Intraindividual difference in estimated GFR by creatinine and cystatin C, cognitive trajectories and motoric cognitive risk syndrome Free

ABSTRACT

Background

Intraindividual differences between estimated glomerular filtration rate (eGFR) based on cystatin C (eGFR_cys) and creatinine (eGFR_cr) can convey important clinical information regarding health status. However, the clinical implications of these differences (eGFR_diff) for risk of cognitive decline and motoric cognitive risk (MCR) syndrome remains unclear. We aimed to investigate the longitudinal associations of eGFR_diff with cognitive trajectories and incident MCR.

Methods

Based on the China Health and Retirement Longitudinal Study, we identified two study subcohorts: one for cognitive trajectory follow-up (6423 participants, 2011–2018) and another for incident MCR follow-up (2477 participants, 2011–2015). The eGFR_diff was defined as eGFR_cys − eGFR_cr. Adjusted ordinal and binary logistic regression models were separately used to assess the associations of eGFR_diff with cognitive trajectories and incident MCR. We also performed discordance analyses for eGFR_diff versus eGFR_cys, eGFR_cr or eGFR based on both creatinine and cystatin C (eGFR_cys-cr).

Results

In the first subcohort, four distinct 7-year cognitive trajectories were identified. Each 1 standard deviation (SD) higher eGFR_diff (value for eGFR_cys − eGFR_cr) was associated with a lower risk of poorer cognitive trajectories {odds ratio 0.909 [95% confidence interval (CI) 0.877–0.942]}. In the second subcohort, 121 participants developed incident MCR after a 4-year follow-up. Each 1-SD higher eGFR_diff (value for eGFR_cys − eGFR_cr) was linked with a 25.3% (95% CI 16.6–33.2) decreased risk for MCR. The above associations persisted in individuals with normal kidney function. Additionally, the risk for cognitive decline and incident MCR was more strongly associated with eGFR_cys than eGFR_cr and eGFR_cys-cr. For the discordance analyses, the ‘discordantly high eGFR_diff/low eGFR’ group but not the ‘discordantly low eGFR_diff/high eGFR’ exhibited a significantly lower risk of poorer cognitive trajectories and MCR compared with the concordant group.

Conclusions

A large negative difference between eGFR_cys and eGFR_cr (eGFR_cys < eGFR_cr) was associated with a higher risk of cognitive decline and incident MCR. The eGFR_diff could capture additional valuable risk information beyond eGFR_cys, eGFR_cr and eGFR_cys-cr.

Graphical Abstract

Open in new tabDownload slide

INTRODUCTION

Due to the aging population, cognitive impairment and dementia have become a serious global public health concern that leads to a reduced quality of life and substantial economic burden [1–4]. The prevalence of mild cognitive impairment and dementia is estimated to be 15.5% and 6.0%, respectively, among adults ≥60 years of age in China [5]. Motoric cognitive risk (MCR) syndrome is considered a pre-dementia syndrome with subjective cognitive complaints and objective slow gait speed [6, 7]. Strong links have been established between MCR and multiple adverse outcomes, such as dementia, falls and disability, as well as all-cause mortality [6, 8–11]. Given that there is no effective disease-modifying treatment for dementia and cognitive impairment [2, 3], early identification of potential risk factors or prodromal biomarkers of cognitive decline and pre-dementia syndromes is crucial to prevent or delay the incidence of dementia and promote healthy aging.

Impaired kidney function, as quantified by estimated glomerular filtration rate (eGFR), has been previously viewed as a potential risk factor for dementia and cognitive impairment [12, 13]. However, several studies have yielded inconsistent results, showing no associations between eGFR and dementia or cognitive decline [14–16]. This inconsistency may be attributable to the different methods used in the estimation of GFR. A previous study suggested that lower cystatin C–based eGFR (eGFR_cys), but not eGFR based on creatinine (eGFR_cr) or a combination of cystatin C and creatinine (eGFR_cys-cr), was significantly associated with dementia [17]. As for the reasons, creatinine is more affected by non-kidney factors, in particular by muscle mass, physical activity and diet, which may lead to overestimation or underestimation of kidney function and confound the relationship between eGFR containing creatinine information and cognitive function [18–20]. Compared with creatinine, cystatin C is not only less influenced by non-kidney factors, but is also unaffected by race or genetic ancestry [18]. When the levels of serum cystatin C and creatinine are disproportionately affected by non-kidney factors, there may be a considerable discrepancy between eGFR_cys and eGFR_cr measured in the same individual [19].

