Body-mass index and obesity in urban and rural China: findings from consecutive nationally representative surveys during 2004–18

1Obesity is a major risk factor for cardiometabolic diseases, renal disease and many cancers,– affecting 670 million adults worldwide in 2016. Globally, the prevalence of obesity in adults has nearly tripled since 1975; in some middle-income countries, mean BMI has increased by 5 kg/m2 and prevalence of obesity by >10 folds. In China, mean BMI and obesity started to increase steadily since the 1980s following a rapid economic development,, but they remained, at least by 2010, well below the levels seen in most other middle- and high-income countries. Despite this, a few national programmes targeting obesity and NCD prevention have been rolled out in China since 2010, including the National Demonstration Areas for Comprehensive Prevention and Control of NCDs and the China Healthy Lifestyle for All Initiative., In 2013 the World Health Assembly also set an ambitious global target for control of non-communicable diseases, including to halt, by 2025, the rise in the prevalence of obesity compared to 2010. Careful monitoring of long-term trends in BMI and obesity in general populations is essential to evaluate the likely success of various national and international initiatives and inform the development of country-specific health policies.

2A few studies have previously reported on the rising trend in BMI and obesity prevalence in China, but they mainly used national survey data collected before 2013,– including some global pooling studies,, and tended to define obesity differently from those used in most international studies. The lack of reliable evidence on China’s recent trends in mean BMI and obesity both nationally and regionally, or by urban-rural locality which would reflect different stages of socio-economic development, led to substantial uncertainty over whether the rise in BMI will continue persistently. Moreover, important questions also persisted on whether the rising trends in BMI may vary significantly between men and women, urban and rural areas, or by socioeconomic status such as education, which could inform more targeted intervention measures.

3To address the evidence gap, we presented detailed analyses of relevant data from six consecutive nationally representative health surveys conducted between 2004 and 2018. The main aims of the study were to: 1) examine the long-term and recent trends in mean BMI and prevalence of obesity among Chinese adults from 2004 to 2018, with specific emphasis on changes before and after 2010 (when various national NCD prevention programs were initiated); 2) assess how the trends may have varied by gender, age, urban-rural locality, and certain socioeconomic status; 3) estimate the number of people who were obese in 2018 versus 2004.

4We used data from China Chronic Disease and Risk Factors Surveillance (CCDRFS) programme, which was established in 2004 with the aim to provide periodic nationwide data on the prevalence of major chronic diseases and the associated behavioural and metabolic risk factors in the general population. The details of the design, objectives and survey methods of CCDRFS have been described elsewhere,– and summarised in the Appendix. In brief, the CCDRFS was incorporated into China’s Disease Surveillance Point (DSP) system, which now covers 324 million people (~24% of total population in China) across all 31 provinces, autonomous regions, and municipalities., Each DSP area covers a rural county or urban district which was selected using multi-stage stratified cluster sampling scheme.

5For each CCDRFS survey, the participants were selected using multi-stage stratified cluster sampling within DSPs (Figure 1,and Appendix for further details). Across all six surveys, a total of 776,571 individuals were invited, and 746,020 (96.1%) participated, including 33,051 (response rate 99.6%) in 2004, 51,050 (99.1%) in 2007, 98,174 (90.5%) in 2010, 189,115 (97.6%) in 2013, 189,754 (97.4%) in 2015 and 184,876 (94.9%) in 2018. The study protocols for each survey were approved by the relevant national and regional ethical review committees. All participants provided written informed consent.

6In each survey, trained health workers collected detailed questionnaire information on demographic and socioeconomic information, and undertook a range of physical measurements including standing height and body weight. In 2004 and 2007, height was measured using wall-mounted stature meters; since 2010, height was measured using mechanical anthropometry stadiometers (TGZ, Bengbu Equipment Inc., China). In 2004, weight was measured using metal body analog weighing scales; since 2007, weight was measured using the same model of electronic body scales, which were calibrated on a regular basis according to standard protocol (see Appendix).

7Overall 74,741 participants aged <18 or >69 years were excluded because these age ranges were not considered in the 2004 survey. We also excluded 26,056 (3.9%) participants with incomplete demographic or height/weight data across different surveys. After these exclusions, a total of 645,223 participants aged 18-69 years remained for the present analyses (Figure 1).

