Study design and population

The cross-sectional study design was performed from June 2019 to September 2021, which selected participants aged 7–18 years from the eastern, western, central, southern and northern regions of Henan Province. This survey used a standardised and uniform protocol in all the selected schools across the selected cities. A three-stage cluster random sampling method was used to select participants. Briefly, in the first stage of sampling, 20 cities were randomly selected from Henan Province using the method of probability proportional to size. In the second stage of sampling, 10 schools were randomly selected from all schools in each selected city using the same method. In the third stage of sampling, 10 classes were randomly selected in each school using the same method. The schools involved were as follows: Xincheng Street primary school of Pingdingshan City, Ziwei primary school of Anyang City, Huanggang primary school of Xinxiang City, Experimental middle school of Luohe City, Huiyang middle school of Zhoukou City, Experimental middle school of Shangqiu City, Xiangcheng high school of Xuchang City, Yuhua experimental high school of Jiaozuo City, Shihe high school of Xinyang City and Qixian high school of Kaifeng City. All of the students from the selected classes in all schools involved were selected as the participants in the survey. A staff member who has not been involved in the survey performed all processes of randomisation. All schools gave permission for questionnaires to be distributed to students. We contacted the school principals, and these principals conveyed this survey information to teachers subsequently. The investigators worked with the school teachers to send text messages to parents of school students and invite the parents to participate in the investigation in which the purpose of the study and the potential benefits for the participants were explained to them. All participants were ensured that their data would remain confidential. The students had the right to decline to participate in the study without any penalty, and they could drop out if they wanted to at any time during the investigation.

The sample size was estimated according to the sample size calculation formula of the cross-sectional study, where α=0.05, and p was the expected prevalence rate. The expected prevalence rates of overweight and obesity in children and adolescents were 12.2% and 7.3%, respectively, so the calculated minimum sample size was 11 059 students. To increase this study’s validity and generalisability, a total sample size of 27 009 students (18 191 boys and 8818 girls) was recruited in this study. All children and adolescents aged 7–18 years old who did not have a history of chronic diseases including cardiovascular, pulmonary, gastrointestinal and renal diseases, as well as autoimmune diseases, diabetes and cancer, as well as those who had not followed any special diets during the past year such as gluten-free, lactose-free or weight-loss diets, and inability to participate in school sport activities were eligible to participate in this study.

Patient and public involvement

Neither patients nor the public were involved in developing this project.

Data collection and questionnaire survey

Data on the children and adolescents aged 7–18 years were collected using a self-reported and standardised questionnaire that was filled in by the participants and one of their parents. The questionnaire obtained information on children and adolescents’ age, gender, residence region, educational stages, height, weight, body mass index (BMI), birth weight, delivery mode, parental age, parental educational level, parental BMI, daily sleep duration, daily physical activity time and daily screen time. Before the implementation of the investigation, all investigators received unified training. Teachers distributed questionnaires to students, asked students to take the questionnaire home and fill in the information in company with their parents, then the teachers collected the completed questionnaires from their students and returned them to the investigator. We emphasised that the first caregiver of the child should answer the questionnaire for he or she knew the child well. Any question or confusion from students was clarified to ensure that every student understood all of the items, and parents were provided with instructions on answering the questions to prevent inaccurate information.29 We sent text messages to parents in advance to remind the students to bring the completed questionnaires when they come back to school. When we collected the questionnaires from the students, we would carefully check whether all of the questionnaires were handed in. The team ensured that students would bring back the answered questionnaire to school. The completed questionnaires were checked for quality control. Those which had logical errors or had too many missing items were defined as invalid questionnaires. After each participant completed the on-site investigation, 5% of the students were randomly selected for repeated investigation, and the results were reproducible to ensure the authenticity of the survey results.

Birthweight collection and categorisation

Data on birth weight of children and adolescents were collected using a standard question: ‘What is your birth weight?’ in the questionnaire. Most of the parents/guardians of children and adolescents reported the birth weight of their child according to the birth certificate that was made by the hospital after birth. For those who do not have the birth certificate, we require parents/guardians to recall the birth weight based on the measurement by themselves. In this study, birth weight was categorised into three groups: when the birth weight was <2500 g, it was defined as LBW; when the birth weight was 2500–3999 g, it was defined as normal birth weight (NBW); when the birth weight was ≥4000 g, it was defined as HBW.30

