Serum Levels of Androgens, Estrogens, and Sex Hormone Binding Globulin and Risk of Primary Gastric Cancer in Chinese Men: A Nested Case–Control Study

Gastric cancer is the fifth most common cancer and ranks the third in mortality worldwide (1). In 2018, an estimated 1,033,701 individuals were diagnosed of gastric cancer and 782,685 patients died of this disease (1), which impose a severe health burden especially in developing countries. In China alone, approximately 390,000 people died from gastric cancer every year, which accounts for 50% of global death cases (1–3). Anatomically, gastric cancer can be classified into cardia and noncardia gastric cancer, which are characterized by distinct risk factors and clinical feature and both of them have a high incidence in Linxian, China (4).

Age-standardized incidence rates of gastric cancer show a male predominance with a male-to-female rate of more than 2:1 in most populations (1, 5). This difference, however, cannot be entirely attributed to the different prevalence of established risk factors, such as smoking (6) and Helicobacter pylori infection (7) between two sexes. Limited studies suggest endogenous sex steroid hormones and sex hormone binding globulin (SHBG, a modulator of hormones bioavailability) may be associated with gastric cancer (8, 9). Independently, epidemiological studies show that men receiving antiandrogen therapies for prostate cancer may have a lower risk of gastric cancer (10), whereas antiestrogen drugs, for example, tamoxifen in women may increase the risk (9). It is biologically plausible that sex steroid hormones involve in the proinflammatory state (11, 12). A decrease testosterone levels has been associated with increasing levels of inflammatory cytokines in aging men (13) and estrogen-related receptors function has been reported as a potential tumor suppressor in gastric cancer (14–16).

However, whether serum sex steroid hormone and SHBG are associated with incident gastric cancer remains unclear. For example, an epidemiologic study reported no association between hormone replacement therapy and gastric cancer in a large prospective study of women (17). In addition, most current evidence comes from Western studies of mostly white subjects and there is no data from Chinese subjects. Cancer rates are strongly population dependent as does sex hormone status that may lead to different associations in different populations (18). Therefore, we investigated whether serum levels of androgens (androstendione and testosterone), estrogens (estradiol and estrone), and SHBG were associated with primary noncardia gastric risk in Chinese males.

We designed a nested case–control study within the Linxian Nutrition Intervention Trial (NIT), including the Dysplasia Trial (19) and the General Population Trial (Supplementary Table S1; ref. 20). The Dysplasia Trial enrolled 3,318 individuals aged 40 to 69 years with esophageal dysplasia in three communes of northern Linxian, China. Subjects were randomized and received either multiple vitamin/mineral supplements (14 vitamins and 12 minerals) or placebos for 6 years, beginning in May 1985 (19). The General Population Trial recruitment commenced in March 1986 and a total of 29,584 healthy villagers of the same age group were randomly assigned to receive four daily vitamin/mineral supplement combinations for 5.25 years from March 1986 to May 1991 in a one-half replicate of a 2⁴ fractional factorial experimental design (20). Each subject was interviewed, given a brief physical examination, and had 10 mL of blood drawn. After collection, serum specimens were separated, aliquoted, and stored frozen at −70°C for future analyses. The study was performed according to the guidelines of the Helsinki declaration. Ethical approval was obtained through the Institutional Review Boards of the NIH and the Chinese Academy of Medical Science (Beijing, China), and all subjects gave written informed consent.

Cases of incident primary gastric cancer was defined using the International Classification of Diseases, 10th edition code for noncardia gastric cancer (i.e., C16.1-C16.8) and were ascertained by the Linxian Cancer and Death Registries Data. Diagnostic materials (case records, endoscopy, and pathology slides) for cases were reviewed and confirmed by a panel of American and Chinese experts (1985–1996) or senior Chinese diagnosticians from Beijing (after 1996) and nearly all cases were confirmed as adenocarcinomas. Loss to follow-up was minimal by 2016 (n = 381, or 1.3%; ref. 21).

Through the 15 years of observation between baseline blood draw and cancer diagnosis (from May 1991 to May 2006; median follow-up time, 12.25 years), 560 (213 female and 347 male) cases of incident primary noncardia gastric cancer were diagnosed in the NIT cohort. For this study, we selected the 328 male cases with sufficient available serum for analysis, including 31 cases from the Dysplasia Trial and 297 cases from the General Population Trial. There were no significant differences on characteristics between the included cases and overall cases. The 1:1 matched controls were male cancer-free subjects sampled from the same trial subjects with a blood sample, matched by age at start of pills (±3 years), date of blood collection (±31 days) and trials.

