A systematic review and meta-analysis of the association between uric acid levels and chronic kidney disease

The function of uric acid (UA) in the genesis and evolution of chronic kidney disease (CKD) has motivated numerous studies, but the results remain inconclusive. We sought to conduct a systematic review and meta-analysis of cohort studies aiming to analyze the association of UA levels with the incidence and progression of CKD. Pubmed/Medline, Lilacs/Bireme and Web of Science were searched to identify eligible studies, following the PRISMA protocol. Data were presented for CKD incidence and progression separately. For the meta-analysis, studies with data stratified by subgroups according to serum UA levels were selected. The inverse variance-weighted random effects model was used to generate a combined effect estimate. Meta-regressions were performed to identify the causes of heterogeneity. The Newcastle–Ottawa Scale was used to assess the risk of bias. The publication bias was tested by funnel plot and Egger’s test. Eighteen CKD incidence studies (n = 398,663) and six CKD progression studies (n = 13,575) were included. An inverse relationship was observed between UA levels and protection from CKD incidence and progression. Lower UA levels were protective for the risk of CKD incidence (RR 0.65 [95% CI 0.56–0.75]) and progression (RR 0.55 [95% CI 0.44–0.68]). UA seems to be implicated both in the genesis of CKD and its evolution.


Selection of studies.
After searching the databases, duplicate studies were excluded. Next, a refining was carried out to select the studies related to the theme addressed through the reading of titles and abstracts.
Studies were included if they met the following criteria: longitudinal cohort studies that evaluated the role of UA in the progression and/or incidence of CKD and had a follow-up of at least one year.
To reduce any confounding variables that may affect the association between CKD and UA, we excluded: studies with animals, children, adolescents, pregnant women and kidney transplant recipients, cross-sectional or case-control studies (to reduce the reverse causation, bias from pre-existing CKD on UA levels), studies addressing acute kidney disease and/or specific types of kidney disease (contrast-induced nephropathy or persistent posttreatment, renal failure after percutaneous coronary intervention, immunoglobulin A nephropathy, autosomal dominant polycystic kidney disease, lupus nephritis), studies on drug effectiveness, studies whose outcome was mortality and studies that assessed risk factors in general for progression of CKD (obesity dyslipidemia, diabetes, hypertension, etc.), whose focus was not to evaluate the role of uric acid.
As they were different measures, the outcomes progression and incidence of CKD were presented separately. Incidence of CKD was defined as individuals who were free of CKD at baseline (glomerular filtration rate (GFR) ≥ 60 mL/min/1.73 m 2 ) but experienced a decline in GFR to less than 60 mL/min/1.73 m 2 during follow-up. Progression was defined as decline in GFR and/or end-stage renal failure, requiring renal replacement therapy. Data extraction. The following data were extracted from the selected studies: author's name, date of publication (year), study design (follow up time), characteristics of the studied population, GFR estimation equation, exposure variable, main outcome, adjustment for possible confounding factors in multivariate analyses and evidence quality analysis.
Evidence quality analysis. The Newcastle-Ottawa Scale (NOS) was used to assess the risk of bias for cohort studies. We used three items to assess study quality: (1) selection of participants (including four domains); (2) comparability of study results (including one domain); and (3) outcome (including two domains) 24 . Each domain had a rating of "yes, " "no", or "unclear. " If there was adequate data against a domain in the included study and met the criteria, it was classified as low risk of bias (yes). Conversely, a domain was classified as high risk of bias if adequate information was not available (no) or not enough data was available to make an assessment (unclear). "Yes" was scored as "1", and "no" or "unclear" was scored as "0. " Scores were tallied up to calculate the final cumulative score. A study was considered high quality if the cumulative score was ≥ 4, and low quality if < 4 25 .

