Diabetes and gastric cancer incidence and mortality in the Asia Cohort Consortium: A pooled analysis of more than a half million participants

1 INTRODUCTION

Asian countries, particularly those in East Asia, have the highest incidence rates of gastric cancer, despite a declined trend in recent years.¹ Conversely, the prevalence of diabetes is increasing globally, particularly in Asia, where more than 60% of those with the disease reside.² Diabetes is linked to increased cancer risk, including liver, pancreatic, endometrial, colorectal, breast, and bladder cancers, potentially due to hyperinsulinemia, hyperglycemia, and chronic inflammation.³ However, evidence linking diabetes and gastric cancer remains inconclusive.

Several studies have found a positive association between diabetes and gastric cancer risk,⁴ whereas some studies reported otherwise.^{5, 6} Some reported the association in both men and women.^{7, 8} whereas others found the association restricted to women.^{9, 10} Only a few have evaluated this association by anatomical subsites of gastric cancer with inconclusive results.^{7, 11} In Asia, studies in Korea, Japan, and Hong Kong have reported the association between diabetes and gastric cancer risk with conflicting findings.^8-12 Limitations of these studies include small number of cases (n < 200),^{11, 12} inadequate control of potential confounders such as demographic and lifestyle factors,⁸ failure to stratify by sex (gastric cancer incidence rates twice as high in men),^{8, 11, 12} and insufficient analysis of anatomical or histological subtypes.^{8-10, 12} Furthermore, although there have been studies incorporating the impact of the diabetes duration,^{8, 11} none has explored its dose–response effect.

Diabetes among Asians shows epidemiological characteristics different from the Western population.¹³ It has been hypothesized that Asians may have an early beta-cell dysfunction and impaired insulin action due to low lean mass.¹⁴ Consequently, understanding the relationship between diabetes and gastric cancer among Asians may help identify differential risks for this population, potentially leading to better prevention strategies and surveillance guidelines for patients with diabetes. Within the Asia Cohort Consortium,^{15, 16} the present study aims to elucidate the association between diabetes and gastric cancer in the Asian population by sex, anatomic subsites, histologic subtypes, and diabetes duration.

2 METHODS

3 RESULTS

A total of 8556 first primary incidence gastric cancer cases were identified (5676 men and 2880 women) during a median follow-up of 14.7 years (77 866 634 person-years; Table 1). At baseline, the median age of the participants was 52 years (range 15–103); they had a mean BMI of 23.1 kg/m² (SD 3.4), 36% ever smoked, and 43% ever consumed alcohol. The overall prevalence of self-reported diabetes was 4.8%.

Individuals with a history of diabetes had a 15% increased risk of incident gastric cancer (HR 1.15, 95% CI 1.06–1.25; I² = 3.00%, p = .41; Figure 1). Among individual cohorts, KMCC showed a significant association (HR 1.50, 95% CI 1.04–2.16), and JACC showed a nonsignificant inverse association (HR 0.85, 95% CI 0.67–1.09).

Diabetes increased the risk of gastric cancer by 12% (HR 1.12, 95% CI 1.01–1.23; I² = 0.00%) in men and 31% in women (1.31, 1.07–1.60; I² = 30.00%, p_{heterogeneity} = .18; Table 2). By country, Korea showed the largest association for overall gastric cancer (HR 1.32, 95% CI 1.01–1.71), although the pattern was similar across the four countries (p_{heterogeneity} = .47). The risk was higher in those who enrolled after 1990 (1990–1999; HR 1.22, 95% CI 1.07–1.39; and 2000–2015; HR 1.23, 95% CI 1.00–1.51), born between 1930 and 1949 (HR 1.19, 95% CI 1.06–1.34), enrolled in their 50 (HR 1.20, 95% CI 1.04–1.39) and diagnosed with gastric cancer in their 60 (HR 1.28, 95% CI 1.10–1.49), although heterogeneity between these subgroups was not found (Table 2).

