Association of kidney function and brain health: A systematic review and meta-analysis of cohort studies

Objective: This study aimed to evaluate the bidirectional association between the kidney dysfunction and the brain health, including structural and functional abnormalities. Design: Systematic review and meta-analysis with network meta-analysis for outcomes with different estimated glomerular filtration rate (eGFR) ranges. Data sources: PubMed, Embase database, Cochrane library and Web of Science (up to Dec. 2021). Eligibility criteria for selecting studies: Longitudinal studies that provided evidence of the impact of kidney function estimated from eGFR and urine albumin-to- creatinine ratio (UACR) or chronic kidney disease (CKD) on structural and functional brain abnormalities, and those that provided evidence of the opposite relationship. Studies with study population mean age under 18 years old were excluded. Main outcome measures: Two independent reviewers screened the included studies, extracted the data, and assessed the risk of bias. We performed a random-effects meta-analysis and a network meta-analysis for outcomes with compatible data. We assessed the risk of bias using the Newcastle – Ottawa Quality Assessment Scale criteria (NOS). Subgroup and sensitivity analyses were conducted to explore heterogeneity in the meta-analyses. Inconsistency analyses using the node-splitting method were performed to confirm the results of network meta-analysis. Results: A total of 53 studies with 3037,357 participants were included in the current systematic review. Among these, 16 provided evidence of structural brain abnormalities, and 38 provided evidence of cognitive impairment and dementia. Analysis of evidence of categorical kidney function showed a positive association between kidney dysfunction and cerebral small vessel disease (cSVD) (relative risk (RR) 1.77, 95% confidence interval (CI) 1.40 – 2.24, I 2 = 0.0%), but such results were not found in the analyses of evidence where the kidney function was measured as a continuous variable. Meanwhile, analysis of 28 prior longitudinal studies with 194 compatible sets of data showed that the worse kidney function as categorical variables was related to a greater risk of global brain cognitive disorder (RR 1.28, 95% CI 1.20 – 1.36, I 2 = 82.5%). Conclusions: In this systematic review and meta-analysis, we found a positive association between CKD and functional brain disorders. However, the relationship between the kidney dysfunction and structural abnormalities in the brain remains controversial. As for the opposite relationship, structural brain abnormalities, especially cerebral microbleeds and silent infarction, but not functional brain abnormalities, are associated with worse renal function. In addition, a higher UACR, but not a lower eGFR, was associated with a higher risk of Alzheimer ’ s disease and vascular dementia.


Introduction
Brain health disorders, including structural and functional brain abnormalities and chronic kidney disease (CKD) are growing public health issues. Neurological disorders were the leading cause of disability-adjusted life years (DALYs, 276 million) and the second leading cause of death (9 million) in 2016 globally according to the Global Burden of Diseases study (GBD, Neurology Collaborators, 2016. Structural brain abnormality, also known as cerebral small vessel disease (cSVD) (Jokinen et al., 2020), is a frequent cause of stroke and the primary subtype of vascular cognitive impairment. The neuroimaging features of SVD include small subcortical infarcts, lacunes, white matter hyperintensities (WMH), enlarged perivascular spaces (EPVS), microbleeds, and brain atrophy (Wardlaw et al., 2013;van Harten et al., 2006). Functional brain abnormalities, including Alzheimer's disease, vascular dementia, and other dementias, are among the largest contributors to neurological DALYs. The age-standardized incidence and mortality rates per 100,000 population were 103.83 (95% UI,), respectively, for dementia in 2019 (Gao and Liu, 2021). Approximately 50 million people worldwide were living with dementia in 2018, and this number will more than triple to 152 million by 2050 (World Health Organization, 2012). CKD is usually characterized by the presence of persistent albuminuria or an estimated glomerular filtration rate (eGFR) of < 60 mL/min/ 1.73 m 2 . Notably, the DALYs due to impaired renal function nearly doubled from 1990 to 2019, reaching 76.5 million, with 3.16 million deaths globally in 2019 (Roth et al., 2020). CKD is increasingly recognized as an elevated risk factor for neurological disorders. Research has shown that the prevalence of brain health disorders is high (approximately 50%) in advanced CKD (Murray et al., 2006;Naganuma et al., 2011;Sarnak et al., 2013;Yokoyama et al., 2005), and microvasculopathy has been proposed to link these two conditions (Ito et al., 2009;Knopman, 2010;Weiner, 2008). Therefore, it is vital to target high-risk patients, define a window of opportunity to prevent CKD-induced brain health disorders, and achieve better surveillance.
No consensus has been reached regarding the relationship between CKD and structural or functional brain dysfunction. A study by Vilar et al. supported a positive association between kidney function and the incidence of cSVD (Vilar-Bergua et al., 2016), but these results were disputed in other studies (Aggarwal et al., 2019;Vemuri et al., 2017). Nevertheless, despite reports (Tamura et al., 2011) that worse renal function is a risk factor for dementia and cognitive impairment, another study (Helmer et al., 2011) also identified a null effect. Given this conflicting evidence, the relationship between CKD and structural and functional disorders in the brain remains unclear and requires further investigation. Here, we conducted a systematic review and meta-analysis of published research to elucidate these associations to provide an up-to-date understanding and refine primary prevention strategies.