Although the eGFR equation including both cystatin C and creatinine has been developed to balance the non-kidney determinants and had greater precision and accuracy than either marker alone [21, 22], the differences between eGFR_cys and eGFR_cr (eGFR_diff, defined as eGFR_cys − eGFR_cr) has been ignored [23]. This may mask important clinical information about an individual's health status provided by these differences. For instance, recent studies demonstrated that eGFR_diff can serve as a biomarker reflecting muscle mass [24] and frailty status [23, 25–27]. Additionally, prior studies also found that lower negative eGFR_diff was associated with a higher risk of all-cause mortality, physical function decline, injurious falls, kidney failure, heart failure and other cardiovascular diseases [18–20, 23, 25]. However, it remains unclear whether eGFR_diff is related to the cognitive function and MCR. Considering that eGFR_diff could capture information about the risk for multiple adverse outcomes related to aging to a certain extent [23, 28], we hypothesized that eGFR_diff would predict the aging trajectory of cognitive function and incident MCR. Finally, whether and to what extent eGFR calculated by different markers affects dynamic cognitive trajectories and the risk of MCR is also an open question.

Thus, based on nationally representative data from the China Health and Retirement Longitudinal Study (CHARLS), we conducted this study among middle-aged and older adults to identify distinct 7-year cognitive trajectories and investigate the association between the baseline eGFR_diff and cognitive trajectories; evaluate the relationship between eGFR_diff and 4-year risk of developing MCR; further determine the independent associations of eGFR_cys, eGFR_cr and eGFR_cys-cr with study outcomes; and perform discordance analyses to explore whether eGFR_diff is superior to these eGFR indicators in identifying individuals at increased risk of poor cognitive trajectories and MCR.

MATERIALS AND METHODS

Study population

CHARLS is an ongoing nationwide cohort study conducted among Chinese adults ≥45 years of age. The baseline survey was carried out between June 2011 and March 2012. A total of 17 707 participants from 150 county-level units distributed in 28 provinces of China were recruited at baseline through multistage probability sampling and the subsequent three follow-up visits were carried out (wave 2 in 2013, wave 3 in 2015 and wave 4 in 2018). Details about the data have been previously described elsewhere [29, 30]. The protocols of CHARLS were approved by the Biomedical Ethics Review Committee of Peking University (IRB00001052-11015). All participants provided written informed consent. We identified two subcohorts from CHARLS.

The first subcohort was constructed to explore the longitudinal association of baseline eGFR_diff and other kidney function measures based on creatinine and cystatin C with 7-year cognitive trajectories. Participants were excluded if they had missing data on gender, age, cognitive function, serum cystatin C and creatinine; had prior diagnoses of memory-related disease (Alzheimer's disease, brain atrophy, Parkinson's disease); or were lost to follow-up from waves 2 to 4. A total of 6423 participants were enrolled for the analyses (Supplementary Fig. S1).

The second subcohort was constructed to follow-up for incident MCR. Considering the availability of data required for the assessment of MCR, the data from 2011 to 2015 were used. We excluded those with an unavailable follow-up survey in 2015; missing information about gender, age, assessments of MCR, serum cystatin C and creatinine, a history of memory-related disease or physical disability; or who were diagnosed with MCR at baseline. Ultimately, a total of 2477 participants were included (Supplementary Fig. S2).

The timeline of the study is presented in Supplementary Fig. S3.

Assessment of exposure

The detailed process of blood sample collection, transportation, storage and analysis is described in a previous study [30]. The particle-enhanced turbidimetric assay method was used to measure serum cystatin C, and creatinine concentration was measured using a rate-blanked and compensated Jaffe creatinine method [30].

eGFR_cys, eGFR_cr and eGFR_cys-cr were calculated using the 2012 Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation [22] and 2021 race-free equations [31] were used in the sensitivity analysis. The eGFR_diff was defined as eGFR_cys − eGFR_cr, with positive values indicating a higher eGFR_cys than eGFR_cr and negative values indicating a higher eGFR_cr than eGFR_cys. The study population was divided into three categories based on a cut-off value of 15 ml/min/1.73 m²: negative group (eGFR_diff <−15 ml/min/1.73 m², with eGFR_cys < eGFR_cr), midrange group (eGFR_diff between −15 and +15 ml/min/1.73 m², with eGFR_cys similar to eGFR_cr) and positive group (eGFR_diff >+15 ml/min/1.73 m², with eGFR_cys > eGFR_cr). These cut-off values were chosen to be consistent with previous studies [18, 19, 23, 26, 32] and 15 ml/min/1.73 m² can represent a clinically meaningful difference in eGFR-determined CKD stages [18, 19].

Assessment of cognitive function

Cognitive performance was evaluated by two measures: episodic memory and executive function [3, 33]. In the episodic memory test, participants were instructed to recall 10 Chinese words just read to them (immediate recall) and 5 min later (delayed recall). We counted the number of words correctly recalled and the episodic memory score was the average value of the immediate and delayed recall scores, ranging from 0 to 10. In the executive function test, participants were shown a picture and asked to redraw it. The score was 1 if the participants succeeded. They also were required to subtract 7 from 100 serially five times and identify the date (year, season, month, week, day). Answers to these questions were summed into the executive function score, ranging from 0 to 11. The global cognitive score was determined as the sum of the episodic memory score and executive function score, ranging from 0 to 21.

