High prevalence of chronic kidney disease in Taiwan

In Taiwan, the incidence rate and prevalence throughout all stages of kidney disease were approximately 27,210 per million person-years and 154,600 per million people in the last decade [1]. According to the United States Renal Data System (USRDS) annual report in 2020 [2], Taiwan has the highest prevalence of end-stage renal disease (ESRD) (3,587 per million population) and the highest prevalence of dialysis (3,429 per million population) worldwide [3, 4]. In addition, the USRDS in 2020 [2] reported that the prevalence of ESRD is highest in the age group of 45–64 years (25,156 per million population), compared to 20–44 years (5,290 per million population), 65–74 years (17,194 per million population), ≥75 years (15,066 per million population). Due to improved management for CKD patients, life expectancy [5, 6] and social burden of care [7] are both expected to increase in Taiwan.

Measures of chronic kidney disease

Renal disease is assessed by the degree of injured kidney filtration rate [8, 9], and CKD is relatively asymptomatic until the end stage [10]. The most commonly used measure is estimated glomerular filtration rate (eGFR) by serum creatinine, which evaluates the degree to which the kidneys remove waste from the blood. According to eGFR, CKD is divided into six categories by the international guideline group Kidney Disease Improving Global Outcomes (KDIGO) [11]: early CKD of stage 1 (eGFR of > 90 mL/min/1.73 m²), stage 2 (eGFR of 60–89 mL/min/1.73 m²), stages 3a (eGFR of 45–59 mL/min/1.73 m²), and 3b (eGFR of 30–44 mL/min/1.73 m²), late CKD stage 4 (eGFR of 15–29 mL/min/1.73 m²), and stage 5 ESRD (eGFR < 15 mL/min/1.73 m²). In addition, eGFR is modified by age, sex, and ethnic group [12–14].

Cognitive impairment in chronic kidney disease

Theoretically, the brain and kidney have similar microvascular structures and are susceptible to hemodynamic fluctuations [15–17]. Recent studies suggested that cognitive impairment is associated with cardiovascular risk factors, such as hypertension, hyperlipidemia [18, 19], and diabetes mellitus [20, 21]. Similar to the brain, the kidney receives a large amount of blood flow: 1.0–1.2 liters per minute per 1.73 m² of body surface area, approximately 20–25% of the total cardiac output [22]. When renal function deteriorates, uremic toxins and waste from blood are not removed, and the build-up of toxins impairs brain function [23]. A recent study [24] showed that CKD mainly affects the cortical synchronization of neurons in electroencephalography. Dementia is recognized as a significantly common phenomenon in ESRD, affecting as high as 40% of patients [25]. Studies have reported that ESRD patients treated with hemodialysis exhibited cognitive impairment in the domains of orientation, attention, memory, construction, and executive function [26, 27]. Additionally, ESRD patients who received kidney transplant showed a higher incidence of dementia [28]. While the vast majority of studies which investigated patients with both renal function injury and cognitive function focused on late CKD and ESRD populations, to date, few studies have evaluated cognitive impairment in early CKD stages [29].

Mild cognitive impairment (MCI), an intermediate cognitive state between normal and full-blown dementia, is characterized by problems with orientation, attention, judgment, and memory [30, 31], and has been reported to be associated with cardiovascular risk factors in the last two decades. Recently, there have been increasing studies supporting the association between CKD and MCI. Two recent meta-analyses [32, 33] showed that impaired eGFR is associated with cognitive impairment and can be present in the early and late stages of CKD. A systematic review [29] included three relevant studies on CKD patients <65-years-old, which found that these patients had slow processing speed [34], impaired verbal learning [35, 36], and impaired working memory [35] in moderate stages of CKD. A survey of Japanese CKD patients [37] demonstrated that female participants with eGFR <90 mL/min/1.73 m² showed a faster rate of cognitive decline than those with eGFR ≥90 mL/min/1.73 m².