Diabetes was significantly associated with the risk of noncardia (HR 1.14, 95% CI 1.02–1.28) but not with cardia gastric cancer (HR 1.17, 95% CI 0.77–1.78) with a nonsignificant heterogeneity between the subsites (p_{heterogeneity} = .91; Figure 2). A positive association was observed for the intestinal (HR 1.22, 95% CI 1.02–1.46) but not for the diffuse subtype (HR 1.00, 95% CI 0.62–1.61; p_{heterogeneity} = .45; Figure 2).

The risk of gastric cancer in participants with diabetes relative to people without diabetes (at cohort entry) increased during the first decade of diabetes (first quartile, median 10.9 years; HR 4.70, 95% CI 3.77–5.86) and then decreased, resulting in a nonlinear inverted V or lambda-shaped curve (∧) (p < .01 for nonlinear relationship; Figure 3). When diabetes diagnosis was limited to diabetes duration assessed at baseline, the risk of gastric cancer increased by 40% in the second quartile (median = 2.7 years; HR 1.40, 95% CI 1.01–1.95) and by 39% in the third quartile (5.6 years; HR 1.39, 95% CI 1.00–1.93) compared to those without a diabetes diagnosis. Within 10 years of follow-up, the risk of gastric cancer was linearly related to the duration of diabetes at baseline (p trend = .01; Table S3). When restricted to newly diagnosed diabetes, similarly, the risk of gastric cancer increased in the second quartile (median = 6.9 years; HR 7.20, 95% CI 4.76–10.89).

The analyses of the association between diabetes and gastric cancer mortality consisted of 580 663 participants and 5084 gastric cancer deaths during a median follow-up of 14.7 years (Table S4). Diabetes was associated with a 15% increased risk of death from gastric cancer (HR 1.15, 95% CI 1.03–1.28; I² = 0.00%, p = .68; Table S5). Analyses by various subgroups did not show any differential associations (Table S6). Despite relatively few cases (N = 97), diabetes was positively associated with cardia (HR 2.18, 95% CI 1.14–4.14) but not with noncardia gastric cancer mortality (1.08, 0.95–1.24; p_{heterogeneity} = .04; Table S7). Gastric cancer mortality in participants with diabetes relative to participants without diabetes increased until ~10 years of diabetes duration but decreased afterwards (Table S8).

In the sensitivity analyses, although the overall association was attenuated when cases occurring within the first 2 years of follow-up were excluded, the magnitude of association between diabetes and gastric cancer incidence (HR 1.11, 95% CI 1.00–1.24) and mortality (HR 1.11, 95% CI 0.96–1.29) remained similar (Table S9). The associations restricted to participants >30 years old at baseline did not change for gastric cancer incidence (HR 1.15, 95% CI 1.06–1.26) and mortality (HR 1.15, 95% CI 1.03–1.28; Table S9). In the competing risk analyses, association for gastric cancer incidence slightly decreased but remained significant (HR 1.12, 95% CI 1.03–1.22); whereas for mortality, the estimate was attenuated and nonsignificant (HR 1.10, 95% CI 0.99–1.23; Table S10). For the diabetes duration, the inverted lambda-shaped curve remained consistent (p_nonlinear = .0; Table S11). In the leave-one-out analysis, none of the cohorts included in the study substantially differed (<4%) in both gastric incidence and mortality analyses (Table S12). The quantitative probabilistic bias analysis found a 0%–34% bias if H. pylori was unmeasured (Table S13).

A total of 26 studies regarding the association between diabetes and gastric cancer risk were included in the meta-analysis (Table S14). Diabetes was associated with a higher risk of overall gastric cancer (risk ratio [RR] 1.24, 95% CI 1.15–1.34) with statistically significant heterogeneity across the studies (I² = 86.60%, p < .001; Table S15). Analyses by various subgroups did not show any differential associations by sex, region, study design, method of diabetes assessment, anatomical subsite, publication year, follow-up time, and adjustment variables (Table S16). The leave-one-out analysis revealed <1% differences by each excluded study (Table S17). The funnel plot demonstrated a symmetrical distribution (Figure S4) with no evidence of publication bias was detected by Egger's regression tests (p ≥ .05).

4 DISCUSSION

In this pooled analysis of Asian populations, we found that diabetes was associated with a higher risk of both gastric cancer incidence and mortality. To our knowledge, this is the largest study to comprehensively assess the association between diabetes duration and gastric cancer in the Asian population.

