Nonsteroidal anti‐inflammatory drugs prevent gastric cancer associated with the use of proton pump inhibitors after Helicobacter pylori eradication

Abstract Background and Aim Proton pump inhibitors (PPIs) are a potential cause of gastric carcinogenesis after Helicobacter pylori eradication. Thus, appropriate management including chemoprevention is required. The aim of this study was to evaluate the association between nonsteroidal anti‐inflammatory drugs (NSAIDs) and the incidence of post‐eradication gastric cancer in PPI users. Methods A multicenter retrospective cohort study was conducted. Patients who used a PPI (≥30 days) after H. pylori eradication between 2014 and 2019 were analyzed in nine hospital databases. Gastric cancer incidence was a primary outcome, and their association with NSAIDs use and clinical factors was evaluated. Hazard ratios were adjusted by age, sex, smoking, and Charlson Comorbidity Index. Results During the mean follow‐up period of 2.38 years, 1.13% (31/2431) of all patients developed gastric cancer. The cumulative incidence of gastric cancer in PPI users was 0.25% at 1 year, 0.51% at 3 years, and 1.09% at 5 years in the NSAID users and 0.89% at 1 year, 2.32% at 3 years, and 3.61% at 5 years in nonusers. NSAIDs were associated with a lower gastric cancer risk (adjusted hazard ratio = 0.28, P = 0.005). No gastric cancer was observed in the cyclooxygenase‐2 inhibitor users (n = 256). NSAID use with high dose and long duration was significantly associated with a lower incidence of gastric cancer. Conclusion NSAIDs were associated with a 60% decrease in the gastric cancer incidence in H. pylori‐eradicated PPI users, with dose and duration response effects. NSAIDs may be effective for chemoprevention against PPI‐related gastric cancer.


Introduction
Gastric cancer is the third leading cause of cancer-related mortality worldwide. 1 Helicobacter pylori is the most important carcinogen for gastric cancer, 2 and H. pylori eradication reduces gastric cancer risk by 47%. However, the annual incidence rate of posteradication gastric cancer is approximately 1.4%, 3 and understanding its pathogenesis and establishing a novel preventive approach remain to be required.
Previous reports revealed that proton pump inhibitor (PPI) use was associated with a 2.4-fold increase in the risk of H. pylori post-eradication gastric cancer (post-eradication gastric cancer). 4,5 Another meta-analysis also supported a possible association between long-term use of PPIs and the risk of gastric cancer. 6 Although PPIs may be a potential risk factor for posteradication gastric cancer, they are among the most commonly used drug groups in the world, and millions of patients need them for peptic ulcers and reflux diseases. Identification of populations at high risk of post-eradication gastric cancer and drugs effective for chemoprevention for the high-risk population would be beneficial for appropriate management of H. pylorieradicated PPI users.
Nonsteroidal anti-inflammatory drugs (NSAIDs) are potential chemopreventive drugs for gastric cancer. NSAIDs were associated with a 58% decrease in gastric cancer incidence in a country with a high rate of H. pylori infection, 7 and a similar effect was found in a country with a low rate. 8 Therefore, we hypothesized that NSAIDs may be preventive for gastric cancer development even in H. pylori-eradicated PPI users. However, NSAIDs increase gastrointestinal (GI) bleeding risk and potentially cardiovascular risk. The balance between the benefits and risks of NSAID use must be considered.
To address these issues, we performed a multicenter retrospective cohort study of patients who used PPIs after eradication and evaluated the associations between NSAID use and both gastric cancer development and adverse events, including upper GI bleeding and cardiovascular diseases. Furthermore, we evaluated the association between gastric cancer and patients' genotype of CYP2C19, which is involved in PPI metabolism. The findings may be useful for identification of optimal PPI users who require additional chemopreventive intervention.

Methods
Study design, setting, and participants. We performed a retrospective cohort study using the diagnostic procedure combination databases from nine hospitals between April 2014 and March 2019. The combined database provided records from all inpatients and outpatients of the University of Tokyo Hospital, Shuto General Hospital, Fukui Prefectural Hospital, Nerima Hikarigaoka Hospital, St. Leuk's International Hospital, Toyonaka Municipal Hospital, Ishikawa Prefectural Central Hospital, and Nagasaki Minato Medical Center and from inpatients of Tonan Hospital. The database included diagnosis, comorbidities, and adverse events using the International Classification of Diseases, tenth revision (ICD-10), and drugs and procedures coded according to the original Japanese code.