8BMI was calculated as the weight in kilogram divided by the square of standing height in meter. We used standard WHO criteria to define overweight (BMI ≥25 kg/m2) and obesity (BMI ≥30 kg/m2). The mean BMI and prevalence of obesity in each survey were calculated after incorporating stratification, clustering and sample weights. To account for natural changes in population structures, multi-stage sampling design, non-response, and post-stratification, standard sample weights were constructed based on 2010 China census population (also see Appendix).

9For mean BMI and prevalence of obesity, calculations were also performed seperatedly by sex (men/women), locality of residence (urban/rural), age (five groups), education (four groups), occupation (five groups), and geographical region (seven groups). When results were not stratified by age, standardised mean and prevalence were calculated by averaging the sex- and age- specific mean and prevalence in each survey weighted by the 2010 China census population. We used the proc surveymeans procedure in SAS to estimate standard erros and 95% confidence intervals, using Taylor series linearisations with finite population correction. All analyses accounted for complex sample design including clustering, stratification, and sample weights.

10We further compared trends before and after 2010 to assess potential impacts on trends in BMI and obesity of new national programmes related to obesity and NCD prevention introduced around 2010. The annual changes in mean BMI were calculated as the absolute difference in mean BMI between the start and end years divided by total number of years covered. The annual relative changes in overweight and obesity prevalence were calculated as the difference in prevalence between the the start and end years divided by the prevalence in the start year annualised by accounting for compounding (see Appendix). We performed the Student’s t-test for trends for each characteristics in Table 1 and Welch’s t-test for comparisons of mean between groups and over time, and to compare trends in mean BMI before and after 2010; we used a simulation-based method for prevalence of obesity and overweight (see Appendix). We did not adjust for possible multiple comparisons involved in our analysis. All analyses were done in SAS 9.4 (SAS Institute Inc., Cary, North Carolina, USA) and R version 3.6.0 (R Foundation for Statistical Computing, Vienna, Austria).

11The funders of the study had no role in the design or conduct of the study, including data collection, management, analysis, or interpretation of the results; preparation, review, or approval of the manuscript; or the decision to submit the manuscript for publication.

12From 2004 to 2018, the number of participants included in each survey increased from 32,793 to 155,413, with increasing proportions of older people (mean age increased from 44 years to 52 years; p-value for trend=0.0070) and urban residents (38% to 46%; p=0.0012) (Table 1). Moreover, a higher proportion of participants performed non-manual works (20% to 38%; p=0.0269), while the converse was true for agriculture-related work (57% to 44%; p=0.0161). However, there was little change in the distribution of sample across regions and education levels over the study period.

13Overall the standardised mean BMI levels rose from 22.7 (95% CI: 22.5-22.9) kg/m2 in 2004 to 24.4 (24.3-24.6) kg/m2 in 2018 (Appendix Table 1), driven expectedly by increase in body weight rather than changes in height (Appendix Figure 1). From 2004 to 2010 the mean BMI rose by 0.17 (0.12-0.22) kg/m2 annually, which was twice the annual increase (0.09 [0.06-0.11] kg/m2) observed between 2010 and 2018 (Table 2). Consistent with increase in mean BMI and prevalence of obesity (i.e. a shift in the BMI distribution to the right), we found that BMI distributions became wider from 2004 to 2018 (Appendix Figure 2), as did their corresponding SDs, but the increase in SD was smaller after 2010 especially among men (Appendix Figure 3). In 2004, men had lower mean BMI and prevalence of obesity than women, but by 2018 this had been totally reversed because on average BMI rose more rapidly in men than in women between 2004 and 2018 (annual change of 0.16 [0.14-0.18] vs 0.09 [0.07-0.11] kg/m2; p<0.0001) (Table 2 and Appendix Table 1).