Quality of life in chronic kidney disease

The term “quality of life” can be traced back to the definition in 1948 by the World Health Organization (WHO) as a state of perception of “physical, mental, and social well-being, and not merely the absence of disease” [38]. Prior meta-analysis [39] assessed the utility-based quality of life with the 12-item or 36-item Short Form Health Survey (SF-12 or SF-36), EuroQol Group’s EQ-5D, and other instruments, which suggested that dialysis is significantly associated with reduced quality of life for ESRD patients. However, few studies have investigated the association between quality of life and the early and moderate stages of CKD. We found that these studies exhibited inconsistent results: (1) a cross-sectional study of ≥60 year-old participants in North California [40] showed no relationship between eGFR and quality of life scores in physical and mental dimensions after adjustment for medical comorbidities; (2) a cross-sectional study including ≥50 year-old Irish participants [41] showed that reduced scores in quality of life are associated with eGFR by serum cystatin rather than creatinine; and (3) a recent cross-sectional study [42] enrolled 2,255 older adults in European countries and found that early CKD at stages 3a and 3b, defined by the BIS equation, is associated with lower quality of life on the EQ-Visual Analogue Scale (EQ-VAS). To our knowledge, there is a lack of relevant research on ethnic Han Chinese as well.

Study rationale and aims

Cognitive impairment or dementia is associated with increased physical disability [43, 44] and worsened caregiver burden [45]. While The World Health Organization (WHO) defined a society in which the proportion of people aged ≥65 years is 7% or higher is known as an “aging society,” 14% or higher is regarded as an “aged society,” and the prevalence of older adults in Taiwan exceeded 14% in 2018. Since Taiwan has become an aged society and has the highest worldwide prevalence of CKD and ESRD [2], assessment of cognition in CKD is of paramount importance. In this study, we hypothesized that the high prevalence of cognitive impairment in older adults with CKD in Taiwan, and perception of the physical and mental well-being of older adults was worsened in the early and moderate stages of CKD. In this cross-sectional study, we aimed to (1) obtain an overview of cognitive impairment and quality of life at each stage of CKD in the Han Chinese population, and (2) assess the attributable effect of poor kidney filtration function, after adjusting for demographic factors, on multiple cognitive domains and dimensions of quality of life among community-based healthy older adults in Taiwan. Additionally, we performed a validation study to examine the robustness of the association between eGFR, cognitive impairment, and quality of life. The existence of collinearity in our data analysis was likely since multiple domains of cognition and multiple dimensions of life quality were assessed. We replaced the conventional regression models with partial least squares (PLS) regressions to deal with potential collinearity and confirm robustness.

Study design and participants

The Nutrition and Health Survey in Taiwan (NAHSIT) 2013–2016 [46] was a nationwide cross-sectional study in Taiwan, which was conducted from January 1, 2013 to December 31, 2016. The primary objective of conducting NAHSIT 2013–2016 by the Health Promotion and Administration, Ministry of Health and Welfare in Taiwan was to determine the nutritional status of the Taiwanese population, which was used as a reference for nutrition and health policymaking in a government entity. The NAHSIT involved a systematic and stratified sampling process that covered all 359 townships and city districts in Taiwan, and was divided into eight strata according to population density and geographical area. The content of the survey included (1) behavioral indicators, such as cognition and quality of life assessment; (2) health outcome indicators, such as cardiovascular disease and kidney disease; and (3) laboratory tests of clinical chemistry, such as serum creatinine, lipid profiles, and blood glucose. Data collection was for the primary objective of the study. All participants were visited once during the study period. Our study was a secondary analysis of data derived from NAHSIT 2013–2016. Data for participants aged ≥65 years who completed the MMSE and QoL questionnaire of the 12-item Short Form Health Survey (SF-12) were collected. Participants who had no laboratory examination of renal function were excluded. This study was approved by the Institutional Review Board of Biomedical Science Research, Academia Sinica, Taiwan (AS-IRB01-13067) and the Research Ethics Committee, National Health Research Institutes, Taiwan (EC1020110). Written informed consent was obtained from all participants.

Cognitive and life quality measures

In Taiwan, the most widely used cognitive assessment tool is the culturally adapted traditional Chinese version of the Mini-Mental State examination (MMSE) [47]. The norms have been validated in the Taiwanese population [48] and other Han Chinese populations [49, 50]. The traditional Chinese version of the MMSE includes questions on orientation to time (5 points), orientation to place (5 points), registration (3 points), calculation (5 points), memory recall (3 points), language (2 points), repetition (1 point), and complex commands (6 points), and the MMSE score ranges from 0 (worst) to 30 (best) [51, 52]. For the Taiwanese population, the norms [48] indicated 2 cutoff scores for determining cognitive impairment: a value of <27 for the literate and another value of <16 for the illiterate, respectively.

