6) Prior major gastrointestinal surgery affecting acid secretion; (7) Inability to discontinue medications known to influence bleeding or gastric mucosal healing (warfarin, clopidogrel, direct oral anticoagulants, steroids, non-steroidal anti-inflammatory drugs, anticholinergics, prostaglandin analogues, sucralfate, or aspirin) for the 8-week study duration; (8) Substantial hepatic or renal dysfunction; (9) Pregnancy or breastfeeding; and (10) Any condition deemed by the investigator to make the patient unsuitable for study participation. In particular, patients requiring continuous antiplatelet or anticoagulant therapy who could not safely discontinue these medications for the 8-week study duration were excluded at screening; accordingly, no enrolled patient received antiplatelet or anticoagulant therapy during the study period.

Eligible patients were prospectively randomized in a 1:1 ratio to the fexuprazan or esomeprazole group using a computer-generated randomization schedule with permuted block randomization (block size of 4). No stratification factors were applied. A double-dummy design was used to maintain blinding, with matching active drugs and placebo tablets once daily for 8 weeks, consisting of fexuprazan 40 mg plus matching placebo (fexuprazan group) or esomeprazole 40 mg plus matching placebo (esomeprazole group). Study medications were dispensed in sealed boxes by an independent medication supervisor according to the randomization schedule to ensure concealed allocation.

ESD

All ESD procedures were performed in hospitalized patients under intravenous sedation. A video endoscope (GIF-HQ290; Olympus, Tokyo, Japan) was used for the procedure. Lesion margins were marked with argon plasma coagulation (VIO 300D; ERBE, Tübingen, Germany), followed by submucosal injection of epinephrine (0.01 mg/mL) and indigo carmine. A dual knife (KD-650Q; Olympus, Tokyo, Japan) or an insulated-tip knife (KD-610 L; Olympus Optical, Tokyo, Japan) was used for incision and dissection, with hemostasis achieved using hemoclips or hemostatic forceps. Prophylactic mucosal closure of the post-ESD defect was not routinely performed at either center, and no patient in this study underwent prophylactic closure. Following ESD, chest and abdominal radiographs were routinely obtained immediately after the procedure and on the following day to monitor for complications, including perforation or pneumonia.

Outcomes

The primary endpoint was the complete healing rate of ESD-induced artificial gastric ulcers at 8 weeks, assessed by follow-up endoscopy. Complete healing, a binary endpoint reflecting the ultimate therapeutic goal, was defined as achievement of scar-stage healing without residual mucosal defect according to the Sakita-Miwa classification[21]. Secondary endpoints included the ulcer shrinkage rate at 4 weeks and 8 weeks, the scar-stage healing rate at 4 weeks, and delayed bleeding within 8 weeks after ESD. Ulcer shrinkage rate, a continuous measure expressed as percent reduction in ulcer area from baseline, captures the degree of healing progression and complements complete healing in characterizing healing kinetics. As an exploratory endpoint, the model-estimated time to a 99% reduction in ulcer area was compared between the two treatment groups. Ulcer size was measured by assuming the mucosal defect to be elliptical; the area (mm²) was calculated as π × (long-axis radius) × (short-axis radius). The initial ulcer size was estimated from the resected specimen dimensions measured immediately after ESD. During follow-up endoscopy at 4 weeks and 8 weeks, ulcer dimensions were determined using open biopsy forceps. The shrinkage rate was calculated as [(initial ulcer area - ulcer area at follow-up) × 100/initial ulcer area]. Delayed bleeding was defined according to the Adverse events in GastRointEstinal Endoscopy classification as bleeding occurring more than 24 hours after the procedure that required endoscopic, radiologic, or surgical intervention; transfusion; or resulted in hematemesis, melena, or a ≥ 2 g/dL decrease in hemoglobin[22]. Safety assessments included vital signs, physical examination findings, clinical laboratory evaluations, and the incidence of treatment-emergent adverse events (TEAEs). TEAEs were defined as adverse events (AEs) occurring after initiation of the study medication and were classified by severity and causality.

