Gastric cancer is the sixth most common cancer according to global cancer statistics in 2020.¹ Some risk factors have different effects on gastric cancer by site: cardiac gastric cancer (CGC) and non-cardiac gastric cancer (NCGC).² A high body mass index (BMI) was associated with CGC but not NCGC in a meta-analysis.² The effect of BMI on gastric cancer has been controversial. Several studies have shown that lower BMI was associated with a higher risk of gastric cancer,^3,4 whereas other studies show no statistical significance.^5,6 Although a meta-analysis suggests a positive association between serum glucose level and risk of hormonally driven cancer, such as colorectal, prostate, endometrial, and liver cancer,⁷ their association has been debated. A small Korean nested case-control study showed no association between gastric cancer and serum glucose level.⁸ However, hyperglycemia and higher hemoglobin A1c (HbA1c) were associated with the increased risk of gastric cancer in Japanese cohort studies.^9,10 Previous studies have suggested the inverse association between high-density lipoprotein cholesterol (HDL-C) and overall cancer.^11,12 Several recent studies have suggested an inverse association of serum HDL-C with gastric cancer.^13,14 However, site-specific effects of hyperglycemia and HDL-C on gastric cancer have not been reported.

We investigated the site-specific effects of HDL-C, glucose, and BMI on gastric cancer using large-scale population-based data. We focused on HDL-C among lipids because serum levels of total cholesterol and low-density lipoprotein can be easily changed by lipid-lowering drugs,¹⁵ and the serum level of triglyceride is strongly associated with recent diet and consumption of alcohol. On the other hand, the effect of lipid-lowering drugs on HDL-C serum level is less prominent, HDL-C is mainly affected by exercise,^15-17 and its variability is relatively low. Additionally, we analyzed the site-specific effects of those factors on gastric cancer using a hospital-based validation set.

Data extraction and content

This study used National Health Insurance Service-National Sample Cohort (NHIS-NSC) data (REQ0000006487) gathered by the National Health Insurance Service System (NHISS). These data included epidemiologic and laboratory data, T40 disease code data (International Classification of Disease, 10th revision), and national gastric cancer screening data. Demographic questions included information on chronic disease, medication, habits, and family history; and the laboratory examination measured obesity indices, blood pressure, and blood chemistry. Blood levels of lipids and glucose were measured after 12 hours of fasting. BMI was calculated as weight/height² (kg/m²). Blood sampling was performed on the morning of gastric cancer screening, and endoscopy was the last examination.

National gastric cancer screening data included gastric cancer screening results and basic questionnaires. The Korean national gastric cancer screening provides gastric cancer screening every other year using endoscopy or gastrography for persons aged 40 years or older.¹⁸ Endoscopic screening reports included endoscopic findings, sites of gastric lesions, pathology results, final diagnosis, previously diagnosed gastric cancer, and family history of gastric cancer (Supplementary Table 1). The endoscopic findings included normal, gastritis, suspected gastric cancer, early gastric cancer, advanced gastric cancer, benign gastric ulcer, gastric polyp, gastric subepithelial tumor, and others. The site of a gastric lesion was classified as antrum, body, cardia, or fundus. The final diagnosis was classified as normal, benign disease, suspected gastric cancer, gastric cancer, and others. A repeat endoscopy was performed for pathologically indefinite gastric neoplasms. Gastric cancer detected on a repeat endoscopy is evident in T40 disease code data (C16).

Study population

Subjects who underwent both national gastric cancer screening and general health screening from January 2011 to December 2011 were included in this study. We excluded subjects who had been diagnosed with gastric cancer before 2011 and subjects with pre-existing gastric cancer based on national cancer screening questionnaires (Fig. 1A). In this study, the presence of gastric cancer was defined as follows: (1) final diagnosis with gastric cancer in national cancer screening (n=7,831) or (2) final diagnosis with suspected gastric cancer in national cancer screening with the detection of gastric cancer (C16) in the T40 disease code (n=2,316). Among the cases of gastric cancer in national cancer screening, those with an unknown gastric cancer site (n=915) were excluded. Absence of gastric neoplasm (n=5.48 million) was defined as normal, benign lesion, or other for national cancer screening final diagnoses with the exclusion of gastric cancer (C16) and gastric adenoma (D131) in the T40 disease code (Fig. 1A). This study was approved by the Institutional Review Board of Kyungpook National University Chilgok Hospital (KNUHC 2016-08-024).

