Longitudinal association between serum urate and subclinical atherosclerosis: the Coronary Artery Risk Development in Young Adults (CARDIA) study

Abstract

Objective

The aim of the present study was to determine whether serum urate (sUA) concentration is positively associated with subclinical atherosclerosis, independent of body mass index (BMI), among generally healthy adults.

Design and setting

The CARDIA study followed 5115 black and white individuals aged 18–30 years in 1985–1986 (year 0). Subclinical atherosclerosis comprised coronary artery calcified plaque (CAC; years 15, 20 and 25) and maximum common carotid intima–media thickness (IMT; year 20). sUA (years 0, 10, 15 and 20) was modelled as gender-specific quartiles that were pooled. Discrete-time hazard regressions and generalized linear regressions were used for analyses.

Results

Mean sUA concentration was lower in women than in men, and increased with age. Adjusting for demographic and lifestyle factors, the highest versus lowest quartile of sUA at year 0 was associated with a 44% [95% confidence interval (CI) 20%, 73%] greater risk of CAC progression from year 15 to 25 (P_trend < 0.001), which was attenuated by adjustment for BMI at year 0 (P_trend = 0.45). A stronger association was found between sUA at year 15 and CAC progression at year 20 or 25 (hazard ratio 2.07, 95% CI 1.66, 2.58 for the highest versus lowest sUA quartile P_trend < 0.001), which was attenuated but remained significant with additional adjustment for BMI at year 15 (P_trend = 0.01). A greater increment in sUA concentration from year 0 to year 15, independent of change in BMI, was related to a higher risk of CAC progression (P_trend < 0.001). Similar associations were found between sUA and IMT, but only in men.

Conclusion

sUA may be an early biomarker for subclinical atherosclerosis in young adults; starting in early middle age, sUA predicts subclinical atherosclerosis independently of BMI.

Introduction

Increased subclinical atherosclerosis burden, indicating early atherosclerosis, may be predictive of a higher lifetime risk of cardiovascular disease (CVD) [1]. Coronary artery calcified plaque (CAC) and common carotid intima–media thickness (IMT), as markers of subclinical atherosclerosis, have been increasingly used to improve CVD risk stratification and the prediction of clinical CVD endpoints [1]. Common carotid IMT represents vascular wall thickening that is generally not accompanied by raised plaque, whereas CAC indicates a later stage of subclinical atherosclerosis with the presence of advanced atherosclerotic plaques [2].

Elevated serum urate (sUA) concentration has been studied as a possible circulating biomarker for CVD risk. Urate is generated by purine catabolism, and elevated sUA concentrations may result under certain pathological conditions (e.g. kidney disease) or from reduced excretion of urate in urine or faeces due to excessive alcohol intake [3, 4]. On the other hand, an increase in sUA concentration can also be induced by enhanced production of xanthine oxidase, a form of xanthine oxidoreductase (XOR) which catalyses the generation of urate [5].

Epidemiological studies have demonstrated a link between elevated sUA concentration and the development of CVD and its risk factors, such as obesity and metabolic syndrome [4]. However, debate continues about whether an elevated sUA concentration predicts CVD risk independently or is merely a biomarker that indicates or reinforces the classical CVD risk factors, such as body mass index (BMI) [4]. The inconsistent evidence in this regard may be attributed partly to differences in study duration and the diverse characteristics of participants across studies [6]. An independent association between sUA concentration and adverse CVD outcomes was supported by 10 of 11 prospective cohort studies recruiting patients at high risk of CVD [6]. This was also evident in a recent cohort study of people aged 70 years and above, in whom elevated sUA concentration not only predicted greater CVD mortality independent of classical CVD risk factors, but also improved the prediction when added to these factors [7]. By contrast, evidence is limited and ambiguous among younger and generally healthy populations; the association between sUA concentration and CVD risk seems to be largely confounded by the well-known CVD risk factors [6]. It was demonstrated in a recent cross-sectional analysis in adults aged 30–45 years that the association between sUA and early atherosclerosis was eliminated after adjustment for BMI [8]. Therefore, investigating the role of sUA on subclinical atherosclerosis in a longitudinal setting can be important for determining the underlying aetiology and addressing the reasons for the conflicting evidence.

In this longitudinal study we utilized data from a cohort of young adults enrolled in the Coronary Artery Risk Development in Young Adults (CARDIA) study who have been followed into middle age. We aimed to clarify the role of sUA in CVD risk by examining the association between sUA concentration and subclinical atherosclerosis (as reflected by CAC and common carotid IMT). We hypothesized that (i) higher sUA concentration is associated with higher risk of subclinical atherosclerosis; and (ii) the association between sUA concentration and subclinical atherosclerosis may be attributed in part to its correlation with BMI, particularly in young adults. We further investigated whether sUA elevation that persists with ageing predicts subclinical atherosclerosis independently of other classical CVD risk factors (e.g. waist circumference), in place of BMI.

Methods

Study population

The details of the CARDIA study have been described elsewhere [9]; briefly, it is a prospective cohort study designed to investigate the evolution of CVD risk, including subclinical atherosclerosis. Participants were recruited from four field centres in Chicago, IL; Minneapolis, MN; Birmingham, AL; and Oakland, CA. At baseline (1985–1986), 5115 young black and white adults aged 18–30 years were enrolled in the study. Informing difference between the race groups is a major aspect of the CARDIA design; race was reported by participants to trained staff via an interviewer-administered questionnaire during a telephone interview. Seven follow-up examinations were conducted subsequently at years 2, 5, 7, 10, 15, 20 and 25, with response rates among survivors of 91% (n = 4624), 86% (n = 4352), 81% (n = 4086), 79% (n = 3950), 74% (n = 3672), 72% (n = 3549) and 72% (n = 3499), respectively. Informed consent was obtained from all participants at each examination and the study was approved by the institutional review board at each field centre.

