Biomarkers of Cardiovascular Stress and Subclinical Atherosclerosis in the Community

Abstract

BACKGROUND

Biomarkers of cardiovascular stress have been associated with incident cardiovascular outcomes. Their relations with measures of subclinical atherosclerosis, as assessed by carotid intima-media thickness, have not been well described.

METHODS

Plasma growth differentiation factor-15 (GDF-15), soluble ST2 (sST2), and high-sensitivity troponin I (hsTnI) were measured in 3111 Framingham Offspring participants who also underwent carotid ultrasonography during the sixth examination (1995 – 1998, mean age 58 years, 54% women). Carotid measurements included maximal internal carotid artery (ICA) intima-media thickness (IMT), plaque presence (defined as ICA IMT > 1.5 mm), and mean common carotid artery IMT. Multivariable regressions for carotid measurements versus biomarkers were carried out using linear and logistic models; P < 0.0056 was deemed statistically significant.

RESULTS

Maximal ICA IMT was significantly associated with plasma GDF-15 (β-estimate 0.04 per 1 unit increase in log-GDF-15 SE 0.01, P < 0.0001). Similarly, the odds of having carotid plaque increased 33% (OR 1.33 per 1-unit increase in log-GDF-15, 95% CI 1.20-1.48, P < 0.0001). In contrast, there was no significant association of maximal ICA IMT or plaque presence with sST2 or hsTnI, and none of the three biomarkers was significantly associated with mean CCA IMT. GDF-15 was a stronger predictor of maximal ICA thickness and plaque presence compared with BNP and CRP when these conventional biomarkers were tested together.

CONCLUSION

Higher GDF-15 concentrations are associated with subclinical atherosclerosis, including maximal ICA IMT and carotid plaque presence. Future studies investigating the role of GDF-15 for screening and management of patients with subclinical atherosclerosis are warranted.

Carotid ultrasonography, via carotid intima-media thickness (IMT) assessment, has been well established not only as a surrogate of generalized atherosclerosis (1, 2) and linked with cardiovascular risk factors (3, 4), but also has been shown to be a predictor for myocardial infarction, stroke, coronary heart disease, and combined cardiovascular outcomes (5–7). In addition, carotid IMT has been shown to track subclinical atherosclerosis and help improve classification of cardiovascular risk in asymptomatic cohorts (4, 8).

Recently, biomarkers of cardiovascular stress, including growth differentiation factor-15 (GDF-15), soluble ST2 (sST2), and high-sensitivity troponin (hsTnI), have emerged as strong predictors of cardiovascular risk and outcomes in various community-based populations (9–12). The exact mechanism by which these biomarkers are linked to cardiovascular disease, particularly in asymptomatic, population-based cohorts are incompletely described to date.

Our group has previously demonstrated the longitudinal association of GDF-15, sST2, and hsTnI to incident cardiovascular outcomes in the community (11). Given that all three biomarkers previously related to incident cardiovascular disease, we hypothesized that these biomarkers would be associated cross-sectionally with measures of subclinical atherosclerosis, as assessed via carotid IMT of the common carotid artery and internal carotid artery. Further characterizing associations with subclinical cardiovascular disease measures may offer further insights into the role of these biomarkers in the development of atherosclerosis.

MATERIALS AND METHODS

Study Sample

The Framingham Offspring Study cohort, initiated in 1971 and comprising offspring (and their spouses) of the original Framingham cohort, underwent examinations every 4 – 8 years. The present investigation included participants who attended the sixth Offspring examination (1995 – 1998) and underwent carotid ultrasonography. From 3380 participants, 37 were excluded for missing carotid IMT data, 21 because of missing covariate data, 173 because of prevalent cardiovascular disease, 4 because of missing sST2 data, and 34 because of missing hsTnI data, leaving 3111 participants for analysis. The Institutional Review Board at Boston Medical Center approved the study protocol and all study participants provided written informed consent.

Risk Factor Assessment

Each examination included a comprehensive history and physical examination. Cardiovascular risk factors pertinent to our study included systolic blood pressure, diastolic blood pressure, antihypertensive drug treatment, diabetes (defined as fasting glucose ≥ 126 mg/dL (≥ 7 mmol/L) or the use of oral hypoglycemic medications or insulin, current smoking (defined as smoking at least 1 cigarette per day over the past year), total and HDL cholesterol concentrations (measured after an overnight fast), and body mass index.