14In 2018, the standardised prevalence of obesity was 8.1% (7.6-8.7), more than twice as high as in 2004 (3.1% [2.5-3.7]) (Appendix Table 1). In absolute number, it was estimated that 85 (70-100) million adults (48 [39-57] million men and 37 [31-43] million women) aged 18-69 years in China were obese in 2018, which was three times as many as in 2004 (total 28 million; 12 million men and 16 million women). Again, the rise in obesity over the study period was somewhat smaller after 2010 (6.0% [4.4-7.6] annual relative increase) compared to before 2010 (8.7% [4.9-12.8]) (Table 2). The slowdown in the rise of obesity prevalence was evident across all regions except in Southern China, although it was distributed unevenly across age groups and education levels (Appendix Table 2). Likewise, for the prevalence of overweight the annual increase was about half as fast after 2010 as before 2010 (2.7% [2.0-3.4] vs 6.4% [4.6-8.4]; p=0.0007) (Table 2).

15The trends in mean BMI and obesity prevalence were largely similar across different age groups for men and women in both urban and rural areas (Figure 2 and Appendix Figure 4), except that in urban areas the pace of rise in mean BMI and obesity slowed down more rapidly in older than younger women (Appendix Table 2). In both sexes the lowest mean BMI was in those aged 18-29 years (Figure 2), while the lowest obesity prevalence was seen at the youngest age group in women but at the oldest age group in men (Appendix Figure 4). Overall among women, mean BMI and obesity prevalence generally increased with age until about 50 years of age, with little further change afterwards; among men there was a more complex pattern, with middle-aged men (30-49 years) having the highest mean BMI and prevalence of obesity at each survey (Figure 2 and Appendix Figure 4).

16From 2004 to 2010, the mean BMI and prevalence of obesity rose similarly in urban and rural areas, by about 0.17 kg/m2 annually for mean BMI and 8-10% annually for obesity (Figure 3 and Table 2). Over this period, urban men and women had higher mean BMI and prevalence of obesity than their rural counterparts (Appendix Table 1). After 2010, however, the trends diverged between urban and rural women. Among urban women, the rise in mean BMI nearly stopped (increasing by 0.03 [-0.01, 0.07] kg/m2 annually), whereas the rise among rural women continued at 0.09 (0.05-0.12) kg/m2 annually (Table 2). Among urban and rural men, however, the rise in mean BMI continued, albeit at a slower rate compared to before 2010 (~0.1 kg/m2 annually). For obesity (as well as overweight), the annual rate of increase was halved after 2010 compared with that before 2010 in both men (6.7% [4.2-9.2] vs 13.5% [7.2-20.9]; p=0.0400) and women (2.7% [0.4-5.1] vs 6.3% [0.8-12.8]; p=0.1546). However, in rural areas, the situation appeared different after 2010, with the rise in obesity only moderately slowed down in men (11.3% to 7.5%; p=0.1600) but continued at a similar pace in women (5.6% to 5.5%; p=0.4983) (Table 2). Consequently, by 2018, rural women had higher mean BMI and obesity prevalence than their urban counterparts (p=0.0045) (Figure 3 and Appendix Table 1); among men the similar urban-rural switchover has not yet taken place, partly because a large increase in obesity among urban men between 2015 and 2018. We did not observe divergent trends in the SD of BMI between urban and rural women, suggesting that the larger rise in obesity in rural women after 2010 was a population wide phenomenon rather than a change among high-BMI individuals.

17From 2004 to 2018, the rise in mean BMI was larger among men and women with lower education than those with higher education. However, there was no clear educational pattern for the rise in prevalence of obesity. The slowdown or plateauing in mean BMI and urban obesity since 2010 happened across all education groups, but the rise in rural obesity accelerated among women with secondary or higher levels of education (Appendix Table 2).

18In 2018, urban and rural men with the highest education had a mean BMI ~1.1 kg/m2 higher, and an obesity prevalence more than 100% higher, than those with the lowest education (Figure 4 and Appendix Figure 5). By contrast, urban and rural women with the highest education had lower mean BMI (by 1.6-1.8 kg/m2) and obesity prevalence (by 20-30%) than those with the lowest education.

19The trends in mean BMI and obesity prevalence were similar across different occupation groups (Appendix Figures 6 and 7). In both urban and rural areas, men with non-manual work had seen a large rise in obesity prevalence between 2015 and 2018, but this was not evident among their female counterparts. Men working in agriculture related profession had the lowest obesity but women working in agriculture had similar levels of obesity to those working in other professions.

