Selection and characteristics of studies
The pertinent articles obtained from the literature search and the reference lists were introduced into software EndnoteX9 for management. A total of 3356 publications were identified (Figure 1). After removing duplicates we excluded 1710 citations, based on the screening of titles and abstracts we excluded 1479 citations, and after a thorough assessment of the full-text, we excluded 58 citations. Among 109 eligible full text articles, 87 papers were excluded because of: – duplicate results from the same group of authors were reported (n =4 ); a retrospective study was carried out (n =7); – with less than 3 months of follow-up (n=8 );– an outcome different from the one of interest was considered (n = 33); – with a sample size of fewer than 100 subjects (n=14 ); – definition of urate cut-offs or range limits were lacking (n =16 ); and – did not report adjusted risk estimates for CVD incidence(n =5). Finally, we identified 22 studies [33–54] that met our inclusion criteria. The methodological quality of all the included studies was moderate to high, with NOS score varying from 7 to 8 and a median of 8. Table 1 display the characteristics and effect sizes of the 22 studies of uric acid level and risk of CVD incidence. 22 prospective studies with a total number of 421802 participants were entered in the meta-analysis. Six studied health check-ups and outpatients[34, 35, 42, 43, 50], and the other sixteen studies were restricted to patients with cardiac vascular disease (CVD) [33, 36–41, 44, 45, 47–49, 51–54]. The datum was removed from 18 cohort studies and 4 cross-section studies[39, 48–50]. And no disagreements between the two reviewers regarding study inclusion. Of the twenty-two trials, three were conducted primarily in the United States[33, 41, 44]. Twelve studies were made in Asian countries [34, 38, 39, 42, 43, 46–52] and seven studies were from European countries [35–37, 40, 45, 53, 54]. The number of participants ranged from 324 in a study by Palazzuoli et al to 197144 in the cohort study by Shin Kawasoe et al . Twenty-one studies included both men and women. One study included only men, and none of the studies included only women. Nine studies utilized a lower cut-off value to define hyperuricemia for women opposed to men [34, 36, 38, 40, 42, 43, 51, 52, 54]. Five studies reported gender-specific outcome for CHD mortality[33, 34, 36, 43, 51]. Thirteen studies provided hazard ratio estimates[24, 35, 37, 38, 40–42, 44, 45, 47, 52–54], eight studies provided odds ratio[36, 39, 43, 46, 48–51] and one provided relative risk ratio.
Hyperuricemia and cardiac event incidence
The relationship between uric acid and cardiac event incidence could be presented in three ways: all-cause mortality (four studies, 11288 patients), mortality from CVDs (ten studies, 34897 patients), and other CVD incidences (eight studies, 375617 patients). All studies provided categorical data and none provided continuous data. Meta-analysis with a random-effects model suggested that there was problems with heterogeneity between hyperuricemia and cardiac event incidence (RR=1.30; 95% CI:1.24-1.37; I2=69.5%, P=0.00,Tau2=0.0062) (Figure 2-1). In mortality from CVDs, the pooled result suggested a consistent relationship between hyperuricemia and CVD mortality (RR=1.37; 95% CI:1.29-1.45; I2=31.4%, P=0.157) (Figure 2-2).
A subgroup analysis of cardiac event incidence needed to be done to investigate the heterogeneity of the studies. The selected subsets including the year of publication (Year before 2010, Year after 2010), sample size (<1000, 1000-10000, 10001-100000, >100000), epidemiological research methods (cohort study, cross-sectional study), risk indicator (HR, OR, RR), study location (USA, Asian, Europe), sample source (patient had CVD versus Healthy population), CVD categories(coronary artery disease,hypertension,heart failure,atrial fibrillation,stroke) ,and outcome indicators (all-cause mortality, cardiovascular disease mortality, other cardiovascular morbidities). As the odds ratios could be deemed to be accurate estimates of risk ratios, we, therefore, utilized RR as the common measure of association across studies. Published year before 2010(RR=1.32, 95% CI: 1.23-1.41; I2=24.8%, P=0.22), sample size less than one thousand(RR=1.24,95%CI: 1.12-1.37; I2=0.0%, P=0.56), lived in Europe (RR=1.39, 95% CI: 1.31-1.48, I2=41.7%, P=0.11), coronary artery disease(RR=1.27, 95% CI: 1.19-1.37, I2=42.2%, P=0.11), hypertension (RR=1.26, 95% CI: 1.19-1.33, I2=30.7%, P=0.22) and cardiovascular disease mortality (RR=1.37, 95% CI: 1.29-1.45, I2=31.4%, P=0.16) subgroups showed that hyperuricemia was associated with an increased risk of cardiac event incidence (table 2).
Meta-regression analysis showed that mean SUA level (OR=1.04, p=0.047, Adj R2=23.95%,tau2 = 0.0053, in the univariate models) and CVD prevalence at baseline (OR=1.005, p=0.09, Adj R2=24.2%,tau2 = 0.005, in the univariate models)
may effect the heterogeneity in this study for analyses of cardiac event incidence. Similarly, meta-regression analysis also showed that Study location and outcome indicators (p=0.002, I2=27.3%,Adj R2=83.72%,tau2 = 0.0011, in the multivariate models)may contribute to the heterogeneity in this study(table 3).
Publication Bias Assessment
Funnel plots for CVD incidence were visually examined (Figure 3-1). No evidence of publication bias for studies was noted in the funnel plot, Egger regression asymmetry test (p=0.15), or Begg's test (p=0.26). Then we undertook a sensitivity analysis using public trim and fill method (Figure 3-2), which conservatively imputes hypothetical negative unpublished studies to mirror the progressive studies that cause funnel plot asymmetry. Four imputed studies for categorical data were needed to produce produce symmetrical funnel plots. The fail-safe number was 3384, indicating that 3384 ''negative'' studies would be necessary to increase the P-value for the meta-analysis of above 0.05.