Protocol registration
We registered the protocol of this systematic review with PROSPERO (CRD42021293834).

Study inclusion
The search for the bidirectional relationship between kidney function and brain health was conducted in the PubMed, Embase, Cochrane Library, and Web of Science databases to identify all relevant studies published up to Dec. 2021. Different types of cSVD, cognitive impairment, and dementia as well as their relationship with CKD were investigated in detail. This analysis included four cSVD: brain atrophy, WMH, cerebral microbleeds, and lacunar infarctions. Functional brain abnormalities were referred to as cognitive impairment and dementia, including all-cause dementia, Alzheimer's disease, and vascular dementia. Additional pertinent articles were supplemented by inspecting the references of the included articles. "This study was conducted according to the guidelines set out by Meta-analysis of Observational Studies in Epidemiology (Stroup et al., 2000) and the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (Knobloch et al., 2011).

Inclusion and exclusion criteria
Studies meeting the following inclusion criteria were included: (1) an original article published in English with participants' mean age over 18 years, (2) defined kidney function using eGFR, urine albumin-tocreatinine ratio (UACR), creatinine clearance (CCI) or other clearly defined measurements, or used physical diagnosis of kidney function (e. g., ICD-9), (3) defined cSVDs or cognitive impairment and dementia clearly, (4) measured the structural brain abnormalities using magnetic resonance imaging (MRI), regardless of scanner resolution (1.5 T, 3 T, 7 T), automated assessment/visual assessment, or sequence of scan, (5) cognitive impairment and dementia were identified using generally agreed tests (including the Montreal Cognitive Assessment, MoCA; Mini-Mental State Examination, MMSE; Clinical Dementia Rating score, CDR; Diagnostic and Statistical Manual of Mental Disorders, 4th edition, DSM-IV etc.), (6) provided quantitative measures of the bidirectional association between kidney function with cSVDs and cognitive impairment and dementia, and (7) prospective cohort epidemiological study designs. The exclusion criteria were as follows: (1) publication was a review, case report, animal study, or letter to the editor, (2) publication did not clearly define clinical outcomes, (3) duplicated data, and (4) the mean age of the study population was under 18 years. For the current analysis, the bidirectional relationships between kidney function and cSVDs, cognitive impairment, and dementia were measured using odds ratios (ORs), relative risks (RRs), incidence rate ratios (IRRs), or hazard ratios (HRs) with 95% confidence intervals (CIs).

Data extraction and quality assessment
Two investigators (XYT and YPH) independently extracted the data from the enrolled studies using the same method concerning study quality, population characteristics, basic diseases, and outcomes. CKD was defined as eGFR < 60 mL/min per 1.73 m 2 or CCl < 60 mL/min, or UACR > 30 mg/g. Dementia and cognitive impairment were defined based on cognitive tests or related guidelines ( Table 1). The data synthesized in our study are available directly from published articles or supplementary materials, and no primary data were collected.
The risk of bias among the included studies was assessed using the Newcastle-Ottawa Quality Assessment Scale (NOS) (Stang, 2010). Following the NOS guidelines, we rated the quality of the studies by awarding stars in each domain. XYT and YPH independently assessed the risk of bias for each study, which was cross-checked using JBZ and MAC. In case of disagreement, the investigators discussed with other authors to arrive at a consensus. The quality assessment results were scored on a scale of 0-9.