Statistical analyses.
For the quantitative analysis, studies with data stratified by groups according to serum UA levels (quartiles or quintiles) were selected. To evaluate the gradient, the groups with the lowest levels of UA were compared to the other groups (For quartiles: Q1 vs. Q2; Q1 vs. Q3 and Q1 vs. Q4. For quintiles: Q1 vs. Q2; Q1 vs. Q3; Q1 vs. Q4; Q1 vs. Q5).
For incidence studies, the absolute number of incident and non-incident cases in each group was used to calculate the relative risk (RR), with a 95% confidence interval and p values. For progression studies, the absolute number of cases that progressed and did not progress in each group was used to calculate the RR, with a 95% confidence interval and p values.
The effect estimates rate ratios were combined in the meta-analysis to calculate the overall risk estimate. A random-effects model was used for analysis to account for the variation of real effects across studies. Heterogeneity was quantified using the I 2 statistic, which describes the proportion of total variation in the study estimates due to heterogeneity 26 . The degree of heterogeneity was assessed from the I 2 statistic using the following thresholds for interpretation: (1) 0% to 30%: marginal heterogeneity; (2) 30% to 50%: moderate heterogeneity; (3) 50% to 75%: substantial heterogeneity; and (4) 75% to 100%: represents considerable heterogeneity 26 . The estimation of each study and the standard error (SE) generated a combined estimate, graphically represented by the forest plot.
Meta-regressions (univariate and multivariate) were performed to identify the causes of heterogeneity. The following variables were investigated: sex (difference in percentage of men between comparison groups), serum

Results
Selection of studies. We identified 6889 publications at the electronic databases. After exclusion by duplicity, 4134 studies remained. In the paired selection through the screening by the titles and abstracts, 4084 more articles that did not fit our scope were eliminated. Among the 50 studies selected for full reading, we excluded 26 that did not meet the eligibility criteria. Finally, 24 studies were included and stratified into two groups: 18 6,15,16,18,20,[27][28][29][30][31][32][33][34][35][36][37][38][39] related to the incidence of CKD in the general population and six 5,17,21,40-42 on CKD progression. For the quantitative analysis, nine incidence studies 6,16,28,30,32,34,35,37,39 and three progression studies 5,17,21 were included. Figure 1 represents the flowchart of identification and selection of studies for systematic review and meta-analysis. All studies were longitudinal observational cohort studies, 18 prospective and six retrospective. Of the total, 14 studies were of Asian origin and 10 were conducted in Western countries between 2008 and 2019. In all, 412,238 participants were identified, 398,663 in incidence studies and 13,575 in progression studies. Regarding the confounding factors used in the adjustments of the analyzed studies, there was a diversity of variables: hypertension and/or blood pressure values; diabetes and/or related tests; dyslipidemia and/or lipidogram; use of medications (antihypertensive, diuretics, UA-lowering and lipid-lowering); proteinuria or albumin-creatinine ratio and GFR. All studies had a cumulative quality score of ≥ four and were thus considered high quality (Tables 1, 2, Supplementary table).
Synthesis of the results. UA levels and CKD incidence. An inverse relationship was observed between UA levels and protection for the CKD incidence (Fig. 2 . In other words, the risk of CKD incidence increased with increasing levels of UA, and overall, the risk for incidence was 1.54. The meta-analysis showed considerable heterogeneity by the I 2 statistic (I 2 = 97.7%). Regarding meta-regression (Table 3), in the univariate analysis, the age difference between the groups was the only variable with a p value < 0.200, therefore, the multivariate analysis was not necessary. We conclude that increasing the age difference between the groups increased the risk ratio between UA and the incidence of CKD (Coeff β 1.19 [95% CI 1.12-1.26]).
Funnel plot analysis qualitatively showed an asymmetric shape (Fig. 3), indicating the possibility of publication bias for the association between UA levels and CKD incidence. However, the Egger test showed no indication of publication bias (p = 0.249).
UA levels and CKD progression. An inverse relationship was observed between UA levels and protection from CKD progression (Fig. 4). Lower UA levels were protective for the risk of CKD progression (RR 0. 55  . In other words, the risk of CKD progression increased with increasing UA levels, and overall, the risk for progression was 1.81. The meta-analysis showed considerable heterogeneity by the I 2 statistic (I 2 = 96.4%). Regarding meta-regression (Table 4), in the multivariate analysis no variable was significantly associated with the risk relationship between UA and CKD progression (Table 4).
Funnel plot analysis and Egger's test were not performed for CKD progression because few studies were selected.