The most well-validated tool for assessment of quality of life for the Taiwanese population is the Traditional Chinese version of the SF-36 [53, 54]. Despite being valid and equivalent for Han Chinese and other ethnic groups, SF-36 is limited by its length and time-consuming characteristics. The traditional Chinese version of SF-12, an abbreviated form of SF-36, is validated as equivalent to SF-36 for assessing the dimensions of physical and mental health [55]. The SF-12 questionnaire includes five dimensions of concepts: role-physical (RP), vitality (VT), social functioning (SF), role-emotional (RE), and mental health (MH). Each concept was transformed into a scale ranging from 0 to 100, with a mean distribution of 50 and a standard deviation of 10. A higher score signifies a perception of a better quality of life [56, 57].

Data collection process

Door-to-door visits were conducted by trained interviewers to collect demographic characteristics; culturally adapted traditional Chinese version of the MMSE [48] and SF-12 were administered. After informed consent was obtained from participants, phlebotomy was conducted for laboratory assessments of blood urea, nitrogen, creatinine, blood glucose, HbA1c, and lipid profiles of total cholesterol, low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), and triglycerides. Blood pressure measurements were performed by trained public health nurses and after participants rested for 5 minutes with the arm placed at the same height as the heart. A standard sphygmomanometer with an appropriate cuff size was used. Blood pressure measurements were recorded three times, and a fourth measurement was performed if two of the three BP measurements varied by ≥ 10 mmHg. Diabetes mellitus was defined as fasting glucose ≥126 mg/dl or HbA1c level ≥6.5%, in accordance with the standard of the American Diabetic Association [58]. Hypertension was defined as systolic blood pressure (SBP) ≥140 mmHg or diastolic BP (DBP) ≥90 mmHg according to the American Heart Association standard [59].

Definition of chronic kidney disease groups

A previous study [60] indicated poor agreement between the modification of diet in renal disease (MDRD) and the Cockcroft-Gault equations for assessing CKD in cognitively impaired older adults. In Taiwan, citizens are mono-ethnic, and approximately 97% of the Taiwanese are Han Chinese [61]. For older Chinese adults, two studies [62, 63] concluded that eGFR derived from the Berlin Initiative Study (BIS) equation achieved superior accuracy and lower rates of misclassification of CKD stages compared to the MDRD and Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equations. Because the enrolled participants were healthy older adults in the NAHSIT 2013–2016, eGFR was obtained by the BIS equation (eGFR = 3736 × creatinine^0.87 × Age^0.95 × 0.82 [if female]) [64, 65]. All participants were categorized into three CKD groups according to the derived eGFR: CKD stage 1 (group 1; eGFR > 90 mL/min/1.73 m²), stage 2 (group 2; eGFR of 60–89 mL/min/1.73 m²), stages 3a and 3b (group 3; eGFR of 30–59 mL/min/1.73 m²), and stages 4–5 (group 4; eGFR of 0–29 mL/min/1.73 m²). Since our study targeted the healthy older population, the number of late CKD stages 4–5 was expected to be small, and therefore Stages 4 and 5 were grouped into one category. In addition to eGFR, we confirmed the CKD risk group stages by urine albumin-to-creatinine ratio (ACR). Compromised renal function allows albumin to pass into the urine. According to the KDIGO [11], urine ACR <30 mg/g is defined as normal to mildly increased albuminuria, urine ACR of 30–300 mg/g as moderately increased albuminuria, and urine ACR >300 as severely increased albuminuria.

