Changes in Individual Weight Status Based on Body Mass Index and Waist Circumference in Hong Kong Chinese

Background Weight change predicted diseases and mortality. We investigate 3-year changes in individual body mass index (BMI) and waist circumference in Hong Kong Chinese adults. Methods In the Population Health Survey, 7084 adults in 2003 (baseline) were followed up in 2006. Longitudinal anthropometric data were available in 2941 (41.5%) for BMI and 2956 for waist circumference. Weight status and central obesity were based on objectively measured BMI and waist circumference using Asian standards. Results Mean BMI (SD) increased from 22.8 (3.62) to 23.1 (3.95) (p<0.001) with 1.3 percentage point increase in prevalence of overweight and obesity (from 44.3% to 45.6%). One in 5 (22.0%) normal or underweight baseline respondents became overweight or obese and a similar proportion (24.8%) of overweight and obese respondents became normal or underweight. Prevalence of central obesity increased from 28.3% to 32.4% (p<0.001) with a non-significantly greater increase in women (30.0% to 38.1%) than men (23.0% to 26.1%) (p=0.63). A higher proportion of centrally obese respondents returned to normal (29.4%) than normal respondents developing central obesity (17.4%). Conclusions This is one of the few studies in Chinese, which found dynamic longitudinal changes (increase/stable/decrease) in individual weight status and waist circumference. Future studies with better follow-up and investigating the causes of such changes are warranted.


Introduction
Obesity is increasingly prevalent in developed and developing countries. In mainland China, the prevalence of general (body mass index, BMI, 25) and central obesity increased from 3.4% and 18.6% in 1993 to 26.4% and 37.4% in 2009, respectively [1]. Assessments on time trends of obesity are based on repeated cross-sectional surveys but comparability of data from surveys in different years using different samples can be problematic due to variation in methods and sampling errors. Moreover, apparent stable or changing prevalence rates could be the result of dynamic changes of individual weight status between obese and non-obese, which can only be revealed using longitudinal data.
Although Hong Kong is the most westernised city in China, the government's Behavioural Risk Factor Surveillance showed a stable prevalence of self-reported general obesity (around 20%) in the past decade (2004-2012) [2]. Compared with general obesity, less is known about central obesity in Chinese people. A cross-sectional study in Hong Kong reported that the prevalence of central obesity increased in men (from 23.0% in 2001 to 26.9% in 2005) and remained stable in women [3]. Central obesity among Chinese adults was associated with risk factors of cardiovascular diseases, hypertension and diabetes [4][5][6]. Changes in individual weight status and waist circumference were not examined in the studies above, although in the West, these are increasingly found to better predict diseases and mortality than static baseline measures [7,8].
Among the few longitudinal studies examining individual weight changes, most were based on Western populations and focused on weight gain among normal weight subjects at baseline [9][10][11][12]. Although modest weight loss (10%) was found to reduce risk of cardiovascular diseases, diabetes and mortality [13,14], less is known about the population pattern on individual weight loss. Weight fluctuation or cycling, meaning repeated weight loss and regain, was associated with some negative health outcomes [15], but the effect on mortality is unclear. Studies on trajectory of weight changes have identified four types of trajectory: normal, overweight, late adulthood obesity and early adulthood obesity, and found most people (98% men and 92% women) had upward sloping trajectories [16,17]. Our review of the literature found no report focusing on individual weight changes in Chinese. In the present study, we describe the individual changes in body weight and waist circumference in Chinese adults in Hong Kong.

Ethical statement
Ethical approval was granted by Institutional Review Board (IRB) of the University of Hong Kong/Hospital Authority Hong Kong West Cluster. Written informed consents were obtained from the adult respondents or parent of respondents aged 18 or below, and the procedure was approved by the IRB.

Baseline survey
The Hong Kong Population Health Survey (PHS) in 2003/2004 aimed to collect population representative data to support health policy. Respondents were selected using probability sampling based on the frame of quarters maintained by the Hong Kong Government Census and Statistics Department. Face-to-face interviews were conducted in 3035 households with 7084 land-based non-institutionalised Chinese-speaking respondents aged 15+. We included respondents aged 15-17 as to strengthen the representativeness of our findings to general population in Hong Kong, which usually covers this age period. The response rate was about 72% at household level and 44% at individual level. A structured questionnaire and anthropometric measures were used by trained interviewers to collect data on patterns of health status, health knowledge, attitudes and behaviours. Details of the survey methods and baseline main results have been published elsewhere [18].

Follow-up survey
The 3-year follow-up survey of all baseline PHS respondents were done in 2006. The households were contacted by mail and telephone to obtain informed consent. Survey methods were the same as in 2003/4 to allow for comparison of time trend. Of 3035 households interviewed at baseline, 573 (18.9%) were invalid at follow-up. These included 489 moved households, 64 unoccupied quarters, 13 deceased respondents and 7 non-residential addresses. Of 2462 (81.1% of 3035) valid households, 1251 were fully enumerated and 266 were partially enumerated, representing 61.6% of the 2462 valid addresses. The refusal rate was 17.0% and non-contact rate was 21.4%. A total of 3131 subjects (44.2% of 7084) were successfully re-interviewed.

Study measures
Body weight, height and waist circumference (WC) were measured at respondents' homes by trained interviewers using valid, reliable and calibrated instruments, namely TANITA precision bathroom scale (Model HA-521), 150cm tailoring ruler, and 3.5m foldable metallic measuring tapes. Information about demographics and other characteristics was collected including sex, age, place of birth, education, income and marital status. In the present study, the main outcome variables were overweight, obesity and central obesity as defined by the World Health Organisation (WHO) Asian standard for BMI and WC [19]. Ethical approval was granted by a local institutional review board.

