The ASSURE study (ClinicalTrials.gov identifier NCT03616769) is a multicenter prospective cohort study. The first cross-sectional survey was conducted in 2011. Eligible participants were followed up from November 2011 to June 2018 (mean follow-up was 68.71 ± 11.35 months). During the follow-up period, 76 subjects had missing data, and 97 were not compliant. Thus, the study sample comprised 3,047 eligible participants (1,625 male and 1,422 female participants) older than or equal to 35 years (mean age 60.2 ± 10.4 years). A total of hospitalized subjects were consecutively enrolled from the Cardiology Department of Beijing University Affiliated and of Shanghai Tongji University–affiliated hospitals. All subjects were under treatment for CVD. The inclusion criterion was HAR. The exclusion criteria were severe congestive heart failure and severe renal failure. Severe congestive heart failure was defined as greater than or equal to cardiac functional classification 3 formulated by the New York Heart Association (NYHA). Severe renal failure was defined as an estimated glomerular filtration rate (eGFR) <30 mL/min per 1.73 m² (Figure 1). All participants provided written informed consent for this study, which was approved by the Ethics Committee of Tongji University.

Hospitalized myocardial infarction (MI) was classified as definite or probable based on chest pain symptoms, cardiac enzyme levels, electrocardiographic (ECG) findings, or angioplasty. Coronary heart disease (CHD) was determined to be present if there was (1) ECG evidence of a prior MI, (2) prior coronary artery bypass surgery or angioplasty, (3) coronary angiography showing CHD, (4) symptoms of angina and ECG revealing myocardial ischemia performance or laboratory tests showing increased cardiac enzymes and excluding other types of disease, or (5) a self-reported history of a physician-diagnosed heart attack. Coronary artery disease (CAD) death was classified as “definite” based on chest pain symptoms, hospital records, and medical history.

Cardiovascular events are composed of cardiac events, including non-fatal MI, unstable angina, and coronary revascularization procedures, during the follow-up period. The exclusion criteria were stable angina (>6 months), revascularization procedure for CAD (>6 months), and MI (>6 months).

In this study, cardiovascular death was the only cardiac event death considered. Medical records and death certificates for all patients who had an event were obtained and validated by a cardiologist. Death was confirmed from hospital records or by contact with participants and their families. All materials were reviewed independently by five senior physicians of the cohort study to confirm the cause of death.

Hyperuricemia refers to consuming a normal-purpurin diet, with two determinations on different days: SUA ≥420 μmol/L or 7 mg/dL (male), ≥ 357 μmol/L or 6 mg/dL (female) (15). Hypouricemia refers to consuming a normal-purpurin diet, with two determinations on different days: SUA ≤178 μmol/L or ≤3 mg/dL (male) and SUA ≤149 μmol/L or ≤2.5 mg/dL (female) (16, 17).

The Framingham risk score (FRS) was calculated based on coronary risk factors, including age, sex, total cholesterol (TC), low-density lipoprotein cholesterol (LDL-C), hypertension, and smoking status, according to the National Cholesterol Education Program Adult Treatment Panel III algorithm (18). The calculated total scores were used to estimate the 10-year CHD risk in participants without previous CVD and when an FRS > 20% or between 10 and 20% was considered HAR (18).

The diagnostic criteria for hypertension were as follows: patients taking antihypertensive medication or systolic blood pressure (SBP) ≥140 mm Hg or diastolic blood pressure ≥90 mm Hg (1 mm Hg = 133 kPa) or both. The criterion for hypertension was a single hypertensive disease.

Similarly, the criterion for dyslipidemia was a single dyslipidemic disease. The definition of dyslipidemia is abnormalities in serum levels of lipids, including overproduction or deficiency. Abnormal serum lipid profiles include high TC, high triglycerides (TGs), low HDL-C, and elevated LDL-C.

Qualified ultrasonographers measured ankle and brachial SBPs and Doppler ultrasound (Nicolet Vascular, Elite 100R, USA) was used to measure SBP in the bilateral brachial, tibial, and dorsal pedal arteries. The ankle–brachial index (ABI) has been shown to be a powerful independent marker of cardiovascular risk, with predictive ability similar to that of the Framingham criteria (19, 20).

A questionnaire was designed to collect from all participants information about general characteristics, diagnosis, medical history and related factors, medical treatment, and biochemical examination.

