Study data and quality control
This study used secondary prospective cohort data from the KoGES [10], conducted by the National Institute of Health of the Korean Centers for Disease Control and Prevention (KCDC). The KoGES was conducted to establish a genomic and epidemiologic database for examining genetic and environmental effects on the prevalence of major noncommunicable diseases, such as diabetes mellitus, hypertension, obesity, and MS in Koreans.
In total, 10,030 adults aged 40–69 years living in two cities―Ansan (urban) or Ansung (rural)―in Korea were recruited at 28 health examination centers as the baseline cohort in 2001–2002 and were followed up biennially to November 2016. All participants completed interviewer-administered questionnaires on demographic information, lifestyle factors (including dietary habits), their health condition, and medical history, and anthropometric measurements were also acquired every two years and biochemical tests were conducted biennially.
Study participants recruited at 28 health examination centers across the country took part in questionnaire-based interviews conducted by trained interviewers to obtain their sociodemographic and health behavioral information, and anthropometric measurements and clinical examinations were performed by trained personnel based on the study’s protocol.
The detailed profile and methods of the construction of the KoGES cohort have been described elsewhere [10]. The KoGES participants were recruited with the cooperation of local organizations such as public health centers, and after completing a written consent form to participate in the study, they received questionnaires and checkups about their health status and lifestyle, which lasted for approximately 2–3 hours. In order to reduce the inter- and intra- observer bias between the two cohorts in Anseong and Ansan, standardization training between and within cohorts was regularly conducted starting from the baseline survey. Standardized education and practice with Anseong and Ansan cohort researchers were conducted with a focus on training researchers to measure items with the potential for errors in test values, such as questionnaire methods, blood pressure measurement techniques, diabetes test methods, and standardization of device use. No significant differences were found among the measured values. In addition, within the cohort, maintenance training was regularly conducted three to four times each year.
The questionnaire data were generated through two steps, in which a surveyor interviewed a participant using a questionnaire and then entered the content into the system. To prevent input errors, the questionnaire results were entered twice and data were saved only if the first and second input results matched.
Over the course of the study, promotional materials were regularly sent to each participant, and several phone calls were made to confirm health information, disease status, death, and cause of death after baseline/tracking investigations through the person and his/her family. In particular, the death information from the National Statistical Office was used as secondary data, in addition to the data collected through the cohort, to reconfirm the occurrence of death, the date of death, and the cause of death.
All collected data were input and managed in the KoGES program of the Cohort Epidemiologic Information System by the Korean National Institute of Health. This process was thoroughly managed through the following separate processes (1) integration of collected data, (2) initial quality control, (3) data cleaning for single variables, (4) data cleaning for related variables, and (5) statistical quality control.
Study subjects
Of these 10,030 subjects, 8,150 individuals were analyzed after excluding those (n = 1,880) who 1) were diagnosed with and prescribed medications for MS-related diseases (diabetes, hypertension, and hyperlipidemia), cardiovascular diseases (coronary artery disease and cerebrovascular disease), or any cancer at baseline (n = 834); 2) were underweight (n = 184); or 3) missed a follow-up survey (n = 862). Subjects with MS-related diseases, cardiovascular diseases, or cancer were excluded to determine more precisely the association between WF and incident MS or its components and to exclude patients with factors known to be associated with unintentional weight loss [21]. Underweight participants were excluded because they comprised an insufficient sample size. We classified the participants into two groups―non-obese and obese―according to their obesity status in the final analysis (Fig. 1).
Ethical considerations
The KoGES study protocol was reviewed and approved by the Institutional Review Board of the KCDC, and all study participants submitted written informed consent before enrollment.
Participants voluntarily completed a self-administered questionnaire, which included questions on their previous medical history and health-related behavior. Anthropometric measurements and laboratory tests were conducted as part of a general health check-up, and participants were informed of the results.
This research was approved by the institutional review board of Jeonbuk National University (JBNU 201803019). The KCDC provided the raw data after reviewing our study’s plan and IRB approval.