Biomarker measurements

Blood samples were obtained from study participants in the morning following an overnight fast. The citrated plasma samples were centrifuged immediately and stored at −80° C until assayed. No freeze-thaw cycles were performed prior to the assays described below. Reference limits and assay characteristics have previously been described (11, 13, 14). In brief, a pre-commercial, automated electrochemiluminiscent immunoassay utilizing a Cobas e 411 analyzer (Roche Diagnostics) at a core laboratory (K.C. Wollert) was used to measure plasma GDF-15 concentrations. The assay had an intraassay coefficient of variation (CV) of 0.8% and interassay CV of 2.3% at low concentrations (1120 ng/L), and intra- and interassay CV of 1.1 and 1.0% at high concentrations (9031 nl/L). sST2 concentrations were measured with a high-sensitivity sandwich immunoassay (Presage™ ST2, Critical Diagnostics) at Critical Diagnostics. This assay had an interassay CV of 7.5% at low (25.6 ng/mL) and 6.0% at high (70.9 ng/mL) sST2 concentrations. Single-molecule counting technology (Erenna hsTnI, Singulex) was used for measurement of hsTnI at Singulex. The assay had an interassay CV of 10.0% at low (4.7 ng/L) and 7.7% at high (19.0 n/L) hsTnI concentrations. B-type natriuretic peptide (BNP) and C-reactive protein (CRP) concentrations were measured as previously described (15). In brief, BNP was measured using a high-sensitivity immunoradiometric assay at Shionogi (Shionogi, Osaka, Japan) with average interassay CV of 12.2%, and CRP was measured at the Framingham Heart Study core laboratory using a Dade Behring BN100 nephelometer with an average interassay CV of 2.2%.

Carotid ultrasonography

A single sonographer, certified by the Registry of Diagnostic Medical Sonographers, performed all carotid ultrasonography. Ultrasound image acquisition was obtained at the end of the R wave of an electrocardiogram to capture end-diastole. Using a standard 45° projection from vertical, data was acquired from images obtained at three levels: the common carotid artery (CCA), level of the carotid artery bulb, and proximal internal carotid artery (ICA). One additional image was taken of the carotid artery bulb and proximal ICA centered on the largest plaque. Intima-media thickness values were calculated from the intima-media interface lines drawn by a certified reader on each acquired image (4). Using a distance of approximately 0.5 cm below the carotid-artery bulb and measuring IMT over a 1 cm long segment (deemed not to have any plaque presence), the mean IMT of the CCA was determined. The largest IMT in either the left or right ICA from 1cm above the carotid sinus to the bulb determined the maximum ICA IMT. Maximum ICA IMT was also examined as a dichotomous variable, defined as plaque presence if ICA IMT > 1.5 mm. Replicate measurements in 37 participants was performed for assessment of reproducibility with Pearson correlation coefficients (between replicate measurements) for the mean CCA IMT and maximum ICA IMT, 0.94 and 0.76, respectively (4).

Statistical Analyses

All biomarkers and two carotid IMT outcomes (maximum ICA IMT, and mean CCA IMT) were natural-logarithmically transformed owing to their right-skewed distributions. In primary analyses, multivariable linear regression was used to examine the associations of maximum ICA IMT and mean CCA IMT with each biomarker (GDF-15, sST2, and hsTnI). Beta-estimates were expressed as relative change in IMT per 1-standard deviation (SD) change in log-transformed biomarker, and were back-transformed for interpretation. For the carotid outcome of plaque presence, multivariable logistic regression modeling was applied, with reported odds ratios per 1-SD change in log-transformed biomarker. Adjustment for age, sex, systolic blood pressure, hypertension treatment, total cholesterol, HDL cholesterol, diabetes, smoking, body mass index, and baseline estimated glomerular filtration rate was used in primary models. To account for testing of multiple biomarkers (n=3) and carotid measures (n=3), results for primary analyses were considered significant at a Bonferroni-corrected P-value threshold of 0.05/9 = 0.0056.

In secondary analyses, we tested effect modification by age and sex for associations of three biomarkers and carotid outcomes. In addition, we further adjusted multivariable models for CRP and BNP, which have previously been shown to be associated with carotid IMT measures (16, 17). A P-value threshold of 0.05 was deemed significant in secondary analyses. Analyses were conducted with SAS version 9.2.