Several mechanisms have been proposed for the link between diabetes and gastric cancer. First, insulin and hyperglycemia may directly promote gastric cancer by overstimulating tumorigenic pathways such as PI3K/AKT/mTOR,²⁷ insulin-like growth factor-1 (IGF-1), strongly linked to insulin resistance, promotes gastric cancer cell proliferation and invasion via the Ras/RAF/MEK/ERK and JAK1/STAT3 cellular signaling pathways.²⁸ Hyperglycemia also enhances gastric carcinoma proliferation and migration via thr Pin1/BRD4 transcriptional pathway, involved in the hyperglycemia-induced inflammatory process.²⁹ Second, H. pylori infection, a known risk factor for gastric cancer, is maintained in hyperglycemia, allowing the bacteria to grow and survive longer in patients with diabetes.³⁰ An in vitro study reported enhanced adhesion, growth, and viability of the H. pylori cytotoxin-associated gene A antigen, an oncoprotein, under higher glucose concentrations.³¹ Finally, some glucose-lowering therapies, such as sulfonylureas and insulin therapy, may increase gastric cancer risk by interacting with insulin and insulin IGF-1R signaling.³² In contrast, metformin, which counteracts insulin resistance, can reduce the risk by activating AMPK and inhibiting IGF-1R, respectively.³³

Our observation of the positive association between diabetes and gastric cancer risk corroborates previous studies.⁴ Our up-to-date meta-analysis, including subgroup analysis by sex, anatomical subsite, type of study, and average follow-up time, confirmed the positive direction with gastric cancer. Compared to previous meta-analyses,^{4, 6} our meta-analysis exclusively focused on studies published after January 2000 to ensure temporal relevance. Furthermore, we conducted a subgroup analysis focusing on specific anatomical subsites (cardia and noncardia gastric cancer), addressing a knowledge gap in previous meta-analyses. Additionally, we extended the follow-up period in the subgroup analysis by categorizing studies into <10 years and ≥10 years of follow-up. This allowed for a better understanding of the effects of diabetes on gastric cancer, considering the long onset time of the disease. Although the heterogeneity among studies was high in ours and other meta-analyses,^{4, 6} mainly due to study quality and effect measures,⁴ the heterogeneity among Asian cohorts included in our analysis was relatively low.

We observed that the magnitude of association was higher in women compared to men, although the difference was not statistically significant, which is also shown in our meta-analysis. Although we cannot dismiss a possible interaction between women's hormonal condition and gastric cancer risk, our meta-analysis supports our findings that diabetes is associated with increased risk of gastric cancer in both men and women.

Our study found no significant difference in the association between diabetes and gastric cancer incidence between cardia and noncardia subsites. Although the results could be affected by the relatively small number of cardia gastric cancer cases, there was a slightly higher positive association between cardia gastric cancer and diabetes. When combined with published literature, diabetes was significantly associated with cardia gastric cancer. H. pylori infection, known to be more associated with noncardia gastric cancers, may interact with diabetes as described here; whereas, obesity, more reportedly linked to cardia cancers, may in tandem work with the effects of diabetes.³⁴ Similarly, despite the absence of statistical difference between the two histological subsites, only intestinal gastric cancer showed a positive association. Although both histological subsites may be associated with H. pylori infection, the link is stronger with intestinal subtypes, which again may synergize with diabetes in the carcinogenesis.³⁵ Further research is warranted in disentangling the effects of H. pylori infection and obesity in the diabetes-gastric cancer subtypes association.