Sample size calculations

For the sample size calculation, assuming an 8-week healing rate of 78% for esomeprazole[18] and a clinically meaningful superiority of at least a 10% absolute difference in healing rates, a total sample size of 107 evaluable patients was estimated to be sufficient with a two-sided significance level of 0.05 and 80% power. Considering a 10% dropout rate, a total of 118 patients were planned for enrollment. Because post-randomization attrition exceeded this assumption, enrollment was extended to 132 patients to secure the pre-specified number of evaluable patients. This decision was based solely on the blinded count of evaluable patients, without any interim efficacy analysis.

Statistical analysis

Two efficacy analysis populations were pre-specified. The modified intention-to-treat (mITT) population included all randomized patients who received at least one dose of the study medication. The per-protocol (PP) population included patients in the mITT population who completed the assigned 8-week treatment and underwent 8-week endoscopic follow-up without major protocol violations. The safety population comprised all randomized patients who received at least one dose of the study medication and was therefore identical to the mITT population. The primary endpoint was analyzed in both populations, and the results of the PP, mITT and full ITT analyses are reported together. Patients without an evaluable 8-week assessment were imputed as treatment failures (non-responder imputation). A full intention-to-treat (ITT) population, including all randomized patients, was additionally analyzed post hoc as a further sensitivity analysis of the primary endpoint. All other secondary and exploratory outcomes were evaluated exclusively in the PP population. Between-group comparisons of continuous variables were performed using the independent two-sample t test when the assumption of normality was satisfied, with results presented as mean ± SD. When the normality assumption was violated, data were summarized as median and interquartile range, and the Wilcoxon rank-sum test was applied. Categorical variables were summarized as n (%) and compared between groups using Pearson’s χ² test or Fisher’s exact test, as appropriate. Risk differences (RDs) are presented with Wald 95% confidence intervals (CIs). The time required to achieve a 99% reduction in ulcer area, which is a model-derived quantity and not an observed endoscopic finding, was assumed to follow an exponential decay process. For each patient, a decay constant was estimated using the baseline ulcer area and the area measured at week 4. Specifically, the reduction process was modeled as Y_t = Y₀ ∙ e^-λt, where Y₀ is the baseline ulcer area, Y_t is the area at week t (in weeks), and λ represents the individual decay constant. For each subject, the decay constant was deterministically calculated via log-transformation of the ratio of the two time points as λ = -ln (Y₄ /Y₀)/4. Based on the estimated decay constant, the expected time to achieve a 99% reduction in ulcer area was calculated as t_99% = -ln (0.01)/λ ≈ 4.605/λ for each individual. In patients whose ulcer area was zero at week 4, the decay constant is undefined; for these patients the week-4 area was set to 0.1 mm² so that the model could be applied. As a post hoc assessment of this model assumption, the ulcer area predicted at week 8 by each patient’s model was compared with the observed week-8 ulcer area (Supplementary Table 1). To explore whether differences in the model-estimated time between the two treatment groups varied according to baseline ulcer area, an optimal cutoff value of baseline ulcer area was identified using the Youden index derived from receiver operating characteristic analysis and mutual information, aiming to discriminate between-group differences in the estimated time to 99% ulcer area reduction. The optimal baseline ulcer area cutoff was identified as 800 mm² with a minimum distance of 0.326 (Supplementary Figure 1). All statistical tests were two-sided, with P < 0.05 considered statistically significant unless otherwise specified. Because the 8-week complete healing rate was the single pre-specified primary endpoint, no multiplicity adjustment was pre-specified for the secondary and exploratory endpoints, which were regarded as supportive rather than confirmatory. Analyses were conducted using SAS version 9.4 (SAS Institute, Cary, NC, United States). The statistical methods of this study were reviewed by Nak-Hoon Son from the Department of Statistics, Keimyung University.