RESULTS

The study sample contained 3111 Framingham Offspring cohort participants, 54% were female with a mean age of 58 years. Twenty-six percent were treated for hypertension, 15% were current smokers, and 9% had diabetes. Clinical, biomarker, and carotid measurements are shown in Table 1.

Table 1.

Baseline characteristics of 3111 participants free of prevalent CVD

	Total sample (n=3111)
Clinical characteristics
Age, years	58 (10)
Female sex, n (%)	1687 (54)
Systolic blood pressure, mmHg	128 (19)
Treatment for high blood pressure, n (%)	809 (26)
Diabetes, n (%)	274 (9)
Body mass index, kg/m2	27.9 (5.2)
Current smoker, n (%)	470 (15)
Total cholesterol, mg/dL	206 (40)
HDL, mg/dl	51 (16)
Biomarkers (median, 25th – 75th percentile)
GDF-15, ng/L	1028 (809–1334)
sST2, ng/mL	20.8 (16.6–26.0)
hs-TnI, ng/L	1.3 (0.9–2.2)
Intima-media thickness (median, 25th–75th percentile)
Maximum ICA IMT, mm	1.11 (0.81–1.78)
Plaque presence (ICA IMT > 1.5 mm), n (%)	1025 (33)
Mean CCA IMT, mm	0.58 (0.51–0.66)

HDL, high-density lipoprotein; GDF-15, growth differentiation factor-15; sST2, soluble ST2; hsTnI, high-sensitivity troponin I; CCA, common carotid artery; ICA, internal carotid artery.

Values are mean (standard deviation) unless otherwise indicated.

To convert cholesterol in mg/dL to mmol/L, multiply by 0.02586.

Associations of carotid measures with GDF-15, sST2, and hsTnI

In age- and sex-adjusted regression models, maximal ICA IMT was significantly associated with each biomarker. After multivariable adjustment (including age, sex, systolic blood pressure, hypertension treatment, total and HDL cholesterol, diabetes, smoking, body mass index) the association remained significant only for log-GDF-15 (β-estimate 0.04, SE 0.01, P < 0.0001, Table 2). Specifically, maximal ICA IMT increased 4% per 1-SD increase in log- GDF-15. Neither sST2 nor hsTnI were associated with maximal ICA IMT after multivariable adjustment.

Table 2.

Association of carotid measures with GDF-15, sST2, and hsTnI

Outcome	Biomarker	Age- and sex-adjusted model		Multivariable-adjusted modela, b
Maximum ICA IMTd		β estimate (SE)b	Pc	β estimate (SE)b	Pc
	GDF-15	0.070 (0.009)	<0.0001	0.04 (0.010)	<0.0001
	sST2	0.025 (0.008)	0.003	0.013 (0.008)	0.10
	hsTnI	0.024 (0.008)	0.004	0.02 (0.008)	0.01
Plaque Presence		OR (95% CI)b	Pc	OR (95% CI)b	Pc
	GDF-15	1.48 (1.34–1.63)	<0.0001	1.33 (1.20–1.48)	<0.0001
	sST2	1.15 (1.05–1.25)	0.002	1.10 (1.00–1.20)	0.05
	hsTnI	1.11 (1.02–1.21)	0.02	1.09 (1.00–1.19)	0.04
Mean CCA IMTd		β estimate (SE)b	Pc	β estimate (SE)b	Pc
	GDF-15	0.014 (0.004)	0.0003	0.004 (0.004)	0.30
	sST2	0.009 (0.003)	0.01	0.003 (0.003)	0.31
	hsTnI	0.009 (0.003)	0.007	0.007 (0.003)	0.03

GDF-15, growth differentiation factor-15; sST2, soluble ST2; hsTnI, high-sensitivity troponin I; CCA, common carotid artery; ICA, internal carotid artery

^aAdjusted for age, sex, systolic blood pressure, hypertension treatment, total cholesterol, high-density cholesterol, diabetes, smoking, body mass index, and estimated glomerular filtration rate

^bβ estimate and OR reflects per 1-SD increase in log-transformed biomarker

^cCorrected P-value threshold P = 0.0056

^dMean CCA and maximum ICA IMT were log-transformed;

In age- and sex-adjusted models, plaque presence (defined as ICA IMT > 1.5 mm) was significantly associated with GDF-15 and sST2. After multivariable adjustment, plaque presence remained significantly related to GDF-15 (odds ratio (OR) 1.33 per 1-SD increase in log-GDF-15, 95% CI 1.20-1.48, P < 0.0001, Table 2). Neither sST2 nor hsTnI were associated with plaque presence after multivariable adjustment.

