Sleep wake cycling in early preterm infants: Comparison of polysomnographic recordings with a novel EEG-based index

Highlights

• Genuine sleep states with polygraphic correlates can be recorded from early preterm babies.

• Temporal coverage of spontaneous activity transients (SAT%) in the EEG shows significantly different levels between sleep stages both at the individual and at the group levels.

• A trend based on SAT% may provide an objective, online marker for sleep wake cycling in early preterm babies.

Abstract

Objective

To document the occurrence of genuine sleep stages in the early preterm babies, and to develop an EEG-based index for following sleep wake cyclicity.

Methods

Twelve preterm babies were recruited from a study that assessed ventilator strategies. We used altogether 18 polysomnography recordings that were collected at mean conceptional age of 29.3 (25.9–32.7) weeks. Spontaneous activity transients (SAT) were detected automatically and their cumulative coverage in each 20 s