Validation set

This study included patients who were diagnosed with gastric cancer from 2005 to 2013 in Kyungpook National University Hospital and who had data of HDL-C, glucose, and BMI at diagnosis (n=3,346). The private health screening population was used as the control group (Fig. 1B). Among health screening persons who underwent endoscopy and laboratory examination, those who had no gastric neoplasm were enrolled as controls (n=19,609). We further excluded individuals younger than 40 years of age in both gastric cancer (n=120) and control groups (n=1,824). The requirement of informed consent was waived because this study reviewed medical records, did not include personal identification information, and had no effect on disease course.

Statistical analysis

Statistics for demographic characteristics, BMI, and serum levels of lipids and glucose by absence of gastric cancer, NCGC, and CGC were provided as mean±standard deviation or number (percentage). Continuous variables were categorized to conduct further analysis. According to Asian-Pacific guidelines, BMI was categorized as low (<18.5 kg/m²), normal (18.5 to 22.9 kg/m²), overweight (23 to 24.9 kg/m²), obesity I (25 to 29.9 kg/m²), or obesity II (≥30 kg/m²). According to the National Cholesterol Education Program (NCEP) and the Adult Treatment Panel (ATP III), HDL-C was classified as low (<40 mg/dL in men and <50 mg/dL in women) or normal (≥40 mg/dL in men and ≥50 mg/dL in women). Triglycerides were categorized as normal (<150 mg/dL) or high (≥150 mg/dL). Age was categorized by 10-year intervals (40–49, 50–59, 60–69, or ≥70 years). The site of gastric cancer was categorized as NCGC (antrum, body, and fundus) or CGC.

The factors associated with gastric cancer were estimated by logistic regression using odds ratios (ORs) and 95% confidence intervals (CIs). Significant variables in chi-square test or t-test were included in multivariate analysis. In the NHIS-NSC set, we adjusted for age, sex, BMI, HDL-C, diabetes, cerebrovascular disease, heart disease, hypertension, smoking status, drinking status, and family history of gastric cancer. In the validation set, we adjusted for age, sex, BMI, HDL-C, and diabetes. All analyses were conducted using SAS software version 9 (SAS Institute). All statistical tests were two-sided, and values of P<0.05 were considered statistically significant.

Demographic and laboratory findings

A total of 5.49 million participants (2.427 million men [44.2%] and a mean age of 55.7 years) was eligible for this study (Fig. 1A). Among 10,147 gastric cancer cases, CGC was observed in 152, and NCGC was noted in 6,764. Gastric cancers were located in the antrum (43.6%, n=3,013), body (48.7%, n=3,366), fundus (5.6%, n=385), and cardia (2.2%, n=152). Most gastric cancers were categorized as early (77.6%), and patients with advanced gastric cancer represented 17.8% of the study group. The remaining patients (4.6%) had an unknown stage. Many demographic factors were associated with NCGC and CGC. In categorized data, low BMI and low serum HDL-C were associated with gastric cancer (Table 1).

Effects of contributing factors by site

Differing effects of many factors on gastric cancer by site were observed (Table 2). BMI was inversely associated with NCGC (OR, 1.37 for BMI <18.5 kg/m², 0.88 for BMI of 23–25 kg/m², 0.84 for BMI of 25–30 kg/m², and 0.80 for BMI ≥30 kg/m²; P for trend <0.001), whereas BMI had no association with CGC (Fig. 2A). A low serum level of HDL-C was strongly associated with CGC (OR, 1.90; 95% CI, 1.34 to 2.71) and was moderately associated with NCGC (OR, 1.41; 95% CI, 1.34 to 1.49) (Fig. 2B). The effect of hyperglycemia was also more prominent in CGC (OR, 1.50; 95% CI, 1.06 to 2.11) than in NCGC (OR, 1.19; 95% CI, 1.12 to 1.25). Men predominance was larger in CGC (OR, 3.28; 95% CI, 2.09 to 5.14) than in NCGC (OR, 1.98; 95% CI, 1.85 to 2.12), and the positive relationship of gastric cancer with aging was more prominent in CGC (OR, 6.28 for ≥70 years of age) than in NCGC (OR, 4.21 for ≥70 years of age). However, smoking, frequent consumption of alcohol, and family history were strongly associated with NCGC but had no association with CGC (Table 2).

External validation set

We confirmed the site-specific associations of BMI, HDL-C, and glucose with gastric cancer in the external validation set. Totals of 3,262 gastric cancer patients and 14,121 private health screening individuals were finally enrolled (Fig. 1B). Demographic and laboratory differences between gastric cancer and controls are provided in Supplementary Table 2. Cardiac cancer (n=107) and non-cardiac cancer (n=3,119) patients were eligible. Early gastric cancer was seen in 70.2% (n=2,291) and advanced gastric cancer was seen in 29.8% (n=935) of the final eligible patients.

