2.1 Study population
This was an historical cohort study of participants who received a medical health-checkup at Asahi University Hospital (the NAGALA (NAfld in Gifu Area, Longitudinal Analysis) study, Gifu, Japan) . The purpose of medical health-checkup was to promote public health by early detection of chronic diseases and their risk factors and about 60-70% examiners received the examinations, repeatedly; thus, the participants represent apparently healthy individuals. Most of the participants of this medical health-checkup were employees of various companies and local governmental organizations in Gifu, Japan, and their consorts. The medical data of all individuals who agreed to participate in the study were stored in a database after removing all personally identifiable information. For the current study, we used the results of individuals who participated in the health-checkup program for at least one year between 2003 and 2016. The exclusion criteria of this study were as follows: the presence of gastric cancer at baseline examination, missing covariate data (body weight, high-density lipoprotein (HDL) cholesterol, and lifestyle factors) and no follow-up health-checkup programs. Informed consent was obtained from each participant. The study was approved by the ethics committee of Murakami Memorial Hospital and was conducted in accordance with the Declaration of Helsinki.
2.2 Data collection
A self-administered questionnaire was used for gathering the medical history and lifestyle factors of participants . In regard to alcohol consumption, participants were asked the type and amounts of alcoholic beverages consumed per week over the past month, and then the mean ethanol intake per week was estimated . For smoking status, the participants were categorized into three groups: never-, ex- and current smokers. In addition, smoking burden was evaluated by pack-years which were calculated by multiplying the number of cigarette packs smoked per day by the number of years of smoking . For exercise, participants were asked to describe the type, duration and frequency of sports or recreational activities . Based on the results, we defined regular exercisers as the participants who performed any kind of sports activity at least once a week on a regular basis . Body mass index (BMI) (kg/m2) was calculated as body weight (kg) divided by height (m) squared. Waist circumference was measured as the abdominal circumference around the navel. Fasting plasma glucose, triglycerides, or HDL cholesterol was measured using the venous blood after an overnight fast. We also performed an upper gastrointestinal series or gastro-esophageal endoscopy and fecal occult blood test. If gastrointestinal cancer was suspected, we contacted and encouraged the participants to receive further examinations to diagnose it. We then collected the medical information about gastrointestinal cancers by sending a standardized letter to the hospital where the subject received the additional examinations. Specialists in the field of gastrointestinal disease checked the collected information and defined each cases as esophageal cancer, gastric cancer, or colorectal cancer. The first standardized questionnaires were sent on Jan 1st 2003; thus, we set the study period as Jan 1st 2003 to Dec 31st 2016. The primary endpoint of this study was hazard risk (HR) of MHO for gastric cancer after adjusting for sex, age, and lifestyle factors including smoking habits, alcoholic consumption and physical activities.
2.3 Definitions of metabolic phenotypes
We used body mass index >25.0 kg/m2 to identify the individual with obesity. This value has been proposed as a cutoff for the diagnosis of individual with obesity in Asian people  and has often been used in Japan [25,26]. Four metabolic factors (fasting plasma glucose, triglycerides, HDL cholesterol and blood pressure) were used to divide participants into metabolically healthy or metabolically abnormal subgroups . Impaired fasting plasma glucose and/or diabetes was defined as fasting plasma glucose >5.6 mmol/L and/or current medical treatment. Hypertension was defined as systolic blood pressure >130 mmHg and/or diastolic blood pressure >85 mmHg or current medical treatment. Elevated triglycerides were defined as triglycerides >1.7 mmol/L or treatment for hyperlipidemia. Low HDL-cholesterol was defined as <1.0 mmol/L in men and <1.3 mmol/L in women. When none of these four metabolic factors were present, we defined the participants as metabolically healthy (MH) and when one or more of these four metabolic factors were present, we defined the participants as metabolically abnormal (MA) . Then, participants were categorized at the baseline examination into 4 phenotypes: metabolically healthy non-obesity (MHNO), metabolically healthy obesity (MHO); metabolically abnormal non-obesity (MANO), and metabolically abnormal obesity (MAO).
2.4 Statistical analysis
The study participants were divided into four groups based on metabolic phenotypes. Continuous variables were expressed as the means ± standard deviation or median (interquartile range) and categorical variables were expressed as numbers. The clinical characteristics at baseline examination of the four groups were compared; continuous variables of groups were evaluated by one-way ANOVA and Tukey’s Honestly Significant Difference Test or Kruskal-Wallis Test and Steel-Dwass Test, and categorical variables of groups were evaluated by Pearson’s Chi-Squared Test. Because of the censored cases and inconsistent follow-up duration, we used the Cox Proportional Hazards Model to calculate the HR of the four groups. We considered five potential confounders as covariates: age, sex, alcohol consumption , pack-years , and exercise . Because alcohol consumption and pack-years were skewed variables, logarithmic transformation was carried out before performing the Cox Proportional Hazard Model analysis.
Furthermore, we used the Cox Proportional Hazards Model to calculate the HR of each metabolic abnormality (hypertension, impaired fasting glucose, hypertriglyceridemia and low HDL-cholesterol).
The statistical analyses were performed using JMP version 13.2 software (SAS Institute Inc., Cary, NC). A p value <0.05 was considered statistically signiﬁcant.