Rebamipide Plus Proton Pump Inhibitor Versus Proton Pump Inhibitor Alone in the Treatment of Endoscopic Submucosal Dissection-Induced Gastric Ulcer: Retracted

Proton pump inhibitor (PPI) was the main prescription for gastric ulcer after endoscopic submucosal dissection (ESD). Some randomized controlled trials showed that a combination of rebamipide and PPI appears to be more efficient than PPI alone for the treatment of ESD-induced gastric ulcer. However, the sample sizes in these trials were limited and the conclusions were underpowered.This meta-analysis was conducted with 5 randomized controlled trials using the combination of rebamipide and PPI for healing ESD-induced ulcer compared with PPI monotherapy. Relevant studies were searched via MEDLINE, PubMed, Embase, and Cochrane Library databases by using terms such as "rebamipide," "proton pump inhibitor," "endoscopic submucosal dissection," "drug therapy," and "gastric ulcer or artificial ulcer."Five studies were included in this meta-analysis. The number of total patients was 626, with 317 patients in the combination group and 309 patients in the PPI alone group. The heterogeneity among these 5 studies was low (I = 22%, P = 0.28). All 5 studies considered scarring stage 1 rate as a primary endpoint, and the scarring stage 1 rate in combination group (115/317) was higher than that in PPI alone group (63/309) (odds ratio 2.61, 95% confidence interval [CI] 1.76-3.88). The mean difference of initial ulcer size between 2 groups was -4.46 (95% CI -266.61 to -257.69, P = 0.97), but it enlarged to 68.38 (95% CI 35.72-101.05, P < 0.00001) in the 4th week.This meta-analysis demonstrates that combination therapy is more efficient than PPI monotherapy in healing ESD-induced gastric ulcer.


INTRODUCTION
E ndoscopic submucosal dissection (ESD) was developed since the 1990s 1,2 and introduced in China in 2006. In recent years, ESD is considered to be more efficient for resection of gastric adenoma or cancer than endoscopic mucosal resection (EMR), especially for those with tumor size >20 mm and those with flatter shape. 3,4 However, ESDinduced gastric ulcer is much deeper and larger than the one induced by EMR. Compared with EMR, patients receiving ESD have a higher risk of perforation and delayed bleeding, 5 which are considered as the main complications of ESD. The frequency of delayed bleeding for ESD-induced ulcer was up to 7%, 6 and the incidence rate of perforation was about 4%. 5,6 Perforation could be typically closed with the aid of endoclips during ESD, because the stomach is clean because of fasting before undergoing this procedure. Delayed bleeding is very common after ESD, whereas it is closely related to tumor size and location. 6 So, healing of ESD-induced ulcer is quite important for protecting patients from delayed bleeding.
For the treatment of ESD-induced gastric ulcer, proton pump inhibitor (PPI) or histamine-2 receptor (H2R) antagonist 5 was mainly prescribed by physicians. Although it was found that PPI was more efficient than H2R, 4 weeks of PPI monotherapy was not enough for healing ESD-induced gastric ulcer. Recently, rebamipide (Mucosta; Otsuka Pharmaceutical, Tokyo, Japan), a mucosal protective drug developed in Japan, has been widely used on Asians and Egyptians. Results of some clinical trials indicated that the combination of rebamipide and PPI was more efficient than PPI monotherapy for ESD-induced ulcer. However, sample sizes were limited and conclusions were underpowered. Therefore, we systematically searched the databases and conducted this meta-analysis that aims to get the evidence for the combined prescription of ESD-induced ulcer.

