Assessment of coronary vascular function with cardiac PET in relation to serum uric acid

Background Elevated serum uric acid (SUA) levels have been independently associated with cardiovascular disease. Stress myocardial perfusion positron emission tomography (PET) allows for measurement of absolute myocardial blood flow (MBF) and quantification of global left ventricular coronary flow reserve (CFR). A CFR <2.0 is considered impaired coronary vascular function, and it is associated with increased cardiovascular risk. We evaluated the relationship between SUA and PET-measured markers of coronary vascular function. Methods We studied adults undergoing a stress myocardial perfusion PET on clinical grounds (1/2006-3/2014) who also had ≥1 SUA measurement within 180 days from the PET date. Multivariable linear regression estimated the association between SUA and PET-derived MBF and CFR. We also stratified analyses by diabetes status. Results We included 382 patients with mean (SD) age of 68.4 (12.4) years and mean (SD) SUA level of 7.2 (2.6) mg/dl. 36% were female and 29% had gout. Median [IQR] CFR was reduced at 1.6 [1.2, 2.0] and median [IQR] stress MBF was 1.5 [1.1, 2.1] ml/min/g. In the adjusted analysis, SUA was inversely associated with stress MBF (β = -0.14, p = 0.01) but not with CFR. Among patients without diabetes (n = 215), SUA had a negative association with CFR (β = -0.15, p = 0.02) and stress MBF (β = -0.19, p = 0.01) adjusting for age, sex, extent of myocardial scar and ischemia, serum creatinine and gout. In diabetic patients (n = 167), SUA was not associated with either CFR or MBF. Conclusions In this cross-sectional study, higher SUA is modestly associated with worse CFR and stress MBF among patients without diabetes.


Introduction
A number of large epidemiologic studies have demonstrated an independent association between serum uric acid (SUA) levels and risk of myocardial infarction (MI), heart failure, stroke, and cardiovascular mortality. [1][2][3] Elevated SUA is hypothesized to cause increased oxidative stress, microinflammation, lipid oxidation, and inhibition of nitric oxide production. In turn, these all contribute to endothelial dysfunction, microvascular ischemia, interstitial fibrosis, and myocardial dysfunction. [4][5][6] Coronary flow reserve (CFR)-the ratio of peak hyperemic myocardial blood flow (MBF) over MBF at rest as measured by positron emission tomography (PET)-is a robust and reproducible clinical measure of the integrated hemodynamic effects of epicardial coronary artery stenoses, diffuse atherosclerosis, vessel remodeling, endothelial dysfunction, and microvascular dysfunction on myocardial tissue perfusion. [7,8] A reduced CFR can be a sign of flow-limiting coronary artery stenoses as it is associated with the overall extent and severity of angiographic coronary artery disease (CAD). [9] Furthermore, a reduced CFR can indicate the presence of coronary vascular dysfunction involving smaller vessels, which increases the severity of inducible myocardial ischemia and sub-clinical myocardial injury beyond the effects of upstream coronary obstruction. [10] Importantly, there is growing, consistent evidence that impaired CFR is independently and incrementally associated with risk for MI, heart failure as well as cardiovascular death. [7,8,11,12] Over the past decades, the association between hyperuricemia and cardiovascular disease has been extensively studied. The association between SUA and increased cardiovascular risk appears to be only partially accounted by traditional coronary risk factors. This suggests that other mechanisms may contribute to the association between the SUA and increased cardiovascular risk. One such mechanism may involve the potential adverse effect of SUA on vascular function and in particular endothelial function, thereby increasing the potential for coronary vasoconstriction and thrombosis. To date, the direct effect of hyperuricemia on coronary vascular function, as assessed by CFR, has not been studied. We therefore conducted a cross-sectional study to evaluate coronary vascular function related to SUA levels in patients referred for stress myocardial perfusion PET. In addition, we examined whether the relationship between SUA levels and coronary vascular function differed by presence of diabetes based on the close relationship between hyperuricemia, gout and diabetes. [13][14][15][16][17]

Study cohort
All patients clinically referred for stress myocardial perfusion PET at the Brigham and Women's Hospital in Boston, Massachusetts, USA between January 2006 and March 2014 were eligible for inclusion. Of those, we selected patients with at least one SUA level measurement during the 180-day period before or after the PET test date. In patients with multiple PET tests during the study period, we included the study closest to the date of SUA level measurement.
The study protocol was approved by the Institutional Review Board of the Brigham and Women's Hospital which granted a waiver of informed consent.