Assessment of MCR

MCR is defined as the presence of cognitive complaints and slow gait without dementia or mobility disability [8]. In CHARLS, subjective cognitive complaints were identified if participants answered ‘fair’ or ‘poor’ (versus ‘excellent’, ‘very good’ or ‘good’) to a self-reported question about memory loss: ‘How would you rate your memory at the present time?’ [34]. For gait speed, each participant was asked to walk 2.5-m at their regular pace twice (there and back) and the average of the values was used. Slow gait was defined as ≤1.0 standard deviation (SD) for age- (≤75, >75 years) and sex-appropriate mean values of walking speed in this cohort [7, 34].

Covariates

Several potentially confounding covariates were considered in our analyses and obtained through structured questionnaires, blood tests and physical examinations. Specifically, sociodemographic characteristics included age, gender (male, female), marital status (married, other), residence (rural, urban) and education level (no formal education, junior high school or below, high school or above). Health-related factors included smoking status (non-smoker, former smoker, current smoker), drinking status (drinking more than once per month, drinking once or less than once per month, no drinking). Depressive symptoms were assessed via the 10-item Center for Epidemiological Studies Depression Scale (CES-D-10) and a cut-off score ≥10 indicated the presence of depressive symptoms [35]. Body mass index (BMI) was calculated as weight (kg)/height² (m²). Restriction on activities of daily living (ADL) was defined as having difficulty in any of the five ADL tasks (bathing, dressing, eating, getting into/out of bed and toileting) [33]. Information on blood urea nitrogen (BUN), fasting blood glucose (FBG), glycated haemoglobin (HbA1c), high-sensitivity C-reactive protein (hs-CRP) and total cholesterol was obtained through metabolic examination. Information on the medication history of hypertension, diabetes and kidney disease was collected via self-reports. Systolic blood pressure (SBP) and diastolic blood pressure (DBP) were measured three times for each participant and the mean value of these measurements was utilized for further analysis. The diagnosis of hypertension was established if participants exhibited an SBP/DBP ≥140/90 mmHg, were taking anti-hypertensive medication or had a self-reported physician diagnosis [36]. Diabetes was identified in individuals with an FBG level ≥7.0 mmol/l, an HbA1c ≥6.5%, the use of any glucose-lowering treatment or a self-reported physician diagnosis [37]. Stroke, kidney diseases, heart diseases, chronic lung diseases and cancer were ascertained by self-reported physician diagnosis. Muscle strength was evaluated based on handgrip strength, and grip strength was normalized to body weight, calculated as grip strength (kg)/body weight (kg) [38]. Finally, physical activity was evaluated among a randomly selected subsample of the study participants (2799 were available in subcohort 1, 1051 were available in sub-cohort 2) and was divided into four categories: ‘none’, ‘mild’, ‘moderate’ and ‘vigorous’, according to previous studies [3, 39].

Statistical analysis

Based on multiple repeated measurements (including baseline and three follow-up visits) of the cognitive scores, trajectories of global cognitive function were plotted over a 7-year follow-up time using group-based trajectory modelling (GBTM). Group-based multitrajectory modelling (GBMTM) was used to determine the jointly longitudinal changes in episodic memory and executive function across four visits. GBTM and GBMTM can identify clusters of individuals following similar change patterns through multiple visits using uni- and multivariables, respectively [3]. Varied models were fitted and we then selected the optimal number of distinct groups and trajectory shape parameters (e.g. linear, quadratic and cubic) based on Bayesian information criteria, Akaike's information criterion, average posterior probabilities of each trajectory group (≥0.70), odds of correct classification for all groups (≥5), adequate sample size in each trajectory group (>5% of the sample) and a degree of clinical interpretation [3, 40]. The best trajectories of cognition were the main outcomes in the first subcohort.

Baseline characteristics were expressed as mean (SD), median [interquartile range (IQR)] or number (percentage). Differences among groups (according to eGFR_diff categories, cognitive trajectories or incident MCR status) were compared using the chi-squared test or Fisher's exact test for categorical variables and one-way analysis of variance or Kruskal–Wallis test for continuous variables, as appropriate.

