Association of chronic kidney disease with cognitive impairment risk in middle-aged and older adults: the first longitudinal evidence from CHARLS

Study population

CHARLS constitutes a representative sample of middle-aged and elderly individuals in China. The study’s methodology and design have been previously described in detail¹⁸. Launched from June 2011 to March 2012, the baseline survey implemented a multistage and stratified sampling strategy, enrolling 17,708 participants aged 45 and above from 450 villages in 28 provinces. Data collection was conducted through structured questionnaires administered in individual interviews. Follow-up assessments were conducted biennially, with physical examinations and blood sample collection every four years. The study received ethics approval from the Biomedical Ethics Review Committee of Peking University, Beijing (IRB00001052-11015), and all participants provided informed consent. By 2018, CHARLS had successfully completed three rounds of follow-up assessments. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement.

The present study represented a secondary analysis of data from the CHARLS, spanning from 2011 to 2018.Participants were exclusively selected from the CHARLS’s Wave 1 cohort. 194 cases with incomplete follow-up information and 56 participants with no follow-up data were excluded from the analysis to minimize potential bias. In addition, 620 individuals without baseline data on chronic kidney diseases were excluded. 323 patients with CI at baseline were excluded prior to the commencement of the study. Consequently, the final analysis encompassed a sample size of 16,515 participants. The participant selection process was depicted in supplementary Fig. 1.

The analysis revealed several potential confounders that necessitated adjustment at baseline, including sociodemographic factors (age, gender, marital status, and educational attainment), lifestyle and health behaviors (smoking and alcohol intake, along with the individual body mass index [BMI]), and diagnoses of various chronic conditions (hypertension, diabetes, hyperlipidemia, cardiovascular disease [CVD], and stroke). Participants were surveyed regarding their medical diagnoses of the abovementioned chronic diseases. If the respondents provided an affirmative response or were currently taking medications for the management of chronic conditions, such as antihypertensive drugs, hypoglycemic agents, insulin injections, or lipid-lowering medications, they were categorized as having the respective chronic diseases.

Participants’ ages were categorized into three groups: 45–54, 55–64, and 65 years and older. Marital status was classified into two categories: married or cohabiting and other (encompassing separated, divorced, never married, widowed, or married but separated individuals). Education was stratified into three levels: below lower secondary, upper secondary or vocational, and tertiary. The smoking status was dichotomized into smoker and nonsmoker. Alcohol consumption was divided into drinkers and non-drinkers. BMI was allocated into four categories: underweight (BMI < 18.5 kg/m²), normal weight (18.5 ≤ BMI < 24.0 kg/m²), overweight (24.0 ≤ BMI < 28.0 kg/m²), and obese (BMI ≥ 28.0 kg/m²).

Identification of the CI and CKD

The primary endpoint of this study was CI, encompassing Alzheimer’s disease, brain atrophy, Parkinson’s syndrome, and other memory-related disorders, with the specific etiology of CI remaining unspecified. CI status was determined through interviews conducted by trained personnel who inquired, “Have you been diagnosed with a memory-related condition, such as Alzheimer’s disease, brain atrophy, or Parkinson’s disease, by a medical professional?” Positive responses prompted further inquiry regarding the timing of diagnosis and current medication use. The presence of CKD was determined at the baseline survey with the query: “Have you been diagnosed with kidney disease, excluding renal tumors and cancer, by a physician?” Affirmative responses resulted in the classification of participants as having CKD, followed by an inquiry into the timing of their diagnosis. The diagnostic criteria for CKD and CI, based on self-reported diagnoses, are consistent with those used in previous studies that have utilized the CHARLS dataset^19,20.

Statistical analyses

We utilized the Kolmogorov-Smirnov test to verify the normality of the continuous variables in our large sample study. Subsequently, Depending on whether the continuous variables exhibit normality, either the t-test or the Mann–Whitney U test was used. For categorical variables, the chi-square test was employed. Missing values were addressed through multiple imputation using chained equations implemented via the “mice” package in R.

The primary endpoint of this study was the incidence of CI, whereas the secondary endpoint was all cause mortality. Survival analysis was performed using the Kaplan–Meier method, and differences in survival were assessed using a stratified log-rank test. The association between CKD and the risk of CI during the follow-up period was evaluated using Cox proportional hazard regression, excluding patients with prevalent CI at baseline. Data were censored at the earliest occurrence of CI, death, or July 2018. Proportional hazard assumptions were confirmed using Schoenfeld residual plots. Hazard ratios (HRs) and their 95% confidence intervals (CIs) were calculated. Three adjusted models were developed: Model 1 included adjustments for age and gender; Model 2 added adjustments for education, marital status, lifestyle habits including smoking, drinking, and BMI; and Model 3 further incorporated comorbidities such as hypertension, diabetes, hyperlipidemia, CVD, and stroke.

Using Stata software, we calculated the incidence rates (IRs) of CKD leading to CI or mortality, expressed as incidents per 1,000 person-years with their corresponding 95% CIs. Laplace regression was applied to model the time to incident CI or mortality in relation to kidney status, which evaluated the variability in time to events. As less than 11% of cohort participants experienced the outcomes, we analyzed the median time (in years) until the initial 10% of the cohort faced CI or mortality. This analysis was conducted using the “laplacereg” command in Stata.

Stratified and interaction analyses were conducted to evaluate the varying impact of kidney disease across distinct subgroups, using baseline characteristics as covariates. The Cox proportional hazard regression model was reapplied after stratification by age, sex, BMI, presence of hypertension, diabetes mellitus, cvd, and stroke. Statistical significance of the interaction was determined by including interaction terms and the execution of likelihood ratio tests.

A series of sensitivity analyses were conducted. To alleviate the impact of confounding variables, inverse probability weights (IPW) were computed for each participant. The effectiveness of the weighting was assessed by evaluating the standardized mean difference (SMD) of covariates between the weighted populations, with SMD < 0.1 indicating an adequate balance of covariates across groups. The main analysis was subsequently conducted using the weighted data. The Fine and Gray competing risk model was applied to assess the competing risk of mortality relative to the occurrence of CI. The main analyses were repeated in a population that excluded individuals who developed CI within the first 2 years of the follow-up period. To mitigate the risk of including CI events attributable to comorbid chronic conditions, the main analyses were also conducted in a cohort that excluded participants with hypertension, diabetes mellitus, hyperlipidemia, CVD, and stroke at baseline.

We employed two-sided P-values, with the level of significance set at α = 0.05. Statistical analyses were performed using R (version 4.0.3) and Stata statistical software (version 17.0).