Estrone, total estradiol, and testosterone were measured by validated radioimmunoassays after organic solvent extraction and Celite column partition chromatography (22, 23). SHBG was measured by direct immunoassays using the Immulite analyzer (Diagnostic Products Corporation). Free estradiol and free testosterone concentrations were calculated from measured estradiol, total testosterone, and SHBG with albumin assumed to be a constant (40 g/L) using the method by Sodergard and colleagues according to the law of mass action (24). Figure 1 depicts metabolism and existence forms of sex steroid hormones examined in this study.

Sex steroid hormones and SHBG measurements were conducted in 18 batches and each batch except for the last one included 41 samples (every 38 samples including 19 cases and 19 controls were accompanied by three pooled serum samples as internal controls). The three blinded quality control samples collected from the same subject were analyzed consecutively as triplets within each batch to minimize the laboratory variation between samples from the same subject. Using nested components of variance analysis, with logarithmically transformed quality control measurements across all batches (25), the estimated overall (intrabatch and interbatch) coefficients of variation for estrone, estradiol, testosterone, and SHBG were 4.3%–20.0%. Laboratory personnel were blinded as to the identity of cases, controls, and quality control samples.

Demographic characteristics included age and body mass index (BMI). We categorized age as ≤56 and >56 years by median to ensure each level had sufficient and similar number of observations when looking at interactions but we used age as a continuous variable when adjusting for confounding. We first defined BMI according to the standard cut-off points by the Work Group on Obesity in China (WGOC) in categories of underweight: ≤18.5 kg/m², normal: >18.5-≤24 kg/m², overweight: >24.0-≤28.0 kg/m², and obese: >28.0 kg/m² (26), and then we reclassified BMI as dichotomous variable by median (≤21.2 kg/m² and >21.2 kg/m²) because almost all (89.63%) subjects had underweight and normal weight. Health-related lifestyle indicators, including alcohol drinking (yes: any alcoholic beverage consumption in the last 12 months, or no: no alcohol consumption in the last 12 months) and smoking status (yes: any regular tobacco consumption including cigarettes or pipes, or no: never smoked), were included for analysis. Helicobacter pylori seropositivity and pepsinogen I/pepsinogen II ratio were measured in some subjects in previous studies (27), but were not considered further due to their incomplete coverage of this subset of the current cohort. Associations between Helicobacter pylori and gastric cancer show modest effect sizes in previous study due to the overall high infection rate. Trials included the Dysplasia Trial and the General Population Trial.

Baseline demographic characteristics were summarized by case and control groups. Continuous variables were reported as medians and interquartile ranges (IQR). Categorical variables were described as number of observations and percentages (%). Differences in potential covariates between cases and controls were assessed by Wilcoxon–Mann–Whitney test for continuous variables and the χ² test for categorical variables. The correlation between exposures was evaluated using Pearson correlation coefficients among log₁₀-transformed values of exposures in the controls. Pearson correlation coefficients were used to examine the magnitude of the correlation, as well as the direction of the relationship between sex steroid hormones and SHBG.

We used unconditional logistic regression to analyze the matched data with no loss of validity and a possible increase in precision, and estimate ORs and 95% CIs for the association of incident gastric cancer with sex steroid hormones and SHBG. We firstly conducted crude logistic regression for each hormone and SHBG, respectively. We then conducted multivariable logistic regression for each sex steroid hormone and adjusted for age (continuous), smoking status, alcohol drinking, BMI (continuous), trial, and other related sex steroid hormones or SHBG. We treated the sex hormone levels as continuous variables, scaled to logarithm base 10 of hormone levels in the model. We also used quartiles metric of sex steroid hormones to assess the association. A very small fraction (approximately 1%) of subjects had extremely high or low values of sex steroid hormones, which were unreliable based on a priori knowledge; thus, the outliers above the logarithm base 10 of sex hormone levels ± 4SD were excluded from analysis.

Age and BMI have been reportedly influencing the levels of sex steroid hormones and involved in gastric carcinogenesis (28–30). Therefore, we conducted subgroup analysis stratified by these two dichotomous variables to explore whether there were interactions between sex steroid hormones and these factors. In addition, we estimated OR for alcohol drinking, smoking, and tertiles of follow-up (≤5 years, >5 to ≤10 years or >10 years). One set of sensitivity analysis restricted to the General Population Trial population was conducted in addition to the primary logistic regression models. For current analysis, two-sided P values <0.05 were considered to be statistically significant. All statistical analyses were conducted in February 2020 using Stata, version 15.0 (StataCorp, LLP).

