A 3-year Longitudinal Study of the Effects of Physical Activity and Sedentary Behaviors on Childhood Obesity in China: The Childhood Obesity Study in China Mega-Cities

Objectives: Examine school children’s physical activity (PA) and sedentary behaviors (SB) during 2015-2017 and study their effects on children’s weight status and their relevant gender differences in China. Methods: This open cohort study included students from 5 major cities (Beijing, Shanghai, Xi’an, Nanjing, and Chengdu) across China. Data were collected from students in 2015, 2016, and 2017 (n = 5,535) and from their parents and school personnel. Children’s weight, height, and waist circumference were measured. SB and PA factors were evaluated using questionnaires. Mixed effects models examined the relationship between weight status and PA/SB-associated factors using this longitudinal data. Results: These children had high rates of obesity (12.4%, 95% CI 11.6%-13.3%) and central obesity (28.1%, 95% CI 26.9%-29.3%) during 2015-2017. Boys were more likely to be obese (16.5% vs. 8.4%) and central obese (36.3% vs. 19.8%) and spent more time in screen viewing than girls (hours/week ± SD: 2015, 1.8 ± 2.5 vs. 1.5 ± 2.0; 2016, 2.0 ± 2.4 vs. 1.8 ± 2.5; 2017, 1.7 ± 2.3vs. 1.4 ± 2.1 hours/week). Those who walked (cid:0) 5 minutes on their average daily walk to school were more likely to be obese (OR: 1.96, 95% CI 1.03-3.73) than those who spent ≥ 15 minutes on walking to school. When stratied by gender, this higher risk was only observed in girls (OR: 3.01, 95% CI 1.09-8.35). Children who spent more time in screen viewing were more likely to be obese (OR: 1.13, 95% CI 1.06-1.21) and have central obesity (OR: 1.05, 95% CI 1.02-1.09). The association for obesity was consistent in boys and girls girls, OR: 1.12, 95% CI 1.00-1.24). Conclusions: More screen time and less active time were risk factors for developing obesity in urban Chinese children. The associations varied by gender.


Introduction
With the rapid development of the social economy and changes in the lifestyle of residents, the prevalence of overweight and obesity (ov/ob) has increased rapidly in China in the past three decades, especially in the mega-cities (like Beijing and Shanghai) [1][2][3][4]. National data show that ov/ob prevalence among school children in China has increased from less than 3% in 1985 to 5.3% in 1995 and approximately 20.5% in 2014, with boys being about twice as likely to be ov/ob as girls [5,6]. These increases are cause for concern, since children who are obese have a higher risk of developing many health problems, including type 2 diabetes, high blood pressure, and sleep disorders [7][8][9].
Childhood physical activity (PA) and sedentary behaviors (SB) are important risk factors for obesity [10].
Studies have documented shifts in PA over the past couple of decades among Chinese adults, revealing decreases in PA and increases in SB [11], but similar analyses have not been done among Chinese children and adolescents. Previous studies conducted in western countries have indicated that less time spent in PA and higher frequencies of SB was associated with unfavorable body composition in children and adolescents [12][13][14][15]. However, limited studies have evaluated the association of changes of PA and SB with the prevalence of obesity in children and adolescents [15], especially in countries with a Confucian heritage culture such as China, Japan, Korea, and Singapore [16,17].
Given that parents are the primary caregivers for children, parents' support has been linked to higher levels of PA among children and adolescents [18,19], though Fogelholm and colleagues found the effect of parental involvement in PA may have gender differences [20]. Some difference was also observed in limited PA at school and in increased SB [21][22][23][24]. In general, boys are more active than girls, especially among adolescents, while SB increases in both genders with age and is higher in girls [24,25]. However, there is no in-depth research based on longitudinal data on the gender-speci c trends in PA and SB and their associations with obesity and central obesity in Chinese children and adolescents.
To ll these gaps, using longitudinal data collected from students and their parents in ve mega-cities across China, this study investigated: 1) trends in children's overweight/obesity, central obesity rates, and related PA and SB; and 2) associations between PA and SB changes and their weight status, and tested related gender differences. These ndings provide useful insights for ghting the childhood obesity epidemic.
Methods And Materials

