U-Shaped Relationship Between Cardiovascular Mortality and Serum Uric Acid May Be Attributed to Stroke- and Heart-Specific Mortality, Respectively, Among Hypertensive Patients: A Nationally Representative Cohort Study

Background

Hypertension is a well-known and strong risk factor for cardiovascular diseases (CVD), which are the major cause of mortality [1,2]. Uric acid (UA), an end-product metabolite of purine nucleotide in great apes and humans, has been considered as one of the cardiometabolic risk factors in CVD, diabetes, and hypertension [3–5]. Increased serum UA is one of the strongest factors for hypertensive development [6]. Experimental results support that raising uric acid levels causes hypertension in rats [7]. Moreover, pilot studies indicate that lowering serum UA levels can decrease blood pressure in hypertensive patients [8–10].

Interestingly, numerous experimental studies demonstrated that uric acid exerts a beneficial role due to its antioxidant properties [3,11]. It has been demonstrated that uric acid explains about 50% of the total antioxidant capacity in vivo [3]. However, under the setting of pathological acidic or hydrophobic milieu, such as atherosclerotic plaque and cytoplasm, UA becomes a pro-oxidant factor that promotes redox disorder and exerts a deleterious effect in the development and progression of cardiovascular disease [3,11].

Currently, the association of UA levels with adverse outcomes is largely controversial. So far, previous studies noted positive [12–14], negative [15,16], and neutral [17,18] relationships between serum UA concentrations and risks of cardiovascular and total mortality in general adult populations, CVD, and renal disease. Insufficient sample size and variations in study populations and analysis strategies may partly account for the heterogenicity. Two recent large population-based cohorts noted a U-shaped association of total mortality with serum UA in general adult populations [19,20]. The increased mortality risk in adults with hypo- or hyper-uricemia suggests the complex biological roles of UA in humans. In particular, the benefit and harm of strict hypouricemic therapy should be considered.

However, the evidence regarding the link between UA levels and cardiovascular mortality in hypertensive individuals is unclear. To identify this association may provide new insights into the management of UA levels in hypertension, especially the intensive anti-uric acid therapy for hypertensive patients with hyperuricemia. This study aimed to assess the association of serum UA with all-cause and cause-specific mortality among 15 583 participants with hypertension.

Material and methods

STUDY POPULATION AND DESIGN:

Our analysis was based on a dataset from a nationally representative study, the National Health and Nutrition Examination Surveys (NHANES) of the United States [21], which is a stratified and multistage probability-sampling study to assess the characteristics of the nationally non-institutionalized population. The protocols of NHANES have been reported previously [21–23]. The dataset has been built since 1999 and is released in 2-year survey cycles. The protocols and procedures of this study were agreed to by the Research Ethics Review Board of the Centers of Disease Control and Prevention of the United States (Protocol Number. 98–12, 2005–06, and 2011–17). All participants provided written informed consent. The NHANES datasets are available to all researchers to reproduce the results. (https://www.cdc.gov/nchs/index.htm) [21].

We performed the primary analysis using the datasets of 8 two-year survey cycles from 1999–2000 to 2013–2014. In those cycles, there were 38 943 participants aged ≥20 years old with serum UA measurements. Hypertension was defined by blood pressure-lowering treatment and systolic/diastolic blood pressure ≥140/90 mmHg at baseline [21]. We excluded individuals without hypertension (n=22 908). Given the possibility of reverse causality, we further excluded individuals with hypouricemic therapy in the preceding month at baseline (n= 440). Further, we excluded individuals with missing information on mortality status (n=12). In total, 15 583 hypertensive individuals were included for analysis. The flow of the study is presented in Figure 1.

SERUM URIC ACID MEASUREMENT:

The exposure variable of interest was serum uric acid concentrations as tested by the Roche Hitachi Model 917/704 multichannel analyzer, Beckman Synchron LX20, or Beckman UniCel® DxC800 Synchron, as described in prior studies [19]. The protocols have been validated. In brief, UA was oxidized with specific enzyme uricase to generate allantoin and H2O2. The H2O2 further reacted with 2,4,6-tribromo-3-hydroxybenzoic acid and 4-aminophenazone to form quinone-imine dye and hydrogen bromide [19]. The intensity of red color was used to quantify results. The coefficient of variation for UA measurements in each cycle was approximately 2%, suggesting good repeatability. All laboratory variables were assessed by validated protocols and procedures. The details were available at https://www.cdc.gov/nchs/nhanes.