Factors were assessed for all subjects, including daily habits, medical histories, and blood samples. These blood samples were used to measure TC, TGs, HDL-C, LDL-C, serum creatinine (Cr), SUA, glucose, and fasting plasma glucose. Blood samples were drawn from an antecubital vein into a Vacutainer using a 19-gauge needle, and serum concentrations were assessed with commercially available kits. The GFR was calculated as GFR (mL/min per 1.73 m²) = 186 × creatinine^−1.154 × age^−0.203 and GFR (mL/min per 1.73 m²) = 142 × creatinine^−1.154 × age^−0.203 for males and females, respectively. Glucose was measured using a Hitachi 717 analyzer (Roche) with enzymatic reagents from Roche. The presence of symptomatic peripheral arterial disease (PAD) was evaluated by the Rose questionnaire (21). A previous MI or ischemic stroke was documented by hospital records. Physical examination data included body mass index (BMI), blood pressure, and ABI. Severe congestive heart failure was defined as greater than or equal to cardiac functional classification 3 formulated by the NYHA. Severe renal failure was defined as an eGFR < 30 mL/min per 1.73 m².

For follow-up, participants were contacted by physicians of the cohort study at annual intervals. Outcomes were obtained based on the annual phone interviews, 6-year follow-up examinations, hospital records, and death records. The primary clinical event endpoints of this study were estimated all-cause mortality and cardiovascular mortality. The secondary endpoints were CHD and CHD risk equivalent, including PAD, stroke, diabetes mellitus (DM), and CVD. Follow-up time was the number of years from the baseline (first) visit. For subjects who had more than one event, all clinical events were considered in the analysis.

All analyses were performed using the R statistical package (version 3.6.2) [http://www.r-project.org (22)]. Continuous variables are expressed as the mean ± SD and categorical variables as percentages. Continuous and categorical variable difference comparisons were made by independent-samples ANOVA (analysis of variance) and the χ² test, as appropriate. The Kruskal–Wallis test was used to evaluate non-normally distributed continuous variables. A P < 0.05 was considered statistically significant. Due to a skewed distribution, TC, TG, HDL-C, LDL-C, Cr, and SUA were logarithm-transformed (log) for analyses. Crude death from all-cause and CVD was assessed by SUA stratification. Cumulative event rates were estimated with Kaplan-Meier survival curves, and probability values were calculated with the log-rank test. Cox proportional hazard analyses were performed to test the association of SUA and death from all causes or CVD. A Cox regression model was adjusted for potential confounders, including age, sex, duration of hypertension, smoking status, dyslipidemia history, chronic renal insufficiency history, DM history, percutaneous coronary angioplasty (PTCA) history, coronary artery bypass grafting (CABG) history, PAD history, MI history, ischemic stroke history, hypertension, ABI, FRS, eGFR, diuretics, center, and year of screening examination. Potential confounding variables with P < 0.10 were adjusted for multivariate analysis. Restricted cubic splines (RCSs) with knots were used to further explore the shape of the dose–response relationship between SUA level and the hazard ratio (HR) of all-cause mortality and CVD mortality. Knots were at the 5th, 50th, 90th, and quartile of the SUA distribution. Missing values were handled by K-means clustering imputation. All P-values were 2-tailed, and a value < 0.05 was considered significant. The optimized prognostic cutoff value of SUA was analyzed with X-Tile Software (version 3.6.1, https://medicine.yale.edu/lab/rimm/research/software). SUA transformed by RCS was added to the Cox regression mode as an independent variable, and all-cause CVD mortality scores were calculated. A survival receiver operating characteristic (ROC) curve was used to determine the predictive effect of SUA level and its variability on all-cause mortality and CVD mortality. The ROC curve and area under the curve (AUC) were implemented with a regression model predicting the score.

A total of 3,220 eligible participants with available baseline data were enrolled. Among the participants, 76 individuals had missing follow-up data because of changing the telephone number or family address during follow-up; 97 subjects were not compliant (Figure 1). Therefore, the study sample actually comprised 3,047 participants. Based on careful calculation, the missing participants did not significantly affect the results. Our research showed hyperuricemia and hypouricemia prevalences of 18.1 and 16.7%, respectively, and the average SUA for the entire cohort was 322.65 ± 33.12 μmol/L (standard deviation). According to SUA level, values were subdivided into 0–242.0, 242.0–312.0, 312.0–386.75, and more than 386.75 μmol/L subgroups. As shown in Table 1, among all variables examined, abnormal SUA was associated with hypertension, a higher proportion of male subjects, older age, a higher level of Cr, diuretic use, and chronic renal insufficiency. Regarding subjects with HAR, there was no significant difference for ischemic stroke, MI history, CABG history, PTCA history, DM, dyslipidemia, and smoking. Of note, significant differences were observed for female participants with FRS risk stratification. However, FRS risk stratification by subgroup was not significant in male participants. In contrast with the reference group, the participants were more likely to be treated with ACEI.