Study Design And Study Sample
The Childhood Obesity Study in China Mega-cities (COCM) was a US NIH-funded longitudinal study aimed at examining the etiology of childhood obesity and chronic diseases in China, especially in major cities, which have been experiencing rapid socioeconomic changes over the past three decades. These transitions have resulted in many dramatic changes in social environments and people's lifestyles, which have led to an increase in obesity and other health problems. Initially, four major cities were included in the 2015 (baseline) survey for the study, including Beijing (China's capital, in North China), Shanghai (the largest and most economically developed city in China, in Southeast China), Nanjing (China's old capital before 1949, the capital of Jiangsu province, in Southeast China), and Xi'an (the capital of Shaanxi province, in Northwest China, which has served as the capital of China for 13 dynasties and over 1,300 years). Chengdu (in Southwest China, the capital of Sichuan province, one of the national central cities) was added in 2016. In each city, two primary schools and two middle schools were included. In each school, a class from each grade (grades 3-6 in primary schools, and grades 7-9 in middle schools) was included. Written informed consent was obtained from parents or children.
Children with missing data on age, gender, weight, height, and PA-and SB-related measurements were excluded from analysis (n = 171). For cross-sectional data analysis, sample characteristics of 5,535 participants were described across the three waves; for PA-and SB-related factors, the children's rst available observation during 2015-2017 (3,268 different children) were presented by gender. For longitudinal analysis of the associations, we included 2,762 students with the needed data from all three waves from 2015 to 2017.

Assessment and measures 1) Anthropometric measures
Height was measured by a Seca 213 Portable Stadiometer Height-Rod with a resolution of 0.1 centimeter; body weight was measured by a Seca 877 electronic at scale with a resolution of 0.1 kilogram. Waist circumference (WC) was measured by a tape with a resolution of 0.1 centimeter.
Children's BMI (body mass index, kg/m 2 ) was calculated as body weight divided by height squared (kg/m 2 ). Overweight and obesity were de ned based on age-gender-speci c BMI curves developed by the Working Group on Obesity in China (WGOC) that correspond to BMI levels of 24 and 28 at age 18 [26], respectively. Abdominal obesity was de ned as WHtR ≥ 0.46 [27].
Considering the in uence of parental support on the associations between PA, SB, and children's weight status, the following measurements were included: mothers' paired PA (yes/no), fathers' paired PA (yes/no).
The SB outcome measures included: (1) screen viewing (hours/week) and (2) homework (minutes/day). Activities in the screen viewing category included watching TV, video games, videotapes, VCDs, DVDs, and movies/videos online, but did not include computer time (e.g., sur ng the Internet) for homework purposes per week. Children reported on homework time in a separate question.

3) Parental and household characteristics
To consider the in uence of family and household characteristics on the associations among PA, SB, and children's weight status, self-reported height and weight of mothers and fathers were used to calculate parental BMI, and overweight and obesity were de ned based on the standards set by the Working Group on Obesity in China (WGOC) (BMI ≥ 24 kg/m 2 ) [22]. Highest maternal education ("middle school or lower", "high and vocational schools", and "college or above"), and family home ownership (rent or share residency with relatives, "own an apartment", and "own a house") were collected using questionnaires.

Statistical Analysis
For participants' characteristics over the survey years, continuous variables were reported as means (standard deviation, SDs) and compared with the Mann-Whitney-Wilcoxon test. Categorical variables were reported as percentages, and the differences were compared with the chi-square test.
Then mixed-effects models were tted to investigate the associations between ov/ob (also central obesity) and PA/SB-related measures while adjusting for other covariates. The models adjusted for age, parental BMIs, maternal highest education level, living environment, the city, and the year. Analysis strati ed by gender was conducted to test gender-based differences in the associations.
All statistical analyses were conducted using Stata 15.0 (Stata Corp, College Station, Texas, USA). The effected size was reported as odds ratios (ORs) and 95% con dence interval (95% CI).

Associations between changes of PA and SB related factors and weight status in longitudinal data analysis
Mixed-effects models were used to assess the longitudinal associations between obesity and PA/SB factors (Table 3 and Supplemental Table 1). Compared with children who spent more than 15 minutes walking time on the way to school, children who walked less than 5 minutes had a higher obesity risk