From the database, we extracted the data of patients who used a 7-day course of clarithromycin-based or metronidazole-based triple therapy for H. pylori infection between April 2014 and March 2019 using the drug codes (Table S1). We excluded patients who had used PPIs for less than 30 days and developed gastric cancer within half a year after H. pylori eradication. Patients who have history of gastrectomy before H. pylori eradication were also excluded. The follow-up period was from the date of eradication drug use to the final visit. The end of followup was March 2019, and loss to follow-up was defined as the date of the final visit. The study was approved by the institutional review boards of the University of Tokyo Hospital (no. 2019161NI).
Outcomes and variables. The primary outcome was the development of gastric cancer, as defined by the ICD-10 codes (C160, C161, C162, C163, C164, C165, C166, C168, and C169) or procedure codes for endoscopic and surgical resection ( The secondary outcomes were upper GI bleeding and cardiovascular diseases, including cerebrovascular diseases and ischemic heart diseases. Upper GI bleeding was defined as performing endoscopic hemostasis for GI bleeding, with a procedure code of K654. Cerebrovascular disease was defined by ICD-10 codes G450-469, H340, I600-639, I64, and I650-699, and ischemic heart disease was defined by ICD-10 codes I210-I229 and I252. We evaluated the following clinical factors: age, sex, smoking, comorbidities, and mediation use. Age was categorized into two groups: >70 and ≤70 years. The following comorbidities were included based on ICD codes: atrial fibrillation, acquired immunodeficiency syndrome, arterial thrombosis, carotid disease, cerebrovascular disease, chronic heart failure, chronic kidney disease (stage 5 or lower), dementia, diabetes mellitus with or without complications, deep vein thrombosis, hemiplegia, dyslipidemia, ischemic heart diseases, liver disorder (mild/severe), malignancy with or without metastasis, pulmonary embolism, peripheral vascular disease, pulmonary disease, rheumatic disease, transient ischemic attack, peptic ulcer disease, unstable angina disease, and valvular disease. The Charlson Comorbidity Index was calculated using these data. 9 The details of ICD-10 codes are shown in Table S2.
Statistical analysis. The primary endpoint, gastric cancer development, was censored at the date of the final visit. The Kaplan-Meier method was used to calculate the cumulative probability of gastric cancer development at 5 years. Univariate and multivariate Cox models were used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs); the multivariate Cox proportional hazards models were adjusted for age and sex. Missing data for sex (n = 1355) and age (n = 1355) were imputed using the fitted values from general linear models for comorbidities.
We also estimated the HRs for gastric cancer development according to NSAID duration and dose using other Cox models. In addition, we performed survival analyses using Cox models to evaluate the associations between NSAID use and the secondary endpoints (GI bleeding and cardiovascular disease). We calculated the posterior probabilities of gastric cancer using the Bayesian model in patients who underwent CYP2C19 genotyping.
A P value <0.05 was considered statistically significant, and HRs with 95% CIs were determined. All statistical analyses were performed using SAS software v. 9.4 (SAS Institute, Cary, NC, USA).
Patient and public involvement. Patients and the public were neither involved in the study design nor the recruitment.
This study received ethics approval from the institutional review boards of the University of Tokyo (ID: 2058-(2)).
PPIs were principally used for gastroesophageal reflux related symptoms. In addition, PPI users had histories of 14.64% of gastroesophageal reflux disease, 8.68% of cardiovascular diseases, and 10.94% of previous GI bleeding. NSAIDs were principally used for orthopedic disease related pain. In addition, NSAID users had histories of 10.83% of cardiovascular disease, 2.01% of rheumatic diseases, and 2.51% of orthopedic disease including osteoarthritis, spinal disc herniation, and spinal stenosis. Incidence of gastric cancer and the factors associated with gastric cancer in the long-term PPI users. During the mean follow-up period of 2.38 years, 31 patients (1.13%) developed gastric cancer (2 in the cardia, 14 in the non-cardia region; data were not available for 15 patients). The cumulative incidence of gastric cancer was 0.24% at 1 year, 0.52% at 3 years, and 1.09% at 5 years in the patients using NSAIDs and PPI, and 0.89% at 1 year, 2.32% at 3 years, and 3.61% at 5 years in the patients using PPI alone (Fig. 2a). NSAID use was significantly associated with decreased gastric cancer incidence (P = 0.006, log-rank test). The cumulative incidence of gastric cancer was 0% at 1, 3, and 5 years in the patients who used COX2 inhibitors with PPI, and 0.69% at 1 year, 1.72% at 3 years, and 2.81% at 5 years in the patients who used PPI but not COX2 inhibitors (Fig. 2b). The use of COX2 inhibitors was significantly associated with a decreased gastric cancer incidence (P = 0.044, log-rank test).