Statistical analysis
Heterogeneity between studies was evaluated using the I 2 metric and the variance between studies by Tau 2 . Random-effects models were used if I 2 > 50%, while the fixed-effects models were used if I 2 ≤ 50%. When necessary, we transformed (Qin et al., 2021;Yang et al., 2018) effect metrics derived from different studies to allow for pooled analysis. In this study, we considered OR, RR, IRR, and HR as RR to conduct pooled analyses of bidirectional associations between kidney function and brain health across all the included studies. If the studies had both unadjusted and covariate-adjusted RRs, the latter was selected. For continuous renal function data, we collected estimates and their standard errors (SEs) reported as "per 10 mL/min/1.73 m 2 decrease" when measured by eGFR and "per mg/g increase" when measured by UACR.
To define potentially confounding factors and risk of bias, we conducted subgroup and sensitivity analyses. Sensitivity analyses were conducted to assess the influence of each result on the pooled estimate. The Egger's asymmetry test (Egger et al., 1997) was conducted to evaluate potential publication bias. According to the Cochrane Handbook version 5.1.0 (Cumpston et al., 2019), as a rule of thumb, tests for funnel plot asymmetry should be used only when enough studies are included in the analysis. Thus, in the pooled meta-analyses of evidence with a small number of included studies, the Egger's asymmetry test was not applied. P values were two-tailed, and statistical significance was set at P < 0.05. Statistical analyses were performed using Stata version 12.0 (Stata Corporation, College Station, TX, USA).
For the network meta-analysis, studies were included if they provided kidney function measurements as categorical variables. To Table 1 Study characteristics and main results from studies assessing the association between kidney function and brain health.          compare the direct and indirect estimates for each comparison, the inconsistency of our results was confirmed using the node-splitting method (Dias et al., 2010) and its Bayesian P value. If the P value was less than 0.05, an inconsistency was considered to be detected. Rank probabilities of the effects of all available eGFR ranges were calculated and are shown visually using the bar plots of ranking probability for the network (Salanti et al., 2011). Network meta-analyses using Bayesian methods (Lumley, 2002) for indirect treatment comparisons were performed using JAGS and R software (version 3.3.3) with the gemtc package (version: 0.8-2) and rjags package (version: 4-6) in a random-effect model, as most of the head-to-head comparisons only included one set of data providing direct evidence. When a meta-analysis was not feasible, we reported the results of the narrative review.

Literature search outcomes and validity assessment
In the analysis of evidence on the bidirectional relationship between kidney function and cSVDs, we identified 534 potentially relevant reports, of which 197 were excluded because they were duplicates. The remaining 337 articles were subjected to title and abstract screening. A further 252 studies were removed, as they did not met the inclusion criteria (reviews, letters, conference abstracts, and independent studies). A total of 85 articles were found eligible for the full-text review and data assessment.  (Fig. 1).
In summary, 53 studies were included in the current study, of which one study (Jiménez-Balado et al., 2020) focused on kidney function changes and their relationship with the progression of cSVDs and cognitive decline. Table 1 provides an overview of the 53 eligible studies, and all studies were scored 5-9 in NOS quality assessment (supplementary material 3, table 1).
Previous studies have established a null association during the examination of the effect of CKD progression as dichotomous variable on cSVD (Peng et al., 2016;Sedaghat et al., 2018) of 3171 participants (RR 1.18, 95% CI 0.85-1.63, I 2 = 70.3%). However, these results might be biased because of the substantial heterogeneity among the studies. The results of subgroup analysis are shown in supplementary material 3, table 5.

The association between kidney function and cerebral volume
Data regarding the association between renal function and cerebral atrophy were incompatible with the meta-analyses, although four studies provided longitudinal evidence. Jiménez-Balado et al. (Jiménez-Balado et al., 2020) reported in 2020 that worsening eGFR had a null effect on deep WMH volume change during a 4.1 years follow-up of 360 Spanish participants (β − 0.21, 95% CI − 0.67 to 0.25). Barzilay (Barzilay et al., 2016) et al. also found that the baseline eGFR and eGFR change in patients with type 2 diabetes mellitus were not associated with any significant differences in the brain volume measurements. However, on the other hand, Khatri et al. (Khatri et al., 2007) reported in 2007 that both the baseline creatinine clearance (15-60 mL/min) and eGFR (15-60 mL/min) were connected to increased log-WMH volume (β 0.322, 95% CI 0.095-0.550; β 0.322, 95% CI 0.080-0.564 respectively). In addition, Vemuri et al. (Vemuri et al., 2017) showed that a 10% worsening in eGFR or in UACR was related to increasing WMH volume (β 2.849, SE 0.94; β 0.323, SE 0.10, respectively).
Three studies (Jiménez-Balado et al., 2020;Scheppach et al., 2020;Tamura et al., 2016) utilized the continuous UACR as a measure of kidney function. Scheppach et al. (Scheppach et al., 2020) found that the UACR increase was linked to a higher incidence of all-cause dementia in both 54-74-year-old midlife and 70-90-year-old elderly participants. In addition, Tamura et al. (Tamura et al., 2016) also indicated that the UACR doubling was associated with the risk of developing cognitive impairment in a national (the United States) sample of 19,399 adults. On the contrary, a study of 976 middle-aged adults from Jiménez-Balado et al. (Jiménez-Balado et al., 2020) revealed a null effect of the UACR incline on mild cognitive impairment (RR 1.54, 95% CI 0.92-2.58). The overall pooled risk ratio of the three studies was 1.02 (95% CI 1.00-1.04, I 2 = 77.1%), and the small number of included studies might have biased these findings (Fig. 6). Further subgroup analysis is shown in supplementary material 3, table 7.
The sub-group analysis is shown in supplementary material 3, table 8.