Measurements of serum uric acid

Participants were asked to fast for at least 12 h and to avoid smoking and heavy physical activity before the examination. For each participant, an overnight-fasting blood sample was collected between 07.00 am and 10.00 am, in vacutainer tubes. Serum samples were prepared and frozen, shipped (in dry ice) to a central laboratory and stored at −70°C until analysis (within a maximum of four months after collection) [10]. sUA concentration was measured at year 0 using the uricase method [11]. A colorimetric assay (i.e. a modified version of the uricase method, in which urate is oxidized to peroxide) [12] was introduced at year 10 and remained in use at years 15 and 20 for measuring sUA concentration. At year 20, sUA was assessed as part of the Young Adult Longitudinal Trends in Antioxidants (YALTA) ancillary study to CARDIA. In year 17, 105 samples from year 0 were re-examined according to year 15 methods, and significant changes in these results were observed, compared with the original baseline data In order to conform to National Institute of Standards Standard Reference Materials and to be able to compare with sUA levels at year 0, the measurements of sUA at years 10, 15 and 20 were recalibrated based on re-runs of approximately 200 frozen samples in each examination. The new values of sUA reported here were recalculated according to the following regressions: at year 10, new sUA = original sUA×1.011846533+0.577150901; at year 15, new sUA = original sUA×0.95185458+0.8662809; and at year 20, new sUA = original sUA×1.021870749+0.440000277.

Measurements of subclinical atherosclerosis

CAC, the calcified component of coronary atheroma in the vessel wall, was measured using the calcium score (Agatston score) at years 15, 20 and 25 by noncontrast cardiac computed tomography (CT) scan with electrocardiogram cardiac gating [9]. Pregnant women and participants whose body weight exceeded the limit of the scanner (about 160–180 kg depending on the CT system) were excluded on an examination-specific basis. The year 15 (2000–2001) and year 20 (2005–2006) CT examinations were conducted using electron beam (C-150, Imatron, Inc. South San Francisco, CA, for Oakland and Chicago) or a multi-detector systems (LightSpeed Qx/I and Plus, GE Healthcare, Milwaukee, WI, for Birmingham; Somatom Sensation 4 or Volume Zoom, Siemens Healthcare, Erlangen, Germany, for Minneapolis). At the year 25 examination (2010–2011), 64-channel multi-detector CT systems were used at all sites (LightSpeed VCT, GE Healthcare for Oakland; Discovery CT750 HD, GE Healthcare for Birmingham; and Somatom Sensation 64, Siemens Healthcare for Minneapolis and Chicago). Images were analysed using software approved by the US Food and Drug Administration (Calcium Score, TeraRecon Inc. San Mateo, CA) by experienced analysts who measured calcification in each coronary artery (i.e. left main, left anterior descending, left circumflex and right coronary artery) and then values were summed to obtain the total calcium score; analysts were blinded to other information for all participants. Some measurements at the 15-year examination had been analysed using different software, but subsequently were re-analysed using Calcium Score to avoid potential bias. Two CT scans were performed for each participant at years 15 and 20 and the two scores were averaged for the analyses whereas, at the year 25 CT examination, a single scan was performed based on the high correlation between the paired scans and in order to reduce radiation exposure. As part of quality control for the reading protocol, an expert in cardiovascular imaging reviewed CT scans (i) that were initially discordant (one with and one without calcified plaque), (ii) that had a score >200 Hounsfield units (HU), (iii) for patients who had a change in CAC status from positive to negative between year 15 and year 20, (iv) for patients who may have undergone surgical intervention (pacemaker, valve replacement, coronary stent or bypass surgery) or (v) if a technical concern was identified by the reader. A positive Agatston score represented the presence of any CAC, which was defined as a calcium score higher than 130 HU and a minimum lesion size of approximately 2 mm² (four adjacent pixels).

High-resolution B-mode ultrasonography was used at year 20 according to a standard protocol to capture images of the bilateral common, bulb and internal carotid arteries (Logiq 700, General Electric Medical Systems, Chicago, IL, USA). The maximum IMT was defined as the mean of the maximal common carotid artery IMT of the near and far wall on both the left and right sides.

Measurement of other covariates

The demographic, lifestyle and medical characteristics of participants (e.g. age, race, sex, smoking status and medication use) were assessed during interviews. A validated CARDIA Physical Activity History Questionnaire was administered, which measured the frequency of performance of 13 different exercise activities in the past year [13]. A physical activity score was calculated by multiplying the frequency of participation by the intensity of activity. At the year 0, 7 and 20 examinations, dietary intake data were collected through an interviewer-administered diet history questionnaire specifically developed for the CARDIA study [14]. Participants were asked to report dietary intake for the past 28 days, including the frequency, amount and preparation methods. The validity and reliability of this diet history questionnaire were evaluated for 12 selected nutrients, and have been described in detail previously [15]. Alcohol consumption was recorded in mL/per week [16]. Daily nutrient intake was calculated using the nutrient database (version 36) developed by the University of Minnesota Nutrition Coordinating Center.

During clinic visits, participants wore light clothing and no shoes when measured for height (m, to the nearest 0.5cm) using a vertical ruler and weight (kg, to the nearest 0.2kg) using a balance beam scale. BMI was calculated as kg/m². Participants rested in the sitting position for 5 min before resting systolic and diastolic blood pressures were measured three times with a random zero sphygmomanometer on the right arm. The second and the third measurements were averaged and recorded.

