*2.3. Statistical Analyses*

The group mean values, standard deviation and 95% confidence intervals (CI) when appropriate were computed. Standard testing of normality of distribution was used for the assessment of normal distribution. Possible intergroup differences were identified with independent samples Students' test and a one-way ANCOVA model. The primary study endpoints were carotid IMT and brachial FMD, whereas other outcome variables were compared after Bonferroni adjustment for multiple comparisons. On the assumption of mean carotid IMT being 0.61±0.14 mm and brachial FMD being 8% ± 1% in the subjects, we estimated that enrolment of 350 subjects in Southern China and 600 subjects in Northern China would result in adequate power (80%) to detect a 18% difference in carotid IMT and an 8% difference in brachial FMD, between the two location groups at 2*p* < 0.05 significance level [29]. Linear multivariate regressions were performed to assess the major determinants of IMT and FMD, including age, gender, smoking status, BMI, metabolic syndrome, LDL-C, PM2.5, southern and northern locations. The variables with significant standardized beta coefficients (beta value with 2*p* < 0.05) as an indicator of the contribution to the model, were identified, and insignificant variables (2*p* > 0.05) were removed subsequently. Group differences with an error probability of 5% (2*p* < 0.05) were considered statistically significant. Analyses were performed with SPSS version 25.