The risk of plaque presence across increasing quartiles of the 3 biomarkers, GDF-15, sST2, and hsTnI are displayed in Figure 1. Increasing quartiles of GDF-15 were associated with increasing risk of plaque in multivariable analyses (P for trend < 0.0001). Specifically, participants in the upper quartile of GDF-15 had a nearly two-fold increased odds of carotid plaque compared with the lowest quartile (OR 1.95, 95% CI 1.44-2.64, P < 0.0001).

Figure 1.

Multivariable-adjusted odds ratios for carotid plaque presence across quartiles of GDF-15, sST2, and hs-TnI. Error bars represent 95% CI.

The associations of the biomarkers, GDF-15, sST2, and hsTnI, were evaluated with the outcome mean CCA IMT (Table 2). GDF-15 but not sST2 or hsTnI were significantly associated with mean CCA IMT in age- and sex- adjusted models, and none of the three biomarkers were significantly associated after multivariable adjustment at the pre-specified statistical threshold of P=0.0056.

Comparison with established biomarkers

After further adjusting multivariable models for BNP and CRP, biomarkers that previously have been linked with carotid measures, GDF-15 remained a significant predictor of maximal ICA IMT, conferring a 1.04-fold increase in carotid ICA thickness per 1-SD increase of log-GDF-15 (P=0.0002). In contrast, BNP was not associated with maximal ICA IMT in the multimarker model (P=0.48), and CRP only nominally so (P=0.04). Both GDF-15 and CRP were significant predictors of plaque presence in multivariable-adjusted models, and the association with GDF-15 appeared more significant (OR per 1-SD unit increase in log-GDF-15 1.32; 95% CI, 1.18-1.47; P < 0.0001; OR per 1-SD unit increase in log-CRP 1.15; 95% CI, 1.04-1.27; P = 0.008) (Table 3).

Table 3.

Association of GDF-15 and established biomarkers with carotid measures

Modela	Maximum ICA IMT		Plaque presence		Mean CCA IMT
	β estimate (SE)b	Pc	Odds ratio (95% CI)b	P	β estimate (SE)b	Pc
Individual biomarker
GDF-15	0.040 (0.01)	<0.0001	1.33 (1.20–1.48)	<0.0001	0.004 (0.004)	0.30
Biomarker combination
GDF-15	0.037 (0.01)	0.0002	1.32 (1.18–1.47)	<0.0001	0.004 (0.004)	0.36
BNP	0.006 (0.009)	0.48	0.98 (0.89–1.08)	0.71	0.003 (0.004)	0.33
CRP	0.019 (0.009)	0.04	1.15 (1.04–1.27)	0.008	0.002 (0.004)	0.53

GDF-15, growth differentiation factor-15; BNP, B-type natriuretic peptide; CRP, C-reactive protein; CCA, common carotid artery; ICA, internal carotid artery

^aAll models adjusted for age, sex, systolic blood pressure, hypertension treatment, total cholesterol, high-density cholesterol, diabetes, smoking, body mass index, and estimated glomerular filtration rate

^bβ estimate and OR reflects per 1-SD increase in log-transformed biomarker

^cCorrected P-value threshold P = 0.0056

For a given CRP tertile, GDF-15 added further information with regards to plaque risk (Figure 2). For example, when examining individuals in the highest tertile of CRP, those in the lowest tertile of GDF-15 had a 1.44-fold increased odds of plaque, compared with a 2.7-fold increased odds in the highest GDF-15 tertile (with those in the lowest tertile for both biomarkers serving as the referent group).

Figure 2.

Multivariable-adjusted odds ratios for carotid plaque presence across increasing tertiles of GDF and CRP. Referent group for odds ratio was GDF-15 tertile 1 and CRP tertile 1.

DISCUSSION

GDF-15, sST2, and hsTnI, have recently emerged as predictors of cardiovascular outcomes in both patients with existing cardiovascular disease, as well as in the community-dwelling population. The present investigation extends these findings and supports the concept that GDF-15 is strongly associated with subclinical atherosclerosis as measured by carotid ultrasonography, even before clinical cardiovascular disease is recognized. Moreover, GDF-15 was associated with carotid plaque independent of established biomarkers of cardiovascular risk, CRP and BNP.