The plasma total cholesterol, HDL cholesterol and triglyceride concentrations were assessed enzymatically [17] at the Northwest Lipid Metabolism and Diabetes Research Laboratories at the University of Washington (Seattle, WA). LDL-containing lipoproteins were precipitated with dextran sulphate/magnesium chloride before the HDL cholesterol level was determined, and then the LDL cholesterol level was estimated using the Friedewald equation for participants with ≤400 mg/dL triglycerides. Fasting glucose and insulin levels were quantified using the hexokinase method and standard radioimmunoassay, respectively [10]. The validity and reliability of duplicate measurements were evaluated [18]. Metabolic syndrome was defined according to the National Cholesterol Education Program Adult Treatment Panel III [19]. Serum creatinine concentration was measured by the Jaffe method and was used to estimate the glomerular filtration rate [20]. Serum C-reactive protein (CRP) concentration was measured at years 7, 15 and 20 using a Siemens Dade Behring BNII nephelometer (Deerfield, IL, USA) that utilizes a particle-enhanced immunonepholometric assay [21].

Statistical analysis

All analyses were conducted using SAS software (version 9.2, SAS Institute, Inc, Cary, NC). All P-values were two-sided and statistical significance was set at P < 0.05. sUA concentration was assessed in 5049, 3870, 3605 and 3147 of the 5115 participants at years 0, 10, 15 and 20, respectively; in addition, CAC data were unavailable for 3043, 3141 and 3189 participants at years 15, 20 and 25, respectively. At year 20, data were available on carotid IMT for 3254 participants.

Sample size

For our primary analyses, overall, we excluded subjects with missing or invalid sUA data at baseline (n = 73; invalid sUA measurements were those <1 mg/dL, n = 7, very likely due to laboratory measurement error), those with missing data for all three CAC measurements at years 15, 20 and 25 as well as carotid IMT measurement at year 20 (n = 1059), and those who had a low glomerular filtration rate (≤60 mL/min per 1.73 m²) at any of years 0, 10 and 15 (n = 73). Participants did not undergo CAC measurement due to pregnancy, severe obesity or lack of attendance on the measurement day (a different day from the clinic examinations). These exclusions resulted in a total of 3964 participants remaining for any of the following analyses. Participants who were included did not differ from those who were excluded, except that the latter group had a lower level of educated and were slightly more likely to be black or a smoker.

The sample size varied across different sUA exposures, subclinical atherosclerosis outcomes and covariates, therefore the final sample size is shown for each specific analysis in the tables and figures. Fig. 1 and Fig. S1 demonstrate the exclusion process and sample sizes for each analysis.

Fig. 1.

Flow chart for generating the appropriate study sample for each part of the primary analyses

Exposures

Because of the potential gender difference in sUA concentrations [22], sUA was modelled as gender-specific quartiles that were pooled for the analyses. Participant characteristics (i.e. in means or proportions) were described by gender-specific sUA quartiles. The primary longitudinal analyses included sUA at baseline (year 0), year 10 and year 15, and the change in sUA from year 0 to year 15 in separate models as study exposures.

Associations between sUA and CAC or maximum carotid IMT

Multiple discrete-time hazard regression models were used to examine the longitudinal association of year 0 and year 10 sUA with the presence of CAC at year 15 to 25 according to when it was first detected (year 15, 20 or 25), as well as the longitudinal association of year 15 sUA and the change of sUA from year 0 to year 15 with the presence of CAC at year 20 to 25. The presence of CAC is an approximate representation of the total progression of CAC (i.e. Agatston score changes from zero to positive, or an increase in the positive score). As reported previously [23], among 1797 participants who underwent CT scanning at years 15 and 20, changes in Agatston score were observed in 329 and almost all of these changes (n = 312) were increases.

We used multiple generalized linear regressions to examine the longitudinal association between sUA and the maximum carotid IMT at year 20. For all regression models, covariates at the year of sUA measurement were selected based on biological relevance and adjusted for in the final models. We focused on BMI as one of the classical CVD risk factors in the primary analyses, but other risk factors, such as waist circumference and metabolic syndrome were also examined.

To examine whether the associations between sUA and subclinical atherosclerosis differed by gender, an interaction term of sex and sUA (quartiles) was tested in models adjusted for demographic and lifestyle factors. A race–sUA interaction was also examined but did not reach statistical significance in any analysis (data not shown). Treating sUA concentration as a categorical variable yielded similar findings to those using continuous variables, indicating reasonable goodness of fit (data not shown).

Sensitivity analyses

We examined the association between hyperuricaemia (i.e. sUA ≥6.8 mg/dL [24, 25]) and the presence of CAC. The results of the sensitivity analyses are shown in the Online Supplement Table S1.

Results

Participant characteristics

The mean [standard deviation (SD)] sUA concentrations at years 0, 10, 15 and 20 in men were 6.15 (1.14), 6.36 (1.23), 6.42 (1.26) and 6.61 (1.30) mg/dL, respectively. The corresponding values in women were lower: 4.47 (0.96), 4.62 (0.99), 4.74 (1.07) and 5.02 (1.21) mg/dL at years 0, 10, 15 and 20, respectively. sUA tracked over examinations, with sex-specific Pearson correlation approximately 0.7 over 5-year intervals, 0.6 over 10-year intervals and 0.5 over the 20-year interval comparing year 0 to year 20. Table 1 shows the median (and range) sUA concentrations by examination years in men and women.

