Transition Patterns of Weight Status and Their Associated Factors among Elementary School Children: A Longitudinal Cohort Study Using Multistate Markov Model

Abstract

Background:

Childhood obesity is a recognized public health concern worldwide. It is essential to study the natural progression of obesity in the interest of prevention. This study aimed to describe the dynamic changes in weight status among elementary school children and identify the significant factors influencing the progression or regression of weight status.

Methods:

This study involved 928 elementary school children who were followed up annually during their elementary school years. Heights, weights, and vital capacity (VC) were measured each school year. A multistate Markov model containing three weight states was fit to longitudinal weight status data.

Results:

Children with healthy weight and obesity tended to stay in their preceding weight state. Children with overweight, in contrast, were more likely to move to the other two states. The mean sojourn time in obesity and in overweight states was 5.15 years (95% confidence interval [CI]: 4.22–6.3) and 2 years (95% CI: 1.76–2.28), respectively. Children in lower grades, those with a lower VC index, those with a higher initial BMI, those with a higher annual weight increment, and boys were at increased risk of progression to overweight or obesity, with a decreased probability of regression.

Conclusions:

Children with obesity were more resistant to recovery than those with overweight. Prevention and intervention measures should be adopted early when abnormal weight onset occurs. The multistate Markov model was an advanced tool to analyze dynamic changes in status and identify significant factors for progression and regression and helped to develop prevention and intervention targeting strategies.

Introduction

Childhood obesity is a recognized public health concern worldwide because there is enough evidence to indicate that childhood obesity leads to adult obesity and chronic diseases,¹ such as cardiovascular and metabolic diseases, in adulthood.^2,3 Some studies have shown that childhood obesity is also associated with poor cognitive function and psychological disorders.^4–6

The rapid increase in the prevalence of childhood overweight and obesity has been observed globally.¹ In China, the prevalence of childhood overweight/obesity has increased significantly for two decades and varies in different regions due to a diversity in diet, ethnicity, and culture.^7–10 Prevalence of childhood overweight/obesity was higher in northern China than in other areas and higher in economically developed areas than in underdeveloped areas.^11,12 Factors for childhood overweight and obesity, including behaviors, economic status, gender, age, ethnicity, and heredity, have been widely investigated.^13–15

However, most surveys in China have been cross-sectional studies that had difficulties elucidating the real reasons for overweight and obesity in children.^9–12 There are only a few longitudinal studies on the development of childhood overweight/obesity.^16,17 Moreover, many studies only focused on the fluctuation of childhood overweight/obesity prevalence during a period,^9,12,18,19 but whether preceding weight status influenced the current weight status and how factors affected the transitions between different weight states remained unclear. Understanding the natural progression of overweight/obesity is very important for clarifying the roles of associated factors that benefit the development of prevention and intervention strategies.

Weight change involves three states: healthy weight, overweight, and obesity. However, an individual's weight status may change continuously during a period, whereas only a specific state at a particular time can be observed. Therefore, the exact times of the state changes are often unobserved and the interval time between observations may vary by individual. In previous studies, the Cox proportional hazards model and logistic regression have been used to identify factors for overweight/obesity,^13,20,21 but these analyses could not provide insight into the dynamic nature of status changes that may occur over time. The continuous time multistate Markov model in is an advanced model to describe a process in which an individual moves through a series of states.²² First, it can identify transition intensities that reflect an instantaneous flow of individuals between any two states (if transition occurs, how large or small proportion flows into and out of the weight status category). Second, the multistate Markov model takes into account the continuous changes of covariate values over time and assesses the effect of factors on the transition between any two states, whereas in the logistic and Cox regression models, only baseline values of covariates are involved and only one terminal event is permitted. Third, the time interval between two consecutive observations is not required to be equal for individuals in the multistate Markov model, which can be more sufficient to analyze the data than the discrete Markov chain.

Dalian is a modern coastal city and the most developed city in northeast China. Elementary school children in Dalian city have a higher chance to enjoy modern vehicles and engage in sedentary behaviors than those in rural areas. The aim of this study was to construct a multistate Markov model to describe the dynamic nature of weight status over time and identify factors affecting changes in weight states for elementary school children. To the best of our knowledge, this was the first longitudinal study to analyze transitions between weight states by the multistate Markov model, which can display the nature development of overweight/obesity well and clarify the roles of factors for the progression and regression of weight status.

