Chronic celecoxib users more often show regression of gastric intestinal metaplasia after Helicobacter pylori eradication


Dr B.-S. Sheu, Department of Internal Medicine, National Cheng Kung University Hospital, 138 Sheng Li Road, Tainan 70428, Taiwan.



Background and Aim

To test whether the chronic users of celecoxib, a selective cyclo-oxygenase-2 inhibitor, had less Helicobacter pylori-related intestinal metaplasia or if such users’ intestinal metaplasia could be prone to disappear after H. pylori eradication.


The study enrolled 150 chronic celecoxib users and 216 non-users who underwent pan-endoscopy to detect H. pylori infection and its related intestinal metaplasia. One hundred and three H. pylori-infected patients with intestinal metaplasia (43 chronic celecoxib users and 60 non-users) received anti-H. pylori therapy and completed the 12-month follow-up to survey the regression of intestinal metaplasia by mean intestinal metaplasia score.


There were no differences in the prevalence of H. pylori-related intestinal metaplasia between the chronic celecoxib users and controls (P > 0.05). On the 12th month of follow-up, chronic celecoxib users had a lower mean intestinal metaplasia score (1.2 vs. 1.8, P < 0.005) and a higher regression rate of intestinal metaplasia (42% vs. 20%, P = 0.027) than non-users.


With H. pylori infection, chronic celecoxib users still showed limited effects to decrease intestinal metaplasia. Nevertheless, celecoxib should be promising to assist H. pylori eradication for the control of gastric intestinal metaplasia and cancer risk.


Helicobacter pylori is a type I carcinogen identified by the WHO and is recognized as an important pathogen closely related to gastric carcinogenesis.1, 2 Gastric cyclo-oxygenase-2 (COX-2) expression is now attributed as an early event during gastric carcinogenesis3 and can be highly upregulated in H. pylori-infected premalignant gastric changes, such as intestinal metaplasia (IM).4–7 Despite H. pylori eradication, patients with pre-existing IM have persistently high levels of COX-2 expression, even after 2-year follow-up.7 This indicates that H. pylori eradication can only partially reverse IM.

Adding some agent, such as COX-2 inhibitor, to suppress gastric COX-2 expression and reverse IM could be promising for future pilot studies. Alternatively, another promising study would be to determine whether the outcome of H. pylori-related IM is different between long-term non-steroid anti-inflammatory drug (NSAID) users and non-users.

This study enrolled chronic users of celecoxib, a selective COX-2 inhibitor, to test whether such patients may have a less prevalence rate of H. pylori-related IM, and to determine whether H. pylori eradication can reverse IM for such patients. The data should be highly valuable in improving regression of IM and could serve as a supplement control to H. pylori eradication for decreasing the risk of gastric carcinogenesis.

Materials and methods

Patients and study design

This study consecutively enrolled 366 dyspeptic patients, including 150 chronic celecoxib users and 216 non-NSAID users who served as controls. The chronic celecoxib users were defined as having past celecoxib use of more than 3 months before enrolment and requiring long-term maintenance for a background disease such as osteoarthritis or rheumatoid arthritis. This study did not include the chronic celecoxib users with concurrent use of steroids, immune suppressors, other NSAIDs and aspirin. Co-mediation history was matched for the two groups, except for the celecoxib use. None of the enrolled patients had a known history of H. pylori eradication. Moreover, to prevent bias, patients were screened for exposure to antibiotics, bismuth salts and antisecretory agents in the previous 4 weeks before gastroscopy.

After obtaining informed consent, each patient underwent gastric biopsy through pan-endoscopy for histology. At least five bits of gastric biopsy, including two from the antrum, two from the lower corpus and one from the high corpus, were sampled.8 Gastric histology of each patient was reviewed for the presence of gastric atrophy (AT), IM and other H. pylori-related features using the updated Sydney's system. In addition, each patient received a 13C-urea breath test (UBT) as in our previous studies.9 In this study, the presence of H. pylori infection was defined by positive results of both histology and the UBT.

Anti-H. pylori therapy (amoxicillin 1 g, clarithromycin 500 mg and omeprazole 20 mg twice daily for 1 week) was uniformly started for H. pylori-infected patients with IM. Patients with allergies to penicillin or clarithromycin were not included. Six weeks after triple therapy, these patients received a 13C-UBT to ascertain the success of H. pylori eradication.9 For those patients with successful H. pylori eradication, endoscopy was repeated on the 12th month to obtain gastric biopsy for H. pylori-related histology.

