The present study involved a post hoc analysis of the STEP trial. The methods and results for the primary outcome have been published elsewhere.8,9 In brief, the STEP trial was a prospective, multicenter, randomized controlled trial performed at 42 clinical centers throughout China. The main inclusion criteria were as follows: a) age of 60 to 80 years, b) a history of hypertension and treatment with antihypertensive medication or systolic blood pressure (SBP) of 140 to 190mmHg in 3 screening visits, and c) a record of the patient's baseline uric acid concentration. The main exclusion criterion was a history of ischemic or hemorrhagic stroke. The detailed inclusion and exclusion criteria were previously described in the study protocol.9 The patients were randomly assigned in a 1:1 ratio to receive 1 of 2 therapies: intensive treatment (target SBP of <130mmHg) or standard treatment (target SBP of <150mmHg).8

This study was a post hoc analysis of the STEP trial, which was approved by the ethics committees of Fu Wai Hospital and all collaborating centers, and all the enrolled patients provided written informed consent. Therefore, no further approval was required in the present study. Clinical trial registration number: NCT03015311.

The office blood pressure measurements were performed by trained personnel (physicians or nurses) using an office blood pressure monitor (OMRON Healthcare, United States). Before measurement, the participants rested quietly in a seated position for at least 5minutes, and their blood pressure was then measured 3 times at 1-minute intervals by trial staff (observed). Both the office blood pressure and laboratory data (including the uric acid concentration, tested in Beijing CIC Clinical Laboratory) were obtained in a standard manner during all baseline and follow-up clinic visits. Details regarding quality control have been provided in previous studies.9

As previously described in the study protocol,8,9 the primary outcome was a composite endpoint including death from cardiovascular causes, stroke, acute decompensated heart failure, coronary revascularization, acute coronary syndrome, and atrial fibrillation. The secondary outcomes were the components of the primary outcome and all-cause death.

In the restricted cubic spline analyses, primary outcome, and secondary outcomes (excluding all-cause death) were analyzed based on the Fine-Gray subdistribution hazard model. In the analysis of all-cause death, the Cox regression model was used. For death from all causes, we used cox.zph() to test the proportional hazards assumption for a Cox regression model. For other endpoints, we used the Fine-Gray subdistribution hazard model to build a dataset, then tested the proportional hazards assumption use cox.zph(). The associations between the uric acid concentration and all endpoints were evaluated on a continuous scale with restricted cubic spline curves. This is a multivariate analysis, including adjusted prognostically relevant variables (age, sex, body mass index, diastolic blood pressure, alanine aminotransferase, aspartate aminotransferase, urea, creatinine, triglycerides, high-density lipoprotein cholesterol, history of diabetes mellitus, estimated glomerular filtration rate).10

We included all participants with an available baseline uric acid concentration and stratified the them according to these concentrations (tertile 1 [T1], <303.0μmol/L; tertile 2 [T2], 303.0 to <375.8μmol/L; and tertile 3 [T3], ≥ 375.8μmol/L). For continuous variables, the mean±standard deviation (SD) were calculated; for categorical variables, the proportion was calculated in each category substratified by the uric acid concentration. The baseline characteristics for each stratum are depicted appropriately and were compared using the most suitable tests (such as analysis of variance, the chi-square test, and the Kruskal–Wallis test).

In the analyses of the primary outcome and secondary outcomes (excluding all-cause death), cumulative incidence was calculated for the 2 trial groups according to different strata using the Fine-Gray subdistribution hazard model and the results are presented as subdistribution hazard ratios.11 In the analysis of all-cause death, the Cox regression model was used and the results are presented as hazard ratios (HR). The intention-to-treat approach was used in the present analysis. Although multiple events were recorded in this study and a single patient could develop more than 1 event, only the first event of any type per patient was used in the analysis. Model 1 was adjusted for potential confounders, which were significantly different between the intensive treatment group and the standard treatment group. P values for interaction in the subgroup analysis, and subdistribution hazard ratios or HR with 95% confidence intervals (95%CI) were used to compare the intensive and standard SBP control within each tertile.