Definition of exposure and outcome
SUA level measured by enzymatic colorimetric assay in 180 days before or after the PET test was the exposure of interest. Hyperuricemia was defined as !7 mg/dl in men and !6 mg/dl in women.
The primary outcomes of interest were stress MBF and CFR quantified using PET. Absolute MBF in milliliter/minute/gram (ml/min/g) was computed from the dynamic rest and stress imaging series with commercially available software (Corridor4DM; INVIA Medical Imaging Solution, Ann Arbor, MI) and previously validated methods. [18,19] CFR was calculated as the ratio of absolute MBF at stress over rest for the entire left ventricle. CFR <2 is known to be associated with worse cardiovascular outcomes in a general referral population. [11] PET imaging Following standard imaging protocols, patients were studied with a whole-body PET/computed tomography scanner (Discovery RX or STE LightSpeed 64, GE Healthcare, Milwaukee, WI) after at least 4 hours of fasting. Patients refrained from caffeine-and methylxanthine-containing substances and drugs for 24 hours before their scans. Briefly, at rest, radionuclide imaging was obtained with an intravenous bolus administration of 13 N-ammonia or 82 rubidium. Then, a standard intravenous infusion of a vasodilator (i.e., dipyridamole, adenosine, or regadenoson) or dobutamine was given for pharmacologic stress. At peak stress, a second dose of 13 N-ammonia or 82 rubidium was injected, and images were recorded in the same manner. MBF was measured during rest and peak stress with 13 N-ammonia or 82 rubidium as a perfusion tracer, as described previously. [18][19][20][21] Heart rate, blood pressure, and 12-lead ECG were recorded at baseline and every minute during and after pharmacological stress. Left ventricular ejection fraction at rest and stress were calculated from gated myocardial perfusion images with commercially available software. In addition, summed rest score, summed stress score, and summed difference score (stress minus rest) were computed, with higher scores reflecting larger areas of myocardial scar, scar plus ischemia, or ischemia, respectively; summed stress scores 3 are generally considered normal. [22][23][24]

Covariates
We assessed a number of pre-defined variables potentially related to hyperuricemia or coronary vascular function based on patient interview and/or medical record data in the Partners Healthcare Research Patient Data Registry from the 180-day period immediately before or after the PET test date. These variables were: age and sex; body mass index; comorbidities including gout, hypertension, diabetes, smoking, coronary artery disease, heart failure, and dyslipidemia; medications including beta blockers, calcium channel blockers, nitrates, angiotensin-converting enzyme inhibitors, diuretics, gout-related medications (i.e., allopurinol, febuxostat and colchicine), and aspirin; and laboratory data including SUA, serum creatinine, LDL cholesterol and C-reactive protein levels.