RESULTS

Baseline characteristics

A total of 132 patients were enrolled and randomly assigned to either fexuprazan (n = 66) or esomeprazole (n = 66). Of these, 4 patients (2 in each group) withdrew consent before receiving any dose of the assigned study medication and were excluded from the mITT analysis, leaving 128 patients (64 in each group) in the mITT population. From the mITT population, an additional 21 patients (13 in the fexuprazan group and 8 in the esomeprazole group) were excluded from the PP analysis for the following reasons: (1) Additional gastrectomy necessitated by non-curative resection (n = 9 in the fexuprazan group, n = 6 in the esomeprazole group); (2) Major protocol violations (n = 4 in the fexuprazan group); and (3) Treatment for another cancer was necessary (n = 2 in the esomeprazole group). Among these 21 patients, only 1 (fexuprazan group) had an evaluable 8-week assessment, which showed complete healing; the remaining 20 (12 in the fexuprazan group and 8 in the esomeprazole group) had no evaluable 8-week assessment and were therefore classified as treatment failures in the mITT and ITT analyses, as were the 4 patients who withdrew consent in the ITT analysis. An evaluable 8-week endoscopic assessment was thus obtained in 52 of 64 patients (81.3%) in the fexuprazan group and 56 of 64 patients (87.5%) in the esomeprazole group. Finally, 51 patients in the fexuprazan group and 56 patients in the esomeprazole group were included in the PP analysis (Figure 2). Characteristics of the patients and the lesions are summarized in Table 1. No significant differences were observed between the two groups. The two groups were balanced for eCura classification of ESD outcomes. In accordance with the exclusion criteria, no enrolled patient received antiplatelet or anticoagulant therapy during the study period. Drug compliance was more than 95% among all the patients (P > 0.05) (Supplementary Table 2).

Figure 2 CONSORT flow diagram. Patients without an evaluable 8-week endoscopic assessment were imputed as treatment failures in the modified intention-to-treat and full intention-to-treat analyses; these patients were among those excluded from the per-protocol population and do not constitute an additional group. ESD: Endoscopic submucosal dissection.

Table 1 Baseline characteristics of the per-protocol population, n (%)/median (interquartile range)/mean ± SD.

Variables	Fexuprazan 40 mg (n = 51)	Esomeprazole 40 mg (n = 56)	Mean difference	P value
Demographics
Age, years	65.69 ± 9.3	63.95 ± 9.8	1.74	0.3481
Sex				0.4602
Male	34 (66.7)	41 (73.2)	-6.55
Female	17 (33.3)	15 (26.8)	6.55
Body mass index, kg/m2	24.23 ± 3.4	25.06 ± 3.7	-0.83	0.2301
Smoking status				0.8863
Current	8 (15.7)	7 (12.5)	3.19
Ex-smoker	0 (0.0)	1 (1.8)	-1.79
Never	43 (84.3)	48 (85.7)	-1.40
Alcohol consumption				0.4682
Current	14 (27.5)	12 (21.4)	6.02
Never	37 (72.5)	44 (78.6)	-6.02
Helicobacter pylori infection				0.7092
Positive	20 (39.2)	20 (35.7)	3.50
Negative	31 (60.8)	36 (64.3)	-3.50
Comorbidities
Diabetes mellitus	9 (17.7)	11 (19.6)	-2.00	0.7912
Hypertension	22 (43.1)	31 (55.4)	-12.22	0.2072
Cardiovascular disease	1 (2.0)	4 (7.1)	-5.18	0.3662
Cerebrovascular disease	3 (5.9)	2 (3.6)	2.31	0.5722
Chronic kidney disease	0 (0.0)	2 (3.6)	-3.57	0.1732
Liver cirrhosis	0 (0.0)	0 (0.0)	N/A	N/A
Medications
Antihypertensive drug	20 (39.2)	29 (51.8)	-12.57	0.1922
Antidiabetic drug	9 (17.7)	11 (19.6)	-2.00	0.7912
Immunosuppressive agents	1 (2.0)	0 (0.0)	1.96	0.2922
Lesion characteristics
Basal mucosa
Atrophic gastritis	28 (54.9)	30 (53.6)	1.33	0.8902
Intestinal metaplasia	19 (37.3)	27 (48.2)	-10.96	0.2532
Longitudinal location				0.6162
Lower third (antrum/angle)	25 (49.0)	32 (57.1)	-8.12
Mid third (low body/mid body)	22 (43.1)	19 (33.9)	9.21
Upper third (upper body/cardia/fundus)	4 (7.8)	5 (8.9)	-1.09
Horizontal location				0.9642
Anterior wall	10 (19.6)	12 (21.4)	-1.82
Posterior wall	15 (29.4)	14 (25.0)	4.41
Greater curvature	9 (17.6)	10 (17.9)	-0.21
Lesser curvature	17 (33.3)	20 (35.7)	-2.38
Initial ulcer area, mm2	706.9 (451.6-962.1)	824.7 (490.9-1246.8)	N/A	0.1854
Histologic characteristics
Pathology				0.5932
Adenoma	30 (58.8)	29 (51.8)	7.04
Adenocarcinoma, differentiated	16 (31.4)	18 (32.1)	-0.77
Adenocarcinoma, undifferentiated	5 (9.8)	9 (16.1)	-6.27
Invasion depth				1.0002
Mucosa	49 (96.1)	53 (94.6)	1.44
Submucosa	2 (3.9)	3 (5.4)	-1.44
Curative resection among adenocarcinoma	21 (100)	27 (100)	0.00	1.0003
eCuraA	19 (90.5)	24 (88.9)	1.60
eCuraB	2 (9.5)	3 (11.1)	-1.60
eCuraC	0 (0.0)	0 (0.0)	0.00