20Trends in mean BMI and obesity prevalence were largely consistent across 7 geographical regions of China (Appendix Figures 8 and 9). A persistent north-south difference in both mean BMI and obesity prevalence was observed in men and women and in urban and rural areas. Overall, people living in Northern China had a mean BMI more than 2 kg/m2 higher than their counterparts in Southern China. Likewise, there was >3-fold difference in obesity prevalence between Northen and Southern China.

21In sensitivity analyses involving just 60 DSPs that were covered in all six surveys with no changes in geographical boundaries, we observed similar results as in the main analysis, including slowdown in urban BMI and obesity, with the exception of urban female obesity which continued to rise at a similar pace since 2010 compared to before 2010 in these 60 DSPs (Appendix Figure 10). Likewise, when using Chinese cut-off points to define obesity (≥28 kg/m2) and overweight (≥24 kg/m2), the results were generally similar to that based on the WHO definition in the main analyses (Appendix Tables 3 and 4).

22Using large-scale data from six nationally representative population surveys covering a period from 2004 to 2018, the present study showed that, while the prevalence of obesity has more than doubled and total number of adults who were obese more than tripled since 2004, there were divergent trends in urban and rural areas, especially among women. In urban China, the rise in adult BMI and obesity may be plateauing in women and slowing down in men, but in rural China they continue to rise steadily, especially amonng women. The divergent trends led to higher mean BMI and obesity prevalence in rural than in urban women in 2018, with narrowing urban-rural difference among men as well. Across all six surveys, the mean BMI was persistently lower in those with higher than lower education among women, but the converse was true among men, and geographically there were persistent north-south gradients in mean BMI and obesity prevalence. It was estimated that in 2018 a total of 85 million (48 million men, 37 million women) adults aged 18-69 years were obese in China, representing >3-fold increase in about 15 years.

23Previous studies predicted a massive epidemic of obesity in China, based mainly on experiences of other low- and middle-income countries (LMICs)., For example, in Mexico, Brazil and Malaysia, mean BMI and obesity increased progressively to eventually surpasse the levels seen in most high-income countries.– Different from these countries, China’s recent trends in BMI, and the plateauing of urban female BMI, as we found, are more in line with the experiences of high-income countries in East Asia (e.g., Japan and South Korea) and Western Europe (e.g., France).,,– As in the present study, many LMICs and most high-income countries also found BMI in rural populations to be rising faster than in urban areas; however, the urban-rural switchover of female BMI, as we reported in China, was only observed in high-income countries but not yet evident in most other LMICs. In many LMICs, including those in South Asia and Southeast Asia, BMI rose faster among women than men, in contrast to our findings in China.

24A few nationwide studies have previously reported on the levels and trends in mean BMI and obesity prevalence in China. Both our early (i.e. pre-2010) and recent results (post-2010) were generally consistent with relevant findings in other studies at particular time point,– and a rising trend in BMI in general adult populations up to 2013.– Existing evidence on national trends beyond 2013 in China was mainly based on statistical models in global studies,, or from studies that used an alternative cut-off to define obesity which limited their comparability. A nationwide study, involving non-random sampling methods reported steady rise in mean BMI and prevalence of obesity (defined as BMI ≥27.5 kg/m2) among adults aged 18-80 years during 1993-2015. In that study, the annual increase in mean BMI (0.17 kg/m2) and obesity prevalence (12%) were similar to our results during 2004-2010. In another study of >15 million married couples aged up to 49 years in rural China who were planning pregnancy,, the reported prevalence of obesity (defined as BMI ≥28 kg/m2) at around 2014 was lower than our findings using the same definition for both men (7% vs 14%) and women (5% vs 14%). The only study that reported trends post 2015 used data from two national surveys involving 426,000 adults aged ≥19 years from 31 provinces in China during 2013-2018. In that study, the reported annual increases in prevalence of overweight (6%) and obesity (12%), using WHO definition, were greater than our study (2% and 5%, respecitively) during the same period. On the other hand, the obesity prevalence reported in that study for 2018 (3.7%) was noticeably lower than our results (8.1%), which was, as acknowledged in the paper, lower than other nationwide surveys, and could be due to inappropriate sampling methods used (e.g., 26% of participants were unemployed). By analysing data from six consecutive surveys during 2004-2018, the present study provided robust new evidence about evolving epidemic of obesity in China, not only overall but also regionally and by urban-rural locality.

