Participants and methods
We extracted a database of subjects from a stratified CKD cluster sampling study performed in Beijing in 2006. A total of 15,370 adults were selected from 53 primary sampling units and invited to participate. Women needed to be oversampled to obtain sufficient data due to the high male to female sex ratio in China. After weighting, the ratio of men to women in the study was 1:1.18. Ultimately, 13,925 subjects completed the survey and physical examination. Overall, the enrollment rate was 90.6% (92.1% of females and 89.6% of males). The ethics committees in Peking university first hospital approved this study. After obtaining informed consent from all subjects, they completed a screening questionnaire to collect information about socioeconomic status (e.g., gender, age, and education), personal and family health history, and lifestyle behaviors (e.g., smoking). Laboratory measurements also were made by well-trained professionals using uniform standards.
Serum creatinine (Scr) was measured by using a kinetic-rate method described by Jaffe. The glomerular filtration rate was estimated by using a modified Modification of Diet in Renal Disease (MDRD) Study formula: estimated GFR (eGFR, mL·min-1·1.73 m-2) = 175 × standard (Scr)-1.234 × age-0.179 × 0.79 (if female).
Urine albumin and creatinine measurements were made from morning spot urine samples. Urine albumin was measured by using an immunoturbidimetric assay (Audit Diagnostics, Cork, Ireland) and urine creatinine was measured by using Jaffe's kinetic method on a Hitachi 7170 chemistry analyzer (Tokyo, Japan). Albuminuria was diagnosed if the urine albumin/creatinine ratio was ≥17 mg·g-1 for men or ≥25 mg·g-1 for women and proteinuria was diagnosed if this ratio was ≥300 mg·g-1 for either men or women.
Hypertension was defined as measurement of systolic blood pressure ≥140 mmHg or diastolic blood pressure ≥ 90 mmHg, having been diagnosed with hypertension at least twice, or taking an antihypertensive medication at the time of the survey.
Diabetes was defined as measurement of a fasting glucose level > 6.9 mmol·L-1, a previous diagnosis of diabetes, or taking insulin or anti-diabetic medication at the time of the survey.
Dyslipidemia was defined as measurement of total cholesterol (TC) ≥6.22 mmol·L-1, triglyceride (TG) ≥2.26 mmol·L-1, high-density lipoprotein (HDL) < 1.04 mmol·L-1, or low-density lipoprotein (LDL) ≥4.14 mmol·L-1; a previous diagnosis of dyslipidemia; or taking lipid-lowering medication at the time of the survey.
Obesity was defined as having a body mass index (BMI) ≥28 kg·m-2 (BMI = weight/(height2)).
Individuals whose eGFR was > 200 mL·min-1·1.73 m-2 (n = 33) or < 15 mL·min-1·1.73 m-2 (n = 57) were excluded from this study. Participants were divided into 3 groups according to their health status: those with albuminuria or eGFR < 60 mL·min-1·1.73 m-2 were placed in the CKD group; those without CKD but with hypertension, diabetes, dyslipidemia, or obesity were placed in the at-risk group; those without CKD or other medical conditions were placed in the healthy group. We also used proteinuria as a definition of CKD and performed a separate statistical analysis as described below because currently there is no consensus about defining CKD by microalbuminuria or macroalbuminuria. Some authors only consider microalbuminuria to be a risk factor for CKD such as (pre)hypertension or diabetes and its reproducibility of microalbuminuria tests is limited .
General characteristics in each health status group
Sociodemographic characteristics (age, highest education level), health indicators (BMI, fasting glucose level, serum lipid levels, eGFR), and lifestyle behaviors (smoking) of each gender were described by descriptive statistics. The differences in these variables among genders were examined by using chi-square statistics for categorical variables and Student's t-test for continuous values.
Gender-specific adjusted eGFRs in each health status group
For each health status group, a linear model was constructed with eGFR as the dependent variable, gender as the independent variable, and confounding factors (systolic blood pressure, diastolic blood pressure, BMI, fasting glucose level, serum lipid levels, highest education level, and smoking) as covariates. This model was used to calculate the least squares means of eGFR (adjusted eGFR) and to compare eGFRs between genders within each health status group.
Gender-specific rates of decline of eGFR in each health status group
A linear model was constructed for each health status group. In these models, the dependent variable was eGFR and the independent variables were age, gender, and interaction of age and gender. These models were used to calculate the gender-specific rates of decline of eGFR and to test gender-specific differences. Subsequently, confounding factors were added to these models as covariates to calculate the adjusted gender-specific rates of decline of eGFR.
Adjusted gender-specific rates of decline of eGFR (referenced to healthy group) in CKD and at-risk groups
A linear regression model was constructed for each gender. In these models, the variables were (1) eGFR (dependent variable); (2) age, health status, and interaction of age and health status (independent variables); and (3) the aforementioned confounding factors (covariates). This model was used to calculate the adjusted gender-specific rates of decline of eGFR (referenced to the healthy group) for the CKD and at-risk groups. In addition, a linear regression model was constructed with the following variables: (1) eGFR (dependent variable); (2) age, health status, gender, and interactions of age and health status, age and gender, health status and gender, age and health status and gender (independent variables); and (3) the aforementioned confounding factors (covariates). This model was used to test for statistically significant differences between men and women for the adjusted gender-specific rates of decline of eGFR (referenced to the healthy group) for the CKD and at-risk groups.
All statistical analyses were performed using Statistical Analysis System software, version 9.0 (SAS Institute Inc., Cary, NC, USA). Tests for association were considered to be statistically significant when the p-values were less than 0.05.