The last word

Abstract

I thank Drs Pagani, Lucini and Porta for their substantive recent correspondence (Pagani et al. 2012). Two further, brief points arise. Firstly, our collective discussion here is in part assumptive. Typical spectral analytic methods with which low-frequency variability in the heart (LF-HRV) is derived are themselves subject to robust criticism, including but not limited to the following points: (a) the role of error correction in the preparation of a heart rate series, and its relationship to the volatility in spectral components (e.g. Berntson & Stowell, 1998); (b) the assumption that the interpolation of an irregularly sampled heart rate series does not interfere with its spectral components (e.g. Clifford & Tarassenko, 2005); and (c) the necessity for either non-parametric tests for the statistical comparison of spectral components or their logarithmic transformation, to meet common statistical assumptions (e.g. Kobayashi et al. 2012). These are not minor points, but serious questions which must be addressed systematically when a heart rate series is recorded, corrected and decomposed into spectra, and are frequently overlooked. We must also outline the problem of ‘how’ along with ‘what’; it will always be somewhat premature to discuss the meaning we can extract from a metric that we cannot measure sufficiently. To this end, I agree absolutely with the call to maximize the informational content of measured physiological variables first outlined by Pagani & Malliani (2000). Secondly, a broader observation: it is significant that all commentators in this series (Goldstein et al. 2011; Heathers, 2012; Prakesh, 2012; Pagani et al. 2012) have explicitly criticized the meaning and use of measures which purport to calculate cardiac sympathetic tone from LF-HRV, not at least without the admixture of further autonomic, circulatory, neurographic or respiratory information and/or more sophisticated methods of analysis. We are by no means the first to do so recently (Elghozi & Julien, 2007; Rahman et al. 2011). However, it seems that the weight of this opinion comes with a problem of translation, because it has had a moderate influence (at best) within applied research. It is still commonplace to find LF-HRV proposed within an overall marker of ‘autonomic tone’ and therefore used as a biomarker of disease state, mood, emotion, attention, exercise capacity/recovery, etc. This work still readily contains the stated or assumptive linearity of the original Cannon (1929) model of autonomic activation, and directly considers LF-HRV power as an index of cardiac sympathetic tone or the derived low-frequency/high-frequency index as a direct reflection on a unidimensional autonomic dimension. The enthusiasm for insight into the autonomic nervous system through an inexpensive and non-invasive method such as HRV is understandable, but should be heavily tempered in light of the concerns outlined. Readers are invited to give their opinion on this article. To submit a comment, go to: http://ep.physoc.org/letters/submit/expphysiol;98/1/348