Network meta-analysis of the association between different eGFR range and cognitive impairment and dementia
Sixteen cohort studies (Bai et al., 2017;Feng et al., 2012;Gabin et al., 2019;Guerville et al., 2020;Helmer et al., 2011;Hiramatsu et al., 2020;Kurella et al., 2005;Kurella Tamura et al., 2020;Kuriyama et al., 2013;Miwa et al., 2014;Sasaki et al., 2011;Slinin et al., 2008;Takae et al., 2018;Tamura et al., 2011;Tseng et al., 2020;Wang et al., 2010) comprising 607909 participants were included in the network meta-analysis to examine 13 different eGFR ranges. Among these studies, 9 (Bai et al., 2017;Feng et al., 2012;Guerville et al., 2020;Kurella et al., 2005;Kurella Tamura et al., 2020;Kuriyama et al., 2013;Slinin et al., 2008;Tamura et al., 2011;Wang et al., 2010), 8 (Gabin et al., 2019Helmer et al., 2011;Hiramatsu et al., 2020;Kurella Tamura et al., 2020;Miwa et al., 2014;Sasaki et al., 2011;Takae et al., 2018;Tseng et al., 2020), 4 (Gabin et al., 2019;Helmer et al., 2011;Miwa et al., 2014;Takae et al., 2018), 4 (Gabin et al., 2019;Helmer et al., 2011;Miwa et al., 2014;Takae et al., 2018) were related to cognitive impairment, overall dementia, Alzheimer's disease, and vascular dementia, respectively (S Table 10). The network plot (Fig. 9) shows the association between different ranges of eGFR and the incidence of Fig. 3. Meta-analysis of evidence on association between continuous kidney function and cerebral small vessel diseases. A. Meta-analysis of evidence on association between continuous eGFR and cerebral small vessel diseases. B. Meta-analysis of evidence on association between continuous UACR and cerebral small vessel diseases. C. Meta-analysis of evidence on association between continuous eGFR change and cerebral small vessel diseases. Abbreviation: eGFR, estimated glomerular filtration rate; UACR, urine albumin-to-creatinine ratio; RR, risk ratio; CI, confidence interval. Where I 2 is the variation in effect estimates attributable to heterogeneity, overall is the pooled random effect estimate of all studies. Subtotal is the pooled random effects estimate of sub-group analysis studies. Weights are from random-effects analysis. %Weight is the weight assigned to each study, based on the inverse of the within-and between-study variance. The size of the grey boxes around the point estimates reflects the weight assigned to each study. Fig. 4. Meta-analysis of evidence on association between categorial kidney function and cerebral small vessel diseases. A. Meta-analysis of evidence on association between categorial kidney function and cerebral small vessel diseases according to different cerebral small vessel diseases. B. Meta-analysis of evidence on association between categorial kidney function and cerebral small vessel diseases according to kidney dysfunction method. C. Meta-analysis of evidence on association between categorial kidney function change and cerebral small vessel diseases according to different cerebral small vessel diseases. Abbreviation: eGFR, estimated glomerular filtration rate; UACR, urine albumin-to-creatinine ratio; RR, risk ratio; CI, confidence interval. Where I 2 is the variation in effect estimates attributable to heterogeneity, overall is the pooled random effect estimate of all studies. Subtotal is the pooled random effects estimate of sub-group analysis studies. Weights are from random-effects analysis. %Weight is the weight assigned to each study, based on the inverse of the within-and between-study variance. The size of the grey boxes around the point estimates reflects the weight assigned to each study.