NCGC was inversely associated with BMI (P for trend <0.001), whereas CGC had a borderline association with BMI greater than 30 kg/m² (OR, 2.46; 95% CI, 1.00 to 6.11; P=0.054) (Fig. 2C). Low serum HDL-C was consistently associated with both NCGC (OR, 2.32) and CGC (OR, 2.80) (Fig. 2D). Hyperglycemia and increased age were strongly associated with both NCGC and CGC. Men predominance was larger in CGC (Table 3).

This is the first study to evaluate the site-specific effects of HDL-C and hyperglycemia on gastric cancer. The effects of low serum HDL-C, high glucose, aging, and men sex were more prominent in CGC than in NCGC. BMI was inversely associated with NCGC, whereas BMI had no association with CGC. Smoking, frequent consumption of alcohol, and family history were strongly associated with NCGC but had no association with CGC. External validation analyses showed that low serum HDL-C and hyperglycemia had a constant strong association with both CGC and NCGC. The effect of BMI on NCGC was similar to the findings of NHISS, whereas its effect on CGC was slightly different from the findings of the NHISS.

In our NHISS analysis, BMI was inversely associated with NCGC, whereas BMI had no association with CGC. An inverse association between BMI and NCGC was also observed in the validation set. Lack of association between BMI and CGC may have been due to bias due to a relatively low number of CGC cases. In the NHISS dataset, CGC had no association with BMI, but the highest BMI (≥30 kg/m²) had a borderline association with CGC in the validation set. Some differences in the effect of BMI on CGC in each study set may have been related to population composition, including factors such as age, sex, and BMI. A previous meta-analysis shows that CGC is increased in obese patients, whereas BMI has no association with NCGC.² However, the results of recent studies are similar to the results of our NHISS study.^4,19 In a study including Swedish men, obesity (BMI >30 kg/m²) had an association with low risk of NCGC but had no association with CGC.⁴ In a small Chinese cohort study (n=29,584), the highest quartile of BMI was associated with a lower risk of NCGC (hazard ratio [HR], 0.65), but BMI had no association with CGC.¹⁹ The biological mechanism of the negative association of NCGC and BMI is unknown. One possible explanation is that high use of tobacco and alcohol in gastric cancer patients may induce a malnourished state and weight loss during the course of their disease.²⁰ The stomach is a major source of circulating ghrelin, which increases food intake and causes weight gain.²¹ Ghrelin level is decreased in intestinal metaplasia²² and extensive atrophic gastritis, known precursors of NCGC.^23,24 Decreased ghrelin in persons with precancerous lesions may induce less food intake and less weight gain. This association can partially explain the inverse association between BMI and NCGC.

In the current study, low HDL-C was associated with both NCGC (OR, 1.41) and CGC (OR, 1.90), and the association was more pronounced in CGC. Low HDL-C was consistently associated with both NCGC (OR, 2.32) and CGC (OR, 2.80) in the validation set. An inverse association of HDL-C with overall cancer risk has been reported in patients with diabetes.¹¹ A previous Korean cross-sectional study from a single center considering Helicobacter pylori showed that high HDL-C (≥60 mg/dL) was inversely associated with gastric cancer (OR, 0.49) compared to low HDL-C (<40 mg/dL).¹³ A small Korean cohort study considering demographic factors and H. pylori showed that low HDL-C by NCEP increased the risk of gastric cancer (HR, 2.67).¹⁴ However, sub-analysis by site was not conducted due to small numbers of gastric cancer cases in previous studies. A recent cohort study also suggested an inverse association between HDL-C and overall gastric cancer in post-menopausal women but did not provide a site-specific association.²⁴ In the current study, we showed a consistent inverse effect of low HDL-C on CGC and NCGC, and the association was more pronounced in CGC. Furthermore, we confirmed this association in a validation set. The anti-inflammatory and antioxidant potential of HDL-C may be associated with its anti-cancer effect. An in vitro study showed anti-inflammatory reprogramming of macrophages by HDL-C in a cell line and animal model.²⁵ A recent study showed that HDL antioxidant activity limits cell proliferation induced by reactive oxygen species in prostate cancer cell lines.²⁶