In subcohort 1, ordinal logistic regression models were used to assess the associations between baseline eGFR_diff (considered as both continuous and categorical variables) and 7-year cognitive trajectories. In subcohort 2, we used binary logistic regression models to assess longitudinal associations between eGFR_diff (considered as both continuous and categorical variables) and incident MCR over 4 years. Remarkably, among the 95 participants in the positive eGFR_diff group, there were no incidents of MCR events. Consequently, we combined the positive group and midrange group into a reference group for subsequent analysis of categorized eGFR_diff (negative group, with eGFR_diff <−15 ml/min/1.73 m² versus the reference group, with eGFR_diff ≥−15 ml/min/1.73 m²), aiming to obtain robust results. The odds ratio (OR) and 95% confidence interval (CI) were obtained. We progressively adjusted the models for potential confounding factors. The final multivariable model was adjusted for age, gender, marital status, education level, residence, smoking status, drinking status, depressive symptoms, BMI, restriction on ADL, hypertension, diabetes, stroke, heart diseases, chronic lung diseases, cancer, total cholesterol, hs-CRP, medication history of kidney disease and hypertension, baseline global cognitive scores and normalized handgrip strength. The models were further adjusted for physical exercise in the subpopulations who underwent physical activity assessments. We also adjusted for eGFR_cr to explore if the observed associations were independent of the most common measure of kidney function in current clinical practice. Then, the dose–response relationships of eGFR_diff with cognitive trajectories and MCR were evaluated using restricted cubic splines (RCSs) with three knots (at the 10th, 50th and 90th percentiles). Subsequently we explored whether the associations between eGFR_diff and study outcomes persisted among participants with normal kidney function (eGFR ≥60 ml/min/1.73 m²) or without self-reported kidney diseases.

Additionally, to gain a greater understanding of the association of eGFR_diff with cognition and MCR, we individually modelled the associations of baseline eGFR_cys, eGFR_cr an eGFR_cys-cr with future cognitive trajectories and incident MCR. Next, we compared the associations of eGFR based on cystatin C versus creatinine with outcomes by including both eGFR values in the same multivariable-adjusted model. To further elucidate whether eGFR_diff could provide more accurate insights into the risk of cognitive decline and MCR than eGFR_cys, eGFR_cr or eGFR_cys-cr alone, we performed discordance analysis using two approaches, as previously described [41, 42]. First, we defined discordance by the percentile distance between eGFR_diff and eGFR. Participants were classified into three groups: concordant (eGFR_diff percentile − eGFR percentile within ±15 percentile units), discordantly high eGFR_diff (eGFR_diff percentile > eGFR percentile by 15 percentile units) and discordantly low eGFR_diff (eGFR_diff percentile < eGFR percentile by 15 percentile units). Second, participants were divided into four groups with a focus on low levels (individuals with the 20% lowest levels) to assess the robustness of the results: both eGFR_diff and eGFR >20th percentile, eGFR_diff >20th percentile but eGFR ≤20th percentile, eGFR_diff ≤20th percentile but eGFR >20th percentile and both eGFR_diff and eGFR ≤20th percentile. The associations between eGFR_diff with eGFR concordant/discordant groups and cognitive trajectories and MCR were also estimated using the same models described above.

In the above analyses, an inverse probability weighting method was used to address potential selection bias [43]. A multivariable logistic regression model was created to obtain the probability of being included in the two subcohorts and model covariates were the same as those used in the primary analysis. The inverse of this probability was then used to create an analytical weight for each included participant. Absolute standardized mean differences were applied to verify the covariate balance between the included and excluded samples, using a Love plot selected for visualization [44].

Several sensitivity analyses were conducted to validate the study findings. First, we repeated our primary analysis based on the original unweighted samples. Second, we used the 2021 CKD-EPI race-free equations to recalculate eGFRs. Third, we repeated our analysis after excluding those with a history of CKD medication. Fourth, the generalized estimating equation (GEE) models were used to examine the longitudinal association between eGFR_diff and follow-up cognitive scores. Fifth, we investigated the associations of changes in eGFR_diff with cognitive trajectories and risk of developing MCR utilizing creatinine and cystatin C data collected at baseline in 2011 and a follow-up survey in 2015. Changes in eGFR_diff were calculated as the eGFR_diff values in 2015 minus those in 2011. The models were additionally adjusted for baseline eGFR_diff. Finally, we reassessed the relationship between eGFR_diff and study outcomes, adjusting for BUN, self-reported kidney diseases, CKD determined through eGFR_cys or CKD determined through eGFR_cys-cr instead of eGFR_cr.

GBTM techniques were implemented using the Proc Traj in Stata version 15.1 (StataCorp, College Station, TX, USA). Other statistical analyses were performed with R version 4.1.0 (R Foundation for Statistical Computing, Vienna, Austria). A two-sided P-value <0.05 was considered statistically significant.