Fig. 5.
Meta-analysis of evidence on association between continuous eGFR and cognitive impairment and dementia. A. Meta-analysis of evidence on association between continuous eGFR and cognitive impairment and dementia. B. Meta-analysis of evidence on association between continuous eGFR and cognitive impairment. C. Meta-analysis of evidence on association between continuous eGFR and dementia. D. Meta-analysis of evidence on association between continuous eGFR and Alzheimer dementia. Abbreviation: eGFR, estimated glomerular filtration rate; RR, risk ratio; CI, confidence interval. Where I 2 is the variation in effect estimates attributable to heterogeneity, overall is the pooled random effect estimate of all studies. Subtotal is the pooled random effects estimate of sub-group analysis studies. Weights are from random-effects analysis. %Weight is the weight assigned to each study, based on the inverse of the within-and between-study variance. The size of the grey boxes around the point estimates reflects the weight assigned to each study. Fig. 6. Meta-analysis of evidence on association between continuous UACR and cognitive impairment and dementia. Abbreviation: UACR, urine albumin-to-creatinine ratio; RR, risk ratio; CI, confidence interval. Where I 2 is the variation in effect estimates attributable to heterogeneity, overall is the pooled random effect estimate of all studies. Subtotal is the pooled random effects estimate of sub-group analysis studies. Weights are from randomeffects analysis. %Weight is the weight assigned to each study, based on the inverse of the within-and between-study variance. The size of the grey boxes around the point estimates reflects the weight assigned to each study.
Local inconsistency was evaluated by one comparison at a time in the node-splitting analysis of our study by separating the direct evidence of that comparison from the network of indirect evidence. In terms of different eGFR ranges, eight nodes were split, and the P values varied from 0.26 to 0.99, which meant that no statistically significant inconsistencies were detected.
The rank possibilities of different eGFR ranges are shown in supplementary material 3, Fig. 6. The eGFR of < 90 mL/min/1.73 m 2 owned the highest first rank possibility (38.78%), followed by ≥ 45 and 30-60 mL/min/1.73 m 2 , while the eGFR of ≥ 75 mL/min/1.73 m 2 showed 50.19% possibility to be the last risk factor of cognitive impairment/ dementia among the other ranges of eGFR (supplementary material 3, table 12). The eGFR range of ≥ 60 mL/min/1.73 m 2 held 25.21% and 29.19% possibility to be the last second and third at risk eGFR range.
3.8. Meta-analysis and literature review of the relationship between the brain health and kidney function

The effect of cSVD on the kidney function
Four studies (Bouchi et al., 2010;Kobayashi et al., 2010;Shima et al., 2016;Uzu et al., 2010) independently reported that the patients with cSVD (supplementary material 3, table 13), especially the silent cerebral infarction or microbleeds, were significantly at risk to develop the renal insufficiency (pooled RR 2.51, 95% CI 1.88-3.36, I 2 = 0.0%) (Fig. 10). Further subgroup analysis is shown in supplementary material 3, table 14. Notably, only few studies are available on evaluation of this opposite relationship, and hence, the robustness of the pooled estimate might be less.

The effect of cognitive impairment and dementia on kidney function
Currently, only one published article (Tamura et al., 2016) examined the effect of cognitive impairment on CKD progression. The recruited population consisted of 3883 participants from the Chronic Renal Insufficiency Cohort (CRIC) study, and the data revealed that the cognitive impairment had no influence on the incidence of developing end-stage renal disease (HR 1.07, 95% CI 0.87-1.03) or the risk of a 50% decline in baseline eGFR (HR 1.06, 95% CI 0.89-1.27). Fig. 7. Meta-analysis of evidence on association between categorial kidney function and cognitive impairment and dementia. A. Meta-analysis of evidence on association between categorial kidney function and cognitive impairment and dementia. B. Meta-analysis of evidence on association between categorial kidney function and cognitive impairment. C. Meta-analysis of evidence on association between categorial kidney function and dementia. D. Meta-analysis of evidence on association between categorial kidney function and Alzheimer dementia. E. Meta-analysis of evidence on association between categorial kidney function and vascular and dementia. Abbreviation: eGFR, estimated glomerular filtration rate; UACR, urine albumin-to-creatinine ratio; CCI, creatinine clearance; RR, risk ratio; CI, confidence interval. Where I 2 is the variation in effect estimates attributable to heterogeneity, overall is the pooled random effect estimate of all studies. Subtotal is the pooled random effects estimate of sub-group analysis studies. Weights are from random-effects analysis. %Weight is the weight assigned to each study, based on the inverse of the within-and between-study variance. The size of the grey boxes around the point estimates reflects the weight assigned to each study.