Statistical analysis
Patient characteristics were compared between the hyperuricemia and normouricemia groups. Statistical significance was assessed with Wilcoxon rank sum tests or two-sided t-tests for continuous variables and Fisher's exact or chi 2 tests for binary variables. Normality was assessed using a combination of the Shapiro-Wilk test and visual inspection of descriptive statistics and histograms. Because data were not normally distributed, we used natural log transformation of SUA levels, CFR and MBF in all regression models. Pearson correlation was checked between log SUA, log CFR and log stress MBF. As diabetes is a strong predictor of coronary vascular function and CAD, [25] we tested for an interaction between SUA (as a continuous variable) and diabetes (yes/no) on CFR or stress MBF. Neither interaction was statistically significant. For primary analysis, we used unadjusted and multivariable linear regression models to examine the association between SUA levels and coronary vascular function in the main cohort. Our final models were adjusted for age, sex, summed stress score (i.e., a strong indicator of myocardial scar and ischemia), serum creatinine, and presence of gout diagnosis. Because prior myocardial scar or ischemia is a major determinant of CFR, we conducted a sensitivity analysis in which we performed multivariable linear regression models only in patients with summed stress scores 3. [22][23][24] We also performed stratified analysis by the presence of diabetes. All analyses were performed using SAS 9.3 Statistical Software (SAS Institute Inc., Cary, NC).

Cohort characteristics
We identified a total of 382 patients including 208 with hyperuricemia and 174 with normouricemia. Table 1 presents patient characteristics. Mean (SD) age was 68.4 (12.4) years and 36% were female. Mean (SD) SUA level was 7.2 (2.6) mg/dl and 29% had gout. Cardiovascular comorbidities were prevalent as 85% had hypertension, 27% CAD and 11% heart failure. Diabetes was present in 44%, and 23% had any use of allopurinol or febuxostat.
The mean (SD) uric acid level in milligram per deciliter was 9.0 (2.1) for the hyperuricemia group and 5.1 (1.2) for the normouricemia group. Hypertension, diabetes, heart failure, gout, and use of calcium channel blockers, nitrates, xanthine oxidase inhibitors, and diuretics were more common in patients with hyperuricemia. ] ml/min/g in the overall cohort. The median CFR and stress MBF were lower (worse) in the hyperuricemia group versus the normouricemia group. There were weak negative correlations between SUA and CFR (r = -0.13, p = 0.015) and between SUA and MBF at stress (r = -0.24, p<0.001) in the unadjusted analysis (see S1 and S2 Figs). At rest, the heart rate and systolic blood pressure were similar between the hyperuricemia and normouricemia groups, but left ventricular ejection fraction was lower in patients with hyperuricemia (Table 1). Table 2 presents the main results from the multivariable linear regression models. In the final multivariable linear regression model adjusting for age, sex, diabetes, summed stress score, serum creatinine and gout, SUA was associated with stress MBF (β = -0.14, p = 0.01), but not with CFR (β = -0.07, p = 0.14). In a sensitivity analysis limiting to 184 patients with a   normal summed stress score ( 3), SUA was associated with both CFR (β = -0.14, p = 0.037) and stress MBF (β = -0.16, p = 0.048).

Stratified analysis
Characteristics of the diabetic (n = 167) and non-diabetic (n = 215) subgroups are presented in Tables 3 and 4 Overall, patients with diabetes had higher prevalence of comorbidities including 38% with gout versus 22% with gout in the nondiabetic subgroup. 41.9% of the diabetic group and 53.5% of the non-diabetic group had a normal summed stress score. Median stress MBF was lower in the diabetic group but median CFR was the same in both diabetic and non-diabetic subgroups. In patients with diabetes (see S1 and S2 Figs), SUA was not correlated with either CFR (r = -0.02, p = 0.80) or MBF at stress (r = -0.13, p = 0.10) even in the unadjusted analysis. No association was noted in multivariable linear regressions (Table 5). However, among patients with no diabetes, SUA (see S1 and S2 Figs) had a negative correlation with both CFR (r = -0.22, p = 0.001) and stress MBF (r = -0.31, p<0.001) in the unadjusted analysis. In multivariable linear regression adjusting for age, sex, summed stress score (a measure of the extent of myocardial scar and ischemia), serum creatinine and diagnosis of

Cardiac function
Heart rate at rest,  Serum uric acid and coronary vascular function