Ascertainment of variables

In this study, we developed self-reported items for participants and parents to collect information on overweight/obesity-related factors, including parental education level, parental BMI, sleep duration, physical activity time and screen time. Socioeconomic status was defined as education levels of children and adolescents’ parents, involving three levels: middle school or below, high school, university or above. Parental BMI was calculated as the quotient of weight in kg and height squared (m²). In this analysis, parental BMI was classified into three following categories (according to WHO definitions): when the BMI was 18.5–24.9, it was defined as normal weight; when the BMI was 25.0–29.9, it was defined as overweight; when the BMI was ≥30, it was defined as obesity. Sleep duration was assessed by one question: on average, how many hours and minutes did you sleep on a typical day during the past 7 days? In this study, the sleep duration was divided into four levels according to the National Sleep Foundation guideline for age-specific sleep recommendations: very short, short, recommended and long sleep duration.31 For students aged 7–14 years, sleep duration <7 hours/day was classified as very short, within 7–8 hours/day as short, within 9–11 hours/day as recommended, and more than 11 hours/day as long. For students aged 15–18 years, sleep duration <6 hours/day was classified as very short, within 6–7 hours/day as short, within 8–10 hours/day as recommended, and more than 10 hours/day as long. Physical activity time was assessed by the question: how many hours each day do you usually spend in physical activity? Based on engaging in physical activity, children and adolescents were divided into two groups (>2 hours/day vs ≤2 hours/day). For screen time, consisting of watching television, playing on phone and surfing on computer, children and adolescents were asked how many days and how many hours were spent on the three separate activities over the past 7 days. Following the calculation of (days×hours for each day)/7 for each activity, the average screen time was the sum of television, phone and computer time. Using the threshold of 2 hours/day proposed by current scientific evidence and guidelines, the respondents were classified as exceeding (>2 hours/day) or not exceeding (≤2 hours/day) the recommended daily time spent on watching TV or playing on phone.29

Anthropometric measurement

Height and weight were measured by the well-trained professional staff according to the standardised procedure. Height was assessed by the portable stadiometer to the nearest 0.1 centimeter with the participants standing straight without shoes. Weight was measured by the portable electronic scale to the nearest 0.1 kg with the participants wearing light clothes. Both the height and weight were measured twice, and the average value was calculated to eliminate measurement error. BMI was calculated from the formula: weight (kg) divided by height squared (m²).29 Children and adolescents were classified into two weight status categorises: normal and overweight/obesity according to the BMI-based age and sex-specific criteria provided by the Working Group on Obesity in China.32 We used the 85th and 95th percentiles to define overweight and obesity in adolescents. Overweight was defined as BMI ≥85th percentile but <95th percentile, relative to age and sex, whereas obesity was defined as BMI ≥95th percentile. It was a kind of age-specific and sex-specific BMI reference standard for Chinese children. This standard is one of the most broadly used ones in China that showed its superiority in both prospectivity and actuality and is consistent with the East Asian ethnic characteristics of body fat growth, which could eliminate the influence of different populations with different growth patterns and fat accumulation.

Statistical analysis

The questionnaire data were entered using Epidata V.3.1 software. Descriptive statistics were calculated for all of the variables, including continuous variables (presented as mean values and SEs) and categorical variables (displayed by number and percentages). The differences in continuous variables were evaluated by t-test between normal weight and overweight/obesity. For categorical variables, the χ² test was used to evaluate statistical differences between the two groups. We also performed the sample descriptive characteristics across different birth weight groups. As the remarkably different proportions of overweight/obesity among different groups were observed, stratified analysis based on gender-specific and educational stage-specific factors were conducted to evaluate whether there were effect modifications of overweight/obesity, respectively.

Logistic regression models were performed to assess the ORs and corresponding 95% CIs for overweight/obesity when participants with NBW were used as the reference group regarding gender and educational stage differences. In addition to the unadjusted model (model 1), model 2 was adjusted for children and adolescents’ height, weight, BMI, delivery mode, parental age and parental educational level. Furthermore, we also carried out the interaction analysis between birth weight and lifestyle factors of overweight/obesity. In the later analysis process, LBW and NBW were merged as non-HBW to achieve power for the interaction analysis. The relative excess risk due to interaction (RERI) and the attributable proportion due to interaction (AP) were evaluated to measure the additive interaction, deriving 95% CI using the delta method. The additive interaction was considered significant when the 95% CIs of RERI and AP both excluded zero. To estimate the multiplicative interaction, the product term was included in the logistic regression model, calculating the interaction of OR (IOR) and its 95% CI. Significant multiplicative interaction was considered to exist if the 95% CI of IOR did not contain one.23 All statistical analyses were conducted with IBM software SPSS (V.22). P values less than 0.05 with two sides were considered to be statistically significant.