Statistical analysis

The population was stratified by the persistence of positive findings in proteinuria during follow-up as no proteinuria, occasional proteinuria for once, persistent proteinuria for two or more times. Continuous variables are presented as mean (SD) and compared among groups by using ANOVA test. If the variable does not satisfy the normal distribution, the median (interquartile range) and Kruskal-Wallis rank sum test will be used. Categorical variables were presented as frequency (percent) and compared by using χ ² tests. The incidence rate of the adverse outcomes was calculated and compared among groups by using the log-rank test. Cox proportional hazards regression model was used to estimate the association between persistence or degree of proteinuria and the adverse outcomes. In order to reflect the changing status of proteinuria, a time-dependent variable was used for degree of proteinuria, i.e. proteinuria measurements from each circle of health examination were used. In regard to the endpoint of MI, the Fine and Gray model was used to treat death as a competing risk, which impeded the occurrence of MI³². Proportional hazards assumptions of the Cox regression model were verified by testing the interaction with time using the likelihood ratio test, which yielded non-significant p-values. HRs and 95% CI were reported. We included covariates investigated at baseline into the Cox regression model step by step. Model 1 was the univariable mode l. Model 2 included age (continuous) and sex (male versus female). Model 3 included covariates in model 2 plus education (high school or above versus below high school level), current smoking (yes versus no), current use of blood pressure lowering agents (yes versus no), current use of glucose lowering agents (yes versus no), BMI (continuous), MBP (continuous), FBG (continuous), natural log-transformed TG (continuous), LDL-C (continuous) and CKD stages (stage 3, stage 2 versus stage 1). All p-values were two-sided, and p-value of statistical significance was 0.05. All statistical analyses were performed by using SAS 9.4 (SAS Institute; Cary, NC).