The results of a Cox model evaluating factors associated with gastric cancer are shown in Table 2 and Figure 3. In the analyzed H. pylori-eradicated PPI users, NSAID use was associated with a lower incidence of gastric cancer compared with nonuse (adjusted HR [aHR] 0.28, 95% CI 0.11-0.67, P = 0.005). The use of other drugs including aspirin, statin, fibrate, and metformin was not significantly associated with gastric cancer incidence. Analysis of other factors showed that a higher incidence of gastric cancer was associated with age >70 years (aHR 3.52, 95% CI 1.51-8.18, P = 0.004), chronic kidney disease stage 5 (aHR 9.42, 95% CI 1.23-72.36, P = 0.041), and malignancy without metastasis (aHR 0.10, 95% CI 0.01-0.84, P = 0.034).
Sub-analysis of the association between the dose or duration of NSAID use and gastric cancer in long-term PPI users. Of the 931 patients using low-dose NSAIDs, six patients developed gastric cancer (0.64%). None of the 66 high-dose NSAID users developed gastric cancer. The associations between dose or duration of NSAID use and gastric cancer incidence were assessed by using a Cox model ( Table 3). The aHR for low-dose NSAID use was 0.29 (95% CI 0.12-0.71) compared with non-use. In terms of duration of NSAID use, aHRs for short-term (<30 days) and long-term (≥30 days) use were 0.44 (95% CI 0.15-1.27) and 0.16 (95% CI 0.037-0.66) compared with the non-use group.
Incidences of cardiovascular diseases and upper GI bleeding in long-term PPI users. Figure 4 shows a Kaplan-Meier plot for the incidences of adverse events including upper GI bleeding and cardiovascular diseases in the NSAID users and nonusers. No patients developed upper GI bleeding in either group (Fig. 4a). Twenty patients (0.82%) developed cardiovascular events. The cumulative incidence of cardiovascular events was 0.22% at 1 year, 0.55% at 3 years, and 1.10% at 5 years in the NSAID use group and 0.49% at 1 year, 1.21% at 3 years, and 1.70% at 5 years in the non-use group (Fig. 4b). No significant difference in the incidence of cardiovascular events was observed between the groups (P = 0.224, log-rank test).
CYP2C19 genotyping. In order to investigate the effects by PPI metabolism in cancer incidence, CYP2C19 genotype was measured in a total of 199 patients. Of these patients, 78.9% were aged >70 years, 72.4% were men. Genotyping results indicated that 77, 90, and 32 patients were CYP2C19 extensive, intermediate, and poor metabolizers, respectively. A total of five patients had gastric cancer. The risk of gastric cancer (aHR 2.47) in the CYP2C19 non-extensive metabolizers, that is, the intermediate and poor metabolizers, was elevated compared with that in extensive metabolizers; however, this difference was not statistically significant (Tables S3 and S4). Nevertheless, a Bayesian model showed the potential risk of gastric cancer in the nonextensive metabolizers. Combined with the uninformative prior distribution and data, the posterior probabilities of a greater hazard in CYP2C19 intermediate and poor metabolizers than in extensive metabolizers were 79.8 and 87.6%, respectively ( Figure S2), suggesting that impaired CYP2C19 metabolizers may encounter prolonged effects by PPI, which potentially further increase the gastric cancer risk.

Discussion
In this multicenter cohort study, we found that NSAID use was associated with a 60% decrease in the post-eradication gastric cancer incidence in PPI users, and dose-and durationresponse relationships were confirmed. NSAID use was not associated with increased GI bleeding or cerebrovascular events. In addition, COX2 inhibitors showed relatively higher chemopreventive effects, and we identified CYP2C19 poor metabolizers as a potential high-risk population for posteradication gastric cancer.
