“Unraveling the Link: How Healthy Living and Obesity Shape Cardiometabolic Health in Rural Chinese Communities”

Population of the study

The Henan Rural Cohort study was initiated between July 2015 and September 2017 across five rural counties spanning various geographical areas of Henan province, China. The context, structure, and execution details of the Henan Rural Cohort study have been elaborated upon in other publications [12]. In summary, a multistage, stratified cluster sampling approach was employed to select participants for the study. All permanent residents in each administrative unit (rural village) from the chosen communities of each rural county were invited. In total, the research involved 39,259 adults aged between 18 and 79 years. The investigation was conducted in accordance with the Declaration of Helsinki. This study received approval from the ethics committee of the life sciences department at Zhengzhou University. Every participant provided signed informed consent.

For the current investigation, we excluded 9846 adults for whom data on body mass index (BMI), healthy habits, blood pressure, blood lipids, or fasting blood glucose were not accessible. Additionally, 3588 participants diagnosed with stroke, coronary heart disease, cancer, or kidney failure were omitted from the analysis. Furthermore, we excluded 24 non-Han nationality participants and 678 individuals identified as underweight (BMI < 18.5 kg/m²). Consequently, 25,123 participants qualified for the current study.

Evaluation of healthy lifestyles

Lifestyle information was gathered using a questionnaire. Current non-smoking was characterized as participants never having smoked at least one cigarette daily for over six months [13]. Current non-drinking was defined as participants never consuming alcohol at least 12 times a year [13]. Physical activity was categorized into low, moderate, and high intensities as per the International Physical Activity Questionnaire (IPAQ) [14]. Moderate or high physical activity levels were deemed adequate exercise. The Pittsburgh Sleep Quality Index was employed to assess overall sleep quality. Healthy sleep was classified as scores on the Pittsburgh Sleep Quality Index ≤ 5 [15]. Participants reported on the frequency and amount of food consumption over the preceding year for dietary evaluation. A healthy diet was defined as participants meeting at least three recommendations of the Chinese Dietary Guidelines 2022 (vegetable intake ≥ 300 g/day, fruit ≥ 200 g/day, dairy products ≥ 300 ml/day, egg ≥ 42.857 g/day, red and white meat ≥ 42.857 g/day, and fish ≥ 42.857 g/day) [16].

Assessment of weight

Height was recorded without shoes and weight was taken while participants wore light clothing. BMI was calculated using the formula weight/height² (kg/m²). A trained investigator utilized a non-elastic measuring tape to assess waist circumference (WC) at a point 1.0 cm above the navel.

Participants were categorized into three groups based on BMI or WC according to reference values for Chinese adults: normal weight (BMI ≥ 18.5 kg/m² and < 24 kg/m²), overweight (BMI ≥ 24 kg/m² and < 28 kg/m²), and obesity (BMI ≥ 28 kg/m²); normal weight (male: WC < 85 cm; female: WC < 80 cm), increased WC (male: WC ≥ 85 cm and < 90 cm; female: WC ≥ 80 cm and < 85 cm), and abdominal obesity (male: WC ≥ 90 cm; female: WC ≥ 85 cm) [17].

Evaluation of outcomes

The equipment and procedure for measuring blood pressure (BP), blood sample collection, and the measurement of biochemical indexes (blood lipids and fasting blood glucose) were thoroughly documented in previous publications [12]. Hypertension was classified as systolic BP/diastolic BP ≥ 140/90 mm Hg or having taken antihypertensive medication within the last two weeks, based on the 2018 Chinese Guidelines for Hypertension Prevention and Treatment [18]. Dyslipidemia was identified as total cholesterol (TC) ≥ 6.2 mmol/L, triglycerides (TG) ≥ 2.3 mmol/L, high-density lipoprotein cholesterol (HDL-C) < 1 mmol/L, low-density lipoprotein cholesterol (LDL-C) ≥ 4.1 mmol/L, or having taken lipid-lowering drugs within the previous two weeks per the 2016 Chinese Guidelines for Dyslipidemia Prevention and Treatment in adults [19]. Hyperglycemia was defined as fasting blood glucose ≥ 7.0 mmol/L, or having taken antihyperglycemic medications in the previous two weeks [20].

Covariates

Data regarding gender, age, marital status (married/cohabiting, widowed, divorced/separated, and single), educational attainment (illiterate, primary school, middle school, high school, and university or higher), and monthly income per capita (< 500, 500–999, 1000–1999, 2000–2999, and ≥ 3000 RMB) were collected through self-reports via a questionnaire.

Analytical statistics

The number of healthy lifestyle factors was computed for each participant, ranging from 0 to 5. Participants were subsequently grouped into three categories: 0–1, 2–3, and 4–5 healthy lifestyle factors. Logistic regression models were utilized to evaluate the independent influence of weight status as defined by BMI and healthy lifestyle scores on hypertension, dyslipidemia, and hyperglycemia, while controlling for gender, age, marital status, educational level, and monthly per capita income. Specifically, weight status by BMI, healthy lifestyle score, and the aforementioned covariates were all included in the Logistic model concurrently. Hypertension, hyperlipidemia, and hyperglycemia were treated as dependent variables in three separate Logistic models. The analyses were duplicated with an alternative weight status as determined by WC.

To evaluate the combined influence of weight status by BMI and healthy lifestyle scores on hypertension, dyslipidemia, and hyperglycemia, participants were arranged into nine categories: individuals with obesity meeting 0–1, 2–3, and 4–5 healthy lifestyle factors; individuals with overweight meeting 0–1, 2–3, and 4–5 healthy lifestyle factors; and individuals with normal weight meeting 0–1, 2–3, and 4–5 healthy lifestyle factors. Multivariable-adjusted Logistic models were also applied to explore this issue further. Additionally, the analyses were repeated utilizing an alternative weight status defined by WC.

Latent class analysis was employed to identify clusters of healthy lifestyle patterns. This statistical technique categorizes individuals into latent groups based on several categorical variables [21, 22]. To determine the most suitable model, we initiated the analysis with two classes and progressively added more. The fit indicators considered included G², Akaike information criterion (AIC), Bayesian information criteria (BIC), consistent AIC (CAIC), and adjusted BIC [21, 22]. The model with the lowest fit indicators, along with simplicity and interpretability, was selected as the best model [21, 22]. A lower fit indicator signified a preferable model in relation to data fitting. Simplicity of the model necessitated fewer latent classes. Interpretability meant each class should be distinguishable from the others, allowing an appropriate label to be assigned. Additionally, we used the described analytical strategy to investigate the combined influence of weight status and clustering of healthy lifestyles on hypertension, dyslipidemia, and hyperglycemia.

We did not consider overadjustment of covariates a concern for this study. First, our sample size was substantial, and only five covariates (gender, age, marital status, educational level, and per capita monthly income) were involved. Second, typically, the odds ratio concerning the link between exposure and outcome showed no significant variances with and without the adjustment for these five covariates in the previously mentioned Logistic models. Third, the outcomes after applying these five covariates in the Logistic models corresponded with the actual reality. Furthermore, we examined multicollinearity through the Variance Inflation Factor (VIF) in all Logistics models and found the VIF for any tested variables was less than 5, signifying absence of multicollinearity.

Statistical analyses were conducted using SAS version 9.4 (SAS Institute, Cary, California, USA) and R version 4.3.2. Two-sided P values < 0.05 were regarded as statistically significant.