Sympathovagal balance from heart rate variability: time for a second round?

Our attention was recently attracted by a letter by Heathers (2012) addressing methodological aspects of autonomic assessment in intact humans. Basically, the letter calls for an obituary of sympathovagal balance from heart rate variability metrics and proposes to consider in addition baroreflex for autonomic assessment. Having contributed to the introduction of sympathovagal balance in the literature (Pagani et al. 1986), we felt committed to provide some comments in order to help clarify the issue for readers' sake, and because our relevant studies were not quoted or discussed in Heathers's letter or in the review paper they refer to (Goldstein et al. 2011). A detailed answer to prior criticisms (Eckberg, 1997) to the concept of sympathovagal balance as derived from classical neurophysiological models has already been published (Malliani et al. 1998; Pagani & Malliani, 2000). In particular, in our previous discussions we point out the importance for a deeper understanding of considering not only multiple signals, such as respiration and arterial pressure, with appropriate multivariate models and algorithms, but also of addressing the key role of underlying coding in interpreting neural autonomic signals. Information can in fact be gained from signals only if we know the rules of the association between neural patterns and the final effect on a given set of cardiovascular variables. Surprisingly, as of 2000, the percentage of papers addressing neural codes in autonomic activity was very low, being about 0.04% in the case of heart rate variability (HRV) and about 1% in the case of neural activity. These percentages are only slightly improved after 12 years (0.2 and 2.8%, respectively). Coding was essentially not addressed in studies of the other autonomic metrics. No wonder that in the absence of sufficient information on such a critical methodological issue debates might continue still nowadays. Correctly, Heathers (2012) seems to express a feeling of uneasiness regarding the improper use by many authors of metrics derived from HRV, frequently obtained with inappropriate methodologies, such as simple heart rate monitors in sport, that do not guarantee the accurate assessment of a faithful tachogram. In addition, investigators frequently rely on commercial, proprietary software tools to extract autonomic indices, using unclearly defined algorithms. Regarding our first description (Pagani et al. 1986), sympathovagal balance, particularly considering low-frequency (LF)/high-frequency (HF) ratio, was suggested in a context of a dynamic rest–tilt protocol. This approach was later shown capable in healthy humans of predicting a difference in posture, from supine to upright, individually with an accuracy better than 80% (Malliani et al. 1997). One additional weakness of the autonomic assessment in humans derives from the intrinsic complexity of the underlying neural substratum, inclusive of central structures, negative feedback (i.e. the baroreflex) and positive feedback circuits (i.e. sympathosympathetic excitatory reflex, e.g. Pagani et al. 1982; Malliani et al. 1991), peripheral interaction at the effector levels, which are also sensitive to humoral factors, and central coupling among autonomous oscillators (respiratory and vasomotor centres), circuits capable of synchronizing/desynchronizing peripheral vasomotor activity (Baselli et al. 2006). In addition, the picture is rendered more complex by the presence of coactivations between vagal and sympathetic branches of the autonomic nervous system. By definition, coactivations defy the concept of sympathovagal balance; identifying the time windows of coactivations might guide the use of the paradigm of sympathovagal balance only when it actually holds. Addressing this complexity in a real-life clinical application requires striking a balance between the complexity of recordings, methods and accuracy of results, driven by specific and variable questions. Combining different methodologies in a system analytical approach (Ahn et al. 2006) with the specific goal of assessing the complexity of the cardiovascular control (Porta et al. 2009) might help to minimize the uncertainty of simpler techniques; for example, using a dynamic protocol (rest–stand) might improve the predictive capacity of simple RR variability. Needless to say, adding the arterial pressure signal and deriving, through various methods, the gain of spontaneous baroreflex might further upgrade the capacity to recognize dysregulation of the autonomic nervous system, e.g. in prehypertension (Lucini et al. 2002) or other conditions, such as paediatric diabetes (Lucini et al. 2009). In paediatric diabetes, arterial blood pressure variability might also add useful information, as we recently discussed in a letter to The Journal of Physiology (Pagani et al. 2012). We basically agree, therefore, with James A. J. Heathers regarding his suggestion to hope for a second generation of spectral analysis of HRV, more respectful of known physiology. However, given the complexity of the self-regulating organization of the autonomic nervous system, this approach should consider all the possible knowledge available about the system (Haken, 1983), while avoiding selective abolitions of specific data not fitting into simplified visions. In this context, we wish to reframe the following two statements. There is a dissociation between peripherally measured sympathetic activity [i.e. muscle sympathetic nerve activity (MSNA)] and LF HRV. We demonstrated a strong coherence between LF (and HF) oscillations of MSNA and similar oscillations in RR and systolic arterial pressure variability (Pagani et al. 1997); however, no correlation was found with MSNA expressed in the usual time domain metrics, such as bursts per minute. The LF component of RR variability results from a baroreflex relationship. While there is an expected mathematical relationship between the absolute power of LF RR and the frequency domain baroreflex index (Pagani et al. 1988), a disconnection is present between LF RR in normalized units and index α. Indeed, from a series of 191 subjects (43 ± 13 years old) we found a strong correlation between LF RR in milliseconds squared and α LF (r = 0.608, P << 0.001), but no correlation between LF RR in normalized units and α LF (r = 0.093, P = 0.209). More detailed discussion of conditions in which LF RR increases independent of the baroreflex are provided elsewhere (e.g. see figure 1 of Pagani & Malliani, 2000). In keeping with our previous consideration (Malliani et al. 1998), the sympathovagal balance is nothing but a concept, a construct that should be gauged against its heuristic value. We totally agree with the wish of James A. J. Heathers for a second generation of methods to assess in humans the autonomic regulation and dysregulation of various physiopathological conditions, particularly in the context of therapeutic strategies for targeting excessive sympathetic activation in humans (Fisher & Fadel, 2009). This would entail some pronounced changes from the current use of spectral analysis of cardiovascular variabilities, particularly new standards of measurements, techniques and clinical applications; however, in the process, as we recently stated (Pagani et al. 2012), we must be careful to avoid throwing away the water and the baby too.