INTRODUCTION
Preclinical studies have indicated that metformin inhibits the proliferation of gastric cancer cells and induces apoptosis [1,2]. In a Korean nationwide cohort study, metformin reduced gastric cancer risk in patients with type 2 diabetes not receiving insulin treatment [3]. A recent meta-analysis also corroborated the potential anticancer effect of metformin in gastric cancer [4]. These data support the hypothesis that metformin may exert chemopreventive effects on gastric carcinogenesis.
However, not all epidemiologic findings have been consistent. A meta-analysis that carefully evaluated immortal time bias reported that metformin had no protective effect against gastric cancer [5], and a Swedish population-based cohort similarly failed to demonstrate a reduced risk of gastric adenocarcinoma among metformin users [6]. Thus, the anti-gastric cancer effect of metformin remains controversial, and it is unclear whether the discrepancies among studies reflect differences in study design, patient populations, concomitant insulin use, or the specific stage of carcinogenesis at which metformin might act.
Metformin treatment activates various signaling pathways that exert anticancer effects at the molecular level [7]. AMP-activated protein kinase plays a key role in the classical pathway, contributing to the inhibition of cell proliferation and modulation of cellular metabolism [8,9]. Metformin has also been reported to have an anti–cancer stem cell effect and to inhibit epithelial–mesenchymal transition (EMT) in gastric cancer, while suppressing oncogenic microRNAs and activating tumor suppressor microRNAs [7,10]. EMT is a crucial factor in gastric tumorigenesis, and changes in the gastric microenvironment caused by Helicobacter pylori infection may accelerate this process [11]. Although a previous study suggested that metformin inhibits H. pylori growth [12], a clear relationship between metformin and H. pylori–induced EMT in human gastric carcinogenesis has not been established.
The adenoma–carcinoma sequence is a well-known model of carcinogenesis in gastric cancer [13]. Chronic inflammation leads to atrophic gastritis, intestinal metaplasia, dysplasia, and, eventually, carcinoma. H. pylori infection is usually related to this chronic inflammatory cascade. If metformin exerts anticancer effects in the stomach, it may interfere with one or more steps in this sequence, either by preventing the development of precancerous lesions, such as gastric dysplasia, or by inhibiting the progression from dysplasia to invasive cancer. However, few studies have directly evaluated the relationship between metformin and histologically confirmed gastric precursor lesions.
Therefore, we conducted a retrospective observational study in patients with diabetes to evaluate whether metformin could reduce gastric cancer or dysplasia. By analyzing endoscopic biopsy specimens, we aimed to clarify whether metformin influences the occurrence of precancerous lesions, invasive carcinoma, or both and explore how concomitant insulin use might modify these associations.
METHODS
Study population
This retrospective observational study reviewed the medical records of patients treated for type 2 diabetes between January 2009 and December 2018 at a tertiary hospital in Daejeon, South Korea. Patients who received antidiabetic medications for at least 6 months were identified and classified according to their medication exposure. Histologic outcomes were ascertained based on endoscopic biopsy findings obtained during the observation period. Patients who did not undergo an endoscopic evaluation with biopsy during the observation period were excluded from the study. Patients with a dosing period <6 months and those aged <19 years were also excluded. The patients were divided into four groups according to their medication (with or without metformin and with or without insulin): group 1, patients who received oral hypoglycemic agents (OHA) other than metformin (patients without metformin and insulin); group 2, patients who took OHA including metformin and not using insulin (with metformin, without insulin); group 3, patients who took both metformin and insulin; and group 4, patients who did not receive metformin but took insulin (Fig. 1). The histologic results were evaluated in these patients. As endoscopic examinations were performed according to clinical indications rather than scheduled surveillance, the prescription duration was used as a surrogate for the observation time. Patients were classified according to their predominant antidiabetic medication regimen during the observation period after at least 6 months of prescription, and exposure status was treated as fixed in the analysis.