Table 1.

Serum urate concentration at years 0, 10, 15 and 20 by gender-specific serum uric acid quartiles

	Quartiles of sUA concentration
	Q1	Q2	Q3	Q4
Y0 (mg/dL) (n = 3964)
Men	4.90 (1.10, 5.40)	5.80 (5.50, 6.10)	6.50 (6.20, 6.80)	7.50 (6.90, 11.20)
Women	3.50 (1.00, 3.80)	4.10 (3.90, 4.30)	4.60 (4.40, 5.00)	5.60 (5.10, 8.80)
Y10 (mg/dL) (n = 3437)
Men	5.03 (3.01, 5.43)	5.84 (5.54, 6.24)	6.65 (6.34, 7.05)	7.76 (7.15, 12.11)
Women	3.61 (2.20, 3.92)	4.22 (4.02, 4.42)	4.83 (4.52, 5.13)	5.74 (5.23, 9.89)
Y15 (mg/dL) (n = 3412)
Men	5.05 (3.34, 5.53)	5.91 (5.63, 6.20)	6.67 (6.29, 7.15)	7.91 (7.24, 11.91)
Women	3.63 (2.10, 4.01)	4.29 (4.01, 4.58)	4.96 (4.67, 5.34)	6.01 (5.44, 11.05)
Y20 (mg/dL) (n = 3082)
Men	5.24 (3.20, 5.65)	6.16 (5.75, 6.47)	6.88 (6.57, 7.29)	8.10 (7.39, 11.58)
Women	3.71 (1.97, 4.12)	4.53 (4.22, 4.83)	5.24 (4.94, 5.65)	6.37 (5.75, 10.56)

Data are presented as median (range).

sUA, serum urate; Y, year; Q, quartile.

As shown in Table 2, participants with elevated sUA concentrations at year 0 had greater BMI and waist circumference, higher levels of blood pressure, LDL cholesterol, triglycerides and creatinine, were more likely to have metabolic syndrome, but had a lower HDL cholesterol level. This adverse pattern of metabolic variables at higher sUA concentrations was also observed at year 15, but with substantial progression (Table 3). For example, the mean BMI of 26.5 kg/m² in the highest sUA quartile at baseline (Table 2) was only slightly higher than the mean BMI of 25.4 kg/m² in the lowest sUA quartile at year 15 (Table 3). The cumulative percentage of participants with metabolic syndrome increased over time (2.2%, 8.4%, 12.7%, 20.1%, 27.8% and 31.7% by years 0, 7, 10, 15, 20 and 25, respectively). The percentage of current smokers was reduced at later compared with earlier examinations; sedentariness also increased with time (data not shown).

Table 2.

Unadjusted characteristics of participants at year 0 by quartiles of sUA concentration at year 0 (n = 3764^a)

	Quartiles of year 0 sUA concentration
	Q1	Q2	Q3	Q4
n	988	882	923	971
Age (years)	25.0±3.6	25.1±3.6	25.1±3.6	25.1±3.6
Women (%)	55.7	52.6	60.5	54.5b
Whites (%)	44.6	53.1	56.0	57.8b
More than high school education (%)	61.8	65.4	65.6	66.1
Former smoker (%)	27.4	25.5	26.1	28.4
Current smoker (%)	13.4	13.0	14.7	14.4
Physical activity score	408±295	406±277	409±291	432±305c
Total energy intake (kcal)	2832±1329	2753±1247	2682±1258	2783±1299
Alcohol intake (mL/day)	8.8±14.8	10.0±15.7	11.6±20.6	14.3±23.2c
Protein intake (g)	103.5±50.4	100.3±46.3	99.0±48.0	102.5±49.9
BMI (kg/m2)	22.9±3.5	23.9±4.4	24.5±4.7	26.5±5.9c
Waist circumference (cm)	74.2±8.3	76.7±10.2	77.3±10.4	82.2±13.1c
Systolic blood pressure (mmHg)	108.7±10.4	110.2±10.5	109.4±10.2	112.0±11.3c
Diastolic blood pressure (mmHg)	67.2±9.0	68.6±9.2	68.2±9.3	69.8±9.9c
Metabolic syndrome (%)	0.6	1.5	1.6	5.2b
Biochemical markers
HDL-C (mg/dL)	55.5±12.1	53.8±12.8	52.9±12.6	50.7±13.7c
LDL-C (mg/dL)	105.7±30.9	109.3±30.0	110.1±30.1	114.0±32.0c
Triglycerides (mg/dL)	60.8±28.7	65.3±54.6	72.3±41.4	88.4±65.2c
Creatinine (mg/dL)	0.77±0.13	0.80±0.13	0.79±0.13	0.82±0.14c

Data are presented as mean±SD or percentage.

sUA, serum urate; Q, quartile; BMI, body mass index; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol.

^aParticipants who were included in Table 4 in the prediction of CAC presence from year 0 sUA.

^bP < 0.05 for the difference across sUA quartiles.

^cP < 0.05 for the trend across sUA quartiles.

Table 3.

