Study design and population
The present study is based on the data requested from the China Kadoorie Biobank (CKB) study, which is a prospective cohort study of chronic disease in China. Details of the study design and characteristics of the study participants have been described previously [
22,
23]. Briefly, 512,891 participants without major disabilities living in administrative units (rural villages or urban residential committees), aged 30–79 years (mean age: 51.5 years), were recruited in the baseline survey from five urban (Harbin, Qingdao, Suzhou, Liuzhou, Haikou) and five rural areas (from Henan, Hunan, Sichuan, Gansu and Zhejiang) in China between 2004 and 2008. The survey sites were selected based on geographic location, population stability, quality of death and disease registries, and local commitment and capacity. Within each site, permanent residents in each of the 100–150 administrative units (rural villages or urban residential committees) that were selected for the study were identified from local records and sent a letter or leaflet inviting them to participate. The participation rate was 33% in rural areas and 27% in urban areas.
To minimize of effect of existing disease conditions and using the “healthy” population for analyses, the current study excluded 172,373 participants (73,938 males and 98,435 females) who have self-reported the following major diseases, including diabetes, coronary artery heart disease, stroke or transient ischemic attack, hypertension, rheumatic heart diseases, tuberculosis, emphysema/bronchitis, asthma, cirrhosis/chronic hepatitis, peptic ulcer, gallstone/gallbladder diseases, kidney disease, fracture, rheumatoid arthritis, psychiatric disorders, neurasthenia, head injury and cancer. Besides, the analyses also excluded people (5903 males and 7664 females) who were at the extremes of the eight anthropometric and body composition measures (i.e. >99.5 percentile or <0.5 percentile of the distribution for all eight measures) to limit effects of any possible measurement error. A total of 326,951 adults (130,418 males and 196,533 females) formed the sample for the present analyses.
Ethical approval was obtained from the University of Oxford, the China National Center for Disease Control and Prevention (CDC) and the institutional review boards at the local CDCs in the 10 regions before the start of the survey. Written informed consent was obtained from all participants.
Measures and variables
A standardized questionnaire was administered by trained interviewers at the baseline survey using a laptop-based data-entry system, with built-in functions to prevent logical errors and missing values. Questions included socioeconomic and demographic status, health condition and medical history, behavioral pattern including smoking, drinking, physical activity and diet. In terms of physical measurements, trained staffs across the 10 survey sites conducted the standardized measurements with a protocol and instrument. All of the utilized devices were regularly maintained and calibrated for consistency in measurements.
Eight main anthropometric and body composition measures variables, either directly measured or derived, were assessed as exposure variables, including standing height, weight, body mass index (BMI), waist circumference (WC), hip circumference (HC), body fat percentage, waist-to-hip ratio (WHR) and waist-to-height ratio (WHtR). Standing height was measured to the nearest 0.1 cm with an audiometer. Weight was measured to the nearest 0.1 kg using a body composition analyzer (TANITA-TBF-300GS; Tanita Corporation), with subtraction of weight of clothing (0.5 kg in summer, 1.0 kg in spring/autumn and 2.0–2.5 kg in winter). BMI was calculated as the weight in kilograms divided by the square of the height in meters (kg/m2). WC and HC were measured to the nearest 0.1 cm with a soft no stretchable tape. WC was measured mid-way between the lowest rib and the iliac crest or, when this was not practicable, 1 cm above the umbilicus (usually against bare skin in both cases, but subtracting 1 cm if on top of undergarments). HC was measured at the maximum circumference around the buttocks (usually over underpants, but subtracting 1 cm if over a skirt, or 2.5 cm if over trousers). Body fat percentage was estimated by the Tanita body composition analyzer using proprietary algorithms, which reflects the fraction of total weight that was estimated to be fat weight. WHR and WHtR were calculated using the above measures. The issues of quality control (QC) was mentioned in one of the reference paper. At the baseline, QC survey data were available for 15,728 participants (3.1%), with the mean length of time between baseline and QC survey being 17 days [standard deviation (SD) = 36 days]. There was good agreement between the baseline and QC survey for several common variables. The height, weight and BMI showed extremely high correlation with baseline measures (0.99, 0.96 and 0.93, respectively), whereas for other measures of adiposity (waist and hip circumferences, and body fat percentage), they ranged from 0.82 to 0.90.
Each exposure variables were also classified into several categories for the analyses: Height was categorized into 7 groups (threshold 157, 160, 163, 166, 169 and 172 cm for male; 146, 150, 153, 156, 159, 162 cm for female), weight into 8 and 7 groups (threshold 52, 56, 60, 64, 68, 72 and 76 kg for male; 48, 52, 56, 60, 64 and 68 kg for female), BMI into 8 groups (threshold 18, 20, 22, 24, 26, 28, 30 kg/m2), WC into 8 groups (threshold 72, 76, 80, 84, 88, 92 and 96 cm for male; 66, 70, 74, 78, 82, 86 and 90 cm for female), HC into 7 groups (threshold 82, 85, 88, 91, 94 and 97 cm for male; 83, 86, 89, 92, 95 and 98 for female), body fat into 7 groups (14, 17, 20, 23, 26 and 29% for male; 23, 26, 29, 32, 35 and 38% for female), WHR into 7 groups (0.82, 0.85, 0.88, 0.91, 0.94, 0.97 for male; 0.79, 0.82, 0.85, 0.88, 0.91, 0.94 for female), WHtR into 7 groups (0.42, 0.45, 0.48, 0.51, 0.54, 0.57).
In this study, the outcome variable was self-rated health, which was an ordinal variable that rated general health on a 4-point rating scale ranging from “excellent” to “poor.” Respondents were asked, “How is your current general health status?” Responses were “excellent,” “good,” “fair,” or “poor”. For those who reported “good” and above were coded as “1”, “poor” and “fair” were coded as “0”. The dichotomies data was then included in the analyses and “0” was used as the reference group.
Other covariates included study area, sex, age category (in decile), the highest education completed (i.e. no formal school, primary school, middle school, high school, or college/university), household income last year in Chinese yuan (i.e. <2500, 2500–4999, 5000–9999, 10,000–19,999, 20,000–34,999, >35,000), marital status (i.e. married, widowed, separated/divorced, never married), smoking status (i.e. never, occasional, former, or current regular), alcohol consumption (i.e. never, occasional, former, or current regular), and total physical activity in metabolic equivalent hours per day (MET-hours/day) [
24].
Statistical analyses
Selected demographic, socioeconomic, behavioral and anthropometric characteristics were described separately for male and female. The distribution of sociodemographic, unadjusted means with standard deviations (SD) for continuous variables and unadjusted proportions for categorical variables were presented.
The association between each anthropometric and body composition measure and self-rated health was analyzed using logistics regression. Odds ratios (OR) of self-rated good health were reported within each category of the anthropometric and body composition measures, adjusting for the covariates, including study area, age category, education level, household income level, marital status, smoking status, alcohol consumption, and MET-hours/day. To analyze the association between per standard increase and the self-reporting better health, each exposure measures were divided by the SD of that particular variable and entered the regression as a continuous variable. ORs of self-rated health were regressed on levels of each anthropometric measure as a continuous variable, with adjustment for covariates as described above.
To evaluate the effect modification of socioeconomic and demographic factors, stratified analyses were performed using logistics regression, and ORs were analyzed in each stratum of socioeconomic and demographic factors. All statistical analyses were performed using SAS version 9.3 (SAS Institute Inc., Cary, North Carolina, USA), and tests results were reported significant at 0.05 levels.