Statistical analysis
Cohen's effect size (w) was used to compare the socio-demographic characteristics, BMI, weight status and WC between respondents and non-respondents with a smaller Cohen's w (<0.3) indicating a small difference [20]. Only subjects with both baseline and follow-up data were included: 2941 for BMI and 2956 for WC, which was 93.9% and 94.4% of 3131. Changes were presented by sample mean BMI and WC, and by proportions of BMI categories (18.4 or below for underweight; 18.5 to 22.9 for normal; 23 to 24.9 for overweight; 25 or above for obese) and central obesity (WC 90cm or above for men, and 80cm or above for women) [19]. The proportions of subjects in each BMI category at baseline and 3-year follow-up were examined to assess the cumulative incidence of obesity. This was contrasted with the proportion of initially overweight or obese subjects who had become normal or underweight at the 3-year follow-up. Net changes in the prevalence of general and central obesity were calculated by subtracting the number of subjects becoming normal from those becoming overweight or obese and divided by the total number of subjects. Repeated analysis was conducted among respondents age 18+ with similar findings and we presented data of all respondents in this paper.

Results
The respondents had similar basic characteristics as non-respondents (Cohen's w <0.3) except for a lower proportion of tertiary education attainment (Cohen's w = 0.31) ( The results on individual changes of weight status showed that two-thirds of baseline normal weight (68.9%) or obese (68.7%) respondents maintained their weight status at 3-year

Discussion
Cross-sectional analysis of baseline and 3-year follow-up data found small but significant increases in mean BMI (0.3 kg 2 /m) and WC (1.5cm), prevalence of overweight or obesity (1.3 percentage points) and prevalence of central obesity (4.1 percentage points). Two repeated cross-sectional studies in adults in mainland China found a similar increase in prevalence of central obesity (4.1-4.6 percentage points) but a smaller increase in mean BMI (0.1 kg 2 /m) in 3-year intervals. [1,21] Longer-term repeated surveys (about 10 years apart) in China showed greater increases in mean BMI (0.2 to 1.2 kg 2 /m) and mean WC (1.4cm), prevalence of overweight or obesity (2.9 to 10.0 percentage points) and central obesity (2.9 to 9.2 percentage points). [22][23][24] However, as previously discussed, using repeated cross-sectional studies to compare changes in BMI and WC over time is problematic. Longitudinal data are preferred, but none were available in Chinese. A 10-year Japanese longitudinal study found the incidence of obesity (BMI 25) ranging from 5.2% to 25.6% in adults in different regions [9]. Although not directly comparable, other non-Asian studies using a higher BMI cut-off point of 30 found higher 10-year incidence rates of obesity: 15% in Australia, 12.0% in northern Sweden and 17.3% in rural areas of the United States [12,25]. More importantly, we have provided the first evidence on the short-term dynamic changes of individual weight status based on BMI and WC in the most westernised city in China. About half the respondents had changed weight status over 3 years. One-fourth had become overweight or obese, while a similar proportion had become normal. In contrast, WC changes were more pronounced for becoming normal (29.4%) than becoming centrally obese (17.4%). These results were consistent with other local findings that weight control behaviours were popular with more than one-fourth (28.8%) of respondents trying to reduce (15.5%), gain (1.3%) or maintain (12.0%) weight in the past 12 months [2]. The changes on weight status and WC may mostly happen among respondents with around-cut-off BMIs and WCs. Further studies are warranted to investigate the factors associating with the changes. Chinese studies have linked socio-demographic (sex, education) and lifestyle (smoking, alcohol, physical activities and diet) factors to weight gain [26,27]. With massive commercial advertising for weight control in Hong Kong, social pressure to be slim is great and over half the adolescents misperceived their weight status as overweight or obese [28]. On the other hand, aggressive promotion of highenergy food and lack of physical activities typical of urban living are conducive to obesity. The apparent stable prevalence of general and central obesity in Hong Kong is therefore likely due to a balance of the above opposing influences. Our findings on short-term dynamic weight changes suggest the need to examine the factors of individual changes in weight status, and assess their impacts on morbidity and mortality in Chinese and other populations. Underweight and obesity were consistently found to be associated with diseases and mortality in other populations but the evidence was inconclusive for weight change [29][30][31][32][33][34]. Short-(3-5 years) [32,33] and long-term (12.5 years) weight changes [30] in studies using measures at 2 time-points without considering weight fluctuations in between were found to predict mortality. In contrast, long-term follow-up studies using multiple time-point assessments found non-significant associations of weight fluctuation and cycling with mortality [31,34,35]. The association with mortality is complex when weight changes occur in specific stage of the life course and when there are reverse impacts of diseases on weight change. As regards morbidity, short-term weight changes were clearly associated with increased risks of cardiovascular diseases and diabetes [36][37][38].
The major limitation of our study was the high attrition rate (55.8%) which may affect the representativeness of findings. However, non-response bias in the present study was unlikely to be large as respondents and non-respondents had similar baseline characteristics including weight status and central obesity. Our prevalence of weight status and central obesity was also similar to government statistics [2]. Loss to follow-up was mainly due to change in address or unsuccessful contact. Refusal was relatively uncommon (17.0%) and unlikely to be directly related to weight status as this was only one of many topics studied. Any non-response due to these common reasons probably would lead to non-differential errors [39]. Changes on weight status and waist circumference may be attributable to age, period or cohort effects. However, as data were only collected at two time point, this limits the analysis using age-period-cohort model. Finally, the analysis of general and central obesity categories were based on arbitrary cut-off values with a very small proportion being grossly obese, constraining further analysis on more extreme changes. Conclusions This is one of the few studies in Chinese, which found dynamic changes (increase/stable/decrease) in weight status based on BMI and waist circumference. Future studies with better follow-up and investigating the causes of such changes of obesity indicators are warranted.