Assessing circadian blood pressure and heart rate changes

A major advantage of ambulatory blood pressure monitoring (ABPM) over conventional blood pressure measurements is its capability to provide information on 24-h blood pressure and heart rate profiles [1–4]. Circadian blood pressure and heart rate changes, including a fall in these variables at the time of night sleep and their rise when waking up in the morning, represent the quantitatively most important components of 24-h blood pressure and heart rate variability [5–7]. These components may be clinically relevant, as shown by a number of studies, making use of ABPM, which have demonstrated how either an impaired or an excessive blood pressure fall at night and a steeper blood pressure rise at the time of morning awakening, when reproducible [8], might be associated with a higher rate of cardiovascular events [9–15]. However, discrepant results have been obtained on this issue, probably because in most of these studies, circadian blood pressure and heart rate changes were assessed with a considerable degree of approximation, that might have prevented an accurate quantification either the blood pressure and heart rate changes precisely occurring at the onset of sleep, or the rate of rise in these parameters at the actual time of morning awakening [3,4,16]. Methods commonly employed to assess day-night blood pressure and heart rate changes Probably the most common approach to the assessment of night-time blood pressure fall is the simple calculation of the difference between daytime and night-time blood pressure average values or the quantification of the night–day blood pressure ratio. However, in either case, daytime and night-time subperiods are often arbitrarily defined based on fixed clock-time intervals, with or without exclusion of so-called ‘transitional periods’ (i.e. the 3–4 h intervals occurring in the evening and in the early morning over which the actual time of going to bed and of waking up, respectively, show a wide degree of dispersion between individual subjects) [17,18]. Even when investigators make their best to identify the time when a given individual goes to bed at night or gets up in the morning, based on the information reported in subjects’ logbook, the precision in defining daytime and night-time subperiods in the absence of objective measurements of the shift between wakefulness and sleep is often quite limited [19]. A second method that specifically considers the cyclic nature of circadian blood pressure and heart rate fluctuations, and which has been widely used by chronobiologists for a number of years, is the Cosinor method [20]. This method fits blood pressure and heart rate 24-h changes to a cosine function, yielding information on the amplitude of these oscillations and on their phase. However, despite its simple application, this approach is undermined by major problems that are related to a number of unverified assumptions. Among them is the hypothesis that blood pressure and heart rate circadian changes occur very smoothly and can be modelled by a single harmonic, which implies that these changes can be satisfactorily summarized by a single sinusoidal waveform with equal duration of day and night periods. This is clearly not the case in real life, and several studies have shown how circadian blood pressure and heart rate changes do occur much more abruptly both in humans [5–7] and in animals [21–23]. Thus, the Cosinor method may yield excessively smoothed blood pressure and heart rate profiles that fail to properly reflect their actual changes between day and night. A third method employed to assess circadian changes in cardiovascular parameters is the square-wave fit approach [24]. This approach appears to fit the changes in blood pressure and heart rate between day and night better than the Cosinor method does. However, its main limitation is related to the underlying assumption that blood pressure and heart rate changes between day and night always occur very abruptly, such that a complete transition between daytime and night-time patterns is taken as occurring over an extremely narrow time window. This assumption fails to reflect what is commonly observed on inspecting 24-h blood pressure and heart rate tracings in which the rate of change in these parameters between day and night is highly variable between individuals. A fourth approach which partially overcomes some of the limitations of the square-wave fit method is the Cumulative Sums analysis (CuSums) [25]. This method provides an assessment of circadian blood pressure and heart rate oscillations through the quantification of the sequential changes in these variables (considering their sign) ‘weighted’ by their duration in time. The parameters quantifying circadian blood pressure and heart rate changes obtained with this approach, such as the Cusum Plot Height (which corresponds to the circadian amplitude of blood pressure and heart rate fluctuations), are reported not to be influenced by short-term oscillations of these variables [25], and should reflect only their main changes between day and night. However, a limitation of this method is that it may lead to an overmodelling of real circadian blood pressure and heart rate fluctuations, resulting in an excessive smoothing of all other components of 24-h blood pressure variability besides those considered by CuSum plots. A fifth approach, which is more commonly employed in studies addressing this issue, and which may better take into account the complexity of blood pressure variations over time, consists of the application of a partial Fourier analysis, based on use of at least four to six harmonics, to the