Proton Pump Inhibitors and Gastric Cancer: A Large-Scale Propensity Score-Matched Analysis Authors: Cheal Wung Huh, Nak-Hoon Son, So Hyeon Gwon, Hee Byung Koh, Hae-Ryong Yun, Da Hyun Jung Published: 2026 ABSTRACT Background/AimsWe aimed to elucidate whether proton pump inhibitor (PPI) users are at increased risk of gastric cancer compared with non-PPI users or histamine-2 receptor antagonist (H2RA) users.MethodsA population-based cohort study was conducted using data from the UK Biobank from 1990 to 2016. We compared the incidence of gastric cancer after a minimum of 1-year drug exposure between PPI users and non-PPI or H2RA users using Cox proportional hazards models. Secondary analyses assessed the duration and dose–response associations. Sensitivity analyses incorporating various matching ratios, extended lag periods, and propensity score stratification were performed.ResultsAfter large-scale propensity score matching, we included 42,732 new PPI users, 42,732 non-PPI users, and 4,762 new H2RA users. During a median follow-up of 6.7 years, PPI users had a 2.32-fold higher risk of gastric cancer than non-PPI users (hazard ratio [HR], 2.324; 95% confidence interval [CI], 1.646 to 3.282), with numbers needed to harm of 488 and 442 at 2 and 4 years, respectively. Compared with H2RA users, PPI users also had a significantly higher incidence of gastric cancer (HR, 5.343; 95% CI, 1.557 to 18.335). The risk of gastric cancer tended to increase with a longer cumulative duration of PPI use and higher cumulative doses of PPIs. Moreover, these results remained robust after adjustments for various lag periods, matching ratios, and propensity score stratification.ConclusionsPPI use was associated with a higher incidence of gastric cancer, albeit with a low absolute risk. Therefore, long-term PPI use should be considered cautiously, even in regions with a low risk of gastric cancer. INTRODUCTION Proton pump inhibitors (PPIs) are medications widely prescribed globally for the inhibition of gastric acid secretion.1 However, the increasing use of PPIs has raised concerns regarding potential adverse outcomes, including bone fractures, micronutrient deficiencies, pneumonia, chronic kidney disease, small intestinal bacterial overgrowth, and Clostridium difficile infection.2 Moreover, hypergastrinemia resulting from hypochlorhydria has been correlated with gastric neoplasm, raising concerns about the risk of gastric cancer following long-term PPI use.3-11 However, the existing literature on the association between PPI use and gastric cancer has yielded inconsistent results.3-11 To date, several observational studies have reported elevated relative risks (RRs) of gastric cancer associated with PPI use, ranging from 1.39 to 2.87.12 However, the majority of these studies had methodological shortcomings, including indication,3-5,7-9,13-16 survival,8,15-17 and time-related biases.3,5,11,13,16,18,19 Additionally, most studies did not fully account for various medications and comorbidities, potentially leading to distorted and exaggerated findings. Furthermore, treating gastric conditions with PPIs may elevate the risk of gastric cancer compared with the general population, potentially resulting in reverse causality if an appropriate lag-time is not considered in the analysis. To address these issues, recent observational studies and meta-analyses have investigated the incidence of gastric cancer among PPI users compared with histamine-2 receptor antagonist (H2RA) users.10-12,20 However, there is still debate over whether PPIs significantly elevate gastric cancer risk compared with H2RAs as the active comparator.9-11,20,21 Moreover, results regarding the duration-dependent relationship between PPI use and gastric cancer risk remain inconsistent.10,20,21 On the other hand, the risk of gastric cancer associated with long-term PPI use varies among different racial populations. A recent meta-analysis reported a significantly increased risk in Asian populations (RR, 2.62; 95% confidence interval [CI], 1.95 to 3.53), whereas the risk was non-significant in Western cohorts (RR, 1.25; 95% CI, 0.92 to 1.68).12 This suggests that Western populations might be less susceptible to the carcinogenic effects of PPIs. However, many researchers remain concerned about the potential contribution of PPI use to a significant proportion of gastric cancer incidence in Western populations,12,22,23 underscoring the need for further evidence. Thus, we aimed to investigate the association between PPI use and gastric cancer in the general population using the UK Biobank data, employing large-scale propensity score matching (PSM) to mitigate confounding factors and ensure robust results. MATERIALS AND METHODS 1. Data source The UK Biobank is a large population-based prospective observational cohort