A multi-marker approach to evaluate correlates of vascular stiffness: The Framingham Heart Study

Abstract

Background

Arterial stiffness increases with age and contributes to the pathogenesis of systolic hypertension and cardiovascular disease in the elderly. Knowledge about pathophysiological processes that determine arterial stiffness may help guide therapeutic approaches.

Methods and Results

We related seven circulating biomarkers, representing distinct biological pathways (C-reactive protein [CRP], aldosterone-to-renin ratio [ARR], N-terminal pro–atrial natriuretic peptide and B-type natriuretic peptide, plasminogen activator inhibitor [PAI]-1, fibrinogen, homocysteine) to 5 vascular function measures (central pulse pressure, carotid-femoral pulse wave velocity, mean arterial pressure, forward pressure wave amplitude [all measures of conduit artery stiffness], and augmented pressure, an indicator of wave reflection) in 2,000 Framingham Offspring Study participants (mean age 61 years, 55% women). Tonometry measures were obtained on average three years after biomarkers were measured. In multivariable linear regression models adjusting for covariates, the biomarker panel was significantly associated with all 5 vascular measures (p<0.003 for all). Upon backwards elimination, the ARR was positively associated with each stiffness measure (p≤0.002 for all). In addition, CRP was positively related to augmented pressure (p=0.0003), whereas PAI-1 was positively associated with mean arterial pressure (p=0.003), central pulse pressure (p=0.001) and forward pressure wave (p=0.01).

Conclusions

Our cross-sectional data on a community-based sample suggest a distinctive pattern of positive associations of biomarkers of renin-angiotensin-aldosterone system activation with pan-arterial vascular stiffness, PAI-1 with central vascular stiffness indices, and of CRP with wave reflection. These observations support the notion of differential influences of biological pathways on vascular stiffness measures.

Stiffening of the large arteries is a sine quo non of vascular aging. Indeed, increased arterial stiffness contributes to the burden of cardiovascular disease in older individuals, being positively associated with systolic hypertension,¹ coronary heart disease,² stroke,² heart failure,³ and atrial fibrillation.⁴ The assessment of arterial waveforms at different sites in the human body by using applanation tonometry permits a detailed noninvasive characterization of stiffness in different vascular beds. Besides the positive relations to age, higher arterial stiffness has been related cross-sectionally to other vascular risk factors, including blood pressure, body mass index, impaired glucose tolerance, and dyslipidemia.⁵

Substantial research suggests that the steady state and pulsatile components of arterial load have a differential impact on cardiovascular disease risk and may have varying determinants. Furthermore, the concomitant increases in central and peripheral vascular stiffness with age are influenced by several mechanical and biological factors that result in altered vasoreactivity and arterial wall remodeling. The molecular events that are associated with remodeling of the large and medium-to-small arteries have also been well characterized, and involve the combinatorial influences of adhesion molecules, integrins, metalloproteinases, the renin-angiotensin axis, and inflammation on the cellular constituents (endothelial cells, vascular smooth cells, fibroblasts and matrix components) of the vasculature.^6,7 Notwithstanding these advances in our understanding, the primacy of one specific biological pathway over others in mediating the arterial remodeling process is not clearly established. The identification of key pathways implicated in vascular remodeling may offer the opportunity of reversing vascular stiffness by using targeted approaches.^8,9 One potential epidemiological method to identify key pathways involved in a multifactorial condition such as increased vascular stiffness is to evaluate the relations of circulating biomarkers from diverse pathways to vascular stiffness measures. In this context, previous studies (including one from our group) have reported positive associations between circulating biomarkers and direct and indirect measures of arterial stiffness,^10–13 but most studies analyzed single biomarkers or biomarkers for a single pathway; none simultaneously examined a comprehensive panel of biomarkers representing distinct physiological pathways. Accordingly, we related a panel of 7 systemic biomarkers representing inflammation, neurohormonal activation and hemostasis for association with measures of arterial stiffness in a large community-based sample.