Search Strategy and Selection Criteria
We searched the databases including MEDLINE, PubMed, Embase, Chinese National Knowledge Infrastructure, China Biology Medicine, VIP Database for Chinese Technical Periodicals, and the Cochrane Library to find studies in which the combination of rebamipide and PPI was Editor: Bulent Kantarceken.
used to treat ESD-induced gastric ulcer compared with PPI monotherapy. We used the terms "rebamipide," "proton pump inhibitor," "endoscopic submucosal dissection," "drug therapy," and "gastric ulcer or artificial ulcer" for searching. We also searched the references listed of the articles.
All the studies included were randomized controlled trials (RCTs). Patients included in these trials had early gastric cancer or gastric adenoma or polyps resected by ESD. Included in this study were ESD-induced artificial gastric ulcers, which were treated by the combination of rebamipide and PPI compared with PPI monotherapy. Main outcomes focused on the ulcer healing stage and the healing quality between the 2 groups. Exclusive criteria were as follows: first, only the abstract but not the full text of the RCT was accessible, and second, the main outcome could not be obtained even after we communicated with the article authors.

Quality Assessment and Data Extraction
The quality of RCTs included was evaluated according to the criteria proposed by the Cochrane Reviewer's Handbook 5.1.0. It consists of 7 aspects that include random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting, and other bias. The quality of articles was labeled as low-risk bias, high-risk bias, or unclear bias.
Two authors reviewed and estimated the RCTs independently according to the assessment system. A third author attended the estimation in case of disagreement and all 3 of them discussed until they got accordance with each other. The data extracted from RCTs included basic characteristics (such as authors, publication year, number of patients, and drug application in trials). The ulcer stage was evaluated by using the classification of Sakita and Fukutomi, 7 described as active (A1, A2), healing (H1, H2), and scarring stages (S1, S2). Moreover, we also collected the data about the ratio of S1 stage, which was considered as the standard of ulcer healing, and the residual ratio of artificial ulcer, which means residual ulcer size or initial ulcer size as the main endpoints.

Statistical Analysis
We used Review Manager (RevMan) software (version 5.0; Cochrane Collaboration, http://ims.cochrane.org/revman/ download) to conduct the meta-analysis. Odds ratio (OR) with 95% confidence intervals (CIs) and the weighted mean difference (MD) were recommended for dichotomous data and continuous data, respectively. Chi-square was used to assess heterogeneity, and the value of I 2 equal to 25%, 50%, and 75% represent low, moderate, and high heterogeneity, respectively. A difference with P value <0.05 was considered significant. If I 2 was <50%, we took the fixed effects model to calculate meta-analysis. If it was >50%, we assumed that the random effects model and subgroup analysis would be explored as dealing with the data.

Study Inclusion and Assessment
Seven hundred eight studies were identified by our search strategy and only 5 prospective RCTs met our criteria for the meta-analysis. The number of total patients in the studies was 626, with 317 patients in the combination group and 309 patients in the PPI alone group. Table 1 shows the characteristics of 5 studies. Four studies [8][9][10][11] were carried out in Japan and 1 in Korea; out of the 5 RCTs, the study in Korea 12 had the most number of patients. We assessed the 5 RCTs using the criteria proposed in the Cochrane Reviewer's Handbook 5.1.0. Figure 1 listed the existing biases in 7 aspects of all RCTs. Lack of blinding of participants and outcome assessment may be the limitations in almost all the RCTs.

S1 Stage Rate
All 5 RCTs collected S1 stage rate as the main endpoints for assessing the healing of ESD-induced artificial ulcer. The heterogeneity among these 5 studies was low (I 2 ¼ 22%, P ¼ 0.28), so fixed effect model was applied in the meta-analysis. As shown in Figure 2, the S1 stage rate in the combination group (115/317) was higher than that in the PPI alone group (63/309) (OR 2.61, 95% CI 1.76-3.88). Only 2 studies compared the S1 stage rate between combination and PPI alone groups according to the diameter of ulcer <40 mm or >40 mm and no heterogeneity was detected in the 2 studies (I 2 ¼ 0%). Figure 3 showed that the ESDrelated ulcer with diameter 20-40 mm had a higher S1 stage rate in the combination group than in the PPI alone group (OR 4.49, 95% CI 2.26-8.91). Moreover, it was confirmed that the ESD-related ulcer with diameter >40 mm also had a higher S1 stage rate in the combination group than in the PPI alone group (OR 12.66, 95% CI 2.04-78.70) ( Figure 4).