H. pylori eradication therapy has been widely encouraged for prevention of peptic ulcers, as well as subsequent gastric cancers. However, a significant population in H. pylori-eradicated patients often need the prolonged use of PPIs, due to refractory acid-reflux symptoms. Given the potential cancer-promoting effects by PPIs in such patients, 4,5   there is a clinical dilemma when seeing H. pylori-eradicated PPI users. Our current finding that NSAIDs efficiently prevent gastric cancer development in PPI users would be helpful for solving this problem, that is, additional chemopreventive use of NSAIDs may be considered for a subgroup of H. pylorieradicated PPI users. Figure 3 Hazard ratios for development of gastric cancer according to clinical factors in long-term PPI users. †HR adjusted for age >70 years, sex, smoking, and Charlson Comorbidity Index. ‡Male was defined as the sex predicted value >0.5, using the fitted values from general linear models for comorbidities. §Age was including the age predicted values, using the fitted values from general linear models for comorbidities. aHR, adjusted hazard ratio; AT, arterial thrombosis; CHF, chronic heart failure; CI, confidence interval; CKD, chronic kidney disease; COMP, complications; CVD, cerebrovascular disease; DL, dyslipidemia; DM, diabetes mellitus; HTN, hypertension; LD, liver disorder; META, metastasis; NSAIDs, nonsteroidal anti-inflammatory drugs; PD, pulmonary disease; PUD, peptic ulcer disease; PVD, peripheral vascular disease; UAG, unstable angina disease. We found that CYP2C19 non-extensive metabolizers who use PPIs are at high risk of gastric cancer. NSAID use as chemoprevention may be beneficial in these patients following H. pylori eradication. In addition, given that past history of gastric cancers and advanced histopathological changes such as gastric atrophy and intestinal metaplasia are known to be a risk factor for post-eradication gastric cancer, patients with these factors may also be a potential target of chemopreventive approach using NSAIDs.
Since NSAIDs are anti-inflammatory drug, their antiinflammatory effects may contribute to a low incidence of gastric cancer. In addition, several in vitro and in vivo studies suggested that COX2 may play a critical role in proliferation and tumor growth in gastric cancers. 10,11 Given the strong cancer-preventing effects by COX2 inhibitors in our cohort and other studies, 12 COX2 pathway may be central in gastric cancer development at least in a subset of chronic gastritis patients.
Contrary to previous reports, 13-15 NSAID use was not associated with increased upper GI bleeding or cardiovascular events in our study. One explanation for the discrepancy might be that all our patients received H. pylori eradication therapy and used PPIs, which theoretically lower the incidence of upper GI bleeding. In our study, the rate of comorbidities associated with a cardiovascular risk, such as carotid, cerebrovascular, peripheral vascular, and ischemic heart diseases, was also fewer (5%) than in previous studies (more than 10%). 14,15 While this discrepancy might be due to shorter follow-up period and fewer dosage of NSAIDs in our study, our data suggest that chemopreventive use of NSAIDs may be feasible for H. pylori-eradicated PPI users.
Our study has several strengths. First, to our best knowledge, this was the first study to evaluate the association between NSAID use and the incidence of gastric cancer after H. pylori eradication in PPI users. Second, this was a large multicenter cohort study. Nevertheless, our study has several limitations. First, this study was retrospective. Second, our diagnostic procedure combination database lacked data on successful eradication of H. pylori. However, the success rate of H. pylori eradication was expected to be greater than 90%, based on the primary use of vonoprazan-based drugs in Japan. 16 Third, the follow-up period was relatively short and insufficient to accurately evaluate gastric cancer events. However, we will expand our database to double the number of hospitals and will perform further studies for another 2 years in near future. Lastly, information on some risk factors, including diet, body mass index, and family history, was limited in our database.
In conclusion, NSAIDs were associated with a decreased risk of gastric cancer with dose-and duration-response relationships among PPI users after H. pylori eradication, without adverse events. NSAIDs may be a candidate for prevention of PPI-related gastric cancer.

Supporting information
Additional supporting information may be found in the online version of this article at the publisher's website: Figure S1. Cumulative incidence of gastric cancer according to (A) CYP2C19 extensive vs non-extensive metabolizers and (B) CYP2C19 extensive vs intermediate vs poor metabolizers. Figure S2. Posterior probability of gastric cancer in CYP2C19 intermediate and poor metabolizers using a Bayesian model Figure S3. Cumulative incidence of gastric cancer in (A) PPI +NSAID high-dose and low-dose users and PPI-users, (B) PPI +NSAID long-term and short-term users and PPI-users. Table S2. ICD10 codes of comorbidities. Table S3. Association between CYP2C19 extensive and nonextensive metabolizers and gastric cancer incidence (n = 199).