assessment of blood pressure and heart rate circadian fluctuations [26]. Although this method (which generates up to 11 derived parameters), appears to represent an improvement compared to the approaches previously described, it is not in itself free from limitations. On the one hand, it is not yet clear what the precise biological meaning of the various parameters generated by this analysis might be. On the other hand, the application of spectral analysis to unevenly distributed blood pressure or heart rate values sampled at a very low frequency (usually every 20–30 min), compared to the actual frequency of beat- by-beat changes in blood pressure and heart rate (approximately 1 Hz or higher), may also lead to data overmodelling, resulting in excessively smoothed blood pressure profiles and, possibly, in an aliasing error when the analysis focuses not only on circadian changes, but also on faster fluctuations [27] (Fig. 1).Fig. 1: Fourier analysis with four harmonics of 24-h blood pressure profiles. (a) Showing the results obtained when analysing a beat-by-beat intra-arterial recording. (b) Showing the results obtained when the same analysis method was applied to a discontinuous non-invasive blood pressure recording simultaneously obtained in the same subject by applying the cuff of a Spacelabs 90207 device (Spacelabs Inc., Richmond, Washington, USA) on the arm contralateral to that in which the intra-arterial catheter was inserted. The dots refer to individual blood pressure measuremements. The continuous lines refer to spectral analysis curves fitting the raw data. (Parati G, Di Rienzo M, Mancia G, unpublished observations). SBP, Systolic blood pressure.Moreover, as in the case of most of the methods described to date, the mathematical modelling used to fit the raw data implies that changes between wakefulness and sleep at night, and those between sleep and wakefulness at arousal are largely symmetrical, a condition that is not necessarily met in real life when considering circadian cardiovascular fluctuations in individual subjects. In this issue of the journal, Head et al. [28] describe a new method for the analysis of 24-h blood pressure recordings, aimed at quantifying the changes in blood pressure and heart rate from wakefulness to sleep periods and vice-versa, by means of a new six-parameter double-logistic curve fitting procedure. As reported by the authors, a theoretical advantage of this approach compared to previously available methods, is that it does not imply any a priori assumption about the abruptness or symmetry of blood pressure and heart rate circadian changes [28]. The authors applied this method to quantify the circadian changes in blood pressure, heart rate and behavioural activity of normotensive Sprague–Dawley rats and of spontaneously hypertensive rats, with signals being carefully recorded via implanted radiotelemetry transmitters. The main finding reported by Head et al. [28] is that this new logistic curve method revealed clearly asymmetric circadian patterns both in blood pressure and heart rate and in physical activity, both in normotensive and in spontaneously hypertensive rats. In particular, the method Head et al. [28] allowed the detection of a greater rate of increase in diastolic blood pressure during the waking period in spontaneously hypertensive rats as compared to normotensive rats, which was not associated with parallel differences in heart rate or activity changes. No major differences in results were found when considering data averaged every 10, 30 and 60 min, respectively, suggesting that even a low sampling rate might be sufficient for this kind of analysis. Finally, the authors compared the results obtained by their method with those provided by the Cosinor, the square-wave fit and the partial Fourier methods. The parameter that diplayed the best fit with these mathematical models was heart rate, with the double-logistic equation and the Fourier method showing a better fit than the square fit and the Cosinor methods. However, all circadian models performed less well in accounting for circadian changes in blood pressure, although the double-logistic, the partial Fourier and the square-wave fit did better than the Cosinor, and the day–night difference in blood pressure was best estimated by the double-logistic method than by all the other approaches. Overall, in the study of Head et al. [28], the double-logistic method provided a better fitting of the biological data than the square-wave fit or the Cosinor methods, and an equivalent fit compared to the partial Fourier method, but with fewer parameters. Taken together, the results of the study by Head et al. [28] indicate that this new method may indeed represent an improvement, as compared to the other methods until now available described above. However, the data reported by Head et al. [28] also indicate that properly modelling circadian blood pressure changes is by no means an easy task, and that the optimal compromise between a faithful representation of raw data and a satisfactory modelling procedure, that summarizes the biological phenomena through a mathematical equation, is still far from being fully achieved. Finally, it should be emphasized that the actual value of this double logistic approach in analysing 24-h blood pressure and heart rate profiles in normotensive and hypertensive humans, and its ability to quantify changes in circadian blood pressure and heart profiles induced by pharmacological treatment, still remains an unexplored issue. Given the evidence on the prognostic value of alterations in these circadian phenomena, such an issue would deserve to be carefully explored by further ad hoc studies.