RESULTS

Estimated cognitive trajectories

The procedures for choosing the optimal group number and shape parameter for the final global cognitive trajectory model are presented in Supplementary Tables S1 and S2. We determined four clusters to reflect the heterogeneous pattern of longitudinal change in global cognitive function over 7 years (Fig. 1): group 1, ‘low-declining’ [n = 854 (13.30%)]; group 2, ‘moderate low-declining’ [n = 1382 (21.52%)]; group 3, ‘moderate high-stable’ [n = 2117 (32.96%)]; and group 4, ‘high-stable’ [n = 2070 (32.23%)]. The four groups showed a trend of increasingly better cognitive trajectories. Additionally, we also determined four groups using GBMTM that plotted the longitudinal joint changes in episodic memory and executive function based on the same criteria and procedure (Supplementary Fig. S4). Further details are shown in Supplementary Tables S3 and S4.

Baseline characteristics

The distribution of baseline characteristics is shown in Table 1. Supplementary Tables S5 and S6 show the detailed characteristics according to eGFR_diff groups. Among 6423 participants included in the first subcohort, the median age was 58.00 years (IQR 51.00–66.00) and 3025 (47.1%) were men. A total of 3175 (49.4%) participants had an eGFR_diff between −15 and 15 ml/min/1.73 m² (midrange group, with eGFR_cys similar to eGFR_cr), 2915 (45.4%) participants had an eGFR_diff <−15 ml/min/1.73 m² (negative group, with eGFR_cys < eGFR_cr) and 333 (5.2%) participants had an eGFR_diff of ≥15 ml/min/1.73 m² (positive group, with eGFR_cys > eGFR_cr). Among 2477 participants included in the second subcohort, the median age was 65.00 years (IQR 61.00–71.00), 1243 (50.2%) were men and 121 (4.9%) developed incident MCR after 4 years of follow-up. A total of 1092 (44.1%) participants had a midrange eGFR_diff, 1290 (52.1%) had a negative eGFR_diff and 95 (3.8%) had a positive eGFR_diff. Compared with the other two eGFR_diff groups, participants in the negative eGFR_diff group tended to be older, experiencing depression, unmarried, male, smokers and from rural areas. Additionally, they had higher hs-CRP levels and lower educational levels, along with poorer cognitive performance. The characteristics of the population according to cognitive trajectory groups and incident MCR are presented in Supplementary Tables S7 and S8. Those who had worse cognitive trajectories or developed MCR were more likely to have lower eGFR_diff, eGFR_cys, eGFR_cr and eGFR_cys-cr values than those with better cognitive performance or who did not develop MCR.

After inverse probability weighting to eliminate selection bias, the differences between included and excluded subjects were clearly smaller compared with the differences between the original unweighted samples (Fig. 2).

Association of eGFR_diff and eGFR with 7-year cognitive trajectories

The results of the adjusted ordinal logistic regression analyses are reported in Table 2. In model 1, each 1-SD higher eGFR_diff was associated with a lower risk of poorer global cognitive trajectories [OR 0.894 (95% CI 0.867–0.922)]. The association remained significant with adjustment for eGFR_cr [OR 0.894 (95% CI 0.866–0.923)]. When considering eGFR_diff as a categorical variable, participants in the negative eGFR_diff group (eGFR_cys < eGFR_cr) had a higher risk of cognitive decline compared with the midrange group [OR 1.216 (95% CI 1.144–1.293)], whereas positive eGFR_diff (eGFR_cys > eGFR_cr) was correlated with reduced risk [OR 0.858 (95% CI 0.745–0.988)]. The above associations still hold after progressive adjustment for baseline cognitive scores (model 2), muscle strength (model 3) and physical activity (model 3 + physical activity, among 2799 participants, as shown in Supplementary Table S9), but the positive eGFR_diff was no longer significantly associated with global cognitive trajectories. A non-linear dose–response relationship between eGFR_diff and risk of poor cognitive trajectories was detected using the RCS model (for non-linearity, P = 0.001) (Fig. 3A). In addition, among those with normal eGFR (eGFR_cys/eGFR_cr/eGFR_cys-cr ≥60 ml/min/1.73 m²) or without self-reported kidney diseases, eGFR_diff remained strongly and significantly associated with trajectories of cognition (Supplementary Table S10).

For three GFR estimates (Table 2), higher eGFR_cys [OR 0.874 (95% CI 0.840–0.910)] and eGFR_cys-cr [OR 0.896 (95% CI 0.859–0.934)] rather than eGFR_cr [OR 0.969 (95% CI 0.930–1.010)] were significantly associated with a lower risk of global cognitive decline after full adjustment for covariates. We also observed significant associations of higher levels of serum cystatin C [OR 1.114 (95% CI 1.074–1.156)], but not creatinine [OR 1.024 (95% CI 0.985–1.065)], with worse global cognitive trajectories. The above associations remained after additional adjustment for eGFR_cr or serum creatinine. Notably, when we further adjusted for eGFR_cys instead of eGFR_cr, eGFR_cys-cr was no longer associated with global cognitive trajectories.