GDF-15 is a divergent member of the transforming growth factor β cytokine family (18) that is upregulated in response to stressors including in macrophages exposed to oxidized LDL in atherosclerotic carotid arteries (19). It is expressed in several cell types, including macrophages, vascular smooth muscle cells, and endothelial cells in response to oxidative, inflammatory, or metabolic stressors (19–21).

Specific to atherosclerosis, GDF-15 has shown predictive abilities of coronary heart disease mortality and composite outcomes in stable and acute coronary syndromes in patients with prevalent cardiovascular risk factors (22–25). In community cohorts, GDF-15 has conferred prognostic abilities for composite cardiovascular outcomes as well as hard mortality outcomes (9–11, 26). Our findings support the concept that GDF-15 is in fact associated with subclinical atherosclerosis as assessed by maximal ICA IMT as well as the presence of carotid plaque.

In contrast to the strong association of GDF-15 with maximal ICA IMT, we found no such association with mean CCA IMT. Previous studies suggest that maximum ICA IMT and mean CCA IMT may reflect different underlying process, the former related to localized atherosclerotic plaque (27), and the latter representing diffuse arterial wall hypertrophy (28). Interestingly, maximal ICA and plaque presence, but not mean CCA IMT improved risk classification for incident cardiovascular disease in a previous study (29). Our findings suggest that GDF-15 is associated with focal atherosclerosis, rather than diffuse inflammatory processes and arterial thickening. Similar results were found in an older community-based cohort, where GDF-15 was associated with carotid plaque but not IMT (10). Our results extend these findings to a population more than a decade younger than the previous study, without known prevalent cardiovascular disease, and directly compare GDF-15 to other emerging and established biomarkers.

We found that the association of GDF-15 and carotid plaque is independent of CRP, a marker of inflammation. Notably, in multi-marker models with CRP and BNP, the association of GDF-15 with maximum ICA IMT and carotid plaque was more robust than either BNP or CRP. This suggests that GDF-15 may reflect an orthogonal pathway associated with cardiovascular disease, the mechanism of which remains unclear. Experimental studies suggest anti-thrombotic and anti-platelet effects (30). However, in vascular tissue studies, there is growing evidence that GDF-15 may promote increased inflammation and atherosclerosis progression, a process that may involve p53 pathway activation (31, 32). Whether GDF-15 is a mediator of cardiovascular disease or upregulated in response to cardiovascular injury remains unclear.

In comparison to GDF-15, sST2 and hsTnI exhibited no significant associations with carotid measures in our study. Soluble ST2, a member of the interleukin-1 cytokine family, has been shown to serve as a decoy receptor, binding IL-33 – a cytokine shown to have protective effects in the cardiovascular system (33, 34). In atherosclerosis, IL-33 administration in murine models of atherosclerosis led to significantly smaller atherosclerotic plaques while sST2 administration led to larger aortic sinus atherosclerotic plaques (35). Although sST2 has shown associations with cardiovascular outcomes in acute coronary syndrome (36, 37), more recent studies have not shown sST2 to be predictive of atherosclerotic plaque in both patients with carotid atherosclerotic disease (38) as well as patients free of vascular disease (39).

There are several limitations to the present study that merit discussion. Ours is an observational cross-sectional study, and causal inferences cannot be drawn. It is also important to note that carotid IMT may incompletely capture the complex process of atherosclerosis in multiple vascular beds (40), and interval carotid IMT measurements may confer more specificity in this process and relation with atherosclerotic outcomes. The use of our Bonferroni corrected P-value threshold may have been too conservative, given that two of the carotid measures (ICA IMT and plaque presence) are not independent of one another. However, even when using a less conservative P-value threshold, the main results did not differ materially. Lastly, our population was predominantly white, limiting generalizability to other ethnic groups.

In summary, we found that GDF-15 was associated with subclinical atherosclerosis, including higher maximum ICA IMT and the presence of carotid plaque in an ostensibly healthy community-based population without prevalent cardiovascular disease. This association was independent of other established biomarkers including CRP and BNP. Future studies are warranted to elucidate the potential use of GDF-15 for screening and management of patients with subclinical atherosclerosis.

Abbreviations

BNP

B-type natriuretic peptide

CCA

common carotid artery

CI

confidence interval

CRP

C-reactive protein

GDF-15

growth differentiation factor-15

hsTnI

high-sensitivity troponin I

ICA

internal carotid artery

IMT

intima-media thickness

OR

odds ratio

SD

standard deviation

sST2

soluble ST2