The systolic–diastolic pressure relationship

Repeated measurements of systolic blood pressure (SBP) and diastolic blood pressure (DBP) frequently show a highly linear relationship. This phenomenon was observed in the Framingham study over a 14-year period using office blood pressure (BP) measurement, in-home BP monitoring over a few weeks, 24-h ambulatory blood pressure measurement (ABPM) and beat-by-beat BP monitoring over a few minutes. The standard statistical measures of this linear relationship are the regression slope and the SBP–DBP correlation coefficient r. The currently used slope-related measures are (a) the regression slope obtained by treating the DBP as an independent variable, hereafter called the S–D slope, after Gavish et al.,1 and (b) one minus the regression slope, calculated by treating the SBP as the independent variable, called the Arterial Stiffness Index (ASI), after Li et al.2 However, the determination of slope-related measures using standard regression leads to artifactual dependence on r; that is, the slope becomes dependent on the degree of data scattering. The reason for this is that the SBP and the DBP are measured simultaneously, and neither variable can be described as ‘dependent’ or ‘independent’. This problem is eliminated by using ‘symmetric regression’, which handles both variables in a symmetrical way.1, 3 Figure 1 demonstrates why this issue becomes important when attempting to estimate slope-related measures.

Figure 1
figure 1

The relationship between the slope-related measures of the ASI and the S–D slope when calculated for individual patients by different regression methods using the 24-h ABPM data of 3703 subjects (analysis performed with the permission of Dr Michael Bursztyn). The dashed-line contours represent the theoretical relationship between the two measures for a given value of the SBP–DBP correlation coefficient r (data were analyzed with the permission of Dr Michael Bursztyn). With symmetric regression, the ASI equals 1−1/(S–D slope), making both measures equivalent.

The link to risk-related clinical measures

Slope-related measures determined from 24-h ABPMs, mainly the ambulatory ASI (AASI), have been demonstrated to be independent predictors of cardiovascular (CV) mortality, all-cause mortality and stroke in the general population.3, 4, 5 The AASI has been independently associated with target organ damage using both ambulatory and home BP measurements.6, 7 Furthermore, these measures display a gradual increase with age, pulse pressure and pulse wave velocity—all known as risk factors for future CV events.2, 3, 4 On the other hand, slope-related measures are independent of the mean arterial pressure3—a property that may be advantageous in the early detection of CV risk in pre-hypertension.

Based on this background, it can be seen that the potential use of home BP monitors for determining slope-related measures with prognostic significance (in addition to standard BP determination) is especially promising: This is because of the popularity of home BP monitors, its long-term use and the increased availability of data communication between the patient's home and the health-care providers.

This issue of Hypertension Research includes a paper by Stergiou et al.8 that compares the ASI in hypertensive patients, determined using 24-h ABPMs (AASI), with the ASI derived from home BP measurements, called the Home ASI (HASI). The findings that the HASI was higher than the AASI; did not correlate with the 24-h, daytime or nighttime AASI; and displayed similar but weaker correlation than the AASI with age, BP and pulse pressure led the authors to conclude that home BP measurements cannot replace ambulatory measurements in the determination of the ASI. By reaching that conclusion, the authors took for granted that the two methods are supposed to represent the same measure, which is not obvious. Furthermore, the study did not attempt to evaluate the prognostic significance of the ASI determined by the two BP measurement methods. However, the study is important for its attracting attention to the need for a better understanding of the effect of using a specific BP measurement method on the magnitude and the diagnostic and prognostic power of the slope-related measures.

Risk-related measures associated with BP data are conceptually different for ambulatory and home BP methods: The ambulatory method samples BP at a constant rate day and night (but at different rates for each period). In contrast, a home BP measurement is taken during predetermined narrow time windows (0600–0900 and 1800–2100 hours in the present study).8 The fact that the first home BP measurement was taken during 0600–0900 hours, which actually overlaps with the timing of the CV risk-related ‘morning surge’,9 could be a testable rationale for the difference between the HASI and the AASI.

The link to BP variability

The slope-related measures estimated using symmetric regression can be expressed as follows: The S–D slope is given by (SBP variability)/(DBP variability) (also called the BP variability ratio) and the ASI is equal to one minus (DBP variability)/(SBP variability), where the variability is estimated by the standard deviation.3 These expressions link the above-mentioned prognostic power of slope-related measures to the BP variability. Increased BP variability has already been shown to represent an independent risk factor for future CV complications using 24-h ABPMs,10 home BP measurements11 or visit-by-visit office BP measurements.12 In fact, the data of Stergiou et al. show that the SBP variability in the home BP measurements was significantly higher than the SBP variability in the ambulatory BP, and the corresponding home DBP variability was somewhat lower than the ambulatory one. These relationships are consistent with the observed AASI and HASI difference. It is not unlikely that the extra home-based SBP variability is associated with day-to-day changes in the physiological variables that determine the slope-related variables.

The link to arterial properties

The interpretation of the AASI as a surrogate index of arterial stiffness is still debatable. However, it is generally believed that slope-related measures do reflect arterial properties. Figure 2 provides a model that expresses the S–D slope by the (systolic stiffness)/(diastolic stiffness), which is associated with ‘curved’ pressure–diameter relationships (non-elastic arteries), suggesting that a unity S–D slope (or zero ASI) reflects elastic arteries having constant stiffness. According to this testable model, the ASI is more an ‘Arterial Stiffening Index’ than an ‘Arterial Stiffness Index’ in that it represents a functional increase in the arterial stiffness during the systole. The physiological interpretation of the slope-related measures and related factors affecting the SBP–DBP correlations should be exciting topics for further investigation.

Figure 2
figure 2

A physiological interpretation of the S–D slope: The first BP reading (filled circles) and a second one (empty circles) are depicted in an arterial pressure–diameter diagram (a) and the SBP–DBP diagram (b). Under the assumption that the (average) pulse diameter is constant, the transition between the first and the second measurements represents a uniform ‘shift1’ along the diameter axis and corresponding ‘shift2’ and ‘shift3’ along the arterial pressure axis in the DBP and the SBP, respectively, in both diagrams. The S–D slope is given by (shift3)/(shift2). For a curved pressure–diameter relationship, (shift3)>(shift2), which corresponds to an S–D slope>1 (or ASI>0). The ratio (shift3)/(shift1) is the systolic stiffness (one of its definitions), and (shift2)/(shift1) is the diastolic stiffness. Thus, within the present assumption, the S–D slope equals (systolic stiffness)/(diastolic stiffness). The diagram also rationalizes the correlation between the S–D slope and the pulse pressure by their common dependence on the pulse diameter.