Assessing the prognostic relevance of blood pressure variability: discrepant information from different indices

Methods to assess blood pressure variability Analysis of 24-h ambulatory blood pressure recordings, obtained in hypertensive patients, has demonstrated that a greater blood pressure variability is associated with either a greater degree of end-organ damage [1–6] and/or a higher rate of cardiovascular events [5–8]. This was often, but not invariably, the case even after accounting for between-subject differences in mean blood pressure levels [2–5,7,8]. In most of these studies, information on the adverse effects of an increased blood pressure variability was obtained by quantifying this phenomenon as the SD of 24-h average ambulatory blood pressure. However, calculation of 24-h blood pressure SD does not provide a fully satisfactory quantification of the features characterizing blood pressure variations over 24 h. The limitations of this index are mainly related to the complexity of the blood pressure variability phenomenon, which includes different components, characterized either by a different speed of blood pressure change, and by a different distribution over time of the beat-by-beat variations in blood pressure [9]. Indeed, the size of the contribution to overall blood pressure variance and the specific properties of any given component of 24 h blood pressure variability can hardly be derived from SD values, which pool together all different types of blood pressure phenomena. As an example, the SD index makes no distinction between the contribution to overall blood pressure variability given by fast changes associated with emotional or physical challenges, by the long-lasting reduction in blood pressure during night sleep and by the more or less steep blood pressure rise on awakening. The importance of a more precise assessment of the size and features of different blood pressure variability components is emphasized by some of them having an opposite prognostic significance compared to that of 24-h blood pressure SD. This is typically the case of nocturnal blood pressure fall. On the one hand an increased 24-h blood pressure SD is associated with a worse prognosis. On the other hand, a more pronounced end-organ damage [10,11] and a higher rate of cardiovascular events [12,13] are associated with a reduction, rather than with an increase, in the amplitude of nocturnal blood pressure fall at night (non-dipping). This has led to the conclusion that a small 24-h blood pressure variability and a significant nocturnal reduction in blood pressure are both clinically favourable phenomena. However, the magnitude of the nocturnal blood pressure fall is significantly and positively related to the magnitude of 24-h blood pressure SD, as demonstrated by Imai et al. [14] in the Ohasama database and as more recently confirmed by analysis of the Milan–Krakow ABPM database [15]. The correlation between 24 h blood pressure SD and nocturnal blood pressure fall is inconsistent with the reported opposite clinical significance of these two parameters. Thus, a failure to account for the contribution carried by nocturnal blood pressure fall to 24-h blood pressure SD might at least partly mask the prognostic power of the latter parameter. In an attempt to avoid this inconvenience, a number of studies have focused on either daytime SD only or on night-time SD only [5–8]. However, an approach focusing on daytime or on night-time blood pressure variability only would unavoidably miss the blood pressure variation occurring during the other neglected recording sub-period, and thus would overlook potentially important prognostic information [5–8]. With the aim of overcoming this problem, Parati et al. [15] recently proposed a new index to assess 24-h blood pressure variability. This index, based on the calculation of a ‘weighted’ 24-h blood pressure SD, allows removal of the specific contribution of nocturnal blood pressure fall from the 24-h blood pressure variance, while preserving the independent information carried by the degree of blood pressure variations occurring during the daytime and the night-time, respectively. A similar attempt was made by Sega et al. [16] through the analysis of the PAMELA (Pressioni Arteriose Monitorate E Loro Associazioni) database, from subjects living in the area north of Milan, Italy. In this population study, the relationship of left ventricular mass index, quantified by echocardiography, to 24-h blood pressure variability, assessed by a Fourier transform of the 24-h blood pressure profiles, was investigated. A correlation between 24-h blood pressure variability and left ventricular mass index was found only when the two largest cyclic components of the 24-h blood pressure variance, including the nocturnal blood pressure fall, were removed, thereby considering the remaining faster components of blood pressure variability only. However, these more recent approaches to the quantification of blood pressure variability are not immune to limitations. There are technical problems in computing blood pressure SD values and/or spectral indices when considering discontinuous and unevenly spaced blood pressure readings [17–20]. Indeed, a reliable evaluation of 24-h blood pressure SD from discontinuous measurements may be obtained only if the interval between two subsequent measurements, evenly spaced, is less than 15 min [17]. Such a short interval is rarely used in the clinical setting due to practical reasons and to minimize patients’ discomfort. Even when this requirement is fulfilled, the SD index still faces other inherent difficulties. These include its capability of only being able to reflect the dispersion of blood pressure readings around an average value, and not of accounting for the shape or the sequence