During this 1-year study period, the dosage of celecoxib was scheduled as at least 200 mg/day. In contrast, the control group refrained as much as possible from using any NSAID. Control patients with a 2-week continuous use of NSAIDs during follow-up were dropped from the analysis and patients taking <75% of the celecoxib during the study period were likewise excluded.

Analysis of H. pylori-related histology

The endoscopic findings were examined by a single pathologist, who was blinded to the groupings, for gastric histology before and after H. pylori eradication for 12 months. The acute inflammatory score (AIS; range: 0–3), chronic inflammation score (CIS; range: 0–3), atrophic change (absence: 0; presence: score 1–3) and IM (absence: 0; presence: score 1–3) were graded using the updated Sydney's system.10 The total AIS and CIS was the sum of the three specimens (range: 0–9). AT was defined as the presence of atrophy in any one bit of gastric biopsy.

The presence of IM was strictly defined as at least one gastric biopsy with score of 1, which should have had metaplastic cells (goblet cells) involving at least 5% of the upper third of the gastric mucosa. Once two or more bits of biopsy with goblet cells were located on <5% of the upper third of the mucosa, the presence of IM was defined due to their global existence.7

In this study, for IM patients, the mean intestinal metaplasia score (MIM) was defined as the sum of all IM scores from the sampled tissues divided by the number of tissues reviewed for the IM score (range: 0–3). Therefore, the MIM was compared before and after H. pylori eradication on the 12th month of follow-up. Total regression of IM in histology was defined as absence of IM (score 0) for all the gastric specimens.

Statistical analysis

This study assumed the 1-year regression of gastric metaplasia would be around 20% in the control, and thus the effect by celecoxib would become clinically worthwhile if the regression rate could be increased by 15%. Accordingly, a sample as around 50 for each arm was required to achieve a power of 80% at a 5% level of significance.

The Student's t-test or paired t-test was applied where appropriate for the parametric differences. The Fisher's exact test and chi-square test were used as appropriate for the non-parametric proportions. The odds ratio and a 95% confidence interval (CI) were also tested between the different study groups. All tests were two-tailed with a P < 0.05 was considered statistically significant. Multivariate logistic regression was used to elucidate the independent risk factor related with the total regression of IM in celecoxib users after H. pylori eradication. The coefficient value and 95% CI were enclosed for the factor proven to have a significant P-value of <0.05.


Prevalence of H. pylori infection and related histology before treatment

There were 366 dyspeptic patients consecutively enrolled, including 150 chronic celecoxib users and 216 non-users as controls. The prevalence rate of H. pylori infection was similar between the chronic selective COX-2 inhibitor users and non-user controls [72% (108 of 150) vs. 66% (142 of 216), P > 0.05]. For H. pylori-infected patients (Table 1), the demographic parameters, endoscopic diagnosis and pre-treatment histological features, especially the prevalence rate of IM, were similar between the 108 celecoxib users and 142 non-user controls (P > 0.05). Moreover, for the 108 chronic celecoxib users with H. pylori infection, there were similarities in age, sex, disease backgrounds and duration of chronic celecoxib use between those with and without IM. Only smoking history was shown to be higher in the chronic celecoxib users with H. pylori-infected IM than in those without IM (35% vs. 14%, P < 0.05).

Table 1.   The difference of histology between Helicobacter pylori-infected patients with and without chronic celecoxib usage
 Chronic celecoxib users (n = 108)Controls (n = 142) P-value
  1. Peptic ulcer: includes both duodenal ulcers and gastric ulcers.

  2. AIS, acute inflammatory score; CIS, chronic inflammatory score; AT, atrophy; IM, intestinal metaplasia.

Mean age (years)52.352.8N.S.
Smoking, % (n)25 (27)26.1 (37)N.S.
Gender (female:male)63:4580:62N.S.
Endoscopic diagnosis (n, peptic ulcer:gastritis)35:7352:90N.S.
AIS (0–9; mean ± s.d.)1.2 ± 0.51.3 ± 0.4N.S.
CIS (0–9; mean ± s.d.)5.9 ± 1.86.1 ± 1.8N.S.
AT, % (n)43.5 (47)40.8 (58)N.S.
IM, % (n)44.4 (48)45.8 (65)N.S.

Regression of H. pylori-related intestinal metaplasia during follow-up

There were 48 chronic NSAID users and 65 controls with H. pylori infection disclosed to have IM in gastric histology. The demographic backgrounds and the endoscopic diagnosis were similar between the two groups (Table 2). All of them were invited to receive the anti-H. pylori therapy. As illustrated in Figure 1, 108 patients (43 chronic NSAID users and 60 controls) completed the 12-month follow-up and achieved successful H. pylori eradication. For the 108 patients who completed the study design, the pre-treatment and post-treatment MIM after H. pylori eradication were compared within each group and between the two groups (Figure 2). There was no difference of the pre-treatment MIM between the chronic celecoxib users and controls (2.46 vs. 2.48, P > 0.05).