Principal findings
In this meta-analysis, we comprehensively examined the association between kidney dysfunction and structural and functional brain disorders by assessing accessible, compatible data from the longitudinal cohort studies. Structural brain disorders or cSVD were found to have no statistically significant association with the kidney function. We also found that the kidney dysfunction is related to cognitive impairment and dementia. These findings did not vary in subgroup analyses of older age (≥65 years) or middle-aged participants, consistent with most currently published studies.
In addition, our synthesized evidence showed results similar to those of a previous study (Deckers et al., 2017) that UACR might better reflect cognitive function in CKD patients than eGFR. In the analysis of categorical kidney function and dementia, the categorical eGFR was reported to be associated with cognitive impairment and all-cause dementia but not Alzheimer's disease or vascular dementia, whereas categorical UACR was found to be related to all these types of cognitive dysfunction. Our results indicate the critical role of UACR as a risk factor for cognitive impairment and dementia compared with eGFR. The feature of GFR estimation could be blamed for this unstable association between eGFR and functional brain health in our meta-analysis. The equation commonly used for GFR estimation is based on creatinine, which is widely used in routine clinical practice but can be substantially influenced by confounding factors, such as unusual muscle mass (Stevens and Levin, 2013), diet with high meat content, or dietary supplements containing creatine. In particular, reduced muscle mass is a concern in elderly patients, which may lead to an overestimation of GFR. Inaccurate determination of the GFR readings (Zaman et al., 2013) may obscure the potential association with dementia.
Intriguingly, evidence of the opposite relationship demonstrated a negative impact of structural brain abnormalities, cerebral microbleeds, and silent cerebral infarction on renal function. However, functional brain abnormalities had a null effect on kidney function in the pooled results of 2 relative longitudinal studies. This might be largely biased by the small amount of synthesized data. Furthermore, our results from the network meta-analysis demonstrated that a lower eGFR had a more statistically significant harmful effect on the brain cognitive function. In Fig. 8. Meta-analysis of evidence on association between categorial kidney function change and cognitive impairment and dementia. Abbreviation: RR, risk ratio; CI, confidence interval. Where I 2 is the variation in effect estimates attributable to heterogeneity, overall is the pooled random effect estimate of all studies. Subtotal is the pooled random effects estimate of sub-group analysis studies. Weights are from random-effects analysis. %Weight is the weight assigned to each study, based on the inverse of the within-and between-study variance. The size of the grey boxes around the point estimates reflects the weight assigned to each study. Fig. 9. Network plot for evidence on association between different eGFR range and cognitive impairment and dementia. Abbreviation: eGFR, estimated glomerular filtration rate. Each circle represents one eGFR range, and the size of circle and the thickness of connected lines indicate number of trials for each comparison. the rank possibility analysis, the results indicated that the riskiest eGFR range would be approximately ≥ 45-60 mL/min/1.73 m 2 . This implies that patients with stage 3 CKD are at a greater risk of developing the brain functional disorders. We also noticed that a few studies with less than three years of follow-up offered irregular findings. This is reasonable because the brain dysfunction usually does not develop within such a short period.

Underlying mechanisms
The pathophysiological mechanisms of the observed associations between the limited kidney function and global brain health are unclear, and a few plausible theories may, to a certain degree, elucidate these connections. First, scientists have concluded that the microvascular disease may be linked to two conditions. Branches of the anterior, middle, and posterior cerebral arteries (Ito et al., 2009;Knopman, 2010;Weiner, 2008) penetrating the brain tissues and juxtamedullary afferent arterioles in the kidneys are strain vessels that are frequently exposed to high pressure to provide a large pressure gradient over a short distance. Therefore, these vessels strictly rely on their autoregulation (Bugnicourt et al., 2013) to manage hemodynamic stress from the large supplying arteries. Studies using transcranial Doppler ultrasonography (Silvestrini et al., 2006) have proven that the cerebral microvascular hemodynamic impairment is commonly involved in patients with cognitive impairment. Meanwhile in CKD patients, studies have shown a greater risk of developing vascular diseases (Saran et al., 2020), such as arterial hypertension, than that in the healthy population. Therefore, simultaneous occurrence of the kidney and cerebral microvascular diseases may play a key role in these relationships.
Second, assumption had arisen that the blood-brain barrier disruption-induced albumin leakage (Knopman, 2010;Wada et al., 2007;Weiner, 2008) may partly explain the development of WMH. However, this theory has only been proven in animal models (Egashira et al., 2015) but not in humans. Meanwhile, CKD shares a similar pathological basis of gradual alterations in the kidney endothelial cells and glomeruli, leading to a glomerular leakage of serum albumin into the urine. Under these circumstances, it is plausible to hypothesize that the same vascular pathophysiological process that causes albumin leakage in both the cerebrovascular and kidney vasculature would be the main culprit in the correlation between the kidney dysfunction and the brain disorders. Moreover, the kidney function can be regarded as an indicator or window for observation of the pathological process in the brain.
Other potential mechanisms, including direct neuronal injury by uremic toxins, may also be responsible for the connection between the kidney function and the brain health. Elevated homocysteine levels due to CKD or uremia are associated with white matter lesions (Wright et al., 2005) as well as Alzheimer's disease (Seshadri, 2006) through direct prothrombotic effect (Seshadri et al., 2008) and endothelial inflammatory reactions (Hassan et al., 2004). Elevated levels of cystatin-C  in CKD and uremic patients have also been reported to be associated with the development of Alzheimer's disease. Other metabolic toxins accumulate in the renal dysfunction, including phosphorusto-fibroblast growth factor 23 (Fliser et al., 2007) and some guanidine compounds (De Deyn et al., 2009), such as creatinine, guanidine, guanidinosuccinic acid, and methylguanidine could also contribute to the brain function decline.
On the aspect of the opposite relation, structural brain disorders, especially the cerebral microbleeds and silent cerebral infarction, may also predispose patients to progression of the kidney disease. A possible mechanism that may contribute to this involves plasma asymmetric dimethylarginine (Pikula et al., 2009), an endogenous inhibitor of nitric oxide synthase, which was shown to be associated with cSVDs in the Framingham offspring study (Pikula et al., 2009). Studies have consistently hypothesized that the impaired cognitive function may also adversely influence the renal function owing to lower usage or incompliance with the CKD risk reduction strategies. It is plausible to assume that the cognition-deficit CKD patients may have difficulties in adhering to the doctor's orders of dietary potassium restriction or daily medication prescription, which could lead to a more rapid renal function loss. However, further studies are required to explore this opposite relationship.