of their time distribution (Fig. 1) [21].Fig. 1In this issue of the journal, Mena et al. [22] propose a new approach to the assessment of blood pressure variability, which aims to account for some of the above technical problems. The index suggested by Mena et al. [22], defined as the Average Real Variability (ARV) index, is the weighted total sum of the subsequent differences in blood pressure between a given measurements at time t and the measurement immediately preceding, at time t − 1. Therefore, this index focuses on the subsequent changes between consecutive blood pressure readings, and is sensitive to the sequential order of the blood pressure readings whose variability has to be quantified. Mena et al. [22] have also tested the possible clinical value of such an index in predicting cardiovascular complications in hypertension. To this aim, the ARV index was applied to assess the prognostic value of blood pressure variability in a population sample of Maracaibo, Venezuela. The authors’ conclusion is that ARV index has a greater prognostic significance than the SD of 24-h blood pressure values, even when measured with a non invasive, intermittent ambulatory blood pressure monitoring system (ABPM). A further conclusion by Mena et al. [22] is that the ARV index is more reliable in reflecting blood pressure variability than the SD of average blood pressure values, and may be less sensitive to the relatively low sampling frequency of the conventionally employed non-invasive ABPM devices. Therefore, ARV index might add prognostic value to ABPM, and might represent a useful tool to assess the validity of therapeutic measures aimed at controlling blood pressure variability. Blood pressure variability and cardiovascular risk in Maracaibo: a problematic assessment The conclusions made by Mena et al. [22] on the prognostic value of the ARV index should be taken with some caution, due to problems that appear to have affected the quantification of cardiovascular risk in the subjects included in their study. The study population, which included relatively medium-low risk Maracaibo inhabitants (aged approximately 66 years, plasma cholesterol levels of approximately 216 mg/dl, 65% hypertensive, 35% smokers, etc.), showed a rate of events of approximately five per 100 patient/years, which appears to be exceedingly high considering their profile of known cardiovascular risk factors. If a calculation of the expected rate of events in a given population is performed according to the risk charts derived from the Framingham study, no more than 2–2.5 events per 100 patients per year should be expected with such a risk profile [23]. Similar results would be obtained also if other risk charts were used [24] (Table 1).Table 1: Predicted incidence of cardiovascular events (per 100 patients per year) in a population with characteristics similar to that enrolled in the study by Mena et al. [22]There are several explanations that may be advocated to explain this higher than expected event rate in the Maracaibo population. It is possible that, in the definition of the type of events to be considered in the study by Mena et al. [22], there are differences compared to the standard risk charts. In particular, the presence of a significant obstruction in the coronary arteries or the need of a revascularization procedure were also considered as an ‘event’, thus contributing to the calculation of the event rate. Alternatively, non-traditional or newer risk factors might also have played a particularly relevant role in this population (e.g. homocisteine, reactive C-protein, fibrinogen plasma levels, circulating adhesion molecules, etc.), leading to a higher frequency of events. Whatever the explanation might be, this discrepancy versus standard recommendations for risk assessment obviously limits the possibility of directly transferring the scientific evidence obtained in this otherwise interesting study to other populations, in which the risk profile may be ‘qualitatively’ different. It should be emphasized that a similar methodological problem appears also to be present in other relevant studies dealing with the prognostic significance of non-traditional risk factors in specific populations, as exemplified by the study of Perticone et al. [25], which strongly supported a prognostic role of endothelial function, as evaluated by strain-gauge plethysmography and intra-arterial infusion of acetylcholine in the forearm, in a population of Mediterranean patients with mild-moderate hypertension. The total cardiovascular event rate in their study was exceedingly high, too, reading 4.9% per year (mean follow-up time of 31.5 months), and only one-fifth of the observed events rate could be predicted on the basis of the expected population risk profile [25,26]. Conclusion In conclusion, the study by Mena et al. [22] proposes an interesting new method to quantify 24-h blood pressure variability, focusing on the pattern of sequential changes in blood pressure between consecutive measurements. It also offers data on the possible predictive power of this new blood pressure variability index in the clinical management of hypertensive patients. However, the actual prognostic value of this new method deserves further assessment in future studies. There is no doubt that, as with any demonstration of the prognostic significance of non-traditional risk factors or new indices of target organ damage, the effort made by Mena et al. [22] in exploring the clinical value of the ARV index represents an important contribution. In spite of this, care is needed in extrapolating data derived from this specific sample of patients to the general population of hypertensive subjects, given that the international recommendations for risk profile assessment have not been strictly fulfilled.