Methods

Study Sample

Details of the design and selection criteria for the Offspring cohort of the Framingham Heart Study have been described elsewhere.¹⁴ In brief, children of participants of the original Framingham cohort and the spouses of these children were enrolled (n=5129) in 1971. Offspring cohort participants are seen in the Heart Study clinic approximately every 4 years, where a targeted physical examination is performed, cardiovascular risk factors are measured and a medical history is obtained focusing on cardiovascular events since the last examination.

For the present analyses, participants attending the seventh examination cycle were eligible if they had data on arterial tonometry performed at that examination and data on the 7 biomarkers evaluated (see below). Of 3537 attendees at the index examination, 877 individuals were not eligible for tonometry, because the examinations were performed at a nursing home (n=204) or because participants attended examination cycle 7 before the tonometry measurements were implemented (n=673)¹³. Of the remaining 2660 attendees, 660 were excluded from the present analysis for the following reasons: inadequate tonometry data (n=367); lacking biomarker data (n=287) or missing data for covariates included in multivariable analyses (n=6). After these exclusions, 2,000 participants (mean age 61 years, 55% women) remained eligible. Excluded participants had higher values (nominal p < 0.05) for systolic and diastolic blood pressure, body mass index,,total/HDL cholesterol, fasting blood glucose and prevalence of anti-hypertensive treatment, diabetes and smoking. The higher cardiovascular risk profile in participants with inadequate non-invasive vascular function or imaging data is well established.^15,16 All participants provided written informed consent and the Institutional Review Board at the Boston University Medical Center approved the study protocol.

Biomarker Measurements

We chose a panel of biomarkers measured at the sixth examination cycle (approximately 3 years before the seventh examination at which tonometry was performed) for the present analyses because these biomarkers represent several distinct pathophysiological pathways, as reported previously.^17,18 Blood was drawn from fasting participants after they had been in a supine position for 5 to 10 minutes (typically between 7.30 AM-9 AM), centrifuged immediately, and stored at −80°C until assays were performed. C-reactive protein (CRP) was assayed with a nephelometer (Dade Behring BN100), renin and aldosterone were determined using an immunochemiluminometric assay (Nichols assay, Quest Diagnostics) and a radioimmunoassay (Quest Diagnostics), respectively. N-terminal pro–atrial natriuretic peptide (NT-ANP) and B-type natriuretic peptide (BNP) were determined with high-sensitivity immunoradiometric assays (Shionogi, Japan). Plasminogen activator inhibitor (PAI)-1 was assayed with ELISA (TintElize PAI-1, Biopool, Ventura, CA). Fibrinogen was measured with the Clauss’ method and homocysteine was assayed using high-performance chromatography with fluorometric detection. The mean inter-assay coefficients of variation for the biomarkers were as follows: CRP, 2.2%; renin, 2.0 (high concentrations),10.0% (low concentrations); aldosterone, 4.0 (high concentrations), 9.8% (low concentrations); NT-ANP, 12.7%; BNP, 12.2%; PAI-1, 7.7%; fibrinogen, 2.6%; and homocysteine, 9%.

Acquisition of arterial waveforms with tonometry and analysis

With the participants in a supine position, tonometry was performed to obtain arterial waveforms from the carotid, brachial, radial and femoral artery (all on the right side) using a commercially available applanation tonometer (SPT-301, Millar Instruments, Houston, Tex). In parallel with the acquisition of the tonometric data, blood pressure was measured using an oscillometric device and an ECG was recorded. The average systolic and diastolic blood pressure values were used to calibrate the pressure waveforms after they have been signal averaged using the ECG R-wave as the fiducial point. The carotid-femoral pulse wave velocity (PWV) was calculated using the transit distances (obtained from body surface measurements and corrected for parallel transmission in the carotid) and transit times (obtain from the timing of the foot of the carotid and femoral tonometry waveforms). From the calibrated carotid pressure waveform, the following variables were derived (definitions): forward pressure wave amplitude (difference of the pressure at the foot of the waveform and the pressure at the first peak or inflection point); augmented pressure amplitude (difference between the central systolic pressure and the forward wave peak pressure); augmentation index (percent increase in the pulse pressure relative to the systolic inflection point) was used in secondary analyses. The calibrated carotid pressure served as a surrogate for central pressure.