COVARIATES:

Demographic and lifestyle factors at baseline, including age, gender, race/ethnicity, smoke, alcohol intake, and physical activity, and family income level were recorded during personal interviews via standardized questionnaires [21,22]. Physical examinations were performed according to standardized protocols and processes at a mobile examination center. Body mass index (BMI) was calculated as weight (kilograms) divided by height (meters) squared [21]. Average systolic blood pressure and diastolic pressure were assessed as the means of 3 measurements. The detection of biosamples was conducted in special central laboratories with validated methods. Laboratory determinations of triglyceride (TG), total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C), glycosylated hemoglobin (HbA1c), and creatinine were detected in each survey cycle [21]. C-reactive protein (CRP) was measured in NHANES 1999–2006. The Chronic Kidney Disease Epidemiology Collaboration method was applied to calculate the estimated glomerular filtration rate (eGFR) [21]. History of diseases and prescription agents in the preceding 30 days was also solicited at a screening interview. Diabetes mellitus was defined as taking diabetic medication or HbA1c ≥6.5% [21]. Emphysema or chronic bronchitis was identified as chronic obstructive pulmonary disease (COPD). Malignant disease was identified according to self-report [21]. Antihypertensive agents were categorized as angiotensin-converting enzyme inhibitor/angiotensin II receptor antagonist (ACEI/ARBs), β-blocker, calcium-channel blockers (CCB), diuretics, and other antihypertensive drugs [4].

OUTCOMES:

The outcomes of this study were total and cause-specific mortality. All participants were linked to the National Death Index in the National Center for Health Statistics of the US [21,22]. Mortality data were available through December 31, 2015. Using the codes of International Classification of Diseases 10th Revision (ICD-10), cause-specific mortality was categorized as death caused by CVD (heart disease: I00–I09, I11, I13, and I20–I51; stroke: I60–I69), and malignant neoplasms (C00–C97) [21].

STATISTICAL ANALYSIS:

The statistical analyses were conducted following the analytical guidelines. Sampling weights, the masked variance of the primary sampling unit, and strata were used to explain the complex study design and acquire nationally representative estimates [21]. Variables are presented as weighted means (standard error, SE) and proportions unless otherwise noted. We used weighted linear regression or logistics regression to assess the difference across serum UA groups, when necessary. Restricted cubic spline based on Cox regression was applied to show the link between serum UA and total mortality, after adjustment for age, sex, race, smoking, alcohol intake, exercise, poverty-to-income ratio (PIR), BMI, cancer, COPD, diabetes, cardiovascular disease, TG, TC, HDL-C, eGFR, lipid-lowering agents, antiplatelet treatment, ACEI/ARBs, β-blocker, CCB, diuretics, and other antihypertensive drugs [21,24]. All hypertensive patients were stratified into 5 prespecified groups according to uric acid levels: ≤3.5, 3.5–5, 5–6, 6–7.5, and >7.5 mg/dL. Hazard ratios (HRs) and 95% confidence intervals (CIs) were calculated by crude and multivariable Cox regression models for the relationship between UA and total or cause-specific mortality, with UA 5–6 mg/dL as a reference. Three multivariable models were applied. Model 1 was adjusted for demographic variables, including age, sex, and race/ethnicity (non-Hispanic white, black, Hispanic-Mexican, or other). Model 2 was further adjusted for PIR, BMI, smoking status, alcohol intakes, physical activity, TG, TC, HDL-C, eGFR, CVD, diabetes, COPD, and cancer. Model 3 was additionally adjusted for lipid-lowering agents, antiplatelet drugs, ACEI/ARBs, β-blocker, CCB, diuretics, and other antihypertensive drugs [21]. In secondary analyses, the association between baseline UA and all-cause, cardiovascular, heart-specific, and stroke mortality was ascertained in subgroups by age (<65 and ≥65 years), sex (female and male), race (non-Hispanic white and non-white), BMI (<30 and ≥30 kg/m2), eGFR (<60 and ≥60 mL/min/1.73 m2), and anti-hypertension medications (yes/no), with the fully adjusted model except for stratification factors [21]. We also assessed whether the main results were altered after additional adjustment for inflammation marker CRP in NHANES 1999–2006. All tests with a 2-sided P value <0.05 were considered significant using Stata (version 15) software.

Results

BASELINE CHARACTERISTICS:

Overall, 15 583 participants with hypertension (7524 men and 8059 women) were identified in the dataset from NHANES 1999–2000 to 2013–2014 cycles (Figure 1). The mean age was 55.8 years, and 47.6% were males in the study population. The median of serum UA level was 5.7 mg/dL (interquartile range [IQR], 4.8–6.7 mg/dL). Baseline characteristics of hypertensive patients across the strata of uric acid (≤3.5, 3.5–5, 5–6, 6–7.5, and >7.5 mg/dL) are shown in Table 1. Compared to those with lower UA, the participants with higher levels of UA were older and mostly males, and had higher alcohol consumption, TG, and CRP levels. The higher-UA groups were more likely to have adiposity, prior CVD, and to be taking antihypertensive agents. They also had less smoking and vigorous exercise, as well as lower eGFR.