Statistical analysis

Demographic characteristics, eGFR, scores on each domain of MMSE, and SF-12 among the four CKD groups were assessed with one-way analysis of variance for continuous variables and with chi-squared or Fisher’s exact test for categorical variables. The P trend for the four CKD groups was estimated using generalized linear models for continuous variables and the Cochran-Armitage trend test for categorical variables. Generalized linear mixed models (GLMMs), containing both fixed and random effects, were applied for data analyses. Because of the study design, which involved a systemic and stratified sampling from the nationwide population in Taiwan and measurements of multiple related factors, a subject-specific model with a random intercept was used. We described this model in the following matrix formula Y_i = X_i β + R_i + E_i, where Y_i denotes the matrix containing a set of measures of all items of the MMSE and SF-12 on the ith older adult participant as dependent variables, X_i denotes the matrix containing a set of intercepts, independent variables of eGFR (per 30 mL/min/1.73 m²), and the covariates, β denotes the vector of estimated regression parameters, R_i denotes an intercept random effect associated with the ith older adult participant, which has a normal distribution and is independent and identically distributed (i.i.d.); and E_i denotes random error. The covariates used in the GLMMs were none (model 1), a set of variables with age, sex, and interaction of age and sex (model 2), a set of variables with age, sex, and interaction of age and sex, and education years (model 3), and a set of variables with age, sex, and interaction of age and sex, education years, and the characteristics that were imbalanced (defined as P < 0.05) among the four CKD groups (Model 4). All statistical analyses were performed using SAS software (version 9.4; SAS Institute, Cary, NC, USA). Applying the Bonferroni correction with the formula p ≤ α/m, where α was defined as 0.05, and the number of groups for multiple comparisons was 4. In our analysis, a two-tailed p value < 0.0125 (0.05/4) was considered statistically significant.

Validation study

To verify the association between renal function injury, cognitive impairment, and quality of life, we swapped the factors of eGFR in older adults and measured all items of the MMSE and SF-12 in the regression analysis. Conventionally, all items of the MMSE and SF-12 were employed as dependent variables (or response variables), and eGFR was used as an independent variable (or explanatory variable). Instead, we conducted partial least squares (PLS) regressions with the following formula: Y = XB + B₀, where Y denotes the response matrix of eGFR, X denotes the factor matrix of all items of the MMSE and SF-12, B denotes the matrix of factor loadings, and B₀ denotes error terms. Spearman’s rank correlation tests were used to examine the correlations among all items of the MMSE and SF-12. One major advantage of PLS regression is that it can easily manage the multicollinearity problem in conditions when there are numerous highly correlated explanatory variables [66, 67]. PLS regressions can help identify the optimal linear combinations of explanatory variables by projecting them to new spaces. The PLS regression was performed using the PROC PLS procedure implemented in SAS 9.4 software, with a factor number of 2 and principal components regression (PCR) to obtain factor loadings. Variables with a factor loading of > |0.3| were considered significant [68].

Participant characteristics

A total of 497 participants were enrolled in this study. Their mean age was 72.2 ± 6.2 years, 213 (42.9%) were female, and the mean score of global MMSE was 26.8 ± 3.4. The prevalence of MCI was 12.7%. The demographic characteristics and laboratory test results of the four CKD groups are shown in Table 1. Of the included participants, 66 (13.3%) had CKD stage 1, 303 (61.0%) had CKD stage 2, 119 (23.9%) had CKD stages 3a and 3b, and 9 (1.8%) had CKD stages 4–5. In groups 1, 2, 3 and 4, the average age evaluations were 68.5 ± 3.8, 71.1 ± 5.2, 76.9 ± 6.6, and 74.3 ± 8.2 years, respectively (P < 0.0001); proportions of female participants were 69.7%, 42.9%, 27.6%, and 44.4%, respectively (P < 0.0001); and eGFR values were 100.3 ± 7.9, 74.2 ± 8.6, 47.7 ± 8.1, and 23.9 ± 4.6 mL/min/1.73 m², respectively. Regarding sociodemographic characteristics, years of education were lower in groups 3 and 4 (P = 0.0023), and individual income was not significantly different among all CKD groups (P = 0.1435). Medical comorbidities, including hypertension, hyperlipidemia, and diabetes, were not significantly different among the four groups. The systolic and diastolic blood pressures of the four CKD groups were within normal limits, and no significant difference was observed in the mean arterial pressure. In serum laboratory testing, no significant difference was observed in the levels of total cholesterol, LDL-C, triglycerides, and fasting glucose. The levels of HbA1c and HDL-C were not homogenous among the four groups of CKD; however, the mean HbA1c and HDL-C levels were within normal limits.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Renal function and mental function health status (N = 497).