Data sharing
We sent our research plan and IRB approval from our affiliated institution to the KCDC, and pledged to use the data appropriately. After that, the raw data were downloaded from the website.
Measurements
Independent variables
The independent variables included WF, with covariates including age at baseline, sex, education (elementary school, middle school, high school, junior college, or university graduation), marital status (divorced, widowed, single/married, or cohabiting), monthly income (10,000 Korean won), drinking status, and smoking status. The health behavioral information on drinking (never drinker in one’s lifetime, ex-drinker, or current drinker) and smoking status (never smoker in one’s lifetime, ex-smoker, or current smoker) was self-reported [10].
WF was calculated using weight at each follow-up using the following formula: (|weight1- weight2| + |weight2- weight3| + ... + |weightn−1-weightn|)/n-1. We calculated WF including the body-weight values from baseline to just before the event (MS) was captured. WF was defined as the sum of the absolute values of the previous weight minus the next weight divided by the number of measurements minus 1. In other words, from 2001–2002 to 2016, a maximum of 8 measurements were performed every 2 years, and subjects who missed more than 3 examinations were excluded, so the sum of at least 4 absolute values of weight differences were divided by 3 or more. A larger WF means a larger weight change during the cohort data collection period.
The body weight and height were measured to evaluate participants’ level of obesity in terms of the body mass index (BMI). Body weight and height were measured to the nearest 0.1 cm and 0.1 kg, respectively, while participants wore light clothes without shoes. BMI was calculated as kilograms per meter squared, and normal-weight and obese subjects were classified as 18–25 kg/m2 and > 25 kg/m2 respectively, in keeping with the cut-off values recommended for Asian populations [7, 10]. The criteria described are widely used in Korea and Japan to compensate for the problem that the universal standard proposed by the World Health Organization [22] is not suitable for Asians [23, 24]. There are no existing findings regarding body-weight fluctuation and mortality in Asian populations [7], and to the best of our knowledge, there is no standard of obesity generally recognized to be suitable for Asians.
Dependent variables
The dependent variables included MS and its five components. MS was defined by the presence of three or more of the following five components according the Adult Treatment Panel using waist circumference for Asians [17]: 1) central obesity (≥ 90 cm for men or ≥ 80 cm for women), 2) low high-density lipoprotein cholesterol (HDL-C) (< 40 mg/dL for men or < 50 mg/dL for women), 3) hypertriglyceridemia (a triglyceride level ≥ 150 mg/dL), 4) elevated blood pressure (BP) as defined by a systolic BP ≥ 130 mmHg or a diastolic BP ≥ 85 mmHg, and 5) raised fasting glucose (≥ 100 mg/dL).
Waist circumference was measured midway between the inferior margin of the last rib and the crest of the ilium in a horizontal plane with units of 0.1 cm. Blood pressure was measured twice on the subject’s dominant arm while in the sitting position with a 5-minute interval between readings using a mercury sphygmomanometer, and the mean of the two measurements was reported.
A blood sample for triglycerides, fasting glucose, and HDL-C was drawn after 12 hours of fasting and laboratory tests were performed using a Hitachi 747 chemistry analyzer at a central laboratory.
Data analysis
Continuous variables were reported as mean ± standard deviation and categorical variables were expressed as numbers and percentages to describe the sociodemographic and health behavioral characteristics of the study subjects. Comparisons between the normal BMI group and obese BMI group were performed using the Student t-test for continuous variables, and the chi-square test for categorical variables. Kaplan-Meier curves were used to calculate cumulative MS incidence rates according to obesity status, and the statistical significance of differences was compared using the log-rank test. Since people’s weight changes over time, time-dependent Cox regression was performed to reflect changes in weight during follow-up on the risk of MS and its components [25]. The time segments corresponded to weight measurements at intervals of 2 years or more. The models were adjusted for sociodemographic and health behavioral variables. The results are presented as hazard ratios (HRs) and 95% confidence interval (CIs). All reported P-values are two-sided, and P-values < 0.05 were considered to indicate statistical significance. All analyses were performed with SAS software, version 9.4 (SAS Institute, Cary, NC, USA).