Table 1 shows the distributions of demographic and potential risk factors by case and control groups of men in this subset of the NIT cohorts. There were no significant differences between the case and control groups for age at baseline, alcohol drinking, smoking status, androstenedione, testosterone, free testosterone, FT_AT, estradiol, estrone, free estradiol, FE2_AE2, or SHBG (Ps>0.05). Compared with controls, subjects who developed gastric cancer were more likely to have a lower BMI (continuous variable, P < 0.01) or BMI≤21.2 kg/m² (dichotomous variable, P = 0.04). Supplementary Table S2 shows the correlations between sex steroid hormones and SHBG. We found that free testosterone was highly correlated with free and bioavailable testosterone (FT_AT; Pearson R = 0.97), and as well as free estradiol and free and bioavailable estradiol (FE2_AE2; Pearson R = 1.00).

Table 2 presents associations between the six sex steroid hormones, SHBG, and the risk of gastric cancer. There were no overall associations between sex steroid hormone concentrations (androstenedione, testosterone, FT_AT, estradiol, estrone, and FE2_AE2) and risk of gastric cancer in men. None of the risk estimates or the P_trend was statistically significant for these sex steroid hormones. Subjects with serum SHBG in highest quartile, as compared with those in lowest quartile, had an increased risk of gastric cancer in a crude model (OR = 1.71; 95% CI = 1.08–2.71; P_trend = 0.09). Results for quartiles of SHBG remained unchanged after adjustment for multiple covariates (OR quartile 4 vs.1 = 1.87; 95% CI = 1.01–3.44), although no monotonic trends were observed (P_trend = 0.12); effect measure was nonsignificant in linear regression (OR = 1.71; 95% CI = 0.85–3.44).

In subgroup analyses by BMI (Table 3), there were inverse associations of gastric cancer risk with serum FT_AT (OR quartile 4 vs.1 = 0.37; 95% CI = 0.19–0.72; P_interaction = 0.01) and estrone (OR quartile 4 vs.1 = 0.51; 95% CI = 0.27–0.97; P_interaction = 0.22) in lower BMI group. For SHBG, subgroup analysis showed similar association across strata, though there was some evidence for a stronger association in men with BMI>21.2 kg/m² in lower middle serum concentration (OR quartile 2 vs.1 = 2.21; 95% CI = 1.13–4.30; P_interaction = 0.03). For other sex steroid hormones, subgroup analyses showed comparable and not statistically significantly different results. We also examined the results stratified by age, alcohol drinking, and smoking and did not observe heterogeneity across groups (Supplementary Table S3–S5). In general, the Wald tests supported a significant interaction between follow-up time and sex steroid hormones and SHBG (Supplementary Table S6). We found a significant negative association of gastric cancer risk and testosterone in events occurring ≤5 years after baseline. There was effect reversal across three strata of follow-up time for other androgens, estrogens, and SHBG (crossover interactions). When we restricted the subjects to those in the General Population Trial (n = 594), the results of sensitivity analysis were similar with the main analysis (Supplementary Table S7).

This is the first study to prospectively examine the association between prediagnostic serum sex steroid hormones concentrations and SHBG and the risk of noncardia gastric cancer in a population-based setting among Chinese men. There were no overall associations between sex steroid hormone concentrations and risk of developing gastric cancer, but we observed some evidence for novel association for higher serum SHBG concentrations. In addition, higher baseline serum estrone and FT_AT concentrations was inversely associated with risk of gastric cancer among subjects with BMI≤21.2 kg/m² and evidence for a quantitative interaction such that the direction of the association was the opposite among heavier men. Given that this interaction was not predicted a priori and the result may be provisional and we need further studies to evaluate whether these associations reflect a real pattern or are spurious. Of note, our population has relatively low distribution of BMIs even for a population within China. A large fraction of men were underweight at the time of blood sample collection, which may directly alter typical levels of both sex steroid hormones and SHBG.

