Arterial stiffness and prediction of cardiovascular risk

Cardiovascular disease remains one of the main contributors to morbidity and mortality in the developed world. Primary prevention of an unselected population has proven to be effective [1]. However, it is also obvious that this approach is not economically viable given the high cost per year of life saved multiplied by the number of possible beneficiaries [2]. This economical reality would only change if either treatment became more efficacious or less expensive [3]. In fact, as these primary prevention studies also demonstrated that risk and benefit is restricted to certain subpopulations, research has concentrated on identifying those subpopulations with higher risk, larger treatment benefits, reduced eligibility and, consequently, dramatically reduced cost to society. Known contributors to risk can be broadly defined as originating from five areas: (i) demographic factors – age [4], gender [4]; (ii) genetic determination – family history [5]; (iii) disease promoters – hypercholesterolaemia [4], hypertension [4], diabetes [4], smoking [4], obesity [6]; (iv) symptomatic disease – myocardial or cerebral ischaemia [7] and (v) asymptomatic disease – left ventricular hypertrophy [4], carotid artery stenosis [8], large artery stiffness [9–11]. Large epidemiological, longitudinal studies have clarified their individual contribution and the factors determined have been incorporated into risk prediction algorithms for asymptomatic individuals [4]. One example is the risk level suggested by the joint task force of European and other societies on coronary prevention [12], as employed by Nürnberger et al. [13] in the present issue of the journal, which was recently validated against an independent population [14]. As this predicts probability of future symptomatic disease based on a disease-free population, it becomes meaningless once disease has become symptomatic. Hence, it should not be applied to symptomatic populations as performed by Nürnberger et al. [13] (consider their Figure 1, which also includes patients with cardiovascular disease). Risk for further morbidity and mortality increases dramatically once subjects are symptomatic [7]. In consequence, these patients benefit most from (now secondary) prevention. A number of interventions have recently been shown to be effective, and these enjoyed a prompt uptake [15]. Currently available algorithms for the prediction of future events are based on observations that pre-date these changes to medical practice and may no longer accurately predict risk in subjects on treatment. This applies to the EPOZ score used in [13], whereas the SMART score used in [13] has not been prospectively validated in its prediction of risk. Current recommendations for secondary prevention do not employ risk scoring, but treat the presence of disease in a dichotomous fashion [12]. The contribution of asymptomatic disease has been studied less. As opposed to a simple demographic factor, such as age or gender, this involves the evaluation of the individual, with a potentially laborious method to detect such asymptomatic disease. However, it is hoped that the evaluation of asymptomatic individuals may help define subpopulations with lower and higher risk of future events. The latter may in turn be eligible for primary prevention at a cost more acceptable to the community. The assessment of arterial stiffness is an attractive candidate for a number or reasons. Arterial stiffness is associated with severity of coronary artery disease [16], severity of carotid artery disease [17], predicts future morbidity and mortality [9–11] and can be assessed automatically and non-invasively [18]. A number of methods are available, and a number of devices are now commercially available. This has the advantage that methods are easily accessible to anyone without special knowledge or association with a researcher specializing in arterial mechanics. In fact, commercial availability promotes applied research, such as that undertaken by Nürnberger et al. [13]. At the same time, the potential exists for scientists turned inventors/salesmen to introduce a commercial conflict of interest to the peer review process, even more than might ordinarily be expected from normal scientific argument or differences between basic/mechanistic and applied scientists. In particular, objective comparisons of the various methods in their predictive power are lacking, and these are unlikely to be forthcoming. Pulse wave contour analysis is one of several methods to assess arterial stiffness. It involves measuring (either invasively or non-invasively) a central arterial pressure wave. A sample of such a measured pressure waveform is given in Figure 1. The contraction of the left ventricle increases left ventricular pressure until the aortic valve opens. After aortic valve opening, left ventricular contraction continues and pressure keeps increasing until reaching a peak. This is followed later in the cardiac cycle by ventricular relaxation, expansion, aortic valve closure