Methods

Subjects and Data Collection

A two-stage sampling technique was adopted to collect basic samples from 148 elementary schools in the four districts of Dalian in 2003. First, two districts were selected randomly. Second, 11 elementary schools were selected from these two districts, with a proportion to the size of school enrollment in 2003.

Subjects' heights, weights, and vital capacity (VC) were measured annually by trained school nurses during the school day in May from 2004 (Grade 1) to 2009 (Grade 6). The interval time of measurement between two consecutive years was ∼1 year. Measurements were always taken in the morning. Heights and weights were measured when subjects were barefoot and wore light clothing. VC was measured twice with a spirometer; the child was asked to take in a full inhale and then exhale to the limit of full expiration, and the higher value was recorded. BMI was calculated as weight (in kilograms) divided by the square of height (in meters). The VC index (VCI) was calculated as the VC (in milliliters) divided by the weight (in kilograms). The annual weight increment (AWI) was the difference in weight between two consecutive years. The annual height increment (AHI) was the difference in height between two consecutive years.

According to the gender- and age-specific cutoff points developed by the Chinese Working Group on Obesity for Children (Group of China Obesity Task Force, 2004),²³ subjects were classified into one of three weight status categories, healthy weight, overweight or obesity, at each assessment.

A total of 1059 pupils were enrolled in this study in 2003. Pupils who had a growth-related disorder and had non-Chinese parents were excluded from the study. The data for 2004 (Grade 1) were regarded as the baseline data. If the individual record in a certain follow-up year was not completed, the record of this year was deleted. Children with only 1 year's record during the study or without weight and height at baseline were excluded, leaving an analysis cohort of 928 samples.

This study was approved by the school district and ethics committee of the School of Public Health, Dalian Medical University, and verbal consent was obtained from the parents of the participants.

Markov Model

The concept of multistate models

The development of a disease often consists of several stages. A subject may be followed up with at arbitrary times for different reasons. Thus, the exact time of disease onset or transition between states is often unknown. A multistate model has the advantage of investigating the evolution of disease by considering continuous status changes.²²

At a time t, the individual is at state S(t). The transition intensity q_rs is used to represent the instantaneous risk of moving from state r to state s at time t: \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} \begin{align*} {q_{rs}} \left( t \right) = \mathop { \lim } \limits_{ \delta t \to 0} P \big ( {S \left( {t + \delta t} \right) = s{ \rm{ \vert }}S \left( t \right) = r} \big ) / \delta t. \end{align*} \end{document}

All possible intensities form a matrix \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} $$Q = \left( {{q_{rs}}} \right)$$ \end{document} , whose sum of each row equals zero, so that the diagonal entries are defined by \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} $${q_{rr}} = - \sum \nolimits_{r \ne s} {{q_{rs}}}$$ \end{document} , and a single period of occupancy (or sojourn time) in state r can be calculated by \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} $$- 1 / {q_{rr}}$$ \end{document} . All entries can be estimated by inserting data into the Markov model. Covariates vector \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} $$z \left( t \right)$$ \end{document} for transition intensity q_rs can be investigated by modeling the intensity with the following function: \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} \begin{align*} {q_{rs}} \left( {z \left( t \right) } \right) = {q_{rs}} \left( 0 \right) exp \left( { \beta _{rs}^Tz \left( t \right) } \right) , \end{align*} \end{document}

where \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} $$\beta _{rs}^T$$ \end{document} is the coefficient vector that represents the effects of covariates on the transition from state r to state s and can be interpreted similarly to those in the ordinary Cox regression model in terms of ordinary hazard ratios (HRs).

Three-state Markov model

In this study, the Markov model includes three states: healthy weight, overweight, and obesity. These three states are mutually exclusive, and there was no absorbing state because there was no death observed during follow-up. Considering the continuous changes of the weight status, the subject may advance into or recover from the adjacent states in the continuous-time multistate model. Transitions are permitted from healthy weight to overweight, overweight to obesity, overweight to healthy weight, and obesity to overweight. We observed that the probabilities of progressing from healthy weight to obesity and from obesity to healthy weight for children over one year were very low (Table 3), so it is reasonable to assume that transitions from healthy weight to obesity and from obesity to healthy weight for a very short time period were impossible. A map of the model is displayed in Figure 1. Gender, grade, VCI, initial BMI value at baseline, AWI, and AHI were considered exploratory variables to determine whether they were significant in the multistate model.

