This study was carried out in a district of Beijing in 2024. The survey object was permanent residents aged 18-79 years old in the district, with a sample size formula of N=(U² p(1-p))/d² deff, taking U = 1.96, p=10.6% (according to the results of the 2017 Beijing Adult Chronic Disease and its Risk Factor Monitoring, the Beijing prevalence of adult diabetes is 10.6%), d=0.02, deff=1.5, considering gender stratification, the response rate was set to 80%, and the sample size was calculated to be 3413.

A multi-stage stratified whole cluster random sampling design was employed. In the first stage, PPS sampling proportional to population size was used to randomly select three communities from each of the nine streets in our district. In the second stage, each community was divided into some residential groups (with at least 127 households in each group), and one residential group was selected from each community by simple randomization. In the third stage, one resident was selected from each household by the KISH table method. A total of 3,635 subjects were enrolled in the actual survey, 95 samples were excluded from the final analysis due to missing values in at least one variable, and listwise deletion was adopted to address the missing data. Consequently, 3,541 samples were ultimately enrolled in the study analysis, with an overall missing rate of 2.61%. Given that this missing rate was far below the critical threshold recommended by methodological guidelines (5%–10%) and that the results of Little’s MCAR test indicated that the missing data mechanism was missing completely at random (MCAR, P = 0.76), the exclusion of these samples would not introduce any bias or impact the validity of the study results.

This study was approved by the Ethics Review Board of the BJCDC (No. 5 of 2017). This project is supported by the Beijing Municipal Finance Program.

The survey includes a questionnaire survey, physical examination, and laboratory testing. The questionnaire survey is conducted by a uniformly trained and qualified investigator who inquires face-to-face about the basic situation of the respondents, including basic personal information, behavioral risk factors (smoking, alcohol consumption, dietary intake, sleep), and the prevalence of major chronic diseases (hypertension, diabetes mellitus, dyslipidemia), etc. The physical examination was conducted by the investigator using a uniform standard to measure the height, weight, waist circumference, and blood pressure of the respondents, with body mass index BMI = weight (kg)/height² (m²). Waist circumference was measured at the horizontal position of the mid-axillary line between the lower edge of the rib arch and the midpoint of the iliac crest line. Blood pressure was measured according to the methods recommended in the Chinese Guidelines for Blood Pressure Measurement, and the average of the three measurements was taken as the final blood pressure value. Laboratory tests required the collection of fasting venous blood from the investigated subjects. Blood samples were collected, centrifuged, and split, and tested for fasting blood glucose, total cholesterol, LDL cholesterol, HDL cholesterol, triglycerides and so on.

Considering the differences in the criteria for determining obesity in different populations and races, this paper uses the diagnostic criteria of the Diabetes Branch of the Chinese Medical Association (7), which can be diagnosed as metabolic syndrome with the following three or more items: (1) Abdominal obesity (i.e., central obesity): waist circumference ≥90 cm for men and ≥85 cm for women; (2) Hyperglycemia: fasting blood glucose ≥6.1 mmol/L or 2-h post glycemic load blood glucose ≥7.8 mmol/L and/or those who have been diagnosed with diabetes mellitus and treated for the disease; (3) Hypertension: blood pressure ≥130/85 mmHg (1 mmHg = 0.133 kPa) and/or those who have been identified and treated for hypertension; (4) fasting triglycerides (TG) ≥1.70 mmol/L; (5) fasting high-density lipoprotein cholesterol (HDL-C) <l.04 mmol/L.

Compared with internationally accepted diagnostic criteria (e.g., the International Diabetes Federation [IDF] 2006 criteria (8) and the National Cholesterol Education Program Adult Treatment Panel III [ATP III] criteria (9)), the 2024 Criteria of the Chinese Diabetes Society (CDS) differ primarily in two dimensions: waist circumference thresholds and glycemic diagnostic criteria. In terms of waist circumference thresholds, tailored to the body fat distribution characteristics of the Chinese population, the CDS 2024 Criteria set the cutoff values for central obesity at ≥90 cm in men and ≥85 cm in women, which are significantly lower than those of the IDF 2006 Criteria (≥94 cm in men and ≥80 cm in women for European populations) and the ATP III Criteria (≥102 cm in men and ≥88 cm in women for American populations).Regarding glycemic diagnostic criteria, the thresholds for fasting blood glucose and 2-hour postprandial blood glucose specified in the CDS 2024 Criteria differ notably from the IDF 2006 Criteria, which define the fasting blood glucose cutoff as ≥5.6 mmol/L.