¹Independent two sample t-test.

²χ² test.

³Fisher's exact test.

⁴Wilcoxon rank-sum test.

N/A: Not applicable.

Ulcer healing and shrinkage rate after ESD

The complete healing rate at 8 weeks, the primary endpoint, was significantly higher in the fexuprazan group (46 of 51, 90.2%) than in the esomeprazole group (42 of 56, 75.0%; P = 0.040; RD: 15.2%, 95%CI: 1.2-29.2) (Figure 3A). In the mITT population, the corresponding rates were 73.4% (47 of 64) and 65.6% (42 of 64; RD: 7.8%, 95%CI: -8.1 to 23.7; P = 0.337) (Figure 3B), and in the full ITT population 71.2% (47 of 66) and 63.6% (42 of 66; RD: 7.6%, 95%CI: -8.4 to 23.5; P = 0.353). The treatment effect was directionally consistent across populations, although statistical significance was reached only in the pre-specified PP analysis. The 8-week shrinkage rate for ESD-induced ulcers was also significantly higher in the fexuprazan group than in the esomeprazole group (100% vs 97.72%, P = 0.049) (Table 2). The 4-week shrinkage rate for ESD-induced ulcers was significantly higher in the fexuprazan group than in the esomeprazole group (95.92% vs 93.48%, P = 0.049). The complete healing rate was similar between the two groups at 4 weeks (Table 2).

Figure 3 Comparison of complete healing rates of the post-endoscopic submucosal dissection ulcer at 8 weeks. Complete healing at 8 weeks, the primary endpoint, was assessed by follow-up endoscopy and was defined as scar-stage healing (S1 or S2) without a residual mucosal defect according to the Sakita-Miwa classification. A: Per-protocol population (fexuprazan, n = 51; esomeprazole, n = 56): 46/51 (90.2%) vs 42/56 (75.0%); P = 0.040; B: Modified intention-to-treat population (n = 64 per group), in which patients without an evaluable 8-week assessment were classified as treatment failures: 47/64 (73.4%) vs 42/64 (65.6%); P = 0.337.

Table 2 Ulcer healing and shrinkage rate after endoscopic submucosal dissection in the per-protocol population, n (%)/median (interquartile range).

Variables	Fexuprazan 40 mg (n = 51)	Esomeprazole 40 mg (n = 56)	P value
Primary endpoint
Complete healing (100%)	46 (90.2)	42 (75.0)	0.040
Risk difference (95%CI)	15.2 (1.2-29.2)
Secondary endpoints
8-week shrinkage rate	100 (100-100)	97.72 (96.54-100)	0.049
4-week shrinkage rate	95.92 (92-98.53)	93.48 (88.69-97.36)	0.049
Healing stage at 4 weeks			0.464
Active (A1, A2)	3 (5.9)	1 (1.8)
Healing (H1, H2)	42 (82.4)	50 (89.3)
Scar (S1, S2)	6 (11.8)	5 (8.9)
Healing stage at 8 weeks			0.040
Active (A1, A2)	0 (0.0)	0 (0.0)
Healing (H1, H2)	5 (9.8)	14 (25.0)
Scar (S1, S2)	46 (90.2)	42 (75.0)

CI: Confidence interval.