Strengths & limitations
The strengths of this meta-analysis include its large sample size, long follow-up period, and the exclusion of all cross-sectional or retrospective studies. This meta-analysis was performed under the protocol in accordance with the PRISMA (Knobloch et al., 2011), with a comprehensive search strategy which was also reproducible. We not merely focused on the global cognitive function, but structural brain abnormalities as well. Moreover, our study added evidence to the difference between UACR and eGFR in revealing CKD patients' brain health. The quality of all studies was rated medium to high, therefore our meta-analysis provided good evidence for the association between kidney function and brain health. Fig. 10. Meta-analysis of evidence on the effect of cerebral small vessel disease on kidney function. Abbreviation: HR, hazard ratio; CI, confidence interval. Where I 2 is the variation in effect estimates attributable to heterogeneity, overall is the pooled random effect estimate of all studies. Subtotal is the pooled random effects estimate of sub-group analysis studies. Weights are from random-effects analysis. % Weight is the weight assigned to each study, based on the inverse of the within-and between-study variance. The size of the grey boxes around the point estimates reflects the weight assigned to each study.
This meta-analysis was limited by the small number of studies available for comparison and the lack of unadjusted original data from each study. Sociodemographic factors such as age, sex, health behaviors, and vascular risk factors were adjusted among the groups for analysis in each included article. Because the original data were not available for extraction, potential confounding factors, such as gender differences, blood pressure, primary cause of kidney function decline and effect of anemia or malnutrition secondary to CKD, probably biased our estimates. Secondly, The main variables of the kidney function in most studies were eGFR and UACR, while the association of other kidney function markers and brain health, such as serum creatinine, creatinine clearance, beta-2-microglobulin, and cystatin C levels, has not been fully explored in currently published articles. Recent evidence (Scheppach et al., 2020) has shown that the novel kidney markers, such as cystatin C and beta-2-microglobulin, which would not be influenced by non-kidney factors, including muscle mass decline, high meat component diet, and creatinine, could indicate kidney function more stably and better depict the association between the renal function and brain health than that by the eGFR. Therefore, additional studies are required in the future. Moreover, dose-response meta-analyses were not conducted in our study because of the small number of available studies that provided compatible data. In addition, we excluded studies published in languages other than English.