Ulcer Healing Quality
Although S1 stage rate was considered as a primary endpoint in all 5 studies, the ulcer healing rate, which was conducted in 2 studies, was collected as a secondary endpoint based on the data of initial ulcer size and ulcer size at 4 weeks. However, one study 12 calculated the ulcer reduction rate as (initial size À ulcer size at 4 weeks)/initial size, whereas another study 10 calculated the ulcer healing rate as ulcer size at 4 weeks/initial size. So, we did metaanalysis of initial ulcer size and ulcer size at 4 weeks of both groups. As shown in Figure 5, the MD of initial ulcer size between 2 groups was not significant (MD À4.46, 95% CI À266.61 to À257.69, P ¼ 0.97), but that of the ulcer size at 4 weeks between 2 groups was significant as shown in Figure 6 (MD 68.38, 95% CI 35.72-101.05, P < 0.00001), which favors the combination group.

Complications
In the study by Shin et al, 12 31 patients dropped out of the study because of delayed bleeding. In the study by Kato et al, 9 perforation and delayed bleeding did not occur in both groups. In their studies, minor bleeding during ESD could be easily treated with heat coagulation under endoscopy. In the study by Fujiwara et al, 10 no adverse events occurred in the combination group and only 1 patient experienced delayed bleeding on the fifth day after ESD, which was finally stopped by soft coagulation of the exposed blood vessel within the ulcer under endoscopy. Kobayashi et al 8