Subgroup analysis by baseline ulcer area is presented in Supplementary Table 3. Among patients with baseline ulcer area ≥ 800 mm², the 8-week complete healing rate was 90.5% with fexuprazan vs 63.3% with esomeprazole (RD: 27.1%, 95%CI: 5.8-48.5; P = 0.029 by χ² test and P = 0.048 by Fisher's exact test). Among patients with smaller ulcers (< 800 mm²), the corresponding rates were 90.0% vs 88.5% (RD: 1.5%, 95%CI: -14.8 to 17.8; P > 0.999). The treatment-by-ulcer-size interaction for 8-week complete healing was not statistically significant (likelihood-ratio test, P = 0.190), and modelling baseline ulcer area as a continuous variable yielded an interaction P value of 0.841. These data therefore do not establish a differential treatment effect according to baseline ulcer size, and this subgroup finding, based on a threshold derived from the present dataset, is exploratory.

Time to complete ulcer healing after ESD

As an exploratory model-based analysis, we estimated the expected time required to reach a 99% reduction in ulcer area. The median estimated time to this endpoint was shorter in the fexuprazan group than in the esomeprazole group (40.31 days vs 47.22 days; P = 0.045) (Figure 4A). Subgroup analysis stratified by initial ulcer size demonstrated a consistent trend: Among patients with large ulcers (≥ 800 mm²), fexuprazan reduced the estimated time to healing compared with esomeprazole (42.89 days vs 47.62 days; P = 0.039), whereas no significant difference was observed for smaller ulcers (< 800 mm²), although a similar trend was noted (41.03 days vs 44.83 days; P = 0.588). Comparative healing trajectories stratified by baseline ulcer size are illustrated in Figure 4B and C. In a post hoc comparison, the model overestimated the observed 8-week ulcer area in 92.5% of patients (Supplementary Table 1), indicating that the derived healing times are conservative rather than optimistic estimates.

Figure 4 Exploratory analysis of the model-estimated time to a 99% reduction in ulcer area according to baseline ulcer area. In this per-protocol analysis, the time to a 99% reduction in ulcer area was estimated for each patient using an exponential decay model based on baseline and 4-week ulcer areas. A: Overall (40.31 days vs 47.22 days; P = 0.045); B: ≥ 800 mm² (42.89 days vs 47.62 days; P = 0.039); C: < 800 mm² (41.03 days vs 44.83 days; P = 0.588), shown for the fexuprazan vs esomeprazole groups. The curves are model-based extrapolations beyond the last observed measurement and do not represent observed times to endoscopic healing.

Safety

Safety was evaluated through TEAEs at 4 weeks and 8 weeks, along with physical examinations, vital signs, and laboratory results. In the safety population, TEAEs occurred in 4 of 64 patients (6.3%) in the fexuprazan group and 5 of 64 (7.8%) in the esomeprazole group (P > 0.999). All events were mild in severity, none was assessed as related to the study drug, and no serious TEAE, no treatment discontinuation due to an adverse event, and no death occurred during the study period (Supplementary Table 4). No significant between-group differences were observed in procedure-related AEs, including delayed bleeding or perforation (P > 0.05) (Supplementary Table 5). Only one patient in the esomeprazole group experienced delayed bleeding after ESD (1 of 64, 1.6%). The event occurred on post-ESD day 12 in a 51-year-old man (Forrest IIa, hemoglobin decreased from 14.0 g/dL to 8.6 g/dL); the bleeding was successfully managed with endoscopic hemostasis without transfusion, and no further complications were observed.