Diabetes and gastric cancer mortality association were previously investigated within the Asian Cohort Consortium as part of an analysis for all-cause and cause-specific mortality.^{36, 37} With a longer follow-up (12.7 vs 14.7 years), we reevaluated the association between diabetes and gastric cancer mortality overall and newly by subgroups and anatomical subsites. Although we found a positive association between diabetes and gastric cancer mortality, the association disappeared in the 2-year lag time analysis and competing risk analysis. Therefore, diabetes may not be a main risk for gastric cancer mortality, as previous studies have suggested.^{36, 37}

We found that within 10 year-term of diabetes duration increases the risk of gastric cancer, supporting the evidence on the potential role of hyperglycemia-hyperinsulinemia in gastric cancer development. We also observed that after 11 years of diabetes duration, the association seems to weaken or even reverse. Our results were similar to a previous prospective study of two cohorts from the United States, which reported that overall cancer risk gradually reduced after 8 years of diabetes duration.³⁸ Although this study did not include gastric cancer as an outcome, the pattern was similarly shown in colorectal, breast, endometrial, and prostate cancers.³⁸ This may imply a harvesting effect, whereby there may be an increase in cancer detection among diabetes patients. Alternatively, hyperinsulinemia in early diabetes may play a more significant role than hyperglycemia in promoting cancer development because, after a particular diabetes stage, the progressive depletion of beta cell function leads to a steady decline in endogenous insulin production.³⁸

The major strength of our study includes a large sample size of Asian populations with a long-term follow-up, which gave us the statistical power to conduct stratified and sensitivity analyses. We also performed lag analyses to minimize recall bias and reverse causality, and competing risk analyses to mitigate survival bias. Furthermore, we had comprehensive data from each cohort at the individual level for covariate adjustments and heterogeneity assessment, avoiding some limitations of meta-analyses based on published results. Additionally, all study participants were selected from the general population, ensuring generalizability in the targeted population. The number of exclusions due to missing data on self-reported diabetes or any other covariates was not particularly large, <13% and 5%, respectively. Furthermore, we had information on gastric cancer diagnosis and mortality by anatomical subsite and histological subtype. All cases were validated by national cancer registries or by active, long-term follow-up.

Our study has some limitations. First, we lacked information about H. pylori infection. However, despite the synergistic effect of hyperglycemia and H. pylori infection on gastric cancer, there is evidence that diabetes increases the risk of gastric cancer even after H. pylori eradication,¹¹ implying different and independent mechanisms on gastric cancer carcinogenesis. Moreover, our quantitative probabilistic bias analysis revealed a relative wide range bias, which may be explained by regional differences in H. pylori prevalence. Second, diabetes diagnosis was self-reported at recruitment, which may have introduced misclassification bias because some participants may have been unaware of their diabetes diagnosis or developed it during the cohort follow-up. However, misclassification should not have occurred differentially between cases and noncases.¹⁹ Also, the subgroup analysis in our meta-analysis also showed that the association of self-reported diabetes and gastric cancer risk did not differ from the medical record-confirmed results. Diabetes type was unavailable; nonetheless, we can assume that type 2 diabetes predominated, as only 0.43% of diabetes cases were diagnosed before age 30, and our findings were unaffected by the exclusion of patients below this threshold. Third, we could not control for other factors related to gastric cancer; therefore, we could not rule out residual confounding. However, we have adjusted our models for known gastric cancer factors such as obesity, alcohol consumption, smoking, and education as a proxy of income to minimize confounding bias. Although we have a large number of cancer cases, some subgroup analyses may have insufficient statistical power. We cannot rule out the possibility of selection bias, particularly in histology analysis, given the limited percentage of cases with this information (50%); nevertheless, the anatomical subsite accounts for ~65%–100% of cases. Finally, we lacked information on the type of diabetes treatment and hypoglycemic drug use. Even with this information, assessing the influence of diabetes treatment on gastric cancer is challenging due to significant heterogeneity in the effects of different hypoglycemic drugs on gastric cancer risk.^{32, 33} Additional long-term prospective cohort studies with detailed information on diabetes treatment and hypoglycemic drugs are warranted to clarify the associations between diabetes/diabetes duration and gastric cancer risk.

In conclusion, we found that diabetes was associated with an increased gastric cancer incidence. The largest gastric cancer risk was found within the first 10 years after a diabetes diagnosis. As diabetes is highly prevalent in Asia, our findings underline the importance of lifestyle and metabolic health maintenance in the relatively healthy general population, and proposes a closer surveillance in the early stages of diabetes for gastric cancer prevention.