Statistical Analyses

Our primary vascular phenotypes included the following traits that reflect the different components of arterial stiffness: central pulse pressure (marker of pulsatile arterial load), mean arterial pressure (marker of steady arterial load), and carotid-femoral PWV (measure of aortic stiffness). We also analyzed the two main components of central pulse pressure, i.e. the forward pressure wave amplitude and the augmented pressure. Biomarkers were natural logarithmically-transformed to normalize their distributions. PWV was also modeled as inverse PWV to improve its normality and similar association results as with log(PWV) were obtained. To reduce the amount of multiple testing, each vascular stiffness measure was related first to the biomarker panel as a whole. If the biomarker panel was associated with the stiffness measure with a p value below 0.01 (0.05 divided by 5, based on testing 5 primary vascular stiffness measures) in a multivariable-adjusted model, backward selection was used to identify a parsimonious set of biomarkers from among the panel that was separately related to each vascular phenotype. A p value of 0.007 was used to define statistical significance at this step (0.05 divided by 7 based on number of biomarkers tested). The multivariable models adjusted for the following 15 covariates that we have previously reported as key correlates of vascular measures in our cohort: age, age², sex, heart rate, height, weight, total/high-density lipoprotein cholesterol, blood glucose, diabetes, smoking, prevalent cardiovascular disease, hormone replacement therapy, hypertension treatment, aspirin (≥3day/week), and lipid-lowering medication.^5,19 In secondary analyses, we related the biomarker panel to the augmentation index. For biomarkers associated with primary tonometry traits, we tested for effect modification by sex, hypertension status, diabetes and use of antihypertensive or lipid-lowering medications. Since no statistically significant interaction with sex was observed for any of the significant biomarkers, sex pooled analyses are presented. Furthermore, after correcting for the performance of multiple interaction tests, no statistically significant interaction of biomarkers with hypertension status, diabetes, or use of antihypertensive or lipid-lowering medications was observed.

The authors had full access to the data and take responsibility for its integrity. All authors have read and agree to the manuscript as written.

Results

The clinical and biochemical features and arterial function characteristics of our study sample are displayed in Table 1 and Table 2. Our sample was middle-aged to elderly, with about a third of the participants on antihypertensive medications.

Table 1.

Clinical and standard biochemical measures of the study sample.

	Men (n=904)	Women (n=1096)
Clinical features
Age, yrs	61.1±9.9	61.2±9.6
Body mass index, kg/m2	28.2±4.0	26.7±4.8
Height, m	1.75±0.07	1.61±0.06
Weight, kg	86.8±13.9	69.3±13.0
Smoking, %	12.2	12.7
Systolic BP, mm Hg	128±18	126±20
Diastolic BP, mm Hg	76±10	72±10
Antihypertensive Medication, %	36.9	29.7
Lipid-lowering treatment, %	23.9	17.8
Aspirin ≥3 days/week, %	39.4	25.3
Prevalent CVD, %	18.1	8.9
Standard laboratory measures
Diabetes, %	15.7	9.7
Glucose, mg/dL	109±29	99±23
Triglycerides, mg/dL	143±107	128±73
Total/HDL cholesterol ratio	4.49±1.35	3.58±1.15

BP, blood pressure; CVD, cardiovascular disease; HDL, high-density lipoprotein. Values are mean±SD for continuous traits and percent for binary traits.

Table 2.

Biochemical and tonometric measures of the study sample.

	Men (n=904)	Women (n=1096)
Biochemical measures
BNP, pg/mL	6.3 (4.0,15.3)	9.8 (4.0,19.6)
N-terminal pro–atrial natriuretic peptide, pmol/L	283 (197,425)	354 (261,496)
CRP, mg/L	1.68 (0.81,3.37)	2.04 (0.88,4.92)
Aldosterone, ng/dL	9.0 (7.0,13.0)	10.0 (7.0,15.0)
Renin, mU/L	14.0 (8.0,25.0)	10.0 (6.0,18.0)
Aldosterone-renin ratio (10 ng/mU)	0.67 (0.38,1.14)	1.00 (0.56,1.71)
PAI-1, ng/mL	25.2 (16.8,35.5)	19.0 (11.6,30.0)
Fibrinogen, mg/dL	320 (285,369)	332 (293,379)
Homocysteine, nm/mL	9.7 (8.3,11.7)	8.3 (6.9,10.1)
Tonometry measures
Central pulse pressure, mm Hg	48.1±15.8	53.4±16.8
Mean arterial pressure, mm Hg	95.4±11.6	88.5±12.1
Forward pressure wave amplitude, mm Hg	40.4±12.4	41.8±13.3
Augmentation pressure, mm Hg	6.4±6.7	10.4±8.2
Carotid-femoral pulse wave velocity, m/s	10.5±3.7	9.8±3.6
Augmentation index, %	10.6±12.3	18.3±12.1
Brachial artery peripheral pulse pressure, mm Hg	51.2±14.0	54.9±15.2