25Although a few studies in China have previously assessed the urban and rural trends seperatly, they did not identify the emerging divergent trends as we did, mainly because of the time period covered., The drivers behind the increase and the divergent trends in urban and rural China are complex, and may reflect differences in stages of economic development, changing levels and patterns of physical activity, and increase in total calorie intake and consumption of animal-based food.– In China, despite an overall increase in leisure-time physical activity since 2000, the total physical activity has declined sharply from 1991 to 2009, especially in rural areas. Our study also showed that the proportion of people working in agricultural and manual occupations decreased from 60% in 2004 to 38% in 2018. Even among those in agricultural jobs, increased use of machines may also have greatly reduced the level of physical activity involved. Throughout the study period, although income has been increasing steadily in both urban and rural China, rural residents may have used disproportionally more of their increased income on foods than did urban residents, having initially had lower incomes. With higher income and a massive rise in packaged and processed foods in both rural and urban areas, the composition of Chinese diet is changing, generally towards higher intake of fat and animal product.– There is evidence that better educated people and those in more urbanised areas had higher score in ‘healthy eating’, while economically more developed urban area had seen greater improvement in diet quality in recent decades. Data from Food and Agriculture Organization also showed large increase in daily calorie supply and the proportion from animal source foods and sugar in China during this period.,

26The contrasting education gradients in BMI and obesity between men and women observed in our study are consistent with two smaller studies in China using non-nationwide samples in multiple provinces., Previous studies in high- or upper-middle-income countries also reported similar findings among women, but mixed results amon men, with little high-quality data from other LMICs. Our study now provided reliable evidence about the contrasting associtatoins of education with BMI and obesity between men and women, but further studies are needed to elucidate the likely contributing factors.

27Apart from large sample size and nationally representative samples covering periods from 2004 to 2018, the main strengths of our study included high response rates, use of largely consistent protocols over time, and ability to assess the trend by sex, urban-rural locality, geographical regions and certain socioeconomic status. However, our study also has limitations. First, the surveys were not conducted annually, which could have allowed more detailed and quantitative analysis of the plateauing or slowdown in trends, but this will be monitored and assessed in future surveys. Second, the number of DSPs covered across different surveys, hence sampling scheme and study size, changed over time, potentially introducing bias in the observed trends. Nevertheless, in the sensitivty analyses of 60 areas that were included in all six surveys, the results were similar to that in our main analyses. Third, although the instruments used for measuring height and weight have not changed since 2007, a different model was used in 2004, which may affect the comparability of its data with later rounds. Despite this, weight can be measured reliably by different devices and there were regular calibrations in field surveys to ensure data quality. Fourth, our sample had a smaller proportion of men (42-47%) compared with the 2010 census population (50%). This is because, as commonly done in household surveys, our surveys only sampled individuals who had been living in the same address for at least 6 months in the past 12 months. Therefore, people who worked away from their home address were not eligible to participate, which would disproportionally affect men compared with women, and affect the representativness of the study populations. To minimise the likely impact, we used the sex-specific 2010 census population to estimate the overall results.

28In summary, the present study provided robust new evidence about the evolving epidemic of obesity among adult populations in China. While the rise in mean BMI among adult population appeared to have slowed down in last decade, there were divergent trends between urban and rural areas, contrasting associations of education with adiposity in men and women, and large variations in obesity prevalence across different regions. As China continues to modernise and becomes more urbanised, obesity and the associated health burdens (e.g., hypertension and diabetes) are likely to become more and more a rural issue in the coming decades. Our study findings highlighted the need to continue to monitor the longer term trends nationally and regionally, and for better understanding of the factors underlying these trends in order to develop more targeted and effective prevention strategies in both urban and rural China.