Finally, consistent results were obtained when using joint progression trajectories of episodic memory and executive function as outcome variables (Supplementary Table S11).

Association of eGFR_diff and eGFR with 4-year MCR onset risk

In the fully multivariable-adjusted model (model 3), each 1-SD higher eGFR_diff was associated with a 25.3% (95% CI 16.6–33.2) decreased risk of MCR (Table 3). The association remained robust after inclusion of eGFR_cr in all adjusted models [OR 0.651 (95% CI 0.574–0.736)]. Compared with those with eGFR_diff ≥−15 ml/min/1.73 m², participants in the negative eGFR_diff group, with eGFR_diff <−15 ml/min/1.73 m², had a 1.412-fold (95% CI 1.151–1.739) greater risk of incident MCR. These findings still held after additional adjustment for physical activity among 1051 participants with physical activity assessments (Supplementary Table S9). The spline function for eGFR_diff confirmed a linear dose–response relationship between eGFR_diff and the risk of developing MCR (for non-linearity, P = 0.589) (Fig. 3B). There was still a significant association between eGFR_diff and MCR among participants without self-reported kidney diseases or with normal eGFR assessed by serum cystatin C, creatinine or their combination (Supplementary Table S10).

In addition, higher serum cystatin C and creatinine and lower eGFR_cys, eGFR_cr and eGFR_cys-cr were all associated with elevated MCR risk when modelled individually in multivariable-adjusted models. However, when eGFR_cys and eGFR_cr or cystatin C and creatinine were both included in the same multivariable-adjusted model, only eGFR_cys [OR 0.596 (95% CI 0.513–0.692)] and cystatin C [OR 1.559 (95% CI 1.402–1.731)] were significantly correlated with incident MCR (Table 3). We did not observe a significant association between eGFR_cys-cr and MCR after adjusting for eGFR_cys rather than eGFR_cr in our multivariable models.

Discordance analysis of eGFR_diff and eGFR

As displayed in Fig. 4, compared with the concordant group, the discordantly high eGFR_diff had a significant decreased risk of poorer cognitive trajectories [OR for discordant eGFR_diff/eGFR_cys 0.853 (95% CI 0.752–0.968); OR for discordant eGFR_diff/eGFR_cr 0.861 (95% CI 0.781–0.950); OR for discordant eGFR_diff/eGFR_cys-cr 0.778 (95% CI 0.693–0.873)] and discordantly low eGFR_diff had the opposite association, i.e. an increased risk. Analogously, we also found that discordantly high eGFR_diff was associated with a decreased risk of developing MCR [OR for discordant eGFR_diff/eGFR_cys 0.485 (95% CI 0.335–0.696); OR for discordant eGFR_diff/eGFR_cr 0.651 (95% CI 0.485–0.872); OR for discordant eGFR_diff/eGFR_cys-cr 0.526 (95% CI 0.377–0.731)]. However, we did not observe a significant association between discordantly low eGFR_diff and incident MCR.

We redefined discordance and divided the population into four groups to assess the robustness of our findings. Compared with individuals with both eGFR_diff and eGFR (eGFR_cys/eGFR_cr/eGFR_cys-cr) >20th percentile, those with ‘only eGFR_diff ≤20th percentile’, but not ‘only eGFR ≤20th percentile’ had significantly higher risks of a decline in cognitive function and MCR (Supplementary Figure S5).

Sensitivity analyses

Our findings were not substantially changed after conducting the primary analysis using the original unweighted samples (Supplementary Tables S12 and S13). The results from our analyses were also unaffected by re-estimating the GFR values using the 2021 CKD-EPI race-free equations (Supplementary Tables S14 and S15) or excluding participants with a history of CKD medication (Supplementary Tables S16 and S17). Moreover, the associations of eGFR_diff and other kidney function measures with cognitive function based on GEE models are shown in Supplementary Tables S18 and S19. Each 1-SD eGFR_diff increment was significantly associated with an increase in global cognitive scores [β = 0.134 (95% CI 0.068–0.200)]. In addition, we also assessed the associations of changes in eGFR_diff with cognitive trajectories and MCR. Subcohort 1 and subcohort 2 included 4539 and 2093 participants, respectively, who had available eGFR_diff data in 2011 and 2015, and we calculated the 4-year change in eGFR_diff for analysis. As shown in Supplementary Table S20, we found that greater increases in eGFR_diff during the follow-up period were associated with a reduced risk of cognitive decline [OR 0.913 (95% CI 0.866–0.963)] and incident MCR [OR 0.815 (95% CI 0.698–0.950)]. Finally, adjusting for BUN, self-reported kidney diseases, CKD based on eGFR_cys or eGFR_cys-cr rather than eGFR_cr in our multivariable models revealed that eGFR_diff remained independently associated with cognitive trajectories and incident MCR (Supplementary Table S21).