DISCUSSION

This is the first double-blind RCT to evaluate the efficacy and safety of fexuprazan, a novel P-CAB, for the treatment of artificial gastric ulcers after ESD. The main findings are as follows: (1) Fexuprazan achieved a significantly higher complete healing rate at 8 weeks than esomeprazole in the PP analysis, whereas neither the mITT nor the full ITT analysis showed a statistically significant difference; (2) Ulcer shrinkage rates at both 4 weeks and 8 weeks were significantly greater in the fexuprazan group, although these secondary outcomes are supportive rather than confirmatory; (3) Safety profiles were comparable between groups, with no serious AEs observed; and (4) Exploratory analysis suggested a shorter model-estimated time to a 99% reduction in ulcer area with fexuprazan, with the estimated time difference most pronounced in patients with larger baseline ulcers. The direction of effect favoured fexuprazan in every analysis performed, but the precision of the estimates was insufficient to establish superiority in the randomized populations, and the PP finding therefore requires confirmation.

Complete healing of ESD-induced artificial ulcers is essential to preventing post-ESD complications such as delayed bleeding or perforation. However, an 8-week course of PPI therapy may be insufficient to achieve complete healing in a subset of patients[23,24], prompting investigations into more intensive acid-suppressive strategies. These have included P-CAB therapy, as well as combination regimens incorporating mucosal protective agents, such as rebamipide or zinc-based formulations[16,19,25-29]. In addition, several meta-analyses synthesizing these heterogeneous approaches have been published[19,26,30]; however, the results remain inconsistent, and convincing evidence defining the optimal duration of acid-suppressive therapy, particularly for P-CABs, is still lacking.

P-CABs directly and reversibly inhibit the gastric H⁺/K⁺-ATPase via potassium-competitive binding, which allows for more potent and sustained acid suppression than PPIs and may overcome their inherent pharmacologic limitations[14,31]. Accordingly, because P-CABs provide stronger and more sustained acid suppression than PPIs, they would be expected to demonstrate superior or at least comparable healing efficacy for ESD-induced ulcers; however, reported outcomes have varied across studies. Systematic reviews and meta-analyses have reported superior healing of ESD-induced artificial ulcers with P-CAB therapy compared with PPIs[16,32], whereas another found no significant difference between the two drug classes[19]. Beyond healing, P-CABs have also been investigated for prevention of post-ESD delayed bleeding, with randomized trials suggesting a reduction in bleeding rates compared with PPIs, including in patients receiving antithrombotic agents[33,34]. Among currently available P-CABs, fexuprazan is a newly developed agent with minimal influence from CYP2C19 polymorphisms[20]. Notably, most previous RCTs evaluating P-CABs for post-ESD ulcers employed open-label designs, which may introduce assessment bias given the subjective nature of endoscopic ulcer evaluation. Our study addressed this limitation by employing a double-blind, double-dummy design, providing more rigorous evidence for the role of P-CABs in post-ESD ulcer healing. In the pre-specified PP analysis, fexuprazan achieved a significantly higher complete healing rate at 8 weeks than esomeprazole, with a comparable safety profile. This finding is consistent with the class-wide pharmacologic advantage of P-CABs, and the double-blind design of this trial strengthens the reliability of these results.

This study had several strengths. First, it was a multicenter, double-blind RCT, addressing a key methodological limitation of prior open-label trials by removing potential performance and detection bias in the subjective evaluation of endoscopic ulcer staging. Second, all ESD procedures and follow-up endoscopies were performed by three expert endoscopists, each with experience exceeding 500 ESD procedures, ensuring procedural consistency and high technical quality. Third, exploratory analysis provided preliminary insight into the time course of ulcer healing under P-CAB therapy, which may inform future trials designed to optimize treatment duration.