DISCUSSION
We included 5 RCTs for this meta-analysis by computer search and manual screening. Moreover, there was low heterogeneity within all 5 RCTs (I 2 ¼ 22%, P ¼ 0.28) as we compared the S1 stage rate that was used as the primary endpoint in all studies. All 5 studies were carried out in Asia: 4 in Japan and 1 in Korea, as rebamipide (Mucosta; Otsuka Pharmaceutical) was mainly applied in Asian countries and Egypt. 10 According to the endoscopic staging of gastric ulcers by Sakita-Fukutomi 7 classification, the S1 stage was used as a standard for ulcer healing. The meta-analysis showed that the S1 stage rate was higher in the combination group than that in the PPI alone group (OR 2.61, 95% CI 1.76-3.88). It was reported that the initial ulcer size of >20 mm was an independent risk factor for delayed bleeding and affected ulcer healing. 13 In this meta-analysis, 2 studies (I 2 ¼ 0%) took this for consideration and compared the S1 stage rate of Fujiwara et al 10 Kato et al 9 Kobayashi et al 8 Shin et al 12 ESD-induced ulcers according to its diameter of <40 mm or >40 mm. So, we compared the S1 stage rate of ESD ulcers according to its size between 20 to 40 and >40 mm. It was also found that no matter whether the diameter of an ulcer is <40 mm or >40 mm, the S1 stage rate was higher in the combination group than that in the PPI alone group. Moreover, the OR of S1 stage rate between 2 kinds of therapies in ESD ulcers with diameter >40 mm was even higher than that with diameter <40 mm (OR 12.66 vs OR 4.49). From the data above, we may get the conclusion that combination therapy is more effective than PPI alone for ESD-induced gastric ulcers, especially for those ulcers with diameter >40 mm. Because the S1 stage rate was qualitative data, another 2 studies collected the initial ulcer size and ulcer size at 4 weeks. The meta-analysis showed that the initial ulcer size between the combination group and the PPI alone group had no significant difference, but ulcer size at 4 weeks of therapy was much smaller in the combination Araki 11 Fujiwara et al 10 Kato et al 9 Kobayashi         group than in the PPI alone group (MD 68.38, 95% CI 35.72-101.05, P < 0.00001), which further confirmed that combination therapy was more effective than PPI alone for ESD artificial ulcer. ESD-induced gastric ulcer is different from pathologic peptic ulcer and it remains unclear about the mechanisms of this artificial ulcer healing. Kakushima et al 14 reported that ESDinduced gastric ulcers would heal within 8 weeks regardless of ulcer size, location, Helicobacter pylori infection, or even the extent of gastric atrophy. Size reduction of ESD-induced gastric ulcer occurs by contraction in the early phase, and then regenerative mucosa covers the remaining mucosal defect within 8 weeks. 15 However, Oh et al 16 found the healing degree of ESD ulcers depended on the initial ulcer size at 4 weeks. This meta-analysis also confirmed that the healing degree of ESD-induced ulcers was lower in ulcers with initial ulcer size >40 mm than that with diameter between 20 and 40 mm, no matter whether in the combination group or the PPI alone group.
Although it was thought that the H pylori infection is not correlated to ulcer healing, Shin et al 12 (1 of the 5 RCTs) found that an absence of H pylori was a predictor of superior ulcer healing. Huang et al 17 also found that the H pylori infection and the presence of pathological ulcer findings within the ESD specimen were significantly related to the risk of ESD ulcer recurrence. Although the frequency is low, there is a possibility of ESD ulcer recurrence in patients with H pylori infection and those who undergo ESD for lesions with ulceration. 17 As in this meta-analysis, several studies did not show the difference of ulcer healing between H pylori positive and negative, we could not figure out the effect of H pylori on ulcer healing.
However, Fujiwara et al 10 found that severe atrophic gastritis might contribute to delayed healing of ESD-induced ulcers even after 8 weeks of PPI administration, because PPI was just an acid suppressor. Although Kakushima et al 14 found that the H pylori status and gastric atrophy did not affect ulcer healing after ESD, their study underestimated the effect of atrophic gastritis on ESD ulcer healing. It was possible that the action of sucralfate (combined with PPI in their study), which was a cytoprotective agent creating a protective barrier between stomach acid and wound tissues, covered the effect of atrophic condition in the stomach.
Not only the atrophic condition in the gastric mucosa but also the PPI refractory should be taken into consideration for PPI monotherapy of ESD-related ulcers. As PPIs are metabolized primarily by cytochrome P450 2C19 (CYP2C19) in the liver, it was reported that CYP2C19 genotype-dependent differences of PPI influenced the cure rates for gastroesophageal reflux disease and H pylori infection in response to PPI-based therapies. 18 So, ESD-induced ulcers may not heal due to nonresponse to PPI monotherapy. Nonetheless, there is no consensus on the dose and duration of PPI needed for ESDinduced ulcers. Some studies reported that a half dose of PPI for 4 weeks was equivalent to a full dose of PPI for 4 weeks, 19 and some studies reported that 4 weeks was not sufficient but 8 weeks was needed to cure ESD-related ulcers by PPI monotherapy. 14,20 Further studies are needed to determine the dose and duration of PPI and to estimate its cost effectiveness.
Rebamipide, as a mucosal protective drug, has several biological activities such as increasing prostaglandin concentration, 21 upregulating EGF and EGF receptors, 22 stimulating angiogenesis, and inhibiting neutrophils activated by H pylori 23 and nonsteroidal anti-inflammatory drugs. 24,25 It was also found that rebamipide aided in the eradication of H pylori infection. 26 So, for patients infected with H pylori, rebamipide can promote ulcer healing not only by reducing the inflammatory response but also by eradicating H pylori. In the combination therapy, rebamipide also makes up for ulcer healing in patients with no response to PPI as they have an atrophic mucosal condition or are genotypedependent PPI refractory. However, in this meta-analysis, we cannot figure out the difference between the combination and monotherapy groups of the status of patients, such as H pylori infection, genotype polymorphysim, and atrophic mucosal condition (as the data is not shown). These factors were also considered in all of the included studies, and further studies on these factors were thought to be necessary by them.
Our study has several limitations. First, only 5 RCTs were included in this meta-analysis; the number is too limited. Second, the main outcome that could be found in all 5 studies for meta-analysis was only the S1 stage rate. Little data were known about the factors that can be used for subgroup analysis.
In conclusion, the combination therapy is more efficient than PPI monotherapy in healing ESD-induced gastric ulcers, especially in ulcers >40 mm.  . Ulcer size at the end of therapy comparing 2 groups.