To our knowledge, few previous studies have explored the association between serum SHBG concentrations and the risk of gastric cancer, and none of the previous data have suggested a positive association prospectively. An Indian study of 50 subjects found a 2.3-fold increased expression in serum SHBG of patients with gastric cancer across 643 proteins (31). Another case–control study of 86 healthy subjects and 98 patients with gastric cancer conducted in Taiwan, China found a consistently higher serum levels of SHBG among gastric cancer patient groups compared with that of the controls (32). Nevertheless, the two studies collected blood samples at diagnosis and thus could reflect reverse causation, whereby the tumor or other disease processes, such as cachexia, could affect sex steroid hormones concentrations (31, 32). Petrick and colleagues examined the association between prediagnostic concentrations of circulating sex steroid hormones and esophageal/gastric cardia adenocarcinoma risk among men from three western cohorts, which have significantly different etiology from noncardia gastric cancer studied here and did not observe an association (8).

SHBG is a plasma glycoprotein that transports androgens and estrogens while in circulation and regulates the bioavailability of free estrogens and androgens (33). Bioavailable estradiol has the potential ability to attenuate tumor malignancy by inducing apoptosis and reducing cell viability in gastric cancer cells (34), providing indirect evidence in accordance with our findings for SHBG and gastric cancer. Research in the past decade suggests that cell membranes of many tissues could express a receptor for SHBG (35). Binding of SHBG to its receptor has been shown to activate cyclic adenosine monophosphate (36), an intracellular signal transduction pathway important for many biological processes including cancer growth, which may independent from its role in hormone transport (37).

Many studies report that SHBG concentrations are negatively associated with BMI (38, 39). This inverse correlation is also observed at in the current study (Pearson R = -0.2064; P < 0.01). This is thought to occur because adipose cells increase the circulating levels of insulin and increase insulin-like growth factor 1 (IGF1) bioactivity (40) and finally result in reduced hepatic synthesis and blood concentrations of SHBG (41), which suggests that the effect measures of SHBG may be more substantial among slim men. Studies of BMI related androgens and estrogens concentrations change among men are inconclusive (42–45). Further work including men with a wider range of BMI is needed to replicate these associations and infer a possible mechanism underlying our observations.

This study has several strengths in design and analysis. It was the first to investigate the association of gastric cancer with prediagnostic sex steroid hormones and SHBG in an Asian population. The study population was highly homogeneous regarding to occupation and socioeconomic status. With the 15 years of post-trial follow-up between baseline blood draw and cancer diagnosis, minimal loss to follow-up rate (1%), and extensive information on potential confounders, we were able to evaluate the associations prospectively and conduct analyses adjusted for and stratified by major risk factors. However, we should also note the limitations of this study, which include a modest sample size. In addition, we do not have complete data on Helicobacter pylori seropositivity, pepsinogen I/II ratio (27, 46) and lack data on other serologic factors (47–48; e.g., fasting insulin, insulin-like growth factor 1 and inflammatory biomarkers), which may confound the associations between sex hormones and risk of gastric cancer. Our previous study found that individuals with low pepsinogen I/II ratios had higher risk of noncardia gastric cancer (27) and another study found that estradiol and androstenedione have the potential to inhibit the growth of Helicobacter pylori (46). A further limitation is that we only had a single blood draw, and misclassification of sex steroid hormones and SHBG could not be excluded.

In conclusion, there were no overall associations between androgens and estrogens and risk of noncardia gastric cancer in men. Higher concentration of SHBG was associated with increased risk of gastric cancer. We found some evidence for associations of sex steroid hormones in men with lower BMIs. Studies of sex steroid hormones and gastric cancer in women are needed to fully explore this hypothesis and further explore whether BMI will modify the association.

No disclosures were reported.

Z. Zhu: Visualization, methodology, writing–original draft, writing–review and editing. Y. Chen: Investigation, writing–review and editing. J. Ren: Conceptualization, methodology. S.M. Dawsey: Writing–review and editing. J. Yin: Writing–review and editing. N.D. Freedman: Writing–review and editing. J.-H. Fan: Writing–review and editing. P.R. Taylor: Writing–review and editing. Y. Liu: Data curation, formal analysis, supervision, writing–review and editing. Y.-L. Qiao: Supervision, funding acquisition, writing–review and editing. C.C. Abnet: Conceptualization, data curation, formal analysis, validation, investigation.

This work was supported in part by eNCI contracts (N01-SC-91030, N01-RC-47701 and N02CP-2017–00047, to Y.L. Qiao) to the National Cancer Center, Chinese Academy of Medical Sciences, in part by the National Cancer Center, Chinese Academy of Medical Sciences (2017-I2M-B&R-03), and in part by the Intramural Research Program of the NCI, NIH.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.