consisting of more than 500,000 participants aged 40 to 69 years who were enrolled between 2006 and 2010 from 22 centers across England, Scotland, and Wales. The UK Biobank has linked primary care data for approximately 45% of its cohort. A total of 204,027 participants with available primary care data were initially screened. While the UK Biobank participant enrollment period was from 2006–2010, the primary care data include records from first-contact healthcare provided by general physicians and nurse practitioners, starting from as early as 1980. However, in the primary care data used for our study, after applying PSM, the index date began in the 1990s. This study was approved by the National Information Governance Board for Health and Social Care in England and Wales and the Community Health Index Advisory Group in Scotland and the North West Multicenter Research Ethics Committee. Data access for this study was approved by the UK Biobank Access Committee (application 73873). All participants provided informed consent. This study was conducted in accordance with the ethical principles of the Declaration of Helsinki. 2. Study design We conducted a nationwide population-based cohort study that retrospectively evaluated the association between PPI use and gastric cancer. The primary analysis aimed to compare the risk of gastric cancer between PPI users and non-PPI or H2RA users using the UK Biobank database in the general population. The secondary analyses assessed the association between the duration of PPI exposure and gastric cancer risk. 3. Study population The target cohort comprised new PPI users with a cumulative defined daily dose (cDDD) of at least 30 days, with the index date defined as the first day of PPI prescription (Supplementary Table 1). The comparative cohort included participants who were prescribed any drug (excluding PPIs) for ≥30 cDDD-days, with the index date set as the first date of each drug prescription. In both cohorts, participants were censored either at the time of gastric cancer identification (Supplementary Table 2) or the date of the last available data for each participant. For the primary care data, the last follow-up was available up to August 31, 2017. A diagram of the cohort construction is presented in Fig. 1. Exclusion criteria for both cohorts included a history of gastric cancer before or within 1 year of cohort entry, an observation period of less than 365 days before cohort entry, a history of gastrectomy before cohort entry, participants who were prescribed both PPIs and H2RAs for ≥30 cDDD-days, and age <19 years. The PPI group was further divided into subgroups based on the duration of PPI exposure (≥30, ≥90, ≥180, ≥365, ≥730, ≥1,095, and ≥1,460 cDDD-days) for subgroup analysis. We also conducted the same analysis using another comparative cohort of patients who were treated with H2RAs for ≥30 cDDD-days in the UK Biobank database (Supplementary Table 3). At the index date, all patients had at least 1 year of medical information in the UK Biobank, which was essential for identifying new PPI and H2RA users. Finally, the cohort was limited to patients with at least 1 year of follow-up after cohort entry to allow for a latency time-window and minimize detection bias and reverse causality. 4. Drug exposure The exposure of interest was long-term use of PPIs and H2RAs, defined as cumulative drug exposure of ≥30 cDDD-days. PPIs included esomeprazole, lansoprazole, omeprazole, pantoprazole, or rabeprazole sodium, and H2RAs included cimetidine, famotidine, nizatidine, or ranitidine, all listed in the United Kingdom national formulary. The DDDs for each prescription were calculated based on the World Health Organization (WHO) methodology, considering the drug's potency and quantity. The DDD system, endorsed by the WHO for drug utilization studies, assigns a standard dose reference for each medication based on its primary indication in adults.24 The DDD per package, which accounts for both the potency and quantity of the drug, was summed over the exposure period included in the study for each individual patient and used as an estimate to calculate the cDDD. The patients were divided into three groups: PPI group (only PPI use; PPI ≥30 cDDD-days; or combined use, PPI ≥30 cDDD-days and H2RA <30 cDDD-days), H2RA group (only H2RA use; H2RA ≥30 cDDD-days; or combined use, H2RA ≥30 cDDD-days and PPI <30 cDDD-days), and non-PPI group (PPI <30 cDDD-days and H2RA <30 cDDD-days). 5. Outcomes The primary outcome was the incidence of gastric cancer at least 1 year following drug exposure. Gastric cancer was identified using the International Classification of Diseases, Tenth Revision (ICD-10) diagnosis codes C16.0–C16.9, or codes G27 and G28 from the Classification of Interventions and Procedures 4. We also compared the outcomes of the target cohort (PPI users) and the comparative cohort (non-PPI and H2RA users) based on the duration of PPI exposure. 