Unadjusted metabolic profile at year 15 by quartiles of sUA concentration at year 15 (n = 3083^a)

	Quartiles of year 15 sUA concentration
	Q1	Q2	Q3	Q4
n	765	785	776	757
BMI (kg/m2)	25.4±4.6	27.2±5.4	29.4±6.5	32.7±7.4b
Waist circumference (cm)	82.0±12.1	85.9±12.7	90.8±14.3	98.2±14.9b
Systolic blood pressure (mmHg)	109.0±12.3	111.1±13.6	113.9±14.7	116.8±15.8b
Diastolic blood pressure (mmHg)	71.1±10.3	73.3±10.7	75.0±10.9	77.5±12.2b
Metabolic syndrome (%)	4.8	10.3	17.0	35.3c
Biochemical markers
HDL-C (mg/dL)	54.5±14.1	51.9±15.0	50.0±14.0	46.6±13.9b
LDL-C (mg/dL)	106.7±29.2	111.4±31.5	116.7±34.1	118.0±32.7b
Triglycerides (mg/dL)	79.1±48.2	93.8±57.3	107.2±113.1	136.6±118.4b
Creatinine (mg/dL)	0.83±0.17	0.83±0.17	0.86±0.18	0.87±0.19b

Data are presented as mean±SD or percentage.

sUA, serum urate; Q, quartile; BMI, body mass index; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol.

^aParticipants who were included in Table 4 in the prediction of CAC presence from year 15 sUA.

^bP < 0.05 for the trend across sUA quartiles.

^cP < 0.05 for the difference across sUA quartiles.

Association between sUA concentration and CAC

Adjusting for baseline age, sex, race, clinic, education level, physical activity, smoking status and intakes of alcohol, protein and total calories (Table 4), baseline sUA concentration was positively associated with the presence of CAC during year 15 to year 25 (P_trend < 0.001). Compared to participants with sUA concentration at year 0 in the lowest quartile, those in the highest quartile had a 44% [95% confidence interval (CI) 20–73%] greater risk of CAC progression from year 15 to year 25. This positive association was substantially attenuated with further adjustment for baseline BMI (P_trend = 0.45). Prediction from sUA at year 10 yielded similar findings. On the other hand, after adjusting for year 15 demographic and lifestyle covariates, the risk of CAC progression at years 20 or 25 was 107% (95% CI 66–158%) greater among participants in the highest versus those in the lowest quartile of year 15 sUA. This positive association was attenuated but remained significant with additional adjustment for BMI at year 15 (P_trend = 0.01). The association between sUA and CAC appeared to become stronger as participants aged (i.e. β coefficients became larger across models for successive sUA measurement years). Controlling for waist circumference, metabolic syndrome (or individual components) and/or CRP, instead of BMI, did not influence the results (data not shown). No interaction between sex and sUA was found in relation to the presence of CAC; stratified analysis by sex yielded similar results to those described above (data not shown). Consistent with the primary analysis, the positive associations between hyperuricaemia status and CAC presence were significantly attenuated at years 0 and 10, but not at year 15, after additional adjustment of BMI (Online Supplement Table S1).

Table 4.

Hazard ratio (95% CI) of the presence of CAC across quartiles of sUA concentrations^a

	Quartiles of sUA concentration				Ptrend	βi ±SE	Pj
	Q1	Q2	Q3	Q4
Y0 sUA in relation to the presence of CAC during Y15 to Y25 (n = 3764)
Y0 sUA concentration (median and range)	3.70 (1.00, 5.40)	4.30 (3.90, 6.10)	4.90 (4.40, 6.70)	6.60 (5.10, 11.20)
No. of CAC cases	247	256	243	319
CAC rates/1000 person-years	10.99	12.85	11.56	14.60
Model 1b	1.00	1.17 (0.97, 1.42)c	1.13 (0.93, 1.37)	1.44 (1.20, 1.73)	<0.001	0.14±0.03	<0.001
Model 1 + Y0 BMI	1.00	1.06 (0.88, 1.29)	0.99 (0.81, 1.21)	1.09 (0.90, 1.33)	0.45	0.05±0.03	0.18
Y10 sUA in relation to the presence of CAC during Y15 to Y25 (n = 3316)
Y10 sUA concentration (median and range)	3.92 (2.20, 5.43)	4.42 (4.02, 6.24)	5.03 (4.52, 7.05)	6.85 (5.23, 12.11)
No. of CAC cases	205	219	244	284
CAC rates/1000 person-years	19.26	20.74	23.34	27.41
Model 2d	1.00	1.03 (0.84, 1.27)e	1.34 (1.09, 1.64)	1.57 (1.29, 1.91)	<0.001	0.18±0.03	<0.001
Model 2 + Y10 BMI	1.00	0.97 (0.79, 1.20)	1.14 (0.93, 1.41)	1.17 (0.95, 1.45)	0.08	0.08±0.03	0.01
Y15 sUA in relation to the presence of CAC during Y20 to Y25 (n = 3083)
Y15 sUA concentration (median and range)	3.91 (2.10, 5.53)	4.48 (4.01, 6.20)	5.24 (4.67, 7.15)	7.05 (5.44, 11.91)
No. of CAC cases	169	217	251	281
CAC rates/1000 person-years	25.07	31.91	37.41	43.70
Model 3f	1.00	1.47 (1.17, 1.84)g	1.68 (1.35, 2.09)	2.07 (1.66, 2.58)	<0.001	0.22±0.03	<0.001
Model 3 + Y15 BMI	1.00	1.35 (1.08, 1.70)	1.41 (1.12, 1.77)	1.51 (1.19, 1.91)	0.01	0.13±0.03	<0.001
Change of sUA from Y0 to Y15 in relation to the presence of CAC at Y20 to Y25 (n = 3083)
Change of sUA (median and range)	−0.86 (−5.73, −0.32)	−0.06 (−0.50, 0.25)	0.53 (0.20, 0.90)	1.37 (0.84, 6.03)
No. of CAC cases	195	217	243	263
CAC rates/1000 person-years	28.8	32.4	36.5	40.1
Model 3	1.00	1.31 (1.05, 1.64)h	1.62 (1.29, 2.03)	1.87 (1.49, 2.35)	<0.001	0.21±0.04	<0.001
Model 3 + Y15 BMI	1.00	1.20 (0.96, 1.51)	1.38 (1.09, 1.74)	1.45 (1.14, 1.85)	0.002	0.12±0.04	0.004
Model 3 + Y0 BMI	1.00	1.24 (0.99, 1.56)	1.49 (1.19, 1.88)	1.69 (1.34, 2.13)	<0.001	0.17±0.04	<0.001
Model 3 + Y0 BMI + change in BMI from Y0 to Y15	1.00	1.21 (0.96, 1.52)	1.40 (1.11, 1.77)	1.54 (1.21, 1.97)	<0.001	0.14±0.04	<0.001