The rationale for adopting the CDS 2024 Criteria in this study is twofold. First, formulated based on epidemiological survey data of the Chinese population, the criteria set waist circumference, glycemic, and other diagnostic thresholds that are more consistent with the obesity characteristics and metabolic disorder prevalence of the Chinese population, thereby effectively avoiding misdiagnosis or underdiagnosis caused by the direct application of criteria developed for European and American populations. Second, widely used in domestic studies on metabolic syndrome (MetS), these criteria facilitate horizontal comparison of the findings of this study with those of other relevant domestic research. They also provide a reference basis for subsequent cross-regional studies on the pathogenesis of MetS among Chinese and foreign populations.

According to the People’s Republic of China Health Industry Standard for Adult Weight Determination (Standard No. WS/T 428-2013) (10), BMI <18.5 is considered to be underweight, 18.5 ≤ BMI <24.0 is considered to be normal, 24.0 ≤ BMI <28.0 is considered to be overweight, and BMI ≥28.0 is considered to be obese. Harmful drinking (11) refers to drinking behaviors in the past 12 months with an average daily alcohol intake of 61g and above for male drinkers; and 41g and above for female drinkers. Inadequate intake of vegetables and fruits (12) refers to the average daily intake of vegetables and fruits less than 400 g. Sleep deprivation (13) refers to an average daily sleep duration of less than 7 hours.

SPSS 25.0 and R 4.4.2 were used for statistical analysis. The independent variables were statistically described using n (%) if they were count data, and the correlation between categorical independent variables and metabolic syndrome was analyzed using the χ2 test; the measured data were statistically described using x¯ ± s if they conformed to a normal distribution, and if they did not conform to a normal distribution, they were described using M (P25, P75), and correlation analyses were conducted using Pearson correlation. All survey respondents were randomly divided into training set (70%) and validation set (30%) by R language Sample function, and correlation test and multivariate logistic regression analysis were performed on the training set, to screen out the statistically significant influencing factors, and to draw the visualized column line graph prediction model. The discriminative ability of the model and the accuracy of the prediction probability were assessed by ROC curves and calibration curves. All statistical tests were performed with P < 0.05, as the difference was statistically significant.

This study included a total of 3,541 respondents, including 2,479 in the training set and 1,062 in the validation set. The prevalence of metabolic syndrome in this study population was 18.4%. All patients in the training set were divided into six groups according to age, of which 70-79 years old had the least number of patients (6.9%); Han Chinese had the highest number of patients (96.3%), 47.6% had higher education; 73% were married/cohabiting; 52.5% were overweight; 20.9% smoked, 10.3% drank alcohol; and 39.2% had insufficient fruit and vegetable intake; The proportion of people who did not get enough sleep was 24.8%; the median intake of salt was 6; and the median intake of cooking oil was 20g (Table 1 for other information).

The results of the correlation analysis between training-focused metabolic syndrome and its risk factors are shown in Table 1. A total of 13 risk factors were included in this study, and the analysis found that the older the residents were, the higher the prevalence of metabolic syndrome was (P < 0.05); the prevalence of metabolic syndrome was higher in men than in women (23.9% vs. 13.2%, P < 0.05); the prevalence of metabolic syndrome decreased with the higher education level (P < 0.05); the prevalence of metabolic syndrome among people with different marital status was different (P < 0.05), and the prevalence of unmarried people (4.9%) was lower than that of other people; the prevalence of metabolic syndrome among people with different occupations differed (P < 0.05), and those who were engaged in agriculture, forestry, animal husbandry, fishery and water conservancy production accounted for the highest percentage (33.3%); the prevalence of metabolic syndrome increased with the increase of BMI (P < 0.05), and the prevalence of it was as high as 40.% among obese patients; the prevalence of current The prevalence of metabolic syndrome was higher in current smokers than in non-smokers (26.6% vs 16.2%, P < 0.05); the prevalence of hazardous drinking was higher than non-hazardous drinking (30.1% vs 17.0%, P < 0.05); the prevalence of metabolic syndrome was higher in people with insufficient intake of vegetables and fruits than in people with adequate intake of vegetables and fruits (19.9% vs 16.5%, P < 0.05); the prevalence of metabolic syndrome was higher in people with an average daily The prevalence of metabolic syndrome was higher in those with less than 7 hours of sleep per day than in those with more than or equal to 7 hours of sleep per day (22.7% vs. 17.0%, P < 0.05). There were no significant differences in ethnicity, salt intake, and oil intake between patients with and without metabolic syndrome (P>0.05).