6. Covariates We conducted large-scale PSM using the OMOP-CDM, version 5.3.1, tool. Covariates used between the target and comparative cohorts included the following: (1) sociodemographic variables: age, sex, race, alcohol consumption, and smoking history; (2) medical history: hypertension, diabetes, dyslipidemia, chronic liver disease, pulmonary disease, venous thromboembolism, atrial fibrillation, congestive heart failure, chronic kidney disease, peripheral vascular disease, gastroesophageal reflux disease (GERD), Barrett's esophagus, peptic ulcer disease, dyspepsia, Helicobacter pylori infection, inflammatory bowel disease, and the Charlson Comorbidity Index; (3) medication use: metformin, nonsteroidal anti-inflammatory drugs (NSAIDs), statins, antiplatelets, anticoagulants, and steroids; and (4) calendar period: 5-year calendar year bands of cohort entry (1990–1994, 1995–1999, 2000–2004, 2005–2009, and 2010–2016). The Charlson Comorbidity Index was calculated based on the ICD-10 codes in the medical treatment database by summing the weighted scores of each comorbidity. H. pylori infection was identified either through a diagnosis or by a prescription for eradication therapy (clarithromycin-based or bismuth-based quadruple therapy), defined as a combination of a PPI, clarithromycin or amoxicillin, with metronidazole or tetracycline. 7. Statistical analysis To describe the participants' baseline characteristics, we presented means and standard deviations for continuous variables and frequencies and proportions for categorical variables. PSM was used to balance differences in baseline characteristics between groups. Propensity scores were estimated using logistic regression as the predicted probability of PPI versus the predicted probability of being in a comparator cohort (non-PPI or H2RAs) within the calendar year bands (1990–1994, 1995–1999, 2000–2004, 2005–2009, and 2010–2016), after adjusting for covariates and cohort entry. Patients in non-overlapping regions of the propensity score distribution were trimmed. Propensity scores were estimated using a logistic regression model selected through 10-fold cross-validation using high-performance computing. PSM applied a greedy algorithm based on the propensity score to achieve the best match, starting with the 8 digits of the propensity score and working down to 1 digit. Matching variables were selected from covariates with baseline differences at a statistical significance level of 0.05. We assessed balance by calculating the standardized mean differences (SMDs) before and after PSM. If the absolute SMD was less than 0.1, we considered the covariate balanced between the two groups.25 Incidence rates in 1,000 person-years were calculated based on the Poisson distribution, and a log-rank test was performed as a survival analysis between the PPI and control groups (non-PPI or H2RAs). Kaplan-Meier curves were used to express cumulative event rates, and Cox proportional hazards regression was used to present hazard ratios (HRs) and 95% CIs for the PPI group. The number needed to harm (NNH) was calculated using the proportion of gastric cancers that occurred during the observation period. All statistical analyses were performed using SAS (version 9.4; SAS Institute, Cary, NC, USA). 8. Sensitivity and subgroup analyses To address the potential impact of the residual confounding suggested by VanderWeele and Ding,26 we performed sensitivity analysis. We performed 11 sensitivity analyses to assess the robustness of our findings. Specifically, we stratified the lag period after the index date into 1, 2, and 3 years to observe whether the timing of exposure affected our results. We also expanded the PSM ratio to 1:2 and 1:4, as well as 1:1, to explore whether different proportions of matched controls per case influenced the incidence and risk of medications and gastric cancer. To account for the different potencies of PPI types, we converted all PPI prescriptions to omeprazole equivalents using the WHO DDD (Supplementary Table 4). Cumulative omeprazole equivalents were then calculated by summing the dose of each prescription from cohort entry until the time of each event-defining risk set. We extended the duration of PPIs to 30, 90, 180, 365, 730, 1,095, and 1,460 cDDD-days and beyond to determine the incidence rate and HR of gastric cancer as the duration of PPIs increases. Additionally, for the group with a PPI cDDD of ≥30, the dose of omeprazole was divided into categories (<1,900 mg, 1,900–2,499 mg, 2,500–3,599 mg, 3,600–12,499 mg, and ≥12,500 mg). These categories were then used to compare the incidence rate and HR of the outcome. RESULTS 1. Study flow and patients Of the 