After multivariable adjustment, Cox regression models revealed that compared with the SUA (242.0–312.0) subgroup, all-cause mortality was 24.5, 19.1, 22.9, and 35.9% in the SUA (0–242.0), SUA (312.0–386.75), and SUA >386.75 μmol/L subgroups, respectively (Table 2); CVD mortality rates were 12.2, 10.8, 13.4 and 18.3%, respectively. HRs of all-cause mortality and cardiovascular mortality were 2.05 (95% CI = 1.35–2.90), 1.85 (95% CI = 1.54–2.76), and 2.11 (95% CI = 1.59–3.07) and 1.95 (95% CI = 1.29–2.90), 1.70 (95% CI = 1.05–2.81), and 2.42(95% CI = 1.61–3.12), respectively, for the above subgroups.

Figures 2A,B illustrate the relationship among the SUA quartile categories of all-cause mortality and CVD mortality, respectively. Kaplan–Meier curves of survival showed the highest all-cause mortality and cardiovascular mortality for the SUA >386.75 μmol/L subgroup, followed by the SUA 0–242.0 and 312.0–386.75 μmol/L subgroups. The SUA 242.0–312.0 μmol/L subgroup had the least mortality. As depicted in Figure 3, after multivariable adjustment, cubic spline models showed a U-shaped association between SUA level with all-cause mortality (A, B) and CVD-cause mortality (C, D) among males and females, respectively.

The results of age- and sex-adjusted survival ROCs (Figure 4) indicated that SUA level can predict all-cause mortality, with an AUC of apparent prediction of 0.706 (95% CI = 0.696–0.727; P < 0.001); when adjusted for age and sex, the AUC was 0.702 (95% CI = 0.692–0.725, P < 0.001). Meanwhile, SUA levels predicted CVD-cause mortality, and the AUC of apparent prediction was 0.716 (95% CI = 0.701–0.747, P < 0.001); when adjusted for age and sex, the AUC was 0.711 (95% CI = 0.697–0.742, P < 0.001).

Figure 5 shows the adjusted proportional HR of mortality according to SUA cutoff value. Compared with the SUA 180.5–370.5 μmol/L subgroup, HRs for all-cause mortality and cardiovascular mortality in males in the SUA 0–180.5 and SUA >370.5 μmol/L subgroups were 2.01 (95% CI = 1.35–2.56) and 2.08 (95% CI = 1.42–2.57) and 1.79 (95% CI = 1.45–2.57) and 2.09 (95% CI = 1.47–2.62), respectively, after adjusting for age, hypertension, stroke, MI history, smoking status, chronic renal disease, PAD history, PTCA history, and CABG history. In females, HRs of all-cause and CVD mortality were 1.98 (95% CI = 1.25–2.53) and 2.04 (95% CI = 1.39–2.61), and 1.89 (95% CI = 1.31–2.67) and 2.02 (95% CI = 1.43–2.74), respectively.

Based on Figures 6A,B, Cox regression models also demonstrated that factors for risk of mortality included sex, age, hypertension, stroke, metabolic syndrome, DM, MI history, PAD history, hyperlipidemia history, smoking status, PTCA history, CABG history, and low eGFR. Of note, among these risk factors, age, hypertension, stroke, DM, lower eGFR, abnormal ABI, MI history, and CABG history occupied major positions.

First, subjects with HAR were enrolled to assemble a study population as homogeneous as possible, as HAR patients have a poor prognosis. Thus, the results cannot be extended to the entire population. Second, as follow-up participants were contacted by annual phone interviews, our results may have information bias. In addition, some patients had poor compliance, and withdrawal bias may exist. Finally, the follow-up time was not long compared with Western country prospective cohort studies. Hence, the data from this research are not comprehensive, and additional studies are needed.

The strengths of our study include its prospective design and reliable assessment of mortality and cardiovascular events. Second, this research was a multicenter prospective cohort registration study with little heterogeneity. This cohort study also had a longer follow-up time and larger sample size than other Chinese studies.