Logistic regression models were constructed using Forward Selection (FSS) with the statistically significant variables in the correlation analysis as independent variables and the presence of metabolic syndrome as the dependent variable. The results of the covariance diagnosis showed that the tolerance of each variable was >0.1, the variance inflation factor was <3, and there was no covariance between the respective variables (Table 2). The results of the multivariate logistic regression analysis showed that age, gender, marital status, occupation, BMI, and alcohol consumption were independent predictors of metabolic syndrome (P < 0.05). The older the age, the higher the risk of metabolic syndrome (OR>1 for 18-29 years old as reference), and the prevalence risk of people aged 70-79 years was 5.601 times higher than that of people aged 18-29 years old; the prevalence risk of men was 1.827 times higher than that of women; and the prevalence risk of metabolic syndrome was significantly higher than that of the normal population for those who were overweight (OR = 4.865) and obese (OR = 11.482); The prevalence rate of harmful alcohol consumption was 1.673 times higher than that of the normal population; the prevalence risk of the widowed, married/cohabiting, and divorced/separated populations was 1.311, 2.146, and 3.034 times higher than that of the unmarried population, respectively; and the prevalence rate of the agriculture, forestry, animal husbandry, fisheries, and water conservancy production occupational groups, and the unemployed population was higher than that of the other occupational groups (for details, see Table 3).

Based on the results of logistic regression analysis, six factors (age, gender, marital status, occupation, BMI, and alcohol consumption) that have an impact on the prevalence of metabolic syndrome were selected to construct a column chart to predict the risk of metabolic syndrome prevalence. Figure 1 illustrates that BMI and age have the most significant influence on the development of metabolic syndrome. The age group of 18-29 years old, women, people with normal BMI, unmarried people, non-drinking people, and professional and technical people are the protective factors for metabolic syndrome, and the rest are risk factors.

The ROC curves and calibration curves were used to validate the column line plots. Figure 2 shows the results of ROC curve analysis: area under the curve of subjects’ work characteristics for the training set ROC curve AUC = 0.815 (95% CI 0.795-0.836), and for the validation set ROC curve AUC = 0.787 (95% CI 0.756-0.818), which indicates that the model predicts good performance with high discriminatory power; the sensitivity of the training set is 0.768 and the specificity is 0.730; the sensitivity of the validation set is 0.889 and the specificity is 0.573.

Figure 3 illustrates the results of the calibration curve analysis, demonstrating good calibration performance in both the training and validation sets. For the training set, the calibration slope derived via the Bootstrap method (with 1000 resamples) was 1.002 (95% CI: 0.987–1.017), and the calibration intercept was 0.005 (95% CI:0.008–0.018). These results indicated an excellent fitting performance of the model for the training data, which was further supported by a non-significant Hosmer-Lemeshow test result (p = 0.127). The Brier score and Mean absolute error (MAE) of the training set were 0.120 and 0.240, respectively. In the validation set, the initial model exhibited slight overfitting, as evidenced by a calibration slope of 0.761 and a calibration intercept of −0.242. After probability calibration using Platt Scaling, the calibration metrics of the validation set were notably improved: the calibration slope was adjusted to 0.92 (95% CI: 0.88–0.96), and the calibration intercept was −0.03 (95% CI: −0.07–0.01). With a Hosmer-Lemeshow test p-value of 0.082, the calibrated model demonstrated favorable calibration performance. The Brier score and MAE of the validation set were 0.136 and 0.256, respectively.

A total of 3,541 participants were enrolled in this study via questionnaire surveys, physical examinations, and laboratory tests, with 13 risk factors evaluated. A nomogram risk prediction model was constructed based on logistic regression, and several independent risk factors were identified. The area under the receiver operating characteristic curve (AUC) was 0.815, indicating high predictive accuracy of the model. Although the prediction error (MAE = 0.240) was slightly higher than the ideal threshold, this was consistent with the multifactorial pathogenic characteristics of metabolic syndrome. A Brier score of 0.120 suggested that the model could effectively distinguish between individuals at high and low risk.

First, it is a retrospective study in a localized area, and the results may be influenced by factors specific to the area (e.g., environmental, socioeconomic, and cultural practices). Meanwhile, this study can only reveal the correlation between risk factors and metabolic syndrome, but cannot confirm the causal relationship between them, which requires further verification through prospective cohort studies.

Second, the study sample was solely recruited from one district in Beijing, with a lack of data from rural areas and populations in other provinces, which largely limits the external validity of the established model. In addition, the glycated hemoglobin (HbA1c) index was not included in the study, which may affect the accuracy of the analysis results of blood glucose-related indicators.

In the future, large-sample, multicenter, multivariate (e.g., biomarkers such as inflammatory factors) prospective studies to analyze the influencing factors of metabolic syndrome will yield clearer findings and improve prediction accuracy. Additionally, multicenter and cross-regional studies are recommended, which should incorporate samples from populations of different geographic regions and both urban and rural areas to verify the applicability of this model in a broader population.