and diastole. The latter events are irrelevant for the present consideration.Fig. 1: The waveform in the centre is a carotid waveform obtained by applanation tonometry scaled to brachial artery mean and diastolic pressure (Dinamap). Diastolic pressure, 75 mmHg; systolic pressure, 142 mmHg; pulse pressure, 66 mmHg. Below is the measured wave decomposed into its forward and backward component (not to scale), based on ascending aortic flow measured with a 3.5-MHz continuous wave Doppler transducer in the suprasternal notch. Above is the fourth derivative of the pressure waveform, which can be used to mathematically define the augmentation point. With this method, the augmentation point is defined as the first zero crossing from positive to negative of the fourth derivative and this occurs 55 ms after the onset of systole. Pressure at this point is 101 mmHg. Hence, the augmentation index is (142 – 101)/66 = 62%.The increase in pressure at the aortic inlet is not transmitted instantaneously throughout the circulation, but travels with a finite velocity, called pulse wave velocity [19]. This pulse wave velocity is dependent on the stiffness and diameter of the vessel and in the adult human aorta measures approximately 5–10 m/s [20]. As the arterial tree branches and tapers, changes in diameter and stiffness occur and this results in reflection of some portion of the pressure wave. As an example, the aortic bifurcation reflects 6–7% of the forward travelling wave [20]. This reflected wave will then travel backwards with approximately the same speed. When considering a pulse contour close to the aortic root, the initial rise, in pressure (solely generated by left ventricular contraction interacting with the stiffness of the aortic root) is rapidly superimposed with a reflected wave returning from the periphery. In Figure 1, an admittedly less than perfect attempt was made to decompose the composite wave into its forward and backward components. The beginning of this returning, reflected wave is visible on the composite (measured) wave as an inflection, or change in the speed of rise, in pressure (dP/dt), which is termed the augmentation point. This relates to the earliest visible source of reflection [21]. In the example, the source is approximately 15–20 cm away, which does not relate to a single anatomical site, but rather to the superimposition of myriad reflected waves to the point where an inflection first becomes noticeable [20]. The augmentation point can be identified visually [22] or by various mathematical methods, most of which make some assumption about its position in the cardiac cycle [23]. In Figure 1, the fourth derivative was used, which is only one of the available methods. The location of the augmentation point in time depends on (i) the distance to the reflection site [21,24,25,26]; (ii) amplitude of reflection [20,27,28] and (iii) pulse wave velocity to the site [20]. Once the augmentation point has been determined in the time domain, the augmentation index can be expressed in the pressure domain as a percentage of pulse pressure [29]. By convention, where the augmentation point is identified later than the peak of pressure, it is given a negative sign [29]. It is mathematically impossible for the augmentation index to be outside the +100 to –100% range; compare Figures 1–4 in Nürnberger et al. [13], in which one subject has an augmentation index > 100%. Due to the assumptions made about timing in the mathematical approach to identify the augmentation point, central augmentation index in adult humans can range between +80% (close to the beginning of systole) and –30% (close to the beginning of diastole) of pulse pressure [20]. Expressing augmentation index in the pressure rather than the time domain is historical [29]. As such, augmentation index becomes dependent both on the timing of the reflected wave (i.e. gender, height, reflection amplitude and stiffness) and the shape of the forward wave. As discussed above, the latter is dependent on left ventricular outflow and the elasticity of the ascending aorta. Slowing left ventricular outflow (e.g. female gender with prolonged ejection [24,26], decreased heart rate [23,30], myocardial damage with reduced contractility, mitral incompetence and aortic stenosis) increases augmentation index. Most of these factors have been demonstrated by Nürnberger et al. [13]. An increase in ascending aortic stiffness could theoretically decrease augmentation index due to an increase in initial dP/dt [31]. However, in practice, augmentation index increases due to an increase in pulse wave velocity and pulse pressure, as demonstrated by Nürnberger et al. [13]. In summary, augmentation index is a composite measure depending on left ventricular outflow (including myocardial damage, gender, heart rate, aortic stenosis), anatomy (including gender, height) and aortic stiffness (including age, blood pressure and the presence and severity of cardiovascular disease). It can be calculated from measurement of a central arterial pressure waveform only. Augmentation