Table 2.   Disease background and endoscopic features of Helicobacter pylori-infected celecoxib users and controls with intestinal metaplasia
GroupCelecoxib users (n = 48) Controls (n = 65) P-value
Mean age (years)51.950.2N.S.
Smoking, % (n)35.4 (17)36.9 (24)N.S.
Gender (female:male)28:2037:28N.S.
Endoscopic diagnosis (n; peptic ulcer:gastritis)32:1640:25N.S.
Figure 1.

 Schematic flowchart of the study design and patient number listed for the two study groups during follow-up. Controls: non-non-steroid anti-inflammatory drug users. Helicobacter pylori-positive and H. pylori-negative indicate the presence or absence of H. pylori infection, respectively. Intestinal metaplasia (IM)-positive and IM-negative indicate the presence or absence of IM, respectively.

Figure 2.

 The difference of mean intestinal metaplasia scores (MIM) before (pre-treatment score) and after (post-treatment score) Helicobacter pylori eradication in each study group. There was no difference in pre-treatment MIM between the chronic celecoxib group and control group (P > 0.05). After H. pylori eradication, there was a significant decrease of MIM with both the chronic celecoxib group and the control group (P < 0.001, by paired t-test). Moreover, the chronic celecoxib group had a significantly lower post-treatment MIM on the 12th month follow-up (* indicated a P-value of <0.005 by the Student's t-test).

After H. pylori eradication, there was a significant decrease of MIM within either the chronic celecoxib group or the control group (P < 0.001, by paired t-test). Moreover, in Figure 2, the chronic celecoxib users had a lower post-treatment MIM than the non-user controls (1.2 vs. 1.8, P < 0.005). On the 12th month of follow-up, the chronic celecoxib users had a higher regression rate of IM than the controls (42% vs. 20%, P = 0.027, relative risk: 2.9, 95% CI: 1.88–6.91).

Factors related with the total regression of IM in chronic celecoxib users

In Table 3, this study further revealed that there were demographic factors significantly related with the total regression of IM, defined as the post-treatment MIM equal to zero, in the celecoxib users after H. pylori eradication. By univariate analysis, the patients with total regression of IM had a lower rate of smoking history (P < 0.001), a younger age (P < 0.005), and a lower pre-treatment MIM score (P < 0.001). Multivariate logistic regression confirmed that only higher pre-treatment MIM was an independent risk factor related to failure of IM total regression after H. pylori eradication in chronic celecoxib users (P = 0.002, coefficient value = 1.8, 95% CI: 1.14–4.63).

Table 3.   Factors correlated with total regression of intestinal metaplasia in the Helicobacter pylori-infected celecoxib users after H. pylori eradication (univariate analysis)
Total regression of intestinal metaplasiaYes (n = 18)No (n = 25) P-value
  1. * The past duration of celecoxib indicated the period of using celecoxib before H. pylori eradication had been longer than 1 year.

Mean age (years)4854.8<0.005
Smoking, % (n)11.1 (2)60 (15)<0.001
Gender (female:male)10:813:12N.S.
Dosage more than 200 mg celecoxib, % (n)27.8 (5)12 (3)N.S.
Past duration of celecoxib >1 year, % (n)*1015N.S.
Mean pre-treatment MIM (range: 0–3)2.22.7<0.001


This study demonstrated that the prevalence rate of H. pylori infection was not different between chronic celecoxib users and non-user dyspeptic controls. In Table 1, the rate of IM was also not different among H. pylori-infected patients with and without chronic celecoxib usage. These data indicated that H. pylori infection was still rather common in dyspepsia patients despite the chronic usage of NSAID, such as celecoxib. Moreover, once the H. pylori infection persisted, medication with NSAID to suppress the COX-2 expression could not lead to a lower IM rate. So, it would be interesting to determine the potential effect of celecoxib in downregulating COX-2 expression to improve the IM regression after H. pylori eradication.

This study enrolled patients with IM to receive anti-H. pylori therapy, which revealed that patients that kept on using celecoxib for 12 months after H. pylori eradication had a 4.7-fold increase in regression of IM, when compared with those who just received anti-H. pylori therapy. Our findings indicated that the use of selective COX-2 inhibitors, such as celecoxib, after H. pylori eradication could be promising to reverse the IM and decrease the risk of gastric carcinogenesis.