AUTHOR CONTRIBUTIONS

Katherine De la Torre and Minkyo Song are joint first authors. Katherine De la Torre contributed to design, methodology and study visualization; led the data analysis and interpretation of the results; and drafted the first draft of the manuscript. Minkyo Song conceptualized the design of the study, methodology, and visualization; supervised data analysis; interpreted the findings; and created the first draft. Sarah Krull Abe and Eiko Saito contributed as project managers. Md. Shafiur Rahman and Md. Rashedul Islam performed data curation. Sukhong Min contributed to meta-analysis and drafted some sections of the original manuscript. Dan Huang verified data onsite and contributed to data analysis. Yu Chen provided critical revision of the manuscript, and clinical input. Prakash C. Gupta, Norie Sawada, Akiko Tamakoshi, Xiao-Ou Shu, Wanqing Wen, Ritsu Sakata, Jeongseon Kim, Chisato Nagata, Hidemi Ito, Sue K. Park, Myung-Hee Shin, Mangesh S. Pednekar, Shoichiro Tsugane, Takashi Kimura, Yu-Tang Gao, Hui Cai, Keiko Wada, Isao Oze, Aesun Shin, and Yoon-Ok Ahn obtained funding, provided resources, verified data analysis, and provided critical revision of the manuscript. Habibul Ahsan, Paolo Boffetta, Kee Seng Chia, Keitaro Matsuo, You-Lin Qiao, Nathaniel Rothman, Wei Zheng, and Manami Inoue contributed to project supervision and provided critical revision of the manuscript. Daehee Kang obtained funding, contributed to the study conceptualization and design, supervision, and interpretation of results and commented on the paper draft. All authors contributed to the critical revision of the manuscript, and approved the final version. Daehee Kang, Minkyo Song, and Katherine De la Torre act as the guarantors of the study and accept full responsibility for the work and/or the conduct of the study, had access to the data, and controlled the decision to publish.

FUNDING INFORMATION

This work was supported by the following grants for study design, data collection and participant recruitment: Mumbai Cohort Study funded by the International Agency for Research on Cancer, Lyon, France; Clinical Trials Service Unit, Oxford, UK; World Health Organization, Geneva, Switzerland; Japan Collaborative Cohort Study (JACC), the National Cancer Center Research and Development Fund, grant-in-Aid for Cancer Research; Grant for Health Services and grant for Comprehensive Research on Cardiovascular and Life-Style Related Diseases from the Ministry of Health, Labour and Welfare, Japan; and grant for the Scientific Research from the Ministry of Education, Culture, Sports, Science and Technology, Japan; Shanghai Women's Health Study (SWHS), the US National Cancer Institute (NCI) (grant numbers R37 CA070867 and UM1 CA182910; Principal investigator: W. Zheng); Shanghai Men's Health Study (SMHS), the US NCI (grant number UM1 CA173640; Principal investigator: X.O. Shu); Japan Public Health Center-based prospective Study (JPHC Study) I and II, National Cancer Center Research and Development Fund [23-A-31 (toku) and 26-A-2; since 2011] and a Grant-in-Aid for Cancer Research from the Ministry of Health, Labour and Welfare of Japan (from 1989 to 2010; Principal investigator: S. Tsugane); Life Span Study Cohort, the Japanese Ministry of Health, Labour and Welfare and the US Department of Energy; Korean National Cancer Screenee Cohort Study, National Cancer Center National Cancer Center Korea Research Grant (grant number 2210990, 24H1080; Principal investigator: J. Kim); 3 Prefecture Aichi Study, The Japanese Ministry of the Environment (formerly, Environment Agency; Principal investigator: K. Matsuo); Takayama Study, National Cancer Center Research and Development Fund (Principal investigator: C. Nagata); Korean Multicenter Cancer Cohort (KMCC), the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIP; No. NRF-2016R1A2B4014552; Principal investigator: S.K. Park; Seoul Male Cancer Cohort, the National R&D Program for Cancer Control, National R&D Program for Cancer Control, Ministry of Health & Welfare, Republic of Korea (0520160-1; Principal Investigator: Myung-Hee Shin); National Cancer Center Japan Research and Development Fund (30-A-15,23-A-31(toku),26-A-2,29-A-4; Principal investigator: M. Inoue).

CONFLICT OF INTEREST STATEMENT

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.