Discussion
The potential causal role of SUA on CAD and other cardiometabolic diseases has been under debate over the past few decades. While a number of epidemiologic studies showed positive associations between SUA and CAD or cardiovascular disease, [1][2][3] Mendelian randomization studies did not find a causal role of SUA in CAD, cardiovascular disease or diabetes. [26,27] In this cross-sectional study of 382 patients with a wide range of SUA levels, we demonstrated an inverse association between SUA and coronary vascular function in patients with no or minimal myocardial scar or ischemia (i.e., summed stress score 3) or those without diabetes, but no association among diabetic patients. The degree of association between SUA and coronary vascular function in non-diabetic patients was modest with a beta-coefficient of -0.15 for CFR and -0.19 for MBF in the multivariable linear regression models. This present study provides one of the most comprehensive evaluations of myocardial perfusion and coronary vascular function in relation to SUA levels. We examined not only the overall relationship between SUA and coronary vascular function in patients with and without prior myocardial scar or ischemia, but also the relationship stratified by the presence of diabetes. Since not all patients with high SUA have gout, an independent risk factor for cardiovascular disease, our analysis was adjusted for the presence of gout. As a result, we noted a modest negative association between SUA and coronary vascular dysfunction in patients without overt CAD (i.e., normal summed stress score) and those without diabetes, but not in the diabetic group. The lack of a significant association in diabetic patients may be explained by the fact that the modest effect of SUA on CFR seen in non-diabetic patients is likely overshadowed by the known strong association of diabetes and associated metabolic abnormalities (hyperglycemia and insulin resistance) with coronary microvascular dysfunction. [25,28] In other words, even if SUA has a modest causal role in determining coronary vascular function, the effect of SUA on coronary vascular function in diabetic patients may be too subtle. The observed difference in the association of SUA with coronary vascular dysfunction by diabetes may be also related to the greater prevalence of other comorbid conditions and/or use of cardiovascular medications in diabetic patients versus non-diabetic patients, which can be directly or indirectly related to coronary vascular dysfunction. While our findings need to be confirmed, it is worth investigating the longitudinal effect of SUA on coronary vascular dysfunction in patients with and without diabetes separately. Another potential explanation for the difference in the association of SUA with coronary vascular dysfunction by diabetes may be related to the difference in the severity of underlying CAD between the diabetic and non-diabetic groups in the study cohort. In the diabetic group, 41.9% had a normal summed stress score (i.e., 3) while 53% did in the non-diabetic group. As seen in the sensitivity analysis limited to those with a normal summed stress score ( 3), SUA unlikely has a role in determining coronary vascular function in patients with established myocardial damages even if SUA is causally related to CAD.
This study has limitations. First, as discussed earlier, this study is based at a single academic center, in which the study patients were referred for a PET test for a clinical reason. Thus, the results may not be generalizable to those with hyperuricemia and no clinical symptoms of CAD or subtle CAD. Second, as this is a cross-sectional study, the causal relationship between SUA levels and coronary vascular function cannot be determined. Some patients may maintain high SUA levels for a long time unless they are treated. However, to reduce exposure misclassification (i.e., SUA level) in the study cohort, we required all the SUA levels to be drawn within 6 months from the PET test. Third, while the final models were adjusted for several important predictors of CAD risk including age, sex, renal function, gout diagnosis, and a summed stress score (i.e., a marker of myocardial scar and ischemia), there may be residual confounding. Fourth, this study is the first and largest study that investigated an association between SUA and coronary vascular function using a cardiac PET, but further confirmation of our results is necessary in a larger and more generalizable setting.
In conclusion, this cross-sectional analysis showed a modest inverse association between high SUA levels and coronary vascular function in patients without diabetes after adjusting for age, sex, serum creatinine, gout diagnosis and the extent and severity of perfusion defects. Such association was not noted in patients with diabetes. While our results need to be confirmed in different settings, this present study suggests that the effect of hyperuricemia on coronary vascular function or CAD differs by diabetes status and may be more evident in patients without diabetes. Furthermore, this study highlights the need for future research on the association between the change in SUA and the change in coronary vascular function over time, particularly in patients without diabetes.