502,419 participants in the UK Biobank cohort, 194,844 met the inclusion criteria. The target cohort (PPI group) included 58,283 participants and the comparative cohort (non-PPI or H2RAs group) comprised 136,561 participants. After excluding unmatched participants through PSM, a total of 90,226 participants were analyzed in both cohorts. The study flow is shown in Fig. 2. The PSM included 29 covariates. The balance of covariates between the target and comparative cohorts is presented in Supplementary Fig. 1. The maximum SMD after PSM was 0.09. Table 1 presents the baseline characteristics of the study population used in the primary analysis. After a large propensity score-matched analysis, all covariates were well balanced (i.e., SMDs were <0.1) (Supplementary Fig. 1). The potential confounders that may influence the risk of gastric cancer, such as medication (aspirin, metformin, NSAIDs, steroids, and statins), indications for acid suppressant drug use (GERD, Barrett’s esophagus, peptic ulcer disease, dyspepsia, and H. pylori infection), and calendar period, were well-matched. Calendar periods were considered in order to adjust for temporal changes in acid suppressant-drug prescriptions and changes in gastric cancer incidence, as well as heterogeneity in covariate definitions during the study period.27,28 The proportion of gastric cancers also consistently remained similar before and after PSM (Supplementary Table 5). The baseline characteristics of patients included in the subgroups according to PPI treatment duration of ≥30, ≥90, ≥180, ≥365, ≥730, ≥1,095, and ≥1,460 cDDD-days are shown in Supplementary Tables 6 and 7. 2. Association of PPI use and gastric cancer The median follow-up duration was 6.5 years in the PPI group and 6.8 years in the non-PPI group. In the target cohort, the median duration of PPI use was 2,370 days (interquartile range, 366 to 3,650 days), with a median interval from PPI exposure to gastric cancer of 4.3 years. During the follow-up period, 101 participants in the PPI group and 48 in the non-PPI group were diagnosed with gastric cancer (PPI vs non-PPI, 101/271,472 person-years vs 48/285,092 person-years; HR, 2.324; 95% CI, 1.646 to 3.282; p<0.0001) (Table 2). In the analysis comparing PPIs and H2RAs, we also found a significant difference in gastric cancer incidence between the two cohorts (PPIs vs H2RAs, 16/40,087 person-years vs 3/40,047 person-years; HR, 5.343; 95% CI, 1.557 to 18.335; p=0.0077) (Table 2). The Kaplan-Meier curve also showed a significantly higher cumulative incidence of gastric cancer in the PPI group than in both the non-PPI and H2RAs groups (Fig. 3). Moreover, large-scale PSM analysis was performed in subgroups with PPI use exceeding 90, 180, 365, 730, 1,095, and 1,460 cDDD-days. The incidence rates of gastric cancer in these subgroups showed an increasing trend corresponding to the duration of PPI use (PPI ≥90 cDDD-days: HR, 2.498; 95% CI, 1.714 to 3.641; PPI ≥180 cDDD-days: HR, 2.252; 95% CI, 1.488 to 3.409; PPI ≥365 cDDD-days: HR, 2.187; 95% CI, 1.415 to 3.379; PPI ≥730 cDDD-days: HR, 2.249; 95% CI, 1.375 to 3.678; PPI ≥1,095 cDDD-days: HR, 2.361; 95% CI, 1.295 to 4.307; and PPI ≥1,460 cDDD-days: HR, 3.521; 95% CI, 1.426 to 8.691) (Table 2). Similarly, the risk of gastric cancer increased with cumulative omeprazole equivalents (Supplementary Fig. 2). The NNH also decreased with an increase in the duration of PPI use (PPI ≥30 cDDD-days, NNH 806; PPI ≥90 cDDD-days, NNH 604; PPI ≥180 cDDD-days, NNH 670; PPI ≥365 cDDD-days, NNH 592; PPI ≥730 cDDD-days, NNH 488; PPI ≥1,095 cDDD-days, NNH 456; and PPI ≥1,460 cDDD-days, NNH 442) (Supplementary Table 8). These results suggested a positive cumulative dose–response relationship between gastric cancer risk and long-term PPI use. 3. Sensitivity analyses A sensitivity analysis comparing PPI use of ≥30 cDDD-days with the non-PPI group was conducted using different matching ratios, extending the lag period, and propensity score stratification. Results from sensitivity analyses with 1:2 or 1:4 matching ratios and 1- to 3-year lag periods were consistent with the main analysis (Table 3). Propensity score stratification results also exhibited consistent findings (1-year lag period: HR, 2.770; 95% CI, 2.171 to 3.536; 2-year lag period: HR, 2.726; 95% CI, 2.083 to 3.567; and 3-year lag period: HR, 3.007; 95% CI, 2.244 to 4.030) (Table 3). DISCUSSION This was a large population-based cohort study investigating the association between PPI use and gastric cancer using data from the UK Biobank. The main findings were as follows: (1) PPI use for ≥30 cDDD-days was associated with a 2.32-fold increased risk of gastric cancer compared with the non-PPI group, with a NNH of 488 and 442 for 2 and 4 years of follow-up, respectively; (2) new users of PPIs (≥30 cDDD-days) had a 