BNP indicates B-type natriuretic peptide; CRP, C-reactive protein; PAI, plasminogen activator inhibitor. Values are mean±SD for continuous traits and percent for binary traits, except for biomarkers, for which median (Q1, Q3) is shown.

Biomarker panel and measures of vascular stiffness

The panel of 7 biomarkers considered together was associated with each of the primary (central pulse pressure, carotid-femoral PWV, mean arterial pressure, forward pressure wave, augmented pressure) and secondary (augmentation index) arterial traits even after multivariable adjustment (Table 3). Stepwise selection identified aldosterone-to-renin ratio (ARR) as a significant positive correlate of each vascular stiffness measure (p<0.002 for all).

Table 3.

Association of the entire biomarker panel and of individual biomarkers with measures of arterial stiffness.

Characteristic	Model R2	Partial R2	Global P*	Biomarker†	Β‡	P-value
Central pulse pressure, mm Hg	0.2683	0.0158	<0.0001	BNP	0.80±0.37	0.03
				ARR	1.54±0.33	<0.0001
				PAI-1	1.24±0.39	0.001
				Fibrin	0.74±0.35	0.04
Carotid-femoral PWV, m/s	0.4665	0.006	0.0025	CRP	0.14±0.07	0.048
				ARR	0.20±0.06	0.001
Mean arterial pressure, mm Hg	0.2075	0.0328	<0.0001	ARR	2.11±0.26	<0.0001
				PAI-1	0.89±0.30	0.003
Forward pressure wave, mm Hg	0.2301	0.0103	0.0004	ARR	1.00±0.27	0.0002
				PAI-1	0.80±0.31	0.01
Augmented pressure, mm Hg	0.2695	0.0119	<0.0001	CRP	0.62±0.17	0.0003
				ARR	0.49±0.16	0.002

^*A test of whether any of the biomarkers differed with respect to arterial stiffness dependent measures. Covariates in the multivariable models included age, age², sex, heart rate, height, weight, ratio of total/high-density lipoprotein cholesterol, blood glucose, diabetes, smoking, prevalent cardiovascular disease, hormone replacement therapy, hypertension treatment, aspirin (≥3day/wk), and lipid-lowering medication.

^†For tonometry measures with a global P<0.01, individual biomarkers related (p<0.05) to vascular function measures after backward elimination are displayed.

^‡β, the regression coefficient, shows a change in vascular function measure per 1-SD increment in log-marker. Thus, an e^SD-fold increase in BNP (original units) results in an increase of 0.80 mm Hg in central pulse pressure.

BNP, B-type natriuretic peptide; PAI, plasminogen activator inhibitor; CRP, C-reactive protein; ARR, aldosterone-to-renin ratio

In addition, CRP was positively associated with the augmented pressure and with carotid-femoral PWV, although the latter did not reach statistical significance after correction for multiple testing. PAI-1 was associated with central pulse pressure and mean arterial pressure (p<0.005), and with the forward pressure wave amplitude, though the p value slightly exceeded the 0.007 threshold that corrected for multiple testing. BNP and fibrinogen were associated with central pulse pressure but neither p-value was significant after correcting for multiple comparisons (Table 3). In secondary analyses, CRP was associated with the augmentation index (p=0.0002). Overall, components of the biomarker panel explained 1–3% of the interindividual variation in vascular stiffness measures (partial R squares in Table 3).

