Saving the brain one heartbeat at a time

In an article in this issue of The Journal of Physiology, Yamaguchi et al. (2018) set out to demonstrate the time course of changes in several widely used heart rate variability (HRV) measures during an exposure to a hypoxic-ischaemic (HI) insult using fetal sheep as a model. Although this study was conducted in a fetal model, the findings may well be useful as a surrogate for the neonate with HI encephalopathy that needs timely diagnosis for effective treatment. This information is difficult to obtain from neonatal models of HI injury due to anaesthetic confounding or from human studies where there are the confounding challenges of the treatment paradox and the lack of non-cooled controls. The team demonstrated a complex temporal profile of three linear time-domain (also referred to as statistical) HRV measures, standard deviation of the normal–normal interval (SDNN), square root of the mean squared differences of n successive RR intervals (root mean square of the successive differences; RMSSD) and skewness, during the evolving HI brain injury. SDNN and RMSSD represent the true beat-to-beat HRV measures widely deployed throughout HRV studies in all ages. In addition, and to account for the current practice of the ultrasound-based fetal heart rate monitoring, the long-term variation (LTV) and short-term variation (STV) metrics were also computed. These two metrics, unique to fetal monitoring, are not true beat-to-beat HRV measures, but, rather, even in the case of STV, derived from an average over roughly 12 fetal heart beats (3.75 s or 1/16th min epoch). Two groups were studied: mild HI, known to be associated with only selective subcortical neuronal injury, and severe HI, associated with a pattern of diffuse white matter injury and severe subcortical neuronal injury seen in preterm babies. A complete umbilical cord occlusion (UCO) lasting 15 min triggered the mild HI, while a severe HI was triggered by 25 min of complete UCO. For the entire first 6 h after the UCO insult, the RMSSD measure of HRV showed no suppression, but rather an increase in the severe HI group compared to both the mild HI group and the controls. The skewness measure of the RR intervals of HRV was increased in the first 6 h after UCO insult in both experimental groups compared to controls. The conventional clinical belief appears to have simplified matters by presuming that severe asphyxia results in a 'global' HRV suppression, albeit no systematic studies have been conducted yet delineating how various HRV measures in different signal-analytical domains behave. The heart rate monitoring implication is that early loss of HRV would be useful for selecting infants with severe injury for potential trials of neuroprotection. In contrast to this understanding, Yamaguchi et al. found that after severe asphyxial brain injury in preterm fetal sheep, HRV measure RMSSD transiently increased in the first 6 h, the latent phase when a therapeutic intervention may succeed. RMSSD only became consistently suppressed from approximately 6 h after the insult. After this time we know that evolving cell death is becoming established, and is unlikely to be treatable. Thus, if as suggested by recent clinical studies in term infants, babies with 'suppression' of HRV were recruited for a trial of some neuroprotective therapy, there would be a significant chance that they would be already outside the ideal window for treatment. The findings by Yamaguchi et al. now suggest that continuous electrocardiogram (ECG)-derived HRV monitoring with online calculation of a set of HRV measures may be useful for early detection and follow-up of an evolving perinatal brain injury. Such monitoring may identify a pathophysiologically correct time window for therapeutic intervention in severely asphyxiated babies during the latent phase of injury. While the ability to identify babies that can benefit from therapeutic intervention is the key translational finding of the work, there are several other important insights to be gained from this study. I focus on these insights because they are the determinants of translational success of this study's findings and because they highlight the opportunity for a deeper understanding of HRV. First, for HRV monitoring to deliver on this promise, non-invasive acquisition of the fetal ECG from the maternal abdomen is needed at the sampling rate of 1000 Hz (Frasch et al. 2017). This forms the basis for a precise reconstruction of fetal heart rate time series from the R–R interval length fluctuations and to compute HRV measures such as RMSSD, SDNN or skewness deployed in the study by Yamaguchi et al. This fetal monitoring technology does not yet widely exist in hospitals across the world. Second, HRV changes can be subtle and multifaceted. Much like a whole genome cannot be referred to as going up or down, but we can meaningfully discuss changes in specific genes or regions, so the HRV changes cannot be referred to as being globally decreased or increased. Sometimes, we may become too consumed by a single set of the so-called linear or non-linear HRV measures, assigning them the putative prognostic merit. But can we truly assume that any single HRV measure or a set of certain HRV measures, linear or non-linear, is bound to become the magic bullet? What lies beyond the non-linear? For a meaningful qualitative and quantitative description of HRV, we need a consensus on how to report it in a way that does not obscure its complexity and limits its diagnostic or prognostic potential. That requires a recognition of the existence and deeper understanding of the various signal-analytical domains of HRV. Such domains partially overlap and, importantly, represent the underlying physiological dynamics that are associated with certain health outcomes, such as an early evolving HI injury. The signal-analytical framework for HRV analysis has evolved over the past four decades, yet we are still wondering whether the bridge will ever be built between the pragmatic (and quite successful!) use of HRV measures for prediction of health outcomes on one hand and the attempt to assign the various HRV measures a robust physiological meaning on the other hand. The present research paper crosses that bridge carefully by attempting to interpret the findings in early HRV rise captured by RMSSD as the adaptive sympathetic or vagal activation patterns in response to the HI brain injury. In doing so, the authors lay down a paradigm for future studies attempting to dissect such relationships further. According to this paradigm, concomitant recordings and analysis of fetal ECG, brain electrical activity (electroencephalogram, EEG), fetal breathing movements, nuchal muscle activity (EMG) and arterial blood pressure permit inferences into the nature of the underlying rhythms contributing to the HRV. The ability to perform such complex instrumentation reliably represents a unique strength of the fetal sheep animal model of human physiology. While this approach has been practiced in this model for four decades, the joint interpretation to deduce origins of the HRV patterns is rather novel and will guide future studies. The truth about the origins of HRV may be more complicated than previously thought. Not only are sophisticated experimental models required to unravel the HRV, but also conceptual paradigm shifts. HRV, in its multidimensional complexity, may not be reducible to strict representations of dedicated physiological rhythms (e.g. sympathetic or vagal activity fluctuations). Such convenient reductionism may be a product of didactic tradition, but may be contrary to the physiology itself. This notion then logically calls for mathematical approaches that preserve this multidimensional complexity throughout the analysis. As our data sets grow bigger and more complex, any subjective dimensionality reduction may result in decreased accuracy of the HRV models (Frasch et al. 2009; Herry et al. 2016). The study by Yamaguchi et al., underscored in their discussion of HRV skewness findings, highlights the need for novel HRV approaches capturing the complex HRV response to HI injury or other insults resulting in perinatal brain injury that can benefit from early therapeutic intervention. Finally, adult neurology has long sought to deploy HRV monitoring as a means to predict the deleterious effects of stroke (Graff et al. 2013). While this remains a work in progress, findings so far indicate the potential of HRV monitoring to identify patients at risk of adverse early neurological outcomes and of a second stroke over years of follow-up (Graff et al. 2013; Bodapati et al. 2017). Similar diagnostic and discriminatory abilities can be expected in the fetal and neonatal context. ECG-derived HRV monitoring remains the most accessible tool of fetal and neonatal surveillance. Thanks to work by Yamaguchi et al., we are beginning to unravel the potential of this technology to save babies' brains one heartbeat at a time. MGF is supported by CIHR. MGF holds a provisional patent and a PCT filing on fetal ECG technologies.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.