To test the trend of uric acid concentrations during the follow-up years, the mixed effect regression model was used by the function lme() of package nlme. In the fixed effect, the dependent variable was follow-up uric acid concentration. The interaction between follow-up years and treatment group provided the real differences between treatment groups. The autocorrelation among repeated measurements was accounted for in the random effect. The results of this analysis were translated into least square means (LS means). As a sensitivity analysis, the interactions between treatment with uric acid as continuous of endpoints were analyzed using Cox regression for all-cause mortality, and Fine & Gray regression for other primary and secondary outcomes. All analyses were performed with R software, version 3.6.3 (R Foundation for Statistical Computing, Austria). A 2-sided P value of <.05 was considered statistically significant.

Table 1 summarizes the characteristics of the overall study population and each uric acid tertile. Patients were stratified into 3 tertiles (< 303.0μmol/L, 303.0 to <375.8μmol/L, and ≥ 375.8μmol/L) according to their baseline uric acid concentration. Overall, the mean±SD age of the sample at baseline was 66.26±4.83 years, the mean uric acid concentration at baseline was 347.26±89.15μmol/L, and nearly half (46.6%) of the patients were male. The mean body mass index was 25.58±3.16kg/m2, and the mean SBP was 146.07±16.65mmHg. Baseline characteristics were well balanced between the intensive and standard treatment groups in each tertile (table 2), which is consistent with the findings of our main study.8 This indicates that the inclusion of patients in the present study was representative.

Some examined characteristics differed by uric acid tertile; for example, patients in T1 were younger, were more likely to be female, were more likely to have a history of diabetes mellitus, had a lower body mass index, baseline diastolic blood pressure, alanine aminotransferase concentration, aspartate aminotransferase concentration, urea concentration, creatinine concentration, triglyceride concentration, and Framingham Score, and had a higher estimated glomerular filtration rate (P <.05 for all) (table 1).

Further comparisons between the intensive and standard treatment groups were examined within each tertile. Except for fasting serum glucose concentration, all baseline characteristics were well balanced between the 2 trial groups by uric acid tertile (table 2).

The overall trends in the different uric acid tertiles agreed well with the overall trends of each treatment group separately (figure 1 of the supplementary data). Additionally, the 2 treatment strategies in different uric acid tertiles led to a similarly rapid and sustained between-group difference in SBP (figure 2 of the supplementary data), which is similar to the findings of our previously published study.8

Overall, the risk of the primary outcome rose as the cumulative uric acid concentration increased. This association was found in both the intensive treatment group and the standard treatment group (figure 1). The line representing multivariable-adjusted subdistribution HR for the primary outcome of patients receiving intensive treatment continued under the counterpart of patients receiving standard treatment (figure 1). However, the figures showed a wide overlap of the confidence intervals. The multivariable-adjusted subdistribution HR for the secondary outcomes except all-cause death showed similar trend (figure 2 and figure 3). In addition, our results showed that uric acid (as continuous and as tertiles) met the proportionality assumption for all endpoints.

Figure 1.

Central illustration. Cubic spline regression curves relating baseline uric acid concentrations as a continuous variable to the primary outcome. The figure shows HR with shadow 95% confidence intervals relating baseline uric acid concentrations to primary outcome under the Fine-Gray subdistribution hazard model in each SBP treatment arm, with baseline uric acid concentrations as predictor variables and covariable adjustment for age, sex, body mass index, DBP, ALT, AST, urea, creatinine, triglycerides, high-density lipoprotein cholesterol, history of diabetes mellitus, and estimated glomerular filtration rate. ALT, alanine aminotransferase; AST, aspartate aminotransferase; DBP, diastolic blood pressure; HR, hazard ratio; UA, uric acid.

(0.11MB).

Figure 2.

A: cubic spline regression curves relating baseline uric acid concentrations as a continuous variable to stroke. The figure shows HR with shadow 95% confidence intervals relating baseline uric acid concentrations to the primary outcome under the Fine-Gray subdistribution hazard model in each SBP treatment arm, with baseline uric acid concentrations as predictor variables and covariable adjustment for age, sex, body mass index, DBP, ALT, AST, urea, creatinine, triglycerides, high-density lipoprotein cholesterol, history of diabetes mellitus, and estimated glomerular filtration rate. B: cubic spline regression curves relating baseline uric acid concentrations as a continuous variable to acute coronary syndrome. The figure shows HR with shadow 95% confidence intervals relating baseline uric acid concentrations to the primary outcome under the Fine-Gray subdistribution hazard model in each SBP treatment arm, with baseline uric acid concentrations as predictor variables and covariable adjustment for age, sex, body mass index, DBP, ALT, AST, urea, creatinine, triglycerides, high-density lipoprotein cholesterol, history of diabetes mellitus, and estimated glomerular filtration rate. ALT, alanine aminotransferase; AST, aspartate aminotransferase; DBP, diastolic blood pressure; HR, hazard ratio; UA, uric acid.