Discussion

Principal Findings

We assessed the joint association of a panel of 7 biomarkers that represent distinct biological pathways with measures of arterial stiffness in a large (n=2000) community-based sample to elucidate the relative contribution of different pathways to aortic stiffness, wave reflection and microvascular function. We observed an interesting pattern of association with vascular function measures for 3 of the biomarkers, i.e, ARR, PAI-1 and CRP. ARR was positively associated with all measures of arterial function tested, including carotid-femoral PWV, forward pressure wave amplitude and central pulse pressure (measures of conduit artery stiffness), as well as with mean arterial pressure and augmented pressure (which are measures of peripheral vascular function). These observations are consistent with a central role of the renin-angiotensin-aldosterone system (RAAS) in determining both steady state and pulsatile components of afterload, and with its influence on pan-arterial function and remodeling. By comparison, CRP and PAI-1 were differentially related to measures of central versus peripheral vascular function. CRP was associated with augmented pressure, a main correlate of peripheral vascular stiffness whereas, PAI-1 correlated mainly (and positively) with measures of conduit artery stiffness. These observations support the notion of varying influences of biological pathways on central versus peripheral vascular stiffness. Overall, the biomarkers explained only a small proportion of the interindividual variability in vascular stiffness measures.

In the Context of the current Literature

Arterial Stiffness and the Renin-Angiontensin-Aldosterone System

Clinical and experimental evidence link the RAAS to arterial stiffness. Circulating components of the RAAS have been related to measures of arterial stiffness in small-to-moderate sized samples. For example, carotid-femoral PWV was noted to be higher in patients with primary aldosteronism as compared with patients with essential hypertension and with healthy controls.²⁰ Serum aldosterone was positively associated with heart to femoral PWV in a series of patients with hypertension (n=438).¹¹ ARR was also positively associated with PWV in 60 healthy subjects.²¹ These observational studies are paralleled by clinical trial data that suggest that pharmacological inhibition of the RAAS reduces arterial stiffness. For instance, angiotensin II receptor blockers, angiotensin-converting enzyme inhibitors and aldosterone antagonists have been reported to reduce PWV in small studies of heterogeneous groups of patients.^8,9,22,23 Finally, genetic variants in the RAAS have been associated with measures of arterial stiffness including aortic PWV²⁴ and carotid distensibility.²⁵

Substantial experimental evidence suggests that aldosterone and angiotensin II affect vascular remodeling.⁷ RAAS activation influences vasoreactivity and structural and functional remodeling via genomic and non-genomic mechanisms. Such activation is characterized by oxidative stress and inflammation due to the interactions of aldosterone and angiotensin II on the mineralocorticoid and AT₁ receptors⁷ and upregulation of epidermal growth factor receptor expression.²⁶ In line with these clinical and experimental data, we observed a consistent and strongly positive association of ARR with all measures of arterial function that were tested.

Arterial stiffness and inflammation

Biomarkers of inflammation like CRP, interleukin-6 and tumor necrosis factor-α have been associated positively with both indirect (e. g. brachial artery pulse pressure¹⁰) and direct measures of arterial stiffness in previous studies in apparently healthy individuals,²⁷ in specific patient groups (e.g. patients with hypertension²⁸) and in community-based samples.^12,29 Most of these studies had modest sample sizes.^27,29 In a recent report focusing on inflammatory biomarkers from our group, we demonstrated positive association of CRP (measured at the same examination cycle as the tonometry) to the augmented pressure amplitude.³⁰ We replicated the same finding in the present investigation. It is conceivable that inflammatory processes within the vessel wall (as part of vascular remodeling) manifest both in increased reflected wave amplitude and higher systemic levels of CRP. Alternatively, excessive wave reflection may contribute to vascular inflammation and remodeling because of associated abnormalities in pressure and flow in the larger arteries. Longitudinal and interventional studies may be required to establish the directionality of these relations.