Discussion
This study explored the in uence of PA and SB measures on children's weight status changes using longitudinal data collected during 2015-2017 from ve mega-cities across China. About 1/3 of children in big cities in China had central obesity, with more boys being obese than girls. A slight decrease was observed from 2015 and 2017 in screen viewing time, but not in homework time. Children who spent more screen time were more likely to be obese and have central obesity, but similar associations were not observed in homework time. The strati ed analysis found that, when spending more time in screen viewing, boys were more likely to have a higher risk of both obesity and central obesity, while girls may have only a higher obesity risk. In addition, children who spent less walking time on the way to school were more likely to be obese, though the association was only observed in girls, but not in boys.
Some systematic reviews have concluded that lower levels of SB were associated with lower levels of body fatness, while the association was less consistent in longitudinal studies in children and adolescents [28]. For example, one study conducted in the UK [14] did not observe any associations between SB and adiposity in a prospective cohort study conducted in children. Another study [29] found increases in SB were associated with higher increases only in BMI above the 50th percentile. Compared with western countries, students in Confucian heritage cultures in some Asian countries like China, South Korea, and Singapore were reported to have longer homework time and less screen time [30][31][32]. Although evidence from cross-sectional studies have suggested positive associations of both homework and screen time with obesity risk [33][34][35][36], previously the longitudinal relationships were not evaluated in Chinese children and adolescents.
Previous studies, predominantly those conducted in industrialized countries like the United States, have been limited by their inability to distinguish between SB time spent screen viewing and SB time spent on homework; our data was con gured to capture this information. Prior evidence has shown that watching television rather than overall SB time is more likely to affect adiposity, which may be due to unhealthy eating messages emanating from TV as well as possible over-consumption during watching TV [37]. Consistently, our study indicated that screen viewing time, but not homework time, was positively associated with obesity and central obesity risks.
A large national survey including 131,859 children and adolescents in China reported that compared with girls, boys spent more time on screen viewing and were less likely to meet the guideline of limiting screen viewing time [38]. The result was consistent with ours. When strati ed by gender, we observed that boys had more screen time and were more likely to have a higher risk of central obesity. Ren and colleagues [39], on the other hand, found homework and screen viewing were positively related to BMI z-score in girls, but not in boys.
Creating active school communities including walking or biking is a useful way to increase school children's PA levels and to prevent obesity [40]. Research suggests that in recent years bicycle usage in Chinese mega-cities has been declining (Beijing, 14%; Shanghai, 10%), although three decades ago the rate was very high [41]. This is re ected in the habits of children as well as adults. In the present study, while 36.8% of children walked to school, only 8.4% of children chose to ride a bicycle go to school.
Recently, a systematic review that evaluated the associations between adiposity and active communities reported mixed results. A limitation of the 38 studies reviewed is that only 5 measured waist circumference, and very few of them conducted longitudinal data analysis [42]. Our study has the advantage of doing both. Another study found that the inverse associations between BMI z-score and active commuting were greatly mediated by moderate to vigorous physical activity levels [43]. In our study, we observed that children who spent less time on walking to school were signi cantly more at risk for obesity, but not for central obesity, which was also consistent with the evidence from some crosssectional studies [43,44]. Still other studies that focused on screen time had also suggested that compared with PA, children's screen viewing time was more closely associated with overweight and central obesity [45,46], an association our study corroborates. These evidences may partly explain the difference in the relationship between obesity and screen viewing time vs. that with walking time.
The school built environment was reported to be closely related to PA, but its associations with ov/ob were inconsistent [47][48][49]. Rashad and colleagues identi ed a signi cant and negative association between access to the sports eld at school and BMI [50], while another study reported that BMI was not associated with PA equipment, or vise versa [51,52]. On the other hand, early-life living conditions are important for obesity development. However, neither parent-paired PA nor family living environment was associated with obesity or central obesity prevalence in the present study. Consistently Zhang et al. [53] have reported that the percentage of parents' encouragement of exercise in non-obese children was higher than for obese children, but the study failed to observe a signi cant relationship between obesity and parents' encouragement of exercise. Although some studies have reported that living in substandard housing and parental attitude were associated with elevated BMI, a recent systematic review summarized that the relevance of home environment to child weight extended beyond a single indicator [54].
The main strengths of our study include the longitudinal data analysis and strati ed analysis to examine time trends in PA and SB and its associations with general and central obesity in Chinese children. The study has two major limitations. First, it only studied children in large cities in China, but not nationwide.
Second, the follow-up period of 3 years is not long.
In conclusion, the prevalence of obesity is high, and much higher in boys than in girls, in big cities across China. Children who had more sedentary behaviors or/and less active time were more likely to become obese, especially girls. Gender differences in behaviors and associated risks should be considered in childhood obesity interventions. Intervention programs aiming to reduce screen viewing time and increase active commuting time are needed for obesity prevention in China.

Declarations
Availability of data and materials The datasets generated and/or analyzed during the current study are not available due to ethics board requirements, but are available from the corresponding author on reasonable request.