5.34-fold increased risk of gastric cancer (HR, 5.343; 95% CI, 1.557 to 18.335) compared with new users of H2RAs (≥30 cDDD-days); (3) the risk of gastric cancer showed an increasing trend with the cumulative duration of PPI use and cumulative PPI dose; and (4) the results from sensitivity analyses using different matching ratios, extending the lag period, and propensity score stratification were highly consistent with the main analysis. An association between PPI use and the development of gastric cancer is biologically plausible and may be mediated by several factors. First, PPIs reduce gastric acid secretion, leading to hypergastrinemia and the proliferation of enterochromaffin-like cells in the oxyntic gland, potentially contributing to gastric cancer development.29 Second, long-term PPI use can accelerate or induce the progression of corpus gastric atrophy or intestinal metaplasia, which are precursors to gastric cancer.30 Finally, long-term PPI use may alter the gut microbiome, reducing microbial diversity, which has been linked to an increased risk of gastric cancer.31,32 Furthermore, the enhanced acid-suppression effect of PPIs can lead to non-H. pylori bacterial overgrowth.33 Taken together, these factors may contribute to gastric cancer development among PPI users. On the other hand, H2RAs, which only block the effects of histamine, are less effective than PPIs and are associated with lower gastrin levels, making them less likely to be associated with an increased risk of gastric cancer from a theoretical biological perspective.34,35 Several observational studies have suggested a potential association between long-term PPI use and gastric cancer risk. However, most of these studies had notable methodological shortcomings, such as indication bias because of baseline characteristic differences between PPI users and non-users;3-5,7-9,13-16 inclusion of prevalent users leading to survival bias and confounding;8,15-17 time-related biases, such as immortal-time bias and time-window bias;3,5,11,13,15,16,18,19 lack of consideration for cancer latency;4,5,19 and failure to fully account for various medications and comorbidities. These biases can distort and exaggerate associations, thereby limiting the conclusions drawn from previous studies. To address these limitations, three recent pivotal cohort studies have been conducted.9,11,28 Two large-scale cohort studies conducted in South Korea,9,11 a region with a high prevalence of gastric cancer, found no significant difference in gastric cancer risk between long-term PPI users and H2RA users,11 while the PPI group showed an increased risk compared with the non-PPI group.9 However, it is important to note that generalizability of these results to other regions, particularly Western populations, may be limited. Additionally, there were limitations to these studies, such as the lack of using a standard measure for drug utilization studies (e.g., cDDD)9 and failure to compare against non-PPI users.11 Importantly, these two studies reached conflicting conclusions regarding the association between long-term PPI use and gastric cancer risk.9,11 On the other hand, a study using the UK Clinical Practice Research Datalink indicated that, compared with H2RA use, PPI use increased the risk of gastric cancer,10 in contrast to the findings of the previous two studies.9,11 However, this study also had some limitations, such as arbitrary definition of medication exposure (PPI or H2RA use defined as incident use of one prescription or more) and absence of comparisons with non-PPI users. Overall, despite recent research efforts, it remains challenging to draw definitive conclusions about the risk of gastric cancer associated with long-term PPI use. Moreover, evidence regarding the association between long-term PPI use and gastric cancer incidence in Western populations is contentious compared with that observed in Asian populations.12 Therefore, additional studies and robust evidence are still required, not only in regions with a high prevalence of gastric cancer but also in those with a lower prevalence, such as Western populations. Thus, we attempted to address these limitations through careful study design and numerous sensitivity analyses using the UK Biobank data with large-scale PSM. This study had several strengths. First, we utilized the UK Biobank database, enabling large-scale PSM with the inclusion of 29 covariates, a feature lacking in previous studies. Through extensive PSM, we achieved well-balanced target and comparative cohorts, accounting for diagnoses, medication use, the comorbidity index, and the calendar period before cohort entry, thereby mitigating the limitations associated with confounding factors. The meticulous matching of potential confounders that may influence the risk of gastric cancer