Ambulatory arterial stiffness index: merits and limitations of a simple surrogate measure of arterial compliance

Everything should be made as simple as possible, but not simpler. Albert Einstein Arterial stiffness, a descriptive term which indicates the pressure required to obtain a given dilatation in an arterial segment or in the whole arterial tree, has been the object of a growing research interest in recent years and is increasingly used in the clinical assessment of patients. Large-artery stiffness is more and more recognized as a major index of target organ damage in hypertension and an important harbinger of future cardiovascular morbidity and mortality in different clinical settings, including essential hypertension [1–3]. The recent European guidelines [4] for the management of arterial hypertension have included for the first time arterial stiffness among the indexes of subclinical organ damage able to influence prognosis of hypertensive patients. Although systemic arterial stiffness can only be estimated from models of the circulation, regional and local arterial stiffness can be measured directly at various sites along the arterial tree. One major problem with arterial stiffness determination is that accurate measurement of a given artery's stiffness necessarily requires simultaneous, accurate measurement of local arterial cross-sectional area (or diameter) and pressure, which is difficult to obtain in vivo with any technique. This can be accomplished invasively with the ultrasound/catheter tip manometer system introduced by Stefanadis et al.[5]. Similar approaches can be applied noninvasively at different sites. Diameter changes, although small, can be measured accurately [6], but there are problems in estimation of pressure changes at the same site. Thus, any proposed index of arterial stiffness necessarily represents a surrogate of ‘true’ stiffness and a trade-off between simplicity and accuracy. A number of surrogate measures of large-artery stiffness have been proposed, but each has important limitations [7,8] (Table 1).Table 1: Selected noninvasive indexes of arterial stiffness/complianceThe dynamic relation between systolic and diastolic blood pressure (BP) is a novel, theoretically attractive means of investigating arterial stiffness. For a given increase in diastolic BP, systolic BP is expected to increase to a limited extent in a compliant artery, whereas the increase will be greater in a stiff artery. The opposite holds for the increase in diastolic BP for a given increase in systolic BP, which can be considered as a measure of arterial compliance. On the basis of these principles, the ambulatory arterial stiffness index (AASI) has been developed [9]. By plotting the individual values of systolic and diastolic BP measurements obtained through 24-h noninvasive ambulatory BP monitoring, the linear regression slope of diastolic BP on systolic BP is assumed as a global measure of arterial compliance, and its complement (1 minus the slope), named AASI, has been taken as a measure of arterial stiffness [9]. AASI has been associated with preclinical target organ damage in hypertension [10,11] and with an increased risk of cardiovascular mortality in hypertensive patients [12] and stroke in the general population [13]. That a measure of arterial stiffness, which does not require specialized equipment and dedicated personnel, is unquestionably expected to have a great success in the scientific and medical community. Whether simplicity goes in parallel with accuracy is questionable. Two articles in the present issue of this journal shed light on the significance and limitations of AASI. Baumann et al.[14] performed 24-h ambulatory BP monitoring and calculated AASI in 106 adult kidney donors, among which 29% were hypertensive patients. Their main conclusion was that AASI has a strong negative relation with nocturnal systolic and diastolic BP reduction (r = −0.55 and r = −0.48, respectively). Moreover, they found that a significant relation between AASI and pulse pressure (PP) as a measure of arterial stiffness was evident only in subjects with a normal nocturnal BP reduction (‘dippers’), but not in nondipper subjects. These data confirm and extend to a predominantly normotensive population the findings of a recently published study [15], in which we demonstrated the strong dependence of AASI on the degree of nocturnal BP fall in a larger cohort of 515 untreated hypertensive patients. The explanation we offered for this finding is that dipper subjects have a large number of nocturnal systolic and diastolic BP values much lower than the corresponding daytime values, and this increases the regression coefficient of diastolic BP on systolic BP. In contrast, nondipper subjects tend to have a narrower range of diastolic BP values throughout the 24 h. As a mathematical consequence, in nondipper subjects the coefficient of regression B of diastolic over systolic BP over the 24 h tends to decrease, and its complement (AASI, or 1 − B) tends to increase. Taken together, these studies [14,15] suggest that the degree of nocturnal BP reduction should always be properly considered in future studies, which will examine the clinical and prognostic values of AASI. Taking advantage of a database of 140 adult subjects, Gavish et al.