Discussion
The present study found a gradient relationship between UA levels and CKD, with lower UA levels were protective for the risk of CKD incidence and progression. In addition, increasing the age difference between groups (with different levels of UA) increased the risk between UA and incidence of CKD.
UA levels are an independent predictor of the development of CKD and its progression. Some studies have shown that the prevalence of hyperuricaemia gradually increases with the decrease in renal function, being 10 times higher in stages 3-5 than in stage 1 43 . This relationship highlights a potential benefit of UA screening in different phases of CKD 44 .
A retrospective cohort involving 13,133 healthy adults (without hypertension, diabetes, obesity or CKD) showed that increasing UA levels doubled the risk of incident CKD 45 . Similarly, a Japanese retrospective cohort with 5,507 adults followed-up for an average period of 4.6 years found a positive association of hyperuricaemia (≥ 7 mg/dl) with the incidence of CKD (adjusted HR = 1.58 [95% CI = 1.21-2.07]), but with no effect on its progression (HR = 1.08 [95% CI = 0.73-1.59]) 46 . Jalal 47 made a critical review of observational and experimental studies on the potential effect of UA reduction therapy on the prevention of the incidence and progression of CKD and concluded that UA participates in inflammation and evolution of CKD. The authors acknowledge, however, that such conclusions are controversial because they are based on small studies and without placebo group.
Thus, two recently published double-blind, placebo-controlled, multicenter randomized trial 48,49 concluded the use of allopurinol was not beneficial to prevent the progression of pre-existing kidney disease, with no statistical difference between using or not using the UA-lowering agents. On the other hand, a recent systematic review and meta-analysis of 11 randomized clinical trials, including 4,277 participants with CKD, suggested that UA-lowering therapy preserves GFR so that xanthine oxidoreductase inhibitors could improve renal outcomes 50 . This finding may indicate a different role for UA in the pathophysiology of CKD in addition to a nitrogen slag.
The definition of the role of UA is still complex, especially due to the coexistence of risk factors such as hypertension, chronic inflammation and pathophysiological peculiarities of CKD. Recent studies have shown that soluble UA exhibits a behavioral duality acting as pro-oxidant within the cell and antioxidant in the extracellular environment 10,51,52 .
In relation to the pro-oxidant effect, the UA stimulates the generation of reactive oxygen species culminating in oxidative stress and mitochondrial damage, endothelial dysfunction with activation of the renal-angiotensin-aldosterone system, reduction of nitric oxide bioavailability and afferent arteriolopathy [53][54][55] . Furthermore, UA induces the activation of dependent or independent pathways of NLRP3 inflamassome with release of www.nature.com/scientificreports/ pro-inflammatory cytokines (IL1β and IL18), necro-inflammation and interstitial fibrosis 56 . Such pathophysiological mechanisms may justify the role of UA in the incidence and progression of CKD. There were several strengths to this systematic review: 1. it is one of the first studies to gather evidence from observational studies in the search for an association of UA levels, not only with the incidence of CKD, but also with its progression to terminal stages, thus contemplating all stages of CKD; 2. the literature search included several large databases with the search criteria designed to identify as many relevant articles as possible; 3. a proportion of the study selection, data extraction and quality assessment were conducted in duplicate by separate reviewers to reduce reporting bias; 4. all studies were classified as high quality and encourages future researches related to the use of UA reduction therapies for preventing CKD in at-risk populations.
There are some limitations to our study that need to be taken into consideration. First, the studies presented heterogeneous sample sizes, which was minimized with the use of appropriate statistical tests in the metaanalysis. Secondly, different definitions were observed for the final outcome (GFR < 60, decline in GFR, rapid decline of GFR) and for the independent variable (UA levels). Subgroup analysis based on the UA levels (quartiles or quintiles) was conducted to assess whether the presentation of exposure contributed to differences in study results. This analysis showed similar results regardless of the method of presenting the exposure. Thirdly, different forms of GFR estimation (MDRD, CKD-EPI, Cockcroft-Gault, Japanese equation, equation based on insulin clearance, iohexol technique) were adopted. Finally, there was some heterogeneity among risk estimates from the included studies, possibly due to some bias at the study level. However, the variables included in the meta-regression were not sufficient to explain the heterogeneity between studies. Other important variables that interfere in the UA metabolism could not be included, as they were not present in all studies, such as gout, basal GFR, diet components, use of UA-lowering agents.
In conclusion, the present study provides evidence that lower UA levels were protective for the risk of CKD incidence and progression. Increasing the age difference between the groups increased the risk between UA and the incidence of CKD.