sUA, serum urate; CAC, coronary artery calcified plaque; Y, year; Q, quartile; BMI, body mass index; CI, confidence interval.

^aAs described in the Methods, very few participants had CAC regression over time. Therefore, the presence of CAC closely represents the progression of CAC (i.e. Agatston score changes from zero to positive, or an increase in the positive value).

^bModel 1: adjusted for year 0 age, sex, race, clinic, education level, smoking status, physical activity and intakes of total calories, alcohol and protein.

^cHazard ratio (95% CI) for the presence of CAC during year 15 to year 25 across Y0 sUA quartiles.

^dModel 2: adjusted for age, sex, race, clinic, education level, smoking status and physical activity at year 10, and average intakes of total calories, alcohol and protein at years 0 and 7.

^eHazard ratio (95% CI) for the presence of CAC during year 15 to year 25 across Y10 sUA quartiles.

^fModel 3: adjusted for age, sex, race, clinic, education level, smoking status and physical activity at year 15, and average intakes of total calories, alcohol and protein at years 0 and 7.

^gHazard ratio (95% CI) for the presence of CAC during year 20 to year 25 across Y15 sUA quartiles.

^hHazard ratio (95% CI) for the presence of CAC during year 20 to year 25 across the quartiles of sUA change.

ⁱAssociation coefficients when using continuous sUA variables.

^jP-values for the association between sUA and CAC when using continuous sUA variables.

Association between sUA concentration and carotid IMT

There was a significant gender difference in the association between baseline sUA and common carotid maximum IMT at year 20 (Table 5, P_interaction < 0.001). Adjusting for baseline demographic and lifestyle factors, the positive association between baseline sUA and IMT was significant in men (P_trend < 0.001), but not in women (P_trend = 0.08). Further controlling for BMI at year 0 eliminated the association in men (P_trend = 0.08). By contrast, sUA at year 15 significantly predicted greater IMT at year 20 in both men and women (P_interaction = 0.09, both P_trend < 0.001), while further controlling for year 15 BMI eliminated the association in women (P_trend = 0.28), but not in men (P_trend = 0.001). Adjustment for waist circumference instead of BMI did not substantially change the findings (data not shown).

Table 5.

Maximum common carotid IMT (mm) at year 20 across quartiles of sUA concentration

		Quartiles of sUA concentration				Ptrend	βg ±SE	Ph
		Q1	Q2	Q3	Q4
Y0 sUA (n = 3108) Pinteraction < 0.001a
Men		4.90 (1.10, 5.40)b	5.80 (5.50, 6.10)	6.50 (6.20, 6.80)	7.50 (6.90, 11.20)
	Model 1d	0.807±0.007c	0.838±0.007	0.828±0.007	0.858±0.008	<0.001	0.015±0.003	<0.001
	Model 1 + Y0 BMI	0.820±0.007	0.840±0.007	0.827±0.007	0.843±0.008	0.08	0.006±0.003	0.07
Women		3.50 (1.40, 3.80)	4.10 (3.90, 4.30)	4.60 (4.40, 5.00)	5.60 (5.10, 8.80)
	Model 1	0.778±0.005	0.780±0.005	0.787±0.005	0.788±0.005	0.08	0.007±0.003	0.008
	Model 1 + Y0 BMI	0.785±0.005	0.783±0.005	0.786±0.005	0.776±0.005	0.22	−0.001±0.003	0.57
Y10 sUA (n = 2787) Pinteraction = 0.21
Men		5.03 (3.01, 5.43)	5.84 (5.54, 6.14)	6.55 (6.24, 7.05)	7.76 (7.15, 12.11)
	Model 2e	0.810±0.008	0.828±0.008	0.841±0.007	0.865±0.008	<0.001	0.018±0.003	<0.001
	Model 2 + Y10 BMI	0.822±0.008	0.833±0.007	0.839±0.007	0.846±0.008	0.02	0.013±0.003	<0.001
Women		3.61 (2.20, 3.92)	4.22 (4.02, 4.42)	4.83 (4.52, 5.13)	5.74 (5.23, 9.89)
	Model 2	0.766±0.005	0.779±0.006	0.790±0.005	0.798±0.005	<0.001	0.008±0.003	0.01
	Model 2 + Y10 BMI	0.775±0.005	0.783±0.005	0.787±0.005	0.784±0.006	0.25	0.003±0.003	0.24
Y15 sUA (n = 2795) Pinteraction = 0.09
Men		5.15 (3.34, 5.53)	5.91 (5.63, 6.20)	6.67 (6.29, 7.15)	8.01 (7.24, 11.91)
	Model 3f	0.799±0.008	0.841±0.008	0.847±0.008	0.870±0.008	<0.001	0.020±0.003	<0.001
	Model 3 + Y15 BMI	0.813±0.008	0.846±0.008	0.847±0.007	0.853±0.008	0.001	0.011±0.003	<0.001
Women		3.63 (2.10, 4.01)	4.29 (4.01, 4.58)	4.96 (4.67, 5.24)	5.91 (5.34, 10.10)
	Model 3	0.763±0.006	0.784±0.005	0.790±0.006	0.802±0.005	<0.001	0.014±0.002	<0.001
	Model 3 + Y15 BMI	0.774±0.006	0.788±0.005	0.786±0.006	0.785±0.006	0.28	0.002±0.003	0.38
Change in sUA, from Y0 to Y15 (n = 2795) Pinteraction = 0.79
Men		−0.97 (−4.29, −0.50)	−0.14 (−0.50, 0.20)	0.53 (0.20, 0.90)	1.48 (0.90, 6.03)
	Model 3	0.813±0.008	0.837±0.008	0.848±0.008	0.860±0.008	<0.001	0.017±0.004	<0.001
	Model 3 + Y15 BMI	0.824±0.008	0.839±0.008	0.847±0.008	0.848±0.008	0.02	0.009±0.004	0.01
	Model 3 + Y0 BMI	0.818±0.008	0.836±0.008	0.848±0.008	0.854±0.008	<0.001	0.013±0.003	<0.001
	Model 3 + Y0 BMI + change in BMI from Y0 to Y15	0.823±0.008	0.837±0.008	0.847±0.008	0.850±0.008	0.01	0.010±0.004	0.006
Women		−0.78 (−5.73, −0.33)	−0.04 (−0.33, 0.23)	0.51 (0.23, 0.80)	1.25 (0.81, 5.23)
	Model 3	0.768±0.006	0.776±0.006	0.790±0.005	0.804±0.006	<0.001	0.016±0.003	<0.001
	Model 3 + Y15 BMI	0.782±0.006	0.777±0.006	0.785±0.005	0.789±0.006	0.21	0.004±0.003	0.26
	Model 3 + Y0 BMI	0.772±0.006	0.774±0.005	0.787±0.005	0.797±0.005	<0.001	0.011±0.003	<0.001
	Model 3 + Y0 BMI + change in BMI from Y0 to Y15	0.779±0.006	0.775±0.006	0.785±0.005	0.791±0.006	0.08	0.005±0.003	0.11