Figure 1.

Possible transitions between states.

Statistical Analysis

Data were recorded and analyzed with SPSS17.0. Means and percentages were used to describe continuous and discrete variables, respectively. The MSM package in R 3.2 was used to construct the continuous multistate Markov models.

Results

A total of 928 individuals were included in this study, and the average length of follow-up was 5.98 years. The main characteristics of the subjects at baseline (Grade 1) are summarized in Table 1.

Table 1.

The Main Characteristics of Pupils in Grade 1

Variable	\documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} $$\bar {\bf X}$$ \end{document} (S) / n (%)
Gender
Boy	512 (55.2)
Girl	416 (44.8)
Age, years, median (range)	7.08 (6.42–8.33)
VC, mL	1259.03 (317.32)
Height, m	1.25 (0.06)
Weight, kg	25.17 (5.42)
BMI, kg/m²	15.9 (2.49)

\documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} $$\bar X$$ \end{document} , mean; S, standard deviation.

VC, vital capacity.

The prevalence of different weight states at baseline and in each follow-up year is displayed in Table 2. The prevalence of healthy weight declined in the lower grades (grades ≤3) and then tended to stabilize, and the prevalence of overweight increased with the length of follow-up. The prevalence of obesity increased in the lower grades (grades ≤3) and decreased in the upper grades (grades ≥4).

Table 2.

The Prevalence of Weight States at Baseline and Follow-Up Years [n (%)]

States	Follow-up years
States	Baseline	1	2	3	4	5
n	928	925	922	925	927	926
Healthy weight	728 (78.4)	701 (75.7)	673 (73.0)	658 (71.2)	664 (71.6)	672 (72.6)
Overweight	103 (11.1)	105 (11.4)	112 (12.1)	127 (13.7)	132 (14.2)	143 (15.4)
Obesity	97 (10.5)	119 (12.9)	137 (14.9)	140 (15.1)	131 (14.2)	111 (12.0)

The frequency of transitions observed during the study is displayed in Table 3. In total, 94.5% of children with healthy weight and 84.0% of children with obesity maintained their preceding weight state during the study. Most children with healthy weight and obesity both tended to stay in their preceding state. However, only 62% of children with overweight remained in their preceding weight state during the study, so children with overweight were more likely to move to other states than children with healthy weight and obesity. The two highest frequencies of transition from one state to another were related to subjects who transited from overweight to healthy weight (20.9%) and overweight to obesity (17.1%). The frequency of transitions from healthy weight to obesity and from obesity to healthy weight were very low, so it was reasonable to assume that instantaneous risks of moving from healthy weight to obesity and from obesity to healthy weight were zero in our model (Fig. 1).

Table 3.

The Observed Transition Frequency During the Study [n (%)]

Original states	Follow-up states
Original states	Healthy weight	Overweight	Obesity
Healthy weight	3235 (94.5)	172 (5.1)	15 (0.4)
Overweight	121 (20.9)	359 (62.0)	99 (17.1)
Obesity	12 (1.9)	88 (14.1)	524 (84.0)

The interval time of the observed transition was ∼1 year.

The transition intensities estimated by the multistate model are presented in Table 4. In the overweight state, the probability of progression to obesity was almost the same as that of recovery to healthy weight (0.228 vs. 0.270), with a mean of 1/0.498 = 2.00 years is spent in the overweight state before recovery or progression. For the obesity state, a mean of 1/0.194 = 5.15 years was spent in the obesity state before recovery.

Table 4.

The Transition Intensities Estimated by the Multistate Model (q_r_s)

Original states	Follow-up states			Mean sojourn time (years) (95% CI)
Original states	Healthy weight	Overweight	Obesity	Mean sojourn time (years) (95% CI)
Healthy weight	−0.063	0.063	0.000	15.87 (13.61–18.22)
Overweight	0.270	−0.498	0.228	2.000 (1.76–2.28)
Obesity	0.000	0.194	−0.194	5.150 (4.22–6.3)

CI, confidence interval.