DISCUSSION

Based on a nationally representative longitudinal cohort study, our findings revealed for the first time a significant association of intraindividual eGFR_diff with future cognitive trajectories and MCR. Higher eGFR_diff (value for eGFR_cys − eGFR_cr) was strongly associated with a lower risk of poorer cognitive trajectories and incident MCR, independent of demographics and clinical factors including kidney function measures, even in individuals with normal kidney function. Furthermore, we found that eGFR_cys appeared to provide more information regarding the risk for cognitive decline and MCR than eGFR_cr. Finally, discordance analyses indicated that those with discordantly low eGFR_diff/high eGFR but not high eGFR_diff/low eGFR had a higher risk of worse cognitive trajectories and MCR. The eGFR_diff could serve as the more reliable indicator to identify the risk for cognitive decline and pre-dementia symptoms.

Impaired kidney function has been increasingly recognized as a potential risk factor for cognitive dysfunction and dementia [12, 17, 45]. Our study examined the long-term dynamic trajectories of cognitive performance rather than relying on dichotomized cognition status, and also considered MCR, a pre-dementia syndrome. We added the first prospective evidence concerning the association of kidney function measures using cystatin C, creatinine or their combination with cognitive trajectories and incident MCR. Additionally, we also tested the associations of kidney diseases determined via self-report and BUN with the study outcomes. We found that individuals who self-reported kidney diseases exhibited a higher risk of cognitive decline [OR 1.290 (95% CI 1.060–1.572)] and MCR [OR 4.114 (95% CI 2.116–7.675)]. Higher BUN levels were also significantly associated with an increased risk of MCR [OR 1.041 (95% CI 1.004–1.079)]. Kidney dysfunction can contribute to cognitive decline and MCR through various mechanisms, such as uraemic toxin accumulation, increased oxidative stress, uraemia-related neuroinflammation, compromised blood–brain barrier and neurotransmitter dysregulation, which may exert direct detrimental effects on the central nervous system [46, 47]. Impaired kidney function may also potentially influence neurodegenerative changes by affecting beta-amyloid clearance [47]. Notably, both our study and previous research demonstrated a stronger link between eGFR_cys and cognitive-related outcomes compared with eGFR_cr and eGFR_cys-cr [17]. This could be attributed to non-kidney factors, such as muscle mass and diet, that confound their association, and these factors may have a tight bond with cognitive function. The differential influence of non-kidney factors on the levels of cystatin C and creatinine is commonly observed as large differences between eGFR_cys and eGFR_cr within the same individual [19]. Nevertheless, the clinical information carried by these differences has been mostly overlooked in prior studies. Thus our study provided the first evidence that intraindividual eGFR_diff is associated with the future trajectory of cognition and risk of MCR, and these associations were independent of kidney function measurements based on cystatin C, creatinine or their combination, as well as self-reported kidney disease and BUN. We also confirmed that eGFR can still convey risk information for cognitive decline and MCR, even in individuals with normal kidney function. Our study underscores the clinical value of eGFR_diff beyond routine kidney function measurements.

There are several plausible biologic mechanisms to explain the observed associations. On the one hand, the influence exerted by some non-GFR factors on creatinine levels appears to be more significant than that on cystatin C levels [48], with muscle mass being a particularly notable factor. Individuals with lower muscle mass may exhibit reduced levels of serum creatinine, leading to a possible overestimation of eGFR_cr and subsequent lower eGFR_diff. Thus the lower negative eGFR_diff could indicate potential sarcopenia [25], and sarcopenia was proposed as a risk factor for cognitive impairment [49]. However, our results remained robust after additional adjustment for muscle strength in our analyses. Additionally, physical activity, nutritional factors and health status also influence creatinine production, whereas the impact on cystatin C is comparatively negligible, leading to a large eGFR_diff. These non-kidney factors may all influence cognitive function. On the other hand, some non-kidney factors may originate from cystatin C, such as diabetes, obesity and inflammation [48]. Nevertheless, the magnitude of the associations between these factors and serum cystatin C levels is weaker than that between non-kidney factors and serum creatinine [18, 19, 48]. In our study, we confirmed that these associations are independent of diabetes, BMI and hs-CRP. Besides non-kidney determinants, shrunken pore syndrome (SPS), characterized by a low eGFR_cys:eGFR_cr ratio can also be a possible explanation [28, 50]. SPS has been reported to be related to the accumulation of atherosclerosis-promoting proteins, such as pro-inflammatory cytokines, and increased mortality [20, 50, 51]. The SPS theory explains to some extent the links between eGFR_diff and some vascular system dysfunction-related diseases, as supported by several studies reporting associations between eGFR_diff and adverse cardiovascular outcomes [19, 20]. Vascular system dysfunction is regarded as a well-established risk factor for cognitive impairment. Thus we inferred that SPS may negatively impact vascular function by promoting atherosclerosis, potentially resulting in cognitive decline and MCR. Finally, based on baseline characteristics (Supplementary Tables S5 and S6), individuals with negative eGFR_diff tended to be older. Aging is considered the primary cause of cognitive decline and MCR [52, 53], which may also be a plausible explanation for these associations. Nevertheless, all multivariate models were adjusted for age, effectively controlling for its influence. This indicates that adjustment for age does not affect the clinical value of eGFR_diff in assessing the risks.