ASSOCIATIONS OF UA WITH TOTAL AND CAUSE-SPECIFIC MORTALITY:

The median duration of follow-up was 7.4 (IQR, 4.1–11.3) years. Over 116 351 person-years of follow-up, 3291 deaths occurred among the 15 583 patients with hypertension, including 639 (19.4%) heart-related deaths and 630 (19.1%) cancer deaths. Among hypertensive participants in NHANES 1999–2006, 152 died due to stroke. The weighted mortality rates per 1000 person-years of follow-up are shown in Table 2, with 21.2 (95% CI, 20.3–22.1) for all-cause mortality, 3.9 (95% CI, 3.5–4.3) for heart-related and 4.1 (95% CI, 3.7–4.5) for cancer-related mortality, and 1.3 (95% CI 1.1–1.6) for stroke mortality. In multiple restricted cubic spline fitting, there was a U-shaped trend between UA and risks of total and cardiovascular mortality (Figure 2).

The relationships between serum UA levels and risks of all-cause, heart disease, stroke, and cancer-related mortality are shown in Table 2, assessed by several weighted multivariable Cox regression analyses with UA 5.0–6.0 mg/dL as reference. The age- and sex-adjusted HRs (95% CIs) of total mortality from ≤3.5 mg/dL to 3.5–5 mg/dL, 5–6 mg/dL, 6–7.5 mg/dL, and >7.5 mg/dL were 1.40 (1.05–1.88), 1.08 (0.95–1.21), 1.00 (reference), 1.14 (1.02–1.29), and 1.74 (1.50–2.02), respectively. The risks remained significant after adjusting for demographics, income, smoking, amateur sports activity, alcohol intakes, BMI, cancer, COPD, diabetes, cardiovascular disease, TG, TC, HDL-C, eGFR, lipid-lowering agents, antiplatelet treatment, ACEI/ARBs, β-blocker, CCB, diuretics, and other antihypertensive drugs. Compared with patients with UA 5–6 mg/dl, those with UA ≤3.5 and >7.5 mg/dL had an elevated risk of all-cause mortality, by 47% and 35%, respectively.

As expected, a similar trend of association of UA with cardiovascular mortality was noted (Table 2). In multivariable-adjusted model, the HRs (95% CI) of cardiovascular mortality from the lowest to the highest group were 1.92 (1.08–3.42), 1.27 (0.88–1.81), 1.00 (ref.), 1.19 (0.92–1.55), and 1.37 (1.00–1.90), respectively. Further, compared with patients with UA 5–6 mg/dl, the increased cardiovascular mortality in patients with UA ≤3.5 mg/dl was mainly due to stroke (HR 4.34, 95% CI 1.66–11.35), while the increased mortality in patients with UA >7.5 mg/dl was mainly due to ischemic heart disease (HR 1.46, 95% CI 1.10–1.95). However, neither lower serum UA nor higher UA was significantly related to cancer mortality.

In prespecified stratification analyses (Supplemenatry Tables 1–6), the associations of serum concentrations of UA with total and cardiovascular mortality were similar in the subgroups of hypertensive patients regarding sex (female versus male), eGFR (≥60 versus <60), and other subgroups. In the further sensitivity analysis, we assessed the relationship between UA and mortality in participants with CRP measurement. After additionally adjusting for CRP, the results did not alter significantly (Supplemenatry Table 7).

Discussion

STRENGTHS AND LIMITATIONS:

Our study has several limitations. First, the residual confounders unmeasured may not be completely ruled out. Second, the causality of this link between serum UA and mortality risk could not be determined because of the observational nature of the study, although prior experimental studies observed that UA-related metabolism activated inflammation and oxidative stress in the process of cardiovascular disease. Third, the enrolled participants in this study were US civilians, so the results extrapolated to other populations need further verification. Fourth, information on distinguishing primary and secondary hypertension was lacking in this community-based cohort, and further investigation in special hospital-based cohorts is warranted. This study may have some strengths. The NHANES study was a national sampling dataset that favored repeatability and generalization. The sample size and long-term follow-up provide favorable statistical effectiveness. We have adjusted for a variety of potential confounders, including lifestyle, laboratory data, chronic disease, and detailed information of antihypertensive agents, and the relationship between UA and all-cause and cardiovascular mortality remained significant.

Conclusions

Our findings suggest a U-shaped relationship between serum UA and risks of total and cardiovascular mortality in hypertensive patients. Both low and high levels of uric acid were independently associated with increased risks of cardiovascular mortality. Nonetheless, low UA may mainly increase stroke-related mortality, and high UA may mainly increase heart-related mortality. Further investigations are needed to elucidate the potential mechanisms and validate the role of uric acid in the progression of hypertension.