Endoscopic evaluation
Endoscopic evaluations were performed using Olympus CV-260 or CV-290 series endoscopy processors (Olympus Korea Co., Ltd.). Endoscopic procedures were conducted by eight endoscopists who had performed over 5000 gastroscopies. The purpose of endoscopic evaluations was to assess gastrointestinal symptoms, including abdominal pain, anorexia, nausea, vomiting, and heartburn, and to screen for malignancies.
Statistical methods
Statistical analyses were performed using SPSS version 18.0 software (IBM Corp.). The chi-square test was used to evaluate the relationship between the patient groups and histologic results. Logistic regression analysis was used to assess the relative risks of gastric dysplasia and gastric adenocarcinoma according to the medication group. The accepted level of statistical significance was set at p<0.05.
RESULTS
Patient characteristics and study population
A total of 43947 patients with diabetes underwent treatment for more than 6 months during the study period. Endoscopic biopsies were performed in 5269 patients (3434 male and 1835 female patients). The median age was 65.09 years (range: 19–95 years). The median prescription period was 37.1 months (range, 6–118 months), and the underlying comorbidities of the patients are listed in Table 1. The histologic findings included gastritis (71.1%), hyperplastic polyps (9.4%), intestinal metaplasia (44.4%), gastric dysplasia (12.8%), adenocarcinoma (21.6%), and gastric malignancies other than adenocarcinoma (1.6%). Duplicates were allowed for histologic evaluation (Table 1).
Association between medication groups and gastric neoplasia
Age-adjusted logistic regression analysis revealed that group 2 (OHA including metformin) had fewer malignancies than group 1 (OHA other than metformin). The proportion of malignancies was also higher in groups 3 and 4 (insulin use group) than in group 1 (p<0.05) (Table 3). However, the incidence of premalignant lesions (gastric dysplasia and intestinal metaplasia) did not differ between the groups. Analysis stratified by prescription duration showed similar results (Fig. 2).
DISCUSSION
Previous studies have reported conflicting results on the anticancer effects of metformin. In a recent population-based cohort study, the anticancer effects of metformin, excluding prostate and pancreatic cancers, showed negative results [14]. However, a systematic review and meta-analysis suggested that metformin use in patients with diabetes reduces the overall cancer risk [15]. Other meta-analyses have implied that metformin could be a useful adjuvant agent in cancer treatment [16]. These discrepant findings indicate that the anticancer effect of metformin depends on the cancer type, concomitant medication, underlying metabolic status, and possibly the stage of carcinogenesis. This study aimed to evaluate the effects of metformin on the risk of gastric cancer in patients with diabetes.
Most studies on the effects of metformin on cancer risk have focused solely on cancer development and have rarely examined precursor lesions. In the present study, we focused on the adenoma–carcinoma sequence in gastric cancer. If metformin inhibits gastric dysplasia, a precancerous lesion of gastric cancer, this may help explain previously reported anticancer effects. Pathological analyses of endoscopic biopsies are essential to confirm this hypothesis. Therefore, we conducted this retrospective observational study to evaluate the pathological outcomes of endoscopic biopsy in patients with diabetes.
In our study, univariate analysis using the chi-square test showed a significant association between the medication group and the histologic results (Table 2). Group 2 (with metformin and without insulin) had the lowest percentages of gastric dysplasia and adenocarcinoma. The insulin-use groups (groups 3 and 4) had a higher proportion of adenocarcinomas than groups 1 and 2. These findings align with previous reports that suggested a protective effect of metformin and a potential tumor-promoting effect of insulin or insulin-related pathways [3,17,18]. However, logistic regression analysis performed with age as a control showed no significant differences in the occurrence of gastric dysplasia between the groups (Table 3), although the risk of malignancy remained lower in the metformin-only group and higher in insulin users.