https://doi.org/10.1371/journal.pone.0264915.t001

Mini-Mental State Examination (MMSE) and life quality assessment

Our assessment of cognitive function with MMSE scores and quality of life for all participants is shown in Table 2. Compared to groups 1 and 2, groups 3 and 4 showed lower scores for each domain of the MMSE. On average, the global MMSE scores for all participants in groups 1, 2, 3, and 4 were 27.4 ± 2.4, 27.4 ± 2.8, 25.3 ± 4.7, and 24.9 ± 3.4, respectively (P < 0.0001 and trend P < 0.0001). The scores for the domains of calculation (P < 0.0001 and trend P < 0.0001), memory recall (P = 0.0018 and trend P = 0.0012), and complex commands (P = 0.0105 and trend P = 0.0047) differed significantly among the four groups. While the scores for the quality of life questionnaire, including dimensions of VT, SF, RE, and MH, did not differ significantly among the four groups, the RP score was significantly impaired in the late stages of CKD. The stratification analysis by sex is shown in S1 Table. Similar to the analysis for all participants, the scores of global MMSE, and the domains of calculation, memory recall, and complex commands were significantly impaired in stages 3a and 3b for both female and male participants. In addition, scores of domains of orientation to place, registration, and memory were significantly impaired in stages 3a and 3b for females.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Mini-Mental State examination and quality of life assessment of all participants (N = 497).

https://doi.org/10.1371/journal.pone.0264915.t002

Association between cognitive impairment and renal function decline with generalized linear mixed models (GLMMs)

The association between impaired eGFR and each dimension of the MMSE and quality of life questionnaire was examined in adjusted GLMMs (Table 3). The relationships among eGFR, age, and sex for global MMSE (Fig 1) and each domain of the MMSE (Fig 2) were plotted. Lowering eGFR per 30 mL/min/1.73 m² was significantly associated with impaired global MMSE score (β = −0.963, standard error [SE] = 0.239, adjusted P < 0.0001), domains of orientation to time (β = −0.161, SE = 0.456, adjusted P = 0.0004), and of calculation (β = −0.397, SE = 0.109, adjusted P = 0.0003). For quality of life assessment, lowering eGFR per 30 mL/min/1.73 m² was associated with impaired scores of RP (β = −2.387, SE = 0.687, adjusted P = 0.0006), but no obvious changes in the dimensions of VT, SF, RE, and MH (Fig 3). The estimated regression parameters β of eGFR in unadjusted and adjusted GLMMs (Models 1–4) are shown in Table 4.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Global Mini-Mental State Examination (MMSE) score by age, sex, and chronic kidney disease (CKD) stages.

https://doi.org/10.1371/journal.pone.0264915.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Each category of Mini-Mental State Examination (MMSE) score by age, sex, and chronic kidney disease (CKD) stages.

https://doi.org/10.1371/journal.pone.0264915.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Quality of life questionnaires by age, sex, and chronic kidney disease (CKD) stages.

https://doi.org/10.1371/journal.pone.0264915.g003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Generalized linear mixed model estimates of Mini-Mental State examination and quality of life (adjusted by age and sex).

https://doi.org/10.1371/journal.pone.0264915.t003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Comparison of different models for the relationship among eGFR estimates, MMSE, and quality of life.

https://doi.org/10.1371/journal.pone.0264915.t004

Validation study with partial least squares procedures

The Spearman’s correlation coefficients for all explanatory variables are listed in S2 Table). While all items in the MMSE were very weakly correlated, except for the weak association between calculation and complex commands (ρ = 0.374), all dimensions in SF-12 were weakly correlated, except for the moderate association between vitality and mental health (ρ = 0.566). The results of the validation study are listed in Table 5. Using PLS regression with PCR, we extracted two principal component factors from the dimensions of the MMSE and quality of life questionnaire. Factors 1 and 2 accounted for 2.3% and 0.5% of the total variation in eGFR (the dependent variable), respectively. Compared with Factor 1, Factor 2 was very weak in explaining the variation in eGFR. Factors 1 and 2 accounted for 23.5% and 18.2% of the total variation in the PCR model effects, respectively. For Factor 1, the independent variables orientation to time (0.337), orientation to place (0.397), registration (0.324), calculation (0.318), and complex commands (0.354) had factor loadings of >|0.3| and were significant in explaining the variation in eGFR. In contrast, for factor 2, the independent variables RP (0.397), VT (0.411), SF (0.339), RE (0.394), and MH (0.409) had high factor loadings, and they were very weak in explaining the variation in eGFR. These results are compatible with our findings from GLMMs.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Validation with partial least squares (PLS) procedures.

https://doi.org/10.1371/journal.pone.0264915.t005