Table 5 displays the obtained estimates from the multistate model. The 95% confidence interval (CI) shows that gender, grade, initial BMI value, VCI, and AWI were the significant factors for some particular transitions. Compared with boys, girls had a 1.14 ( \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} $$1 / 0.468 - 1$$ \end{document} )-fold greater probability of recovery from overweight to healthy weight. The risks of progression to worse states (healthy weight to overweight and overweight to obesity) were lower in upper grades than lower grades (HR = 0.528 and 0.420, respectively). Children with a higher BMI at baseline had a greater risk of progression to worse states (healthy weight to overweight and overweight to obesity) (HR = 1.555 and 1.390, respectively) and had a lower probability of recovery to better states (overweight to healthy and obesity to overweight) (HR = 0.713 and 0.784, respectively). If the VCI increased 1 U, the probability of progression from healthy weight to overweight decreased by 2.0% (HR = 0.980), and the probabilities of recovery (overweight to healthy weight and obesity to overweight) increased by 4.0% (HR = 1.040) and 4.1% (HR = 1.041), respectively. Each kilogram increase in the AWI was associated with a 21% (HR = 1.210) increase in the risk of progression from healthy weight to overweight.

Table 5.

The Effect of Different Covariates on Transitions

Variable	HRs and 95% CIs
Variable	Healthy weight → overweight	Overweight → healthy weight	Overweight → obesity	Obesity → overweight
Gender^a	1.459 (0.946–2.250)	0.468 (0.302–0.724)^*	0.875 (0.520–1.473)	0.701 (0.438–1.139)
Grade^b	0.528 (0.356–0.782)^*	0.973 (0.642–1.476)	0.420 (0.246–0.717)^*	1.217 (0.748–1.979)
Initial BMI	1.555 (1.402–1.725)^*	0.713 (0.573–0.886)^*	1.390 (1.077–1.793)^*	0.784 (0.672–0.915)^*
VCI	0.980 (0.961–0.999)^*	1.040 (1.013–1.068)^*	1.019 (0.989–1.049)	1.041 (1.007–1.076)^*
AHI	0.966 (0.872–1.070)	0.980 (0.860–1.117)	1.053 (0.895–1.238)	0.928 (0.779–1.105)
AWI	1.210 (1.111–1.317)^*	1.021 (0.911–1.144)	0.997 (0.862–1.153)	1.046 (0.941–1.162)

HR of gender is boy over girl.

HR of grade is upper grades (grades ≥4) over lower grades (grades ≤3).

Significant HR at the 0.05 level.

AHI, annual height increment; AWI, annual weight increment; HR, hazard ratio; VCI, VC index.

Discussion

Some studies have shown fluctuations in the prevalence of overweight/obesity in elementary school children. However, whether preceding weight status affected current weight status and how factors affected transitions between different weight stages remained unclear. This was the first study to use a multistate Markov model to analyze dynamic changes in children's weight status with longitudinal data.

The transition matrix (Tables 3 and 4) revealed that children with overweight were more likely to move to the other two states than children with healthy weight and obesity. Children in the overweight state had an almost equal chance to move to the healthy weight and obesity state (transition intensities were 0.270 and 0.228, respectively). Furthermore, the mean sojourn time in the obesity state was 5.15 years, which was much longer than the 2.00 years observed in the overweight state (Table 4). A longitudinal study of 1494 elementary students in the United States showed that elementary school children tended to stay in their initial weight status group, except for the initially underweight group,¹⁷ which was partly consistent with our results. Our results indicated that preventive and control measures should be taken early in the overweight stage to avoid advancing to the obesity state, which was relatively resistant.

Gender played an important role in childhood obesity in previous studies.^15,17,24 Gender also affected one transition significantly in our results. Boys were less likely to recover from an overweight state to a healthy state. This is consistent with a longitudinal study in the United States, in which a different Markov model, the discrete Markov chain, was used.¹⁷ Previous studies in China also showed that the overall prevalence of obesity and overweight was higher in boys than in girls during elementary school.^9,25 Traditional Chinese values favor heavier boys.²⁶ Chinese parents are more sensitive to girls' weight and place much stricter supervision on girls because girls with good shape are more popular in society, whereas they think that males in family should be strong, so they tend to misperceive overweight boys as being strong and healthy.²⁷ Chinese girls also accept these traditional gender values and thus pay more attention to body shape than boys. In contrast, studies in different areas reported that boys had unhealthier diets than girls. A survey in Shanghai, China, found that boys had more unhealthy behaviors leading to obesity than did girls,²⁸ and it can be speculated that the same is true in the city of Dalian.