sUA, serum urate; IMT, intima–media thickness; Y, year; Q, quartile; BMI, body mass index.

^aSex difference was tested in basic models without adjusting for BMI.

^bMedian and range of sUA concentrations for all such values.

^cMean and standard error of maximum common carotid IMT for all such values.

^dModel 1: adjusted for age, race, clinic, education level, smoking status, physical activity and intakes of total calories, alcohol and protein at year 0.

^eModel 2: adjusted for year 0 sUA, year 10 age, race, clinic, education level, smoking status and physical activity and average intakes of total calories, alcohol and protein at years 0 and 7.

^fModel 3: adjusted for year 0 sUA, year 15 age, race, clinic, education level, smoking status and physical activity and average intakes of total calories, alcohol and protein at years 0 and 7.

^gAssociation coefficients when using continuous sUA variables.

^hP-values for the associations between sUA and IMT when using continuous sUA variables.

Greater increment in sUA over time predicted higher risk of subclinical atherosclerosis

As demonstrated in Tables 4 and 5, a greater increment in sUA concentrations from year 0 to year 15 was associated with a higher risk of CAC progression at years 20 and 25, as well as with greater carotid IMT at year 20 (all P_trend < 0.001). The associations remained significant with further adjustment for BMI at year 15 (P_trend = 0.002 for CAC, and P_trend = 0.02 for IMT in men) or the change in BMI from year 0 to year 15 (P_trend < 0.001 for CAC, and P_trend = 0.01 for IMT in men).

Discussion

Among this young to middle-aged adult cohort with 25-year follow-up, a higher sUA concentration was associated with an increased risk of subclinical atherosclerosis, including CAC progression (i.e. Agatston score increases from zero to positive, or an increase in the positive score) and greater carotid IMT. The prediction of subclinical atherosclerosis based on sUA concentration in young adults was largely attributable to the correlation between this serum factor and classical CVD risk factors, such as BMI. However, as individuals reached early middle age at the time of the CARDIA year 15 examination (ages 33–45 years), an independent role of sUA in prediction of subclinical atherosclerosis became evident. Thus higher sUA concentration in early middle age, perhaps coupled with other metabolic disorders that have developed by that time, may carry independent risk for subclinical atherosclerosis.

Despite some possible antioxidant capacity of sUA [3, 26], evidence from animal models and epidemiological studies have clearly demonstrated an association between increases in sUA concentration and adverse CVD risk factors [4], including several inflammatory biomarkers such as CRP and IL-6 [27]. Metabolic syndrome and hypertension are prevalent among individuals who have high sUA concentrations [4], as also observed in this study. Similarly, in a recent longitudinal study among 245 participants, sUA concentration was found to be a significant predictor of both current and future components of metabolic syndrome [23, 28]. Moreover, in two meta-analyses of prospective cohort studies among adults, Kim et al. reported that hyperuricemia was related to significantly higher risk of the incidence and mortality of stroke and coronary heart disease (CHD) [29, 30]. In this context, our findings of a positive association between sUA concentration (as well as hyperuricaemia status) and subclinical atherosclerosis were consistent with these previous observations.