Arterial stiffness and natriuretic peptides and homocysteine

Previous studies reported associations of natriuretic peptide levels and homocysteine with different measures of arterial stiffness including carotid-femoral and carotid-radial PWV.^13,31,32 In the present investigation, using a multi-marker approach we observed weaker positive associations of BNP with central pulse pressure that did not reach statistical significance upon correction for multiple testing. Homocysteine levels were not associated with vascular stiffness measures when modeled with other biomarkers. It is important to note that our observations (based on a statistical procedure to select from among a panel of biomarkers) do not exclude an important role for the natriuretic peptides or homocysteine in vascular remodeling. Many of the biomarkers tested are correlated and the fact that certain markers drop out of the model during the backward elimination process (likely because other markers are more closely related to arterial function due to additional direct or indirect effects) do not rule out an important role of these biomarkers in arterial physiology.

Arterial stiffness and hemostasis

PAI-1 has been associated positively with PWV³³ and to aortic stiffness in previous studies.³⁴ In our analyses, PAI-1 was positively related to measures of conduit artery stiffness, i.e. mean arterial pressure, central pulse pressure, and forward pressure wave amplitude, even after multivariable adjustment. Epidemiological evidence further supports that PAI-1 modulates vascular remodeling. In the ARIC cohort, PAI-1 levels were positively associated with intima media thickness.³⁵

Experimental data suggest that PAI-1 inhibits the activity of the fibrin degrading enzyme plasmin and also has direct effects on the vessel wall.³⁶ PAI-1 is bound to vitronectin in the extracellular matrix where it inhibits migration of vascular smooth muscle cells.³⁷ Thus, clinical and experimental data suggest that higher PAI-1 levels may contribute to stiffening of the aorta over the life-course, which might explain its association with forward pressure wave and blood pressure traits like central pulse pressure and mean arterial pressure.

Fibrinogen was related to carotid-femoral PWV in hypertensives (n=229),³⁸ and to pulse pressure/stroke index in a large sample of American Indians.³⁹ However, no association was observed between fibrinogen and PWV in patients with end-stage renal disease⁴⁰ and in apparently healthy middle-aged women.⁴¹ Our results are consistent with the latter studies. Fibrinogen was weakly associated with central pulse pressure, but this association was no longer significant after considering the multiple comparisons performed. All other measures of arterial function were not related to fibrinogen in this multi-marker approach.

Limitations

Our cohort consists of middle-aged to elderly Americans of European descent, which limits the generalizability of our findings. There was a time lag of about 3 years between the tonometry and biomarker measurements, which might have diminished our ability to detect associations between some biomarkers and the tonometry traits. However, we have shown in previous studies in our cohort that biomarkers (measured at the sixth examination cycle) were powerful correlates of endothelial function (measured at the seventh examination cycle) even if the traits are not measured contemporaneously.⁴² Furthermore, a significant proportion of attendees had to be excluded due to missing or inadequate tonometry and biomarker data. This is a well known but unavoidable limitation of large epidemiological cohort studies that may bias toward the null hypothesis because of loss of cases that presumably had more extreme values for the analyzed variables.^15,16 An additional limitation specific to the homocysteine concentrations was the fact that folic acid fortification of enriched cereal grain products was introduced during the course of the sixth examination cycle. Consequently, the homocysteine concentrations related to the measures of arterial stiffness may reflect recently reduced levels for many of the participants and not their usual homocysteine exposure before fortification. Finally, we cannot infer causality from these observational data.

Conclusions

We have observed both specific and generalized relations between various biomarkers of distinct biologic pathways and a comprehensive but concise group of vascular function measures. We observed a consistent positive association of the ARR with various measures of arterial stiffness, wave reflection and microvascular function. This convincing association is consistent with the notion that pharmacological inhibition of the RAAS may revert or reduce the progression of arterial stiffness and associated abnormalities in wave reflection and mean arterial pressure. In addition, the pleiotropic effects of the RAAS pathway on various physiologically and anatomically distinct vascular measures may contribute to the observed interplay between large and small artery function.¹⁹ CRP and PAI were significantly associated with certain arterial stiffness measures, indicating that inflammation and hemostasis also modulate (or are modulated by) peripheral and central arterial stiffness, respectively. Our findings suggest pathways that should be studied further in order to elucidate the pathophysiology of large artery stiffening. Such studies offer the potential to identify much needed interventions that specifically limit or reverse stiffening of the large arteries.