However, there was a negative finding by rofecoxib, another selective COX-2 inhibitor, to improve the regression of IM after H. pylori infection.11 The reason why chronic celecoxib users, but not rofecoxib, had a higher rate of IM regression than non-NSAID users was unknown. The first possibility could be the case selection of more severe IM in our series, when compared with the rofecoxib study. In this study, the MIM score was around 2.5 (Figure 2) and higher than that of the rofecoxib study's level. The lower initial score might have dampened the effect of treatment.

The second possibility could be due to the different suppression levels of COX-2-related pathway between celecoxib and rofecoxib. Moreover, there should be some other pathways, such as the COX-2-independent ones, to account for the different results between celecoxib and rofecoxib on the regression of gastric IM. In an elegant review,12 the COX-2-independent anticarcinogenic effects of different selective COX-2 inhibitors were compared. The review showed that celecoxib could be more advantages than rofecoxib in several pathways achieving anticarcinogenesis, including mechanisms involved in the inhibition of cell cycle progression, induction of apoptosis and inhibition to angiogenesis and metastasis.12

Recently, a novel animal study disclosed that celecoxib could significantly decrease gastric tumour volume, but had no significant change in COX-2 and prostaglandin signals under COX-2 inhibition.13 The animal study findings in gastric cancer, as well as our clinical findings in the precancerous lesion like IM, implied that certain NSAID, especially celecoxib, should be involved in the COX-2-independent pathway, such as NF-κB, and IκB. Thus, they can counteract the progression of gastric carcinogenesis.13–16 Further study would be necessary to trace the exact mechanism of celecoxib to counteract gastric carcinogenesis, especially as early as the precancerous step.

Eradication of H. pylori should be protective against the progression of premalignant gastric lesions.2, 17, 18 However, there were persistent, even progressive, premalignant changes in more than 50% of subjects during a 5-year follow-up study.17 This data support the need of early H. pylori eradication to prevent progression of IM. Moreover, aside from H. pylori eradication, the study implied that there should be something added to prevent further IM progression or even more aggression. Our study revealed that the regression of IM could be more evident in chronic celecoxib users after H. pylori eradication. Up to now, there were no other COX-2 selective or non-selective NSAID confirmed to address positive effect on the regression of such premalignant lesions before. Therefore, our study will be highly original in implicating selective COX-2 inhibitor, especially the celecoxib to serve with a protective role in counteracting gastric carcinogenesis.

Nevertheless, in our study, more than 50% of patients still had persistence of IM despite 1-year selective COX-2 use in limited doses (200-mg celecoxib) after H. pylori eradication. The clinical trial with an increase of follow-up period or an add-up of celecoxib dosage should be promising for such patients. Furthermore, the tentative study design just focused on patients with the long-term need of NSAIDs for their underlying diseases. It would be more promising to conduct a pilot or a large-scale-controlled trial to start celecoxib in patients with persistent IM even several years after H. pylori eradication. Such data will potentially answer the optimal dosage and duration of COX-2 inhibitors to serve as potential chemoprevention and reverse earlier precancerous lesions, thereby stopping gastric carcinogenesis.

Our study also revealed that smoking might be positively correlated with the presence of IM. In Table 3, we elucidated even for the chronic celecoxib users that the presence of smoking and old age could be negative factors in decreasing the rate of IM regression after H. pylori eradication. These data were compatible to previous findings by Leung et al., who disclosed the factors relevant to the regression of IM after H. pylori eradication in non-NSAID users.17 Moreover, in this study, we disclosed that the pre-treatment MIM could also be an independent risk factor to determine the success of total regression of IM in the chronic celecoxib users. Therefore, further studies would be needed to show the efficacy in IM regression among patients with older age or higher pre-treatment MIM but who quit smoking or add up the dosage or duration of celecoxib.

There is a concern about a higher dosage of celecoxib of more than 400–800 mg/day, which may increase cardiovascular risk.19, 20 However, in our study, we applied a low dosage of 200 mg/day. Therefore, cardiovascular risk will not be increased if such dosage is applied for long-term prevention. This hypothesis is supported by a long-term chemoprevention trial for the prevention of spontaneous adenomatomous polyps, which illustrated no increased of cardiovascular events.21

In conclusion, chronic celecoxib users had similar prevalence of H. pylori and related precancerous potentials in having IM. Use of celecoxib can be a promising supplemental treatment to decrease the risk of gastric cancer in H. pylori-infected patients with IM.


Declaration of funding interests: this study was supported by grants NSC93-2314-B-006-011 and NSC95-2314-B-006-029-MY3 from the National Science Council, Taiwan.