index has been shown to be predictive of future cardiovascular and all cause mortality [32]. Nürnberger et al. [13] are the first to relate augmentation index to classical risk scores in subjects with and without symptomatic cardiovascular disease. As both relate in a similar fashion to the same underlying independent variables (e.g. risk increases with age and augmentation index increases with age), it would be surprising if no relation was found. It is certainly worth further investigation to confirm or dismiss the consistency of the results independently obtained to date. Indeed, the main findings are that the augmentation index is significantly related to risk scores in individuals both with and without symptomatic cardiovascular disease. Importantly however, there was substantial overlap in augmentation index between risk levels assessed using all three risk scores (Figures 1–3 in Nürnberger et al. [13]), such that augmentation failed to stratify individuals in relation to risk. This fact would suggest that augmentation index does not provide value in risk prediction beyond that obtained from conventional risk markers in a population such as the one studied by Nürnberger et al. [13]. This is interesting in the light of mortality results from a very high risk population of 180 dialysis patients, 70 of whom died during 4 years of follow-up, and where mortality was predicted by age, pulse wave velocity, diastolic blood pressure, pre-existing disease, medication with an angiotensin-converting enzyme inhibitor and augmentation index [32]. In fact, in this high risk population, some classical risk factors were less prominent predictors of future mortality. This shows that an association study such as that performed by Nürnberger et al. [13], in which poor stratification performance for augmentation was observed, does not mean that augmentation index is not a strong predictor. This particularly applies where classical risk factors are poor predictors, and where augmentation index outperforms. There are a number of additional factors to be considered with regard to interpretation of the data presented by Nürnberger et al. [13]. The authors recruited patients from an outpatient clinic (and an additional 15 volunteers). This would suggest that these individuals had some form of disease and to refer to them as ‘healthy’ may not be entirely correct. In fact, 34 ‘healthy’ patients had arterial hypertension, certainly had unusually high risk scores, and some of them were medicated. Thus, the division between ‘healthy’ patients and ‘patients with disease’ is likely to be somewhat blurred. The study included 216 individuals, which is a small sample size to assess the independent impact of disease status, age, gender and medication profile on augmentation index. This is further subdivided into groups with (2/3) and without symptomatic disease (1/3). A questionnaire was used to assess the presence of atherosclerotic disease which was deduced from a wide range of definitions. A more homogenous group defined by precise criteria (angiographic coronary disease severity) would permit a more detailed analysis regarding the factors influencing augmentation index. It is not surprising that in a small population (72 subjects) with a higher variation and contribution of unmeasured or difficult to assess variables (such as changes in contractility after myocardial infarction, potentially some degree of mitral regurgitation or aortic stenosis, and medication), multiple regression will yield a smaller amount of explained variation. This is demonstrated by weaker relationships in the symptomatic subjects for all variables tested in Table 2 presented within Nürnberger et al. [13], not only for age. In contrast, the demonstration of a consistent effect in such a small group in the presence of unmeasured error would suggest a rather strong influence. Furthermore, the gender balance between age groups and disease status varied substantially. In particular, the disease group contained fewer females in the highest age group compared to the disease-free group. In addition, the sample size was too small to adequately account for the effect of medication on augmentation index. This is evident in the authors’ conclusion that the augmentation index was not influenced by vasoactive drugs when there is published evidence to the contrary [20,27,28]. In conclusion, the study by Nürnberger et al. [13] confirms previous studies regarding the dependence of augmentation index on age, height, heart rate and blood pressure. The main new finding was that augmentation index related to composite cardiovascular risk scores (independent variables included age, gender, heart rate and blood pressure) both in individuals with and without symptomatic atherosclerotic disease. The major question remains: what is the best measure of arterial properties to predict future morbidity and mortality, in addition and after adjustment for classical risk factors? This would involve a longitudinal cohort assessment of their relative predictive values, as well as their cost.