35 Years Later Than Grossman et al. (1990): Respiratory-Sinus-Arrhythmia Quantification Strategy Is Still Not the Problem — A Critique of Lewis et al. (2012)

Diverse methods have been developed to quantify respiratory sinus arrhythmia (RSA), including peak-valley, spectral analytic and moving-polynomial-filtering estimations. Although these three procedures were found—some 35 years ago—to provide functionally equivalent indices of RSA (Grossman, van Beek & Wientjes, 1990), Lewis et al. (2012) attempted a partial replication of the earlier study in order to compare degree of agreement of these measures, as well as to examine the levels of influence of respiration and heart-period nonstationarity upon each RSA index. This paper continues to be both influential and controversial. However, several limitations in their report call several of the study’s conclusions into question : 1) The investigators failed to apply natural logarithm (ln) transformation to two of the three measures before pursuing subordinate analyses, even though the unaltered RSA estimates violate assumptions of normality and require ln transformation in order for inter-individual parametric statistics to be meaningful. 2) Instead of first comparing their own already ln-transformed measure (RSAp-b; derived from moving-polynomial detrending) with the other ln-transformed measures (peak-valley RSA and spectral-analysis-derived RSA), they mostly performed ln transformation of the latter measures only after already detrending each of these with their moving polynomial filter method. This sequence of analysis confounded the detrending procedure with the ln transformation and did not permit meaningful comparisons between their own moving-polynomial-filtered estimate and the other two. 3) Methods and findings of previous, highly pertinent studies were either misrepresented or not reported. 4) Instead of examining the critical issue of vulnerability of the different methods to violations of stationarity, evidence was simply presented that nonstationarity occurred in baseline time series of heart period and that detrending methods, as designed to perform and as expected, indeed, served to detrend the original time series. 5) The authors overinterpreted their findings regarding the correlations with and moderation by respiratory parameters. 6) Although the senior author performed analyses reported in a much earlier study that revealed very strong respiratory moderation of RSAp-b, these findings failed to be addressed in Lewis et al. (2012).