There is growing acknowledgement that substantial differences between eGFR_cys and eGFR_cr can convey risk information for various adverse clinical outcomes. Previous studies have established that higher eGFR_diff was associated with a lower risk of frailty, injurious falls, cardiovascular diseases, end-stage kidney disease and mortality, and the associations remained significant across categories of eGFR [18–20, 23, 25]. Our study similarly demonstrated that eGFR_diff can inform individuals with normal kidney function about their risk for cognitive decline and incident MCR. We further performed discordance analysis and highlighted that the eGFR_diff can embed more important clinical insights related to cognitive function over eGFR based on cystatin C, creatinine or their combination. With the growing utilization of cystatin C in clinical settings, clinicians can simultaneously obtain both cystatin C and creatinine levels. Our study sheds light on optimizing the utilization of these markers to improve clinical practice. Despite the fact that using the combination of cystatin C and creatinine is more accurate than either marker alone for estimating GFR [21, 22], relying solely on eGFR_cys-cr may be insufficient in capturing the complete risk information for cognitive impairment and MCR. Evaluating intraindividual eGFR_diff should continue to be strongly emphasized in clinical practice.

Strengths and limitations

To our knowledge, this is the first study to comprehensively reveal the longitudinal associations of eGFR_diff and kidney function measures based on creatinine or cystatin C with long-term cognitive trajectories and incident MCR, and we conducted multiple sensitivity analyses to confirm the robustness of our results. Our findings underscore the potential clinical value of eGFR_diff over eGFR_cys, eGFR_cr and eGFR_cys-cr alone, which may optimize the risk assessment and clinical management of cognitive impairment and MCR. However, several limitations also exist in our research. First, our study included only a Chinese population. This might restrict the generalization to other races and ethnicities. However, we applied race-free equations to re-estimate eGFR to verify our conclusion. Second, although we adjusted for several potential confounders, other unmeasured or residual confounders may still affect our findings. Furthermore, we used a relatively simple and limited assessment tool for cognitive function, which lacked the cut-off values of cognitive impairment, but the moderate–high validity of this test was reported compared with the Mini-Mental State Examination and the Clinical Dementia Rating Scale [54]. Finally, we included measurements of kidney function using cystatin C, creatinine, calculated eGFR, BUN and self-reported kidney disease. However, we did not collect data on albuminuria or other CKD-related indicators. This limitation could potentially affect a more accurate assessment of CKD.

CONCLUSIONS

This longitudinal study indicated that intraindividual differences between eGFR_cys and eGFR_cr were associated with long-term cognitive trajectories and incident MCR among middle-aged and older adults. The eGFR_diff could convey additional valuable information regarding the risk of cognitive decline and MCR beyond eGFR_cys, eGFR_cr or eGFR_cys-cr, which should be considered in clinical translation. Monitoring eGFR_diff may provide significant clinical utility for the early prevention of cognitive impairment and MCR, thus promoting healthy aging.

ACKNOWLEDGEMENTS

The National School of Development at Peking University provided the data in the China Health and Retirement Longitudinal Survey. We thank the staff of the China Health and Retirement Longitudinal Study team for their invaluable contributions.

FUNDING

Our work was funded by the National Natural Science Foundation of China (82073668 and 81872708 to L.T.). The funder had no role in the design and conduct of the study; in the collection, analysis and interpretation of the data; or in the preparation, review or approval of the manuscript. Each participant included in this study signed a written informed consent form before taking the survey. Ethical approval for all the CHARLS waves was granted from the Institutional Review Board (IRB) at Peking University. The IRB approval number for the main household survey, including anthropometrics, is IRB00001052-11015.

AUTHORS’ CONTRIBUTIONS

J.W., X.G. and L.T. contributed to the study conception and design. J.W., Z.W., Z.H. and Z.X. contributed to the data collection. J.W., Y.L., R.J. and X.Z. contributed to the data analysis and interpretation. J.W., Y.L., and L.T. contributed to manuscript writing and reviewing. X.G. and L.T. contributed to study supervision.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available in China Health and Retirement Longitudinal Study (http://charls.pku.edu.cn/). Materials are available upon request to the corresponding author.

CONFLICT OF INTEREST STATEMENT

The authors declare that they have no competing interests.

REFERENCES