Two hypotheses may explain this phenomenon. First, a larger-scale study may be required to demonstrate the modest chemopreventive effects of metformin on dysplasia. A randomized controlled trial involving fewer than 5000 patients did not demonstrate the anti–gastric cancer effects of metformin, whereas a large observational study demonstrated a protective effect [3,19,20]. Similarly, our study might have been underpowered to detect small differences in dysplasia after adjusting for age and other factors. Second, metformin may inhibit the transition from dysplasia to cancer rather than the initial formation of dysplasia. A Japanese study revealed an important genetic alteration in the transition from gastric dysplasia to carcinoma in the form of APC/TP53 mutations [21]. While no study has evaluated the relationship between metformin and these genetic changes in gastric cancer, metformin has been shown to modulate cyclin D1 and p53 expression in a cervical cancer cell line, leading to inhibition of cell proliferation and induction of apoptosis [22]. Collectively, these findings support the hypothesis that metformin might preferentially act at a later stage of gastric carcinogenesis, where dysplastic cells are already genetically unstable and more susceptible to apoptotic signals. Our observation that metformin was associated with fewer malignancies but not with a lower frequency of dysplasia is consistent with this stage-specific effect.
In this study, insulin appeared to interfere with the anticancer effects of metformin. A Korean nationwide cohort study revealed that metformin treatment without insulin decreased gastric cancer risk, although no reduction was observed in the incidence of gastric cancer in patients receiving both metformin and insulin [3]. Similar results were observed in a systematic review and meta-analysis, where the use of insulin and its analogs increased cancer risk [17]. However, a large population-based study in Taiwan revealed conflicting findings: while patients with diabetes may have a 14% higher risk of gastric cancer, insulin use itself was not associated with an increased risk [18]. Other studies suggest that insulin resistance and hyperinsulinemia, rather than exogenous insulin alone, are related to cancer incidence [23]. Thus, it is unclear whether the increased proportion of gastric cancer in patients treated with insulin is due to insulin itself, the severity of diabetes and insulin resistance, or the complex metabolic milieu in which insulin is used.
The effects of metabolic syndrome may have confounded our results. A Chinese study showed that metabolic syndrome increases the risk of gastric cancer [24]. However, a recent meta-analysis revealed that gastric cancer risk or survival is not significantly associated with metabolic syndrome overall, except in Western women [25]. These data suggest that metabolic disturbances, including obesity and insulin resistance, contribute to gastric carcinogenesis in a population- and sex-specific manner. In our cohort, we were unable to fully assess metabolic syndrome or insulin resistance. These unmeasured factors may have influenced the observed associations.
Our study had several limitations. First, this was a retrospective, single-center study. The relatively small sample size and short follow-up duration may have been sources of bias. Second, the patients were classified according to metformin and insulin treatment. However, hypoglycemic agents other than metformin may exert different effects on cancer risk, and this potential influence has not been fully evaluated. Additionally, changes in antidiabetic medications over time could not be fully accounted for. Furthermore, detailed measures of diabetes severity and glycemic control, such as HbA1c levels and microvascular complications, were not available, which may have further limited the interpretation of the results. Third, the number of patients in each group was limited, and the proportion of patients using insulin (groups 3 and 4) was relatively high. As our center is a referral center, patients with more advanced diabetes, who have longer disease duration and more comorbidities, may have been included. Finally, other factors that may affect the occurrence of gastric cancer, such as alcohol consumption, smoking history, H. pylori infection, and concomitant medications (including statins and aspirin), could not be evaluated owing to the limitations of the retrospective design, and residual confounding factors cannot be excluded.
Despite these limitations, this study has several strengths. Unlike many previous large-scale studies based on insurance claims data, our study conducted a comparative analysis of the occurrence of histologically confirmed precancerous lesions and gastric cancer according to the use of diabetes medication. A detailed analysis of histologic data from endoscopic biopsies is difficult in extensive population-based studies. Our findings provide complementary information on how metformin and insulin differentially influence various stages of gastric carcinogenesis.
In conclusion, metformin may reduce the risk of gastric cancer in patients with diabetes, particularly in those who do not receive insulin; however, it does not appear to reduce the occurrence of gastric dysplasia. These findings support the possibility that metformin exerts a stage-specific effect, acting mainly on the progression from dysplasia to carcinoma, rather than on the initial development of dysplasia. Well-designed prospective studies with sufficient statistical power and detailed assessments of H. pylori status, metabolic factors, and molecular alterations are needed to confirm our conclusions and further elucidate the mechanisms by which metformin influences gastric carcinogenesis.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Download Citation