In the analysis of NHISS data, the effect of high glucose was also more prominent in CGC (OR, 1.50) than in NCGC (OR, 1.19). Hyperglycemia was also strongly associated with NCGC (OR, 2.55) and CGC (OR, 2.29) in the validation set. Previous studies of 2,400 and 2,600 Japanese subjects have shown that hyperglycemia and HbA1c increase the risk of gastric cancer (risk ratio, 2.1 to 3.1).^9,10 Although the association between hyperglycemia and gastric cancer has been reported, no site-specific effect of hyperglycemia on gastric cancer has been reported. A possible explanation for the hazardous effect of hyperglycemia on gastric cancer is the increases in insulin resistance, insulin, and insulin-like growth factor 1 (IGF-1). IGF-1 is known to inhibit apoptosis, leading to cancer.²⁷ Higher IGF-1 and lower IGF-binding protein-3 have been shown to increase the risk of colorectal cancer in cohort studies.^28,29 Another plausible explanation is that high blood sugar can act as a direct carcinogen. Findings that high glucose level increased DNA damage in cultured cells support this possibility.³⁰

This is the first study to evaluate the site-specific effects of HDL-C and glucose on gastric cancer. Low serum HDL-C level and high fasting glucose level were strongly associated with CGC in both the NHISS and the validation set. BMI had no association with CGC in the NHISS, wherein the highest BMI was associated with CGC in the validation set. This result suggests that obesity-related metabolic changes, such as low HDL-C and hyperglycemia, rather than obesity itself, may be related to CGC.

Interestingly, family history, current smoker, and frequent consumption of alcohol were strongly associated with NCGC, whereas they had no association with CGC. Those with first-degree relatives with gastric cancer were 1.5-fold more likely to have NCGC in this study. Family history was also positively associated with overall gastric cancer (OR, 1.5 to 3.5) in previous case-control studies.³¹ The results that NCGC was associated with past (OR, 1.6) and current smoking (OR, 1.56) are similar to those in a case-control study from Japan.³² The finding that frequent consumption of alcohol was positively associated with NCGC is similar to that of a meta-analysis (pooled risk ratio, 1.20 for heavy drinkers versus nondrinkers).³³ Men predominance was also more marked in CGC (OR, 3.82) than in NCGC (OR, 1.98). A strong men predominance in CGC may be related to a stronger effect of sex hormones on CGC than on NCGC. This relationship needs further evaluation. The positive association between gastric cancer and age, with a dose-dependent pattern, was also more pronounced in CGC than in NCGC, particularly in the oldest age group.

This study has several strengths. First, we used a large data set that included many epidemiologic and clinical factors, and the large size allowed us to investigate the site-specific effects of various metabolic components on gastric cancer. To the best of our knowledge, this is the first study to evaluate the site-specific effects of HDL-C and hyperglycemia on gastric cancer. Second, this population-based study has less selection bias than is present in case-control studies. Third, we performed an external validation study to confirm these associations. Last, the quality of demographic and laboratory data is high because we included subjects who underwent both gastric cancer screening and general health examinations.

However, this study has several limitations. First, H. pylori infection was not considered in this study, although a previous small study considering H. pylori and metabolic factors has shown an inverse association between HDL-C and overall gastric cancer.¹⁴ Second, the small proportion of CGC cases may be insufficient to estimate the effect of BMI on CGC. Last, cases of gastric cancer in the validation set were retrospectively collected from a hospital, and epidemiologic variables such as alcohol and smoking are incomplete.

In conclusion, low serum HDL-C and hyperglycemia were associated with both CGC and NCGC, and these effects were more pronounced in CGC. NCGC was inversely associated with BMI. CGC appears to be strongly related to metabolic changes, such as low HDL-C and hyperglycemia, rather than obesity itself. Far different site-specific effects of many contributing factors on gastric cancer suggest the need for different approaches to predict and prevent them. The consistent positive relationship between low HDL-C and hyperglycemia with gastric cancer regardless of site suggests that those factors need to be considered as risk factors, similar to H. pylori infection, precancerous lesions, and age.

Supplementary materials can be found online at https://doi.org/10.7570/jomes22045.

The authors declare no conflict of interest.

This work was supported by the Korean Society of Gastrointestinal Cancer Research (Research Award 2018) (Su Youn Nam). The funders had no role in the design and conduct of the study; collection, management, analysis, or interpretation of the data; or preparation, review, or approval of the manuscript.

Study concept and design: SYN; acquisition of data: SYN and SWJ; analysis and interpretation of data: SYN, JJ, and SWJ; drafting of the manuscript: SYN; critical revision of the manuscript: SYN, JJ, and SWJ; statistical analysis: SYN and JJ; obtained funding: SYN; administrative, technical, or material support: SYN; and study supervision: SYN and SWJ.