Is hyperuricemia an independent risk factor for new-onset chronic kidney disease?: a systematic review and meta-analysis based on observational cohort studies

A literature search was performed using electronic databases Medline Ovid/Medline (1948 to present), PubMed/Medline, Embase, and ISI Web/Web of Science for published studies from January 1970 to September 2013. Key words used were: “hyperuricemia”, “uric acid elevated”, “hyperuricaemia”, “uricacidemia”, and “uricacidaemia”; and “CKD” or “chronic kidney disease” or “chronic renal dysfunction” or “chronic renal insufficiency”. The “related items” function was used to broaden the search. The electronic search was up to September 2013 with no limitations regarding the type of publication. Studies were screened by two reviewers (L.L. and C.Y.) independently with the following inclusion criteria: (1) report of the association between serum uric acid (SUA) and new-onset CKD; (2) studies of observational cohort design; and (3) report of the hazard ratio (HR) or odds ratio (OR) in cohort studies with 95% confidence intervals (95% CI) or sufficient information to calculate these figures. Exclusion criteria were: (1) studies published in languages other than English; and (2) studies reporting on patients with end-stage renal disease or acute renal impairment or patients requiring dialysis. The definition of CKD was: eGFR <60 mL/min per 1.73 m² for more than 3 months or the presence of albuminuria (kidney damage) for more than 3 months, according to the National Kidney Foundation Kidney Disease Outcomes and Quality Initiative (KDOQI) CKD guidelines [12]. Studies that used other definitions for CKD, such as serum creatinine levels, were excluded. If more than one study reported duplicated material, the most recent study was included in the systematic review. Reference lists of all included studies were reviewed for any additional relevant studies. The whole process of this systematic review was conducted in strict accordance with the criteria of the moose checklist to ensure the quality of this study (Additional file 1).

KDOQI CKD guidelines were used for the definition of CKD severity: Stage 1, GFR >90 mL/min per 1.73 m² with microalbuminuria or macroalbuminuria; Stage 2, GFR 60–89 mL/min per 1.73 m² with microalbuminuria or macroalbuminuria; Stage 3, GFR 30–59 mL/min per 1.73 m²; and Stage 4 GFR 15–29 mL/min per 1.73 m²[12]. Two reviewers (L.L. and C.Y.) retrieved the following data from the included studies independently: author, year published, country where study was performed, material source, cut-off level for hyperuricemia, sample size (number of patients and controls), SUA levels in CKD and non-CKD groups (mean ± SD), variables adjusted for calculation in the analysis, and adjusted OR or HR estimates with corresponding 95% CI. If the study classified participants according to quartiles of uric acid level, the highest quartile was taken to indicate hyperuricemia. Data reporting on the OR or HR of hyperuricemia and new-onset CKD or CKD risk were retrieved and combined.

The Newcastle–Ottawa Scale (NOS) was used for meta-analysis of observational studies in epidemiology for assessment of quality [13]. Risk of bias was assessed among the cohort studies on three aspects: (1) selection of participants (containing four domains); (2) comparability (one domain); and (3) outcome measure (containing three domains). Each domain was rated as “Yes”, “No”, or “Unclear”. Each quality domain was categorized as low risk for bias (Yes) when adequate data were reported to assess the study quality or the study meeting the criteria, and high risk (No) when the study reported adequate data but did not meet the criteria for that quality domain. Studies that did not report data to assess quality were categorized as “Unclear” and thus potentially at high risk of bias. “Yes” was scored 1 and “No” or “Unclear” were scored “0”. A quality bar was plotted for each domain to examine the limitations of the studies. Studies of high quality were defined as score >5 points.

Statistical analysis was performed in accordance with The Cochrane Library handbook and the Quality of Reporting of Meta-Analyses guidelines [14, 15]. Different measures for risk estimates from multiply analyses, such as HR (Cox regression) or OR (logistic regression), were extracted from studies; only the most adjusted risk estimates that examined associations between SUA level or hyperuricemia and new-onset CKD were extracted and pooled. We used the random effects model for pooled analysis of adjusted odds ratios (AOR) because of anticipated statistical heterogeneity, but the fixed effects model was also used to ensure robustness of the model chosen and susceptibility to outliers. Heterogeneity among studies was assessed using the χ ² test. P-values <0.05 were considered statistically significant.

Meta-regression analysis was conducted to explore heterogeneity among studies. Possible sources of heterogeneity included participant-specific characteristics such as age, sex, location, and source of participants; and study-specific characteristics such as data type for calculating HR or OR (continuous variables or dichotomous variables), sample size, study quality, and definitions used for hyperuricemia. Funnel plots and Egger regression tests were used to investigate publication bias. We used RevMan version 5.1 (The Cochrane Collaboration, The Nordic Cochrane Centre, Copenhagen, Denmark) and Stata version 11.0 for Windows (Stata Corp., College Station, TX, USA) for data analysis and graph formation.