[16] disclosed a previously unrecognized relationship between AASI and the correlation coefficient between systolic and diastolic BP. They suggest that a low correlation coefficient between systolic and diastolic BP values over the 24 h tends to artificially decrease B (and increase AASI) when a standard asymmetrical regression is used. In other words, they showed that calculation of AASI by standard regression produces a considerable overestimation in the diastolic-on-systolic slope, which is inversely dependent on the correlation coefficient between systolic and diastolic BP. No such artifactual relationship is found when AASI is derived from a regression model, which handles both variables in a symmetrical way. As symmetrical regression is currently not performed easily by most statistical packages, the authors also provide the following simplified calculation of symmetrical AASI, which is based on standard systolic-on-diastolic BP slope, as well as on the correlation coefficient between systolic and diastolic BP values: It should be recognized that the approaches developed by us [15] and Gavish et al.[16] essentially unravel the same mathematical relation. The extent of the nocturnal BP fall has two major implications on the calculation of AASI. As reported above, individuals with a small reduction in diastolic BP from day to night (‘nondippers’) tend to have a narrower range of diastolic BP values throughout the 24 h. Since diastolic BP is the dependent variable in the regression model upon which AASI is based, non dippers tend to have a lower diastolic-on-systolic BP slope and a higher AASI. Incidentally, the artifactual dependence of AASI on nocturnal BP reduction is also able to explain why those studies in which a low number of BP measurements was collected at night show greater average AASI values than the studies with uniform frequency of measurements during the 24 h [17,18]. Second, nondipper subjects tend to have a lower correlation between systolic and diastolic BP and higher AASI values. Given that nocturnal declines in systolic and diastolic BP are fairly well correlated with each other, individuals with a large nocturnal diastolic BP fall also tend to have a large systolic BP fall. This leads to higher correlation coefficients between systolic and diastolic BP in dipper subjects than in nondipper subjects. As shown by Gavish et al.[16], in a standard, ‘asymmetrical’ linear regression approach, a low correlation coefficient artificially reduces B and increases AASI in nondipper subjects. This statistical artifact is abolished when using a symmetrical regression model. The above considerations make the symmetrical regression model proposed by Gavish et al.[16] theoretically more appropriate than the standard regression proposed by Li et al.[9] in estimating slope-related parameters. It is worth mentioning that symmetrical AASI has an exponential relationship with age, which is in agreement with the nonlinear relation between many stiffness measures and age, with a transition at about 55 years of age [19]. In contrast, this exponential relation disappears when AASI is calculated following a standard regression model. Still, it remains to be established whether AASI values obtained with the former approach keep the clinical and prognostic impact, which has been attributed to AASI calculated with standard regression. Currently, it is controversial to what extent AASI is a true measure of arterial stiffness. Baumann et al.[14] found a relationship between AASI and brachial PP, which has been proposed as a simply obtainable measure of arterial stiffness. Indeed, a high PP mostly reflects a high aortic stiffness, at least under some circumstances, including advanced age and isolated systolic hypertension. Brachial PP is, however, also affected by a number of factors unrelated to stiffness, including the extent of amplification from aorta to the brachial artery, the degree of vasodilatation, heart rate, cardiac contractility, and stroke volume [20]. Unfortunately, a more direct measure of arterial stiffness was not available in the article by Baumann et al. [14]. We have recently reported that the relation between AASI and a widely accepted measure of aortic stiffness, such as pulse wave velocity (PWV), is weak and importantly affected by other factors [15]. Interestingly, Gavish et al.[16] suggest that AASI could reflect dynamic arterial stiffening (i.e., the relative increase in arterial stiffness from diastolic to systolic values), rather than ‘average’ arterial stiffness tout court. In conclusion, the analysis of the dynamic relation between systolic and diastolic BP over a wide range of values appears to be a promising new approach in evaluating cardiovascular risk. Evidence is accumulating that AASI is strongly influenced by factors unrelated to arterial stiffness, including nocturnal BP reduction and the correlation coefficient between diastolic and systolic BP values. Clearly, more work is needed to establish the physiological determinants of AASI, as well as its clinical significance. On the contrary, a wider implementation of arterial stiffness measurement in clinical practice and at the population level is presently limited by the need of dedicated and rather expensive instrumentation and trained observers, and new simpler auscultatory [21] and oscillometric [22] methods have recently been proposed in this regard. Clearly, the development of better and easy-to-use techniques is of paramount importance for increasing the availability of arterial stiffness measurement also in general practice, in the daily management of patients in whom assessment of arterial properties may contribute to their cardiovascular risk stratification. Acknowledgement No funding was obtained for this commentary. The authors have no disclosures.