This study had several limitations. First, the sample size was relatively small and powered to detect a difference in the complete healing rate rather than in clinical events such as delayed bleeding. Complete ulcer healing was selected as the primary endpoint as a surrogate for clinical outcomes, and detecting a difference in delayed bleeding would have required a substantially larger trial. Complete healing has nonetheless served as a commonly used endpoint in pivotal P-CAB trials in this indication, although the correspondence between healing kinetics and clinical events remains to be established. Delayed bleeding was included as a secondary endpoint, and only one event occurred, precluding meaningful between-group comparison. Second, the primary endpoint was statistically significant in the PP but not in the mITT analysis, and the post-randomization exclusions were numerically imbalanced (13 fexuprazan vs 8 esomeprazole). Although the principal reason for exclusion was unrelated to the assigned treatment, attrition bias cannot be excluded. In addition, the primary endpoint, the ulcer shrinkage rates, and the exploratory time-to-healing analysis all yielded P values close to the conventional significance threshold without multiplicity adjustment, and the CI for the absolute difference in complete healing was wide (95%CI: 1.2-29.2). These considerations, together with the limited sample size noted above, indicate that the present findings require confirmation in a larger randomized trial. Third, the trial was conducted exclusively at two Korean referral centers, and fexuprazan is currently approved in a limited number of countries, which may limit the direct generalizability of these results to other populations. Fourth, although randomization was performed using permuted blocks of size 4, no stratification factors - such as baseline ulcer size or center - were pre-specified. A numerically larger initial ulcer area was observed in the esomeprazole group, and although this difference did not reach statistical significance, it may have introduced residual imbalance in the overall treatment-effect estimate. The greater apparent benefit in the ≥ 800 mm² subgroup is hypothesis-generating, as the treatment-by-ulcer-size interaction was not significant, and future trials should prespecify baseline ulcer size as a stratification factor. Fifth, esomeprazole 40 mg once daily was used as the comparator, consistent with the approved dose for gastric ulcer treatment in Korea; however, the standard dose differs in some countries, where 20 mg is used, which should be considered when interpreting the results. Sixth, all patients received intravenous esomeprazole during the first 48 hours after ESD because intravenous fexuprazan is not currently available, in line with standard post-ESD management[7,10]; consequently, the immediate post-ESD acid-suppressive effect of fexuprazan could not be evaluated. This trial therefore evaluates fexuprazan initiated approximately 48 hours after ESD, and the identical early exposure may have attenuated rather than exaggerated the observed between-group difference. Seventh, the time-to-healing analysis was exploratory and model-based; because the decay constant was derived from only two ulcer-area measurements per patient, the estimate is sensitive to measurement error and to the assumption of an exponential trajectory. This study was not powered to determine the optimal treatment duration, and the model-estimated time to a 99% reduction in ulcer area was extrapolated from endoscopies performed only at weeks 4 and 8. More frequent endoscopic assessments and a pre-specified longitudinal modeling strategy would be required to precisely determine the earliest time point at which scar formation is achieved. Eighth, this study did not include a formal cost-effectiveness analysis. Although fexuprazan and esomeprazole are similarly priced in Korea, P-CAB pricing varies considerably across countries, and whether the observed healing benefit translates into cost-effectiveness in other settings warrants dedicated pharmacoeconomic evaluation. Finally, the study medications and matching placebos were supplied free of charge by the manufacturer of fexuprazan; although the company had no role in its design, conduct, analysis, interpretation, or reporting, this involvement should be taken into account when interpreting the results.

CONCLUSION

In conclusion, in this multicenter, double-blind RCT, fexuprazan 40 mg achieved a significantly higher complete healing rate of post-ESD gastric ulcers at 8 weeks compared with esomeprazole 40 mg in the pre-specified PP analysis, with a comparable safety profile; the difference was directionally consistent but not statistically significant in the mITT analysis. The benefit was numerically larger in patients with large baseline ulcers (≥ 800 mm²), in whom acid-suppressive therapy has historically been most challenging, although this subgroup finding is hypothesis-generating. Exploratory analysis further suggested a shorter model-estimated time to a 99% reduction in ulcer area with fexuprazan. Because the randomized-population analyses did not reach statistical significance and the PP result is susceptible to post-randomization exclusion bias, the evidence of superiority remains inconclusive. Future large-scale, multicenter RCTs with more frequent endoscopic assessments and with clinical endpoints such as delayed bleeding are warranted to confirm these findings.

ACKNOWLEDGEMENTS

The authors thank the research nurses and endoscopy unit staff at Yongin Severance Hospital and Severance Hospital for their dedicated assistance with patient enrollment, study coordination, and endoscopic follow-up. The study medications and matching placebos were supplied free of charge by Daewoong Pharmaceutical Co., Ltd., Seoul, Republic of Korea, the manufacturer of fexuprazan. No financial support was received for this trial.