We identified 1662 relevant citations in the combined search of online electronic databases (partial overlap) and 21 additional citations from the reference lists of relevant articles. After removing duplicated records, 790 studies were screened by examining titles and abstracts, and 129 relevant studies after primary selection were reviewed by reading the complete text for eligibility. There were 41 studies that reported on associations between SUA and renal disease, and these were assessed in detail. Articles with a cross-sectional or case–control style, a different definition of CKD or insufficient data were excluded. Finally, 13 studies that met inclusion criteria were included in our meta-analysis (Figure 1).

The characteristics of the 13 studies are detailed in Table 1. A total of 190,718 participants were included, with sample sizes ranging from 324 to 94,422. Four studies included the United States or populations of Western countries, and nine studies were conducted in Asian populations. Nine studies reported the relationship between increasing SUA level and the development of new-onset CKD during follow-up of non-CKD participants (AOR calculated on continuous variables per 1 mg/dL), and seven studies presented risk estimates for associations between hyperuricemia and the incidence of CKD (AOR calculated on dichotomous variables). The NOS quality assessment of the included cohort studies is listed in the Additional file 2. Eight cohort studies showed a comparatively high quality (NOS >5).

The main limitation observed in the majority of studies was unclear reporting of dropout rates, as many studies did not mention uncompleted follow-ups.

Four studies provided the AOR for uric acid level as a risk factor for the development of CKD in males and females separately. Three studies grouped participants in quartiles according to uric acid level (the OR calculated for the highest quartile compared with the lowest quartile as a reference was retrieved for pooled analysis). Eleven studies defined CKD as the development of eGFR <60 mL/min per 1.73 m², while the remaining two used albuminuria for the definition.

With regard to choice of participants, 11 studies used population-based participants who underwent health examinations, while the source in the other two studies was hospital-based participants with diabetes mellitus [18, 24].

Different cut-off levels were used for defining hyperuricemia (Table 1). An SUA level >7.0 mg/dL in men or >6.0 mg/dL in women was used in the majority of studies.

Adjustments made for potential confounders of at least three factors were mentioned in all 13 studies.

Seven studies provided the adjusted OR of increasing risk of new-onset CKD per 1 mg/dL SUA elevated. The results of pooled analysis revealed a significant positive association between SUA and incidence of CKD (summary OR, 1.06; 95% CI, 1.04–1.08) with evidence of significant heterogeneity among studies (Q = 34.76, P < 0.001, I ² = 83%) (Figure 2).

On examining study variability using meta-regression analysis, we found that the population age was the cause of heterogeneity. In one study that used an elderly population [17], the OR was comparatively higher than the other studies. By excluding this particular study, we observed that a clear positive association was retained (summary OR, 1.05; 95% CI, 1.02–1.07) while heterogeneity was greatly reduced (Q = 6.93, P = 0.23, I ² = 28%) (Figure 3).

Six studies provided risk ratios for the association between hyperuricemia and new-onset CKD in non-CKD populations during follow-up. Results of summary analysis of these six studies showed a significant association between hyperuricemia and increasing risk of the incidence of CKD (summary OR, 2.59; 95% CI, 2.14–3.13) in a random effects model. Moderate heterogeneity was found among studies (Q = 9.55; P = 0.09; I ² = 61.2%) (Figure 4).

Results of meta-regression analysis indicated that heterogeneity was caused by follow-up year and sex proportion. One study from Japan only included male volunteers and had a follow-up of 18 years and reported a comparatively higher OR and contributed 44% of heterogeneity [20]. After excluding this study, no significant heterogeneity across the remaining studies was detected (Q = 0.480, P = 0.31, I ² = 17%). For these five other studies, the association between hyperuricemia and the onset of CKD was still significantly positive (summary OR, 2.11; 95% CI, 1.70–2.61).

Results of subgroup analysis according to sources of participants, geographical area, sex and follow-up year are shown in Table 2.

For geographical location, positive results were observed between SUA and the development of new-onset CKD in both Asian countries (summary OR, 1.05; 95% CI, 1.03–1.08) and Western countries (summary OR, 1.17; 95% CI, 1.11–1.22).

Three studies provided results for males and females, and the results of pooled analysis by sex indicated a similar significant positive association between hyperuricemia and new-onset CKD in both males (summary OR, 1.43; 95% CI, 1.05–1.94.) and females (summary OR, 1.21; 95% CI, 1.04–1.41) (Figure 5). A study by Mok et al. reported that a stronger association beween hyperuricemia and CKD for in females than males, a finding contrary to results from the other studies, contributed to the heterogeneity of pooled analysis [21].

The subgroup of follow-up year revealed that the longer time period for observation, the greater the risk of the development of new-onset CKD in patients with elevated uric acid levels.

Funnel plots revealed no evidence of publication bias for associations between increasing SUA and new-onset CKD and the relationship between hyperuricemia and new-onset CKD (Figure 6a, Figure 6b). Results of the Begg adjusted rank correlation test for detecting publication bias were negative (SUA, P = 0.576; hyperuricemia, 0.452). The Egger regression asymmetry test showed no significance for either SUA (P = 0.176) or hyperuricemia (P = 0.977).