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 [4–6]. 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 [9–12]. 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.

Methods

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) (Table 1). The mean BMI significantly increased by 0.3 percentage point from 22.8 (95% CI 22.6–22.9) at baseline to 23.1 (95% CI 23.0–23.3) at follow-up (p <0.001) (Table 2). Only few respondents (3.4% at baseline and 4.3% at follow-up) were grossly obese (BMI ≥30). Overall prevalence of overweight or obesity increased from 44.1% to 45.4% (i.e. 1.3 percentage points). The prevalence of underweight (10.0% to 8.1%, i.e, 1.9 percentage points) and overweight (20.0% to 18.9%, i.e, 1.1 percentage points) decreased but the prevalence of obesity (24.1% to 26.5%, i.e, 2.4 percentage points) and normal weight (45.8% to 46.5%, i.e, 0.7 percentage point) increased. Corresponding significant increase in mean WC (cm) by 1.5 from 79.1 (95% CI 78.8–79.5) to 80.6 (95% CI 80.2–80.9) was observed. Overall, the prevalence of central obesity increased from 28.3% to 32.4% (i.e. 4.1 percentage points) (p <0.001) with women (30.0% to 38.1%) having a non-significantly greater increase than men (23.0% to 26.1%) (p = 0.63).

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Comparison of baseline characteristics of respondents and non-respondents at 3-year follow-up

https://doi.org/10.1371/journal.pone.0119827.t001

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Proportions of BMI and waist circumference categories at baseline and 3-year follow-up, cross-sectional analysis.

https://doi.org/10.1371/journal.pone.0119827.t002

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 follow-up compared with only one-third of overweight (35.3%) and less than half of underweight (44.8%) respondents (Table 3). One in 5 (22.0%) normal or underweight respondents became overweight or obese (i.e. incidence of overweight or obesity), and a similar proportion (24.8%) of overweight or obese respondents became normal or underweight. This yielded 1.3% net increase in the prevalence of overweight or obesity. Relatively more overweight respondents became obese (27.5%) than obese respondents became overweight (16.4%) (p <0.001).

Regardless of baseline categories of WC, about half the women (48.2% to 55.9%) remained in the same category at follow-up. In contrast, the proportion of men maintaining their baseline WC categories increased from 28.6% for WC <70cm to 67.4% for WC ≥90cm (p for linear trend <0.001) (Table 4). Among respondents with normal WC at baseline, 17.4% (21.1% women and 13.8% men, p <0.001) developed central obesity at follow-up. Conversely, among respondents with central obesity at baseline, 29.4% (32.6% women and 27.4% men, p = 0.11) reduced their WC to normal at follow-up. This yielded a 4.1% net increase in the prevalence of central obesity. Overall, 59.3% maintained normal WC, 12.4% developed central obesity, 8.3% resumed normal WC and 20.0% remained centrally obese (data not shown in table).

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Changes in weight status from baseline to 3-year follow-up (row %).

https://doi.org/10.1371/journal.pone.0119827.t003

Download:

• PNG
  larger image
• TIFF
  original image

Table 4. Changes in waist circumference from baseline to 3-year follow-up (row %).

https://doi.org/10.1371/journal.pone.0119827.t004

Discussion

Cross-sectional analysis of baseline and 3-year follow-up data found small but significant increases in mean BMI (0.3 kg²/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 kg²/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 kg²/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).[22–24] 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 high-energy 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 [29–34]. 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 [36–38].

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.

Acknowledgments

We would like to thank the Department of Health (HKSAR) and Policy 21 Ltd. of HKU for data collection at baseline, and Mr Steve Chan for help with fieldwork arrangement.