is noteworthy, including medication (aspirin, metformin, NSAIDs, steroids, and statins) and indications for acid suppressant drug use (GERD, Barrett’s esophagus, peptic ulcer disease, dyspepsia, and H. pylori infection). We also adjusted for the calendar period, as calendar-time specific propensity scores may result in more appropriate confounding control than that obtained via a single propensity score model.36,37 Second, our study design aimed to minimize potential biases, and we conducted various sensitivity analyses, all of which yielded consistent results with the main analysis. Third, we adopted a new-user design to mitigate immortal time bias, restricting inclusion to individuals first exposed to either the target or comparative drugs during a continuous 1-year observation period.17,18 Finally, to address indication bias, we conducted an additional analysis with a control group of H2RA users as an active comparator for PPI users, which also yielded statistically significant results. This study also had some limitations. First, we might not have completely eliminated the impact of residual confounding from inadequately measured or missing factors, such as H. pylori infection, which is not routinely screened for by general practitioners in the UK. To overcome this concern, we verified the H. pylori status not only by diagnosis but also by assessing the combination of medications used for H. pylori eradication. Nevertheless, in the present cohort, H. pylori infection status could be identified in only approximately 1% of participants. Considering that the prevalence of H. pylori infection in the United Kingdom (or Europe) has been reported to be approximately 35%,38 this likely represents substantial under-ascertainment. Recent studies have reported that long-term PPI use is significantly associated with an increased risk of gastric cancer, particularly among patients who have undergone H. pylori eradication, suggesting a potential interaction between prior H. pylori infection and prolonged acid suppression in gastric carcinogenesis. However, further evidence across different regions and ethnic populations is still required to clarify this relationship. Second, PPIs and H2RAs are available over the counter in the United Kingdom, potentially resulting in some missing prescription information. However, patients requiring long-term PPI or H2RA use often receive prescriptions from their general practitioner rather than purchasing drugs over the counter owing to financial incentives. Third, although the overall association between cumulative PPI exposure and gastric cancer risk was positive, the dose–response pattern based on cumulative omeprazole-equivalent categories was not strictly monotonic. This may reflect variability introduced by dose standardization across different PPI agents and reduced statistical stability in higher exposure categories with fewer events. In addition, cumulative duration and cumulative dose represent related but distinct exposure dimensions, which may partly explain minor differences in trend patterns. Fourth, stratifying based on the type of gastric cancer (cardia vs non-cardia) was not feasible as this information is not consistently documented in the UK Biobank. Fifth, the primary outcome of our study was based on ICD-10 codes. Thus, issues regarding diagnostic accuracy and overestimation or underestimation of gastric cancer incidence may have occurred. However, the gastric cancer incidence rate reported in this study (18 per 100,000 person-years) is similar to previous estimates of gastric cancer incidence in Europe, which average approximately 16 per 100,000 person-years.39 Sixth, the UK Biobank is well recognized to exhibit a “healthy volunteer” selection bias, with participants generally being healthier, more health-conscious, and of higher socioeconomic status than the general UK population.40 Because of this, the baseline incidence of gastric cancer and the distribution of risk factors may differ from that of the general population. Finally, the results of our study could not establish a causal relationship between PPIs and gastric cancer owing to its retrospective observational cohort design, as opposed to a randomized controlled trial design. Despite being based on a large-scale cohort with extensive efforts to mitigate various biases, our results only indicate a potential association between PPI use and gastric cancer. In summary, long-term use of PPIs was associated with a higher incidence of gastric cancer than that with non-PPI or H2RA use, albeit with a low absolute risk. The incidence of gastric cancer showed an increasing trend parallel to the duration of PPI use or cumulative PPI dose. Therefore, long-term PPI use should be approached cautiously, even in regions with a low risk of gastric cancer.