Future implications
Most of the studies included in our analysis concerning the renal function and the brain cognitive status used eGFR or UACR as categorical measures, and few studies used the continuous renal function markers. In this case, it would be plausible to expect that a nonlinear model might better mirror the relationship between the renal function and the brain cognitive status. Therefore, additional studies are needed to confirm whether the non-significant results in those studies using continuous eGFR/UACR were truly null.
Both the brain and the kidney consist of abundant vascular tissues and also have similar hemodynamic properties that can respond to diseases such as hypertension and diabetes mellitus (O'Rourke and Safar, 2005) in similar ways at the microscopic level. Although analyzed in subgroups of these illnesses, our results could not fully rule out these confounding factors because of the lack of unadjusted original cohort data. Thus, false-positive results might occur in the cause-effect relationship between the kidney function and brain health, as found in our study, due to the existence of probable indirect confounders between the exposure and outcome. As well, other possible confounding factors are of importance to be analyzed in future researches including gender differences, blood pressure, primary cause of kidney function decline and the potential effect of anemia or malnutrition secondary to CKD.etc. Mendelian randomization analysis may be helpful in future studies.
Last but not least, currently commonly used methods or questionnaires for cognitive evaluation in CKD patients may not fully examine the domains of impairment that are frequently found in these patients. The most commonly used cognitive test among our included studies was the MMSE, where 9 (Bai et al., 2017;Feng et al., 2012;Helmer et al., 2011;Koop-Nieuwelink et al., 2019;Kuriyama et al., 2013;Li et al., 2018;Miwa et al., 2014;Sasaki et al., 2011;Wang et al., 2010) of the 37 studies evaluated cognition ability. The MMSE was well validated and thoroughly designed, but may be potentially disadvantaged in detecting early and mild cognitive impairment. However, the results of this meta-analysis showed that the cognitive impairment is more closely related to digressed renal function than dementia. Thus, the use of the MMSE may not properly supervise the mildest stages of cognitive deficits, as it is often found in patients with CKD. The utilization of the MMSE in patients with CKD is also concerning because the MMSE does not assess any aspect of executive function (Hachinski et al., 2006), which is the domain that has been proven to be more strongly associated with CKD progression than the other cognitive domains because its deficits are linked to vascular risk factors and are often noted in patients with vascular dementia. The trial-making test forms B (TMT B) (Llinàs-Reglà et al., 2017) and the Category Fluency test (Welsh et al., 1994), and tests of executive function are therefore recommended for use in examinations of patients with CKD. However, to date, only one prospective cohort study has provided evidence of TMT B (Tollitt et al., 2021). Further research is needed to identify the most suitable cognitive test to assess cognitive ability in patients with CKD.

Public Health Implications
As our results proposed, decreased renal function may, to a certain extent, worsen brain cognitive function; thus, our study suggests that the management of modifiable factors in the pathway from the kidney function progression to functional neurodegenerative change is among the keys for prevention. Consequently, preservation of the kidney function to prevent dementia could be just as vital in the clinical care of older patients as it is for middle-aged patients. Given the large burden of structural and functional brain abnormalities in patients with CKD, cognitive screening and brain MRI should be performed for all patients with CKD. It is important for clinicians to select appropriate cognitive screening tests for CKD patients to avoid missing cases, and executive function should be considered as one of the domains assessed in adult CKD patients with suspected cognitive impairment.

Conclusion
Our meta-analysis is among the most extensive studies examining the association between the kidney dysfunction and functional and structural brain abnormalities. We conclude that worsening kidney function is associated with a decline in cognitive function, but not with the overall brain structural disorders. In addition, a higher UACR, but not eGFR, was associated with a higher risk of Alzheimer's disease and vascular dementia. Individuals with stage 3 CKD have a higher risk of developing cognitive brain dysfunction. Structural brain abnormalities, especially cerebral microbleeds and silent infarctions, might have harmful effects on the renal function. Further, the effects of brain abnormalities on the kidney function remain unclear. Hence, future longitudinal studies are still needed to elucidate the causal relationships and explore the window of prevention for the occurrence of these brain abnormalities in CKD patients.

Ethical approval
Ethical approval for this evidence synthesis was not required.

Funding
This work was supported by the National Natural Science Foundation of China (grant numbers 82270866, 82070851), the Beijing Municipal Administration of Hospitals' Youth Program (grant number QML20170204), and Excellent Talents in the Dongcheng District of Beijing. The authors are solely responsible for the design and conduct of this study, all study analyses, the drafting and editing of the manuscript, and its final contents.

Contributors
XT and YPH contributed equally to this work. JBZ is corresponding authors. XT, YPH, YHC, HJG, HX, IP, YSQ and JYZ have full access to all the data in this study and take full responsibility as guarantors for the integrity of the data and the accuracy of the data analysis. XT and YPH contributed to studies selection, data extraction, data analyses, and manuscript drafting. XT, YPH contributed to data analyses, data interpretation, and manuscript drafting. JBZ, XT and YPH contributed to study design, data interpretation, and final approval of the manuscript.
MAC revised this manuscript. The corresponding author attests that all listed authors meet authorship criteria and that no others meeting the criteria have been omitted.

Copyright/license for publication
The authors plan to disseminate these findings to appropriate audiences such as academia, clinicians, policy makers, and the general public through various channels including press release, social media, or enewsletter.

Transparency statement
The lead author, JBZ, affirms that the manuscript is an honest, accurate, and transparent account of the study being reported; that no important aspects of the study have been omitted.