Several possible mechanisms have been proposed to explain the adverse effect of elevated sUA. For example, the adverse effect may be attributed to the oxidant by-products of sUA generation, leading to further reaction with endogenous nitric oxide (NO), decreasing the bioavailability of NO, and inducing endothelial dysfunction [31]. XOR has been widely linked to atherosclerosis-related pathological conditions and cardiac tissue damage [5]. Thus it has been suggested that elevated sUA concentration may merely serve as a marker of XOR activity. This hypothesis was supported by the findings of studies using either allopurinol (an inhibitor of XOR that suppresses sUA generation) or probenecid (a uricosuric agent that reduces sUA concentration by increasing uric acid excretion in urine); endothelial function was dramatically improved by allopurinol, but not by probenecid [32, 33]. On the other hand, it has been shown that elevated sUA directly stimulates vascular smooth muscle cell proliferation and infiltration via the mediation of several important proinflammatory factors and signalling cascades in the pathogenesis of atherosclerosis [34–36]. In addition, hyperuricaemia per se was shown to induce endothelial dysfunction by inhibiting the synthesis and release of NO [37], and the renin–angiotensin system in vascular endothelial cells (a hormonal vasoconstriction system) was found to be activated by elevated sUA [35]. Elevated sUA concentration may be a result of impaired renal function, which in turn may accelerate renal and vascular damage, mediate vasoconstriction and lead to weight gain and hypertension [4]. Several studies also demonstrated an inverse association between insulin resistance and renal excretion of uric acid [38, 39].

Given these findings and proposed mechanisms, it is possible that elevated sUA could lead to atherosclerosis independently of established CVD risk factors, e.g. BMI, hypertension, hyperlipidaemia and insulin resistance [4, 40]; however, epidemiological evidence is still controversial. In a meta-analysis of prospective cohort studies including 164,542 disease-free adults (mean age of 50 years at baseline; mean follow-up of 10.5 years), complete adjustment for established CVD risk factors eliminated the positive relation between sUA and the prevalence and incidence of CHD that was found in the age- and sex-adjusted model [41]. In addition, in a cohort study following 13,504 middle-aged to old men and women for 8 years, no independent association was observed between increased sUA and risk of CHD [42]. By contrast, some studies have demonstrated that sUA was a strong predictor of CVD events, with significance sustained after adjustment for conventional CVD risk factors [43, 44]. For example, consistent with our findings, a longitudinal study with follow-up for about 6 years showed that sUA predicted CAC progression among subjects with type I diabetes without renal disease, independent of conventional CVD risk factors, such as the metabolic syndrome; the baseline mean sUA concentrations in men and women were approximately 6 and 5 mg/dL, respectively [45]. The heterogeneity of study methodologies, including differences in study duration, sample sizes and characteristics of study populations, may contribute partially to the conflicting evidence [6]. In addition, the association between sUA and CVD has been widely reported among the middle aged and elderly, but the evidence for this relationship is limited among young and generally healthy populations. Relatively few clinical CVD events occur in young populations compared with the elderly, which reduces the statistical power for assessing the relationship. In this regard, subclinical atherosclerosis may be of particular interest, in addition to its significant added value for improving the prediction of clinical CVD outcomes.

To our knowledge, only a few studies have examined the longitudinal association between sUA and subclinical atherosclerosis, and the conclusions from cross-sectional studies have been varied. For instance, in two studies among the same population, Neogi et al. demonstrated that sUA was positively related to the presence of carotid plaque but not CAC [46, 47]. Oikonen and co-workers reported that the association between sUA and early atherosclerosis was eliminated after adjustment for BMI among adults aged 30–45 years [8]. Recently, Krishnan et al. reported a subset of the data presented herein, namely the cross-sectional CARDIA year 15 examination results [48]. They found that sUA was associated cross-sectionally with CAC, assessed in all participants at 33–45 years of age. Our additional investigation of the role of sUA in a longitudinal setting adds important new information for determining the underlying aetiology and addressing the controversy in the existing evidence. Of particular clinical importance is that assessment of sUA in young adulthood could be a harbinger of future subclinical atherosclerosis, considering the strong and independent associations measured in early middle age between sUA and CAC, and the progressive increase in sUA from young adulthood into middle age.

The strengths of this study included the prospective study design with 25-year follow-up of people from young adult to middle age, the general good health of the study sample, the repeated measurements of both exposures and outcomes and the assessment of two aspects of subclinical atherosclerosis. These strengths facilitated the examination of the temporal association between sUA and subsequent subclinical atherosclerosis, although causality remains undetermined as for all observational studies. Although residual confounding may not be ruled out, the current analyses carefully tested and controlled for various factors that may influence the association between sUA and CVD. However, our findings may not be generalized to older populations, given that so far we have only followed participants up to middle-age. We found that women had lower mean sUA concentrations than men, which may be due to the action of the hormone oestrogen [22], and that the magnitude of the longitudinal association between sUA and IMT differed by gender.

In conclusion, we observed a significantly positive correlation between sUA concentration and subclinical atherosclerosis among young to middle-aged adults during a 25-year follow-up. Our data suggest that sUA is responsive to metabolic factors such as BMI during young adulthood, but the association of sUA with subclinical atherosclerosis strengthened with age and became independent of BMI by early middle age.. Thus we hypothesize that higher sUA concentrations indicate the presence of pathways that are linked to other metabolic disorders commonly emerging by middle age. Given the high correlation between sUA measurement in young adulthood and across many years, a relatively high sUA concentration measured in young adulthood may alert a clinician to future risk; thus increased attention to sUA and other metabolic factors is warranted as a patient moves into middle age.