(0.21MB).

Figure 3.

A: cubic spline regression curves relating baseline uric acid concentrations as a continuous variable to cardiovascular death. The figure shows HR with shadow 95% confidence intervals relating baseline uric acid concentrations to the primary outcome under the Fine-Gray subdistribution hazard model in each SBP treatment arm, with baseline uric acid concentrations as predictor variables and covariable adjustment for age, sex, body mass index, DBP, ALT, AST, urea, creatinine, triglycerides, high-density lipoprotein cholesterol, history of diabetes mellitus, and estimated glomerular filtration rate. B: cubic spline regression curves relating baseline uric acid concentrations as a continuous variable to all-caused death. The figure shows HR with shadow 95% confidence intervals relating baseline uric acid concentrations to the primary outcome under a Cox regression model in each SBP treatment arm, with baseline uric acid concentrations as predictor variables and covariable adjustment for age, sex, body mass index, DBP, ALT, AST, urea, creatinine, triglycerides, high-density lipoprotein cholesterol, history of diabetes mellitus, and estimated glomerular filtration rate. ALT, alanine aminotransferase; AST, aspartate aminotransferase; DBP, diastolic blood pressure; HR, hazard ratio; UA, uric acid.

(0.21MB).

During the median follow-up period of 3.34 years, a total of 336 primary outcome events occurred in 144 of 4132 patients (3.5%; 1.0% per year) in the intensive treatment group and in 192 of 4162 patients (4.6%; 1.4% per year) in the standard treatment group, with a subdistribution HR of 0.74; 95%CI, 0.60-0.92; P=.007] (table 2 of the supplementary data). Therefore, intensive treatment considerably reduced the incidence of primary outcome events when compared with standard treatment, with an absolute difference of 1.1 percentage points.

The incidence of primary outcome events was significantly lower in T1 of intensive SBP intervention when compared with standard treatment (subdistribution HR, 0.55; 95%CI, 0.36-0.86; P=.007) (table 3 and figure 4A). Additionally, no significant benefit was derived from intensive treatment in patients in T2 and T3 when compared with standard treatment (table 3 and figure 4B,C). The Interaction P value between SBP control and uric acid stratification was .29. The results for most of the secondary outcomes were similar to those for the primary outcomes in different tertiles (table 3 of the supplementary data).

Figure 4.

Cumulative incidence for the primary outcome by stratification of uric acid. Cumulative hazards over time are depicted for tertile 1 (A), tertile 2 (B), tertile 3 (C) associated hazard ratios (HRs). The lines depict the intensive and standard arms. The primary outcome was a composite of stroke, acute coronary syndrome, acute decompensated heart failure, coronary revascularization, atrial fibrillation, or death from cardiovascular causes. The hazard ratio, 95% confidence interval (CI), and P value for the primary outcome were calculated with the use of the Fine-Gray subdistribution hazard model for the competing risk of death. The inset shows the same data on an enlarged y-axis.

(0.59MB).

As a sensitivity analysis, the interaction between treatment with uric acid as a continuous variable revealed no significance (table 4 of the supplementary data).

The results of the mixed effect regression model were translated into LS means (table 4). In the standard group, the uric acid concentrations at baseline, first visit, second visit and third visit were 347μmol/L, 342μmol/L, 341μmol/L, and 333μmol/L. In the standard group, the follow-up uric acid concentrations were 348μmol/L, 344μmol/L, 345μmol/L, and 337μmol/L, respectively. The P values revealed no significant differences over time between the 2 treatment groups (table 4).

The main limitation of this study is that much of the reported variation in HR between the subgroups was caused by chance and that, in the absence of a statistically significant interaction, the best estimate of the effect of the intervention was given by the study-wide effect estimate, including all patients. In addition, the post hoc analysis of randomized controlled trials is subject to potential confounding factors. In this study, this disadvantage presented as large overlap of confidence intervals both in cubic spline regression analyses and tertile analyses.