The VCI reflects pulmonary function. Some studies showed that the VCI had a negative correlation with BMI,^29,30 whereas others claimed that there was a positive correlation or no correlation between them.^31,32 In this study, children with a higher VCI were more likely to recover from overweight and obesity than those with a lower VCI, and children with a lower VCI tended to progress from healthy weight to overweight. It is well known that regular physical exercise can effectively improve the VCI.^33,34 Children with a higher VCI might have performed more physical exercise than those with a lower VCI, and physical exercise can help to control and lose weight, which may explain the effect of the VCI on transitions.

In this study, initial BMI when entering elementary school significantly affected the transitions between weight states. The risks of progression from healthy weight to overweight and from overweight to obesity increased with a high initial BMI, and the probability of recovery from obesity decreased with a high initial BMI. In previous studies, children who were already overweight or obese when entering kindergarten tended to stay overweight or obese.^35,36 A longitudinal study of U.S. elementary children showed that the probability of becoming overweight or obese was highest for initially obese children followed by initially overweight and healthy weight children,¹⁷ which was consistent with our results. These results suggested that more attention should be paid to children who are already overweight and obese at the beginning of the elementary school. In China, a sizeable percentage of preschoolers are raised by grandparents who hold the traditional belief that heavy young children are a sign of good nutritional status and wealth, and early childhood overweight or obesity tends to be ignored. Therefore, a balanced and healthy approach to feeding children should be advocated to maintain healthy weight for children in early childhood. This is an effective way to prevent and control obesity for elementary school children.

The association between age and prevalence of overweight/obesity varies in different regions. Some previous studies showed that the prevalence of childhood overweight/obesity increased with age,^37,38 whereas others showed that the prevalence was associated with age nonlinearly.^9,39 A cross-sectional survey in Shanghai, China, showed that the highest prevalence of obesity was in the age group of 6 years (Grade 1 group), and the highest prevalence of overweight was in the age group of 9–12 years.⁹ This result was partly similar to ours. In this study, the prevalence of obesity peaked in the fourth grade (9–10 years old) and that of overweight increased with grade. Furthermore, children in upper grades were less likely to become overweight and to change from overweight to obese than those in lower grades. It has been reported that weight perception, not actual weight, was a significant correlate of lifestyle.^30,40 A recent study among elementary children in Guangzhou, China, reported that children who perceived themselves as overweight had a higher intention to change their weight.⁴⁰ Children in upper grades may be more concerned with body shape or appearance than those in lower grades; such perceptions may motivate them to control and reduce weight.

AWI was significantly associated with overweight onset. However, the AHI was not significantly associated with any transition. These results revealed that an increase in height would not decrease the risk of progression to worse states or increase the chance to recover to better states. Although the increase in height for children is often accompanied by an increase in weight, this kind of increment in height cannot change children's BMI value greatly; in other words, it cannot influence children's weight status. However, an increase in weight for healthy weight children can usually increase a child's BMI, especially when their weight increases faster than their height due to overnutrition. Obesity is a negative factor for producing growth hormone in children,^41,42 so fast weight gain is not good for children's height growth. Some Chinese parents think that obese children will become normal as their height increases, but it is impractical for children to change from overweight or obese to normal weight as their height increases. Therefore, controlling weight gain is the key to controlling obesity in children.

This study had limitations. First, this study did not take into account hereditary factors, for example, parents' weight status and chronic disease history. Second, behavioral factors, such as dietary habits, physical activity, and sleep time, were not included. Future research should investigate more variables to explore the important factors for the progression and regression of weight states.

Conclusion

The multistate Markov model is a useful tool to describe the dynamic change in weight status and determine significant factors for transitions between states. Children with obesity were more resistant to recovery than those with overweight. Prevention and intervention strategies should be adopted early when abnormal weight onset occurs. Children in lower grades, those with a lower VCI, those with a higher initial BMI, those with a higher AWI, and boys were at an increased risk of progression to overweight or obesity, with a decreased probability of regression.

Footnotes

Acknowledgments

We thank all the pediatricians, nurses, and other participants involved in the elementary students' health checkup program in Dalian city.

Author Disclosure Statement

No competing financial interests exist.

References

, Fleming

, Robinson

, et al. Global, regional, and national prevalence of overweight and obesity in children and adults during 1980–2013: A systematic analysis for the Global Burden of Disease Study 2013. Lancet, 2014; 384:766–781.

Dietz

. Health consequences of obesity in youth: Childhood predictors of adult disease. Pediatrics, 1998; 101:518–525.

Al-Goblan

, Al-Alfi

, Khan

. Mechanism linking diabetes mellitus and obesity. Diabetes Metab Syndr Obes, 2014; 7:587.

Lakshman

, Elks

, Ong

. Childhood obesity. Circulation, 2012; 126:1770–1779.

Esposito

, Gallai

, Roccella

, et al. Anxiety and depression levels in prepubertal obese children: A case-control study. Neuropsychiatr Dis Treat, 2014; 10:1897.

Reeves

, Postolache

, Snitker

. Childhood obesity and depression: Connection between these growing problems in growing children. Int J Child Health Hum Dev, 2008; 1:103–114.

, Cai

, Wang

, et al. The trend analysis of overweight and obesity in Chinese students during 1985–2010. Chin J Prev Med, 2012; 46:776–780.

Huang

, Chen

, Lin

, et al. Association between acculturation and body weight status among migrant children in Guangzhou, China: A cross sectional study. BMJ Open, 2018; 8:e018768.

Zhang

, Guo

, Tu

, et al. Further increase of obesity prevalence in Chinese children and adolescents cross sectional data of two consecutive samples from the city of Shanghai from 2003 to 2008. Pediatr Diabetes, 2012; 13:572–577.

10.

, Pan

, Liu

, et al. Comparisons of obesity prevalence in children aged 7 to 12 years old in Liaoning and Guangdong Province. Chin J Sch Health, 2017; 38:173–176.

11.

, Sun

. Geographic and population difference of BMI in Chinese school-age youth. Chin J Pediatr, 2004; 42:328–332.

12.

, Yu

, Li

, et al. Prevalence of obesity among primary students between 2009 to 2014 in China: A meta-analysis. Nutr Hosp, 2014; 31:5348–5352.

13.

Williams

, Ge

, Petroski

, et al. Socioeconomic status and other factors associated with childhood obesity. J Am Board Fam Med, 2018; 31:514–521.

14.

Pawar

, Choksey

, Jain

, et al. Prevalence of overweight and obesity in 4 schools of South Mumbai. J Clin Diagn Res, 2016; 10:1–2.

15.

Zhang

, Feng

, Zhai

, et al. Clustering of unhealthy lifestyle behaviours and associations with perceived and actual weight status among primary school children in China: A nationally representative cross-sectional study. Prev Med, 2018; 112:6–14.

16.

Song

, Hu

, Xu

, et al. Dynamic anthropometric changes in children as shown by longitudinal data for 2004–2009 from the city of Dalian, China. Ann Hum Biol, 2012; 39:511–515.

17.

Chen

, Baranowski

, Moreno

, et al. Obesity status transitions across the elementary years: Use of Markov chain modeling. Pediatr Obes, 2016; 11:88–94.

18.

, Tran

, Eriksson

, et al. Prevalence and incidence of overweight and obesity among Vietnamese preschool children: A longitudinal cohort study. BMC Pediatr, 2017; 17:150.

19.

O'Dea

, Chiang

, Peralta

. Socioeconomic patterns of overweight, obesity but not thinness persist from childhood to adolescence in a 6-year longitudinal cohort of Australian schoolchildren from 2007 to 2012. BMC Public Health, 2014; 14:222.

20.

Lai

, Qi

, Tao

, et al. Associations of grandparental diabetes mellitus with grandchild BMI status. BMC Public Health, 2019; 19:172.

21.

Paz-Graniel

, Babio

, Mendez

, et al. Association between eating speed and classical cardiovascular risk factors: A cross-sectional study. Nutrients, 2019; 11. DOI:10.3390/nu11010083

22.

Jackson

. Multi-state models for panel data: The MSM package for R. J Stat Softw, 2011; 38:1–28.

23.

Group of China Obesity Task Force. Body mass index reference norm for screening overweight and obesity in Chinese children and adolescents. Chin J Epidemiol, 2004; 25:97–102.

24.

Leung

, Fu

, Agaronov

, et al. Diet-related determinants of childhood obesity in urban settings: A comparison between Shanghai and New York. Public Health, 2015; 129:318–326.

25.

Liu

, Li

, et al. 20-year trends in prevalence of overweight and obesity among children aged 0–6 in Harbin, China: A multiple cross-sectional study. PLoS One, 2018; 13:e0198032.

26.

Shi

, Lien

, NirmalKumar

, et al. Perceptions of weight and associated factors of adolescents in Jiangsu Province, China. Public Health Nutr, 2007; 10:298–305.

27.

Wen

, Hui

. Chinese parents' perceptions of their children's weights and their relationship to parenting behaviours. Child Care Health Dev, 2011; 37:343–351.

28.

Zhang

, Feng

, Zhai

, et al. Astell-Burt T, Shi X. Clustering of unhealthy lifestyle behaviours and associations with perceived and actual weight status among primary school children in China: A nationally representative cross-sectional study. Prev Med, 2018; 112:6–14.

29.

Park

, Kim

, Jung

, et al. Predicted normal values of pulmonary function tests in normal Korean children. Allergy Asthma Respir Dis, 2014; 2:187–193.

30.

Chen

, Shu

, Han

, et al. Analysis of vital capacity weight index of 1806 elementary school students from Haidian District of Beijing. Chin J Gen Pract, 2016; 14:1707–1709.

31.

Sun

, Chen

. The relationship between obesity and forced vital capacity among university students. Nutr Hosp, 2015; 31:2202–2204.

32.

Leone

, Courbon

, Thomas

, et al. Lung function impairment and metabolic syndrome: The critical role of abdominal obesity. Am J Respir Crit Care Med, 2009; 179:509–516.

33.

, Ki

, Sunghwun

, et al. Correlation between basic physical fitness and pulmonary function in Korean children and adolescents: A cross-sectional survey. J Phys Ther Sci, 2015; 27:2687–2692.

34.

, Wang

, Liu

, et al. Physical activity and lung function growth in a cohort of Chinese School Children: A prospective study. PLoS One, 2013; 8:e66098.

35.

Moreno

, Johnston

, Woehler

. Changes in weight over the school year and summer vacation: Results of a 5-year longitudinal study. J Sch Health, 2013; 83:473–477.

36.

Smith

, Bartee

, Dorozynski

, et al. Prevalence of overweight and influence of out-of-school seasonal periods on body mass index among American Indian schoolchildren. Prev Chronic Dis, 2009; 6:20.

37.

Ogden

, Carroll

, Curtin

, et al. Prevalence of overweight and obesity in the United States, 1999–2004. JAMA, 2006; 295:1549–1555.

38.

García-Alvarez

, Serra-Majem

, Ribas-Barba

, et al. Obesity and overweight trends in Catalonia, Spain (1992–2003): Gender and socio-economic determinants. Public Health Nutr, 2007; 10:1368–1378.

39.

Nakano

, Sei

, Ewis

, et al. Tracking overweight and obesity in Japanese children; a six years longitudinal study. J Med Invest, 2010; 57:114–123.

40.

, Zhang

, Ma

, et al. Self-perception of weight status and its association with weight-related knowledge, attitudes, and behaviors among Chinese children in Guangzhou. J Epidemiol, 2017; 27:338–345.

41.

Garten

, Schuster

, Kiess

. The insulin-like growth factors in adipogenesis and obesity. Endocrinol Metab Clin North Am, 2012; 41:283–295.

42.

Misra

, Bredella

, Tsai

, et al. Lower growth hormone and higher cortisol are associated with greater visceral adiposity, intramyocellular lipids, and insulin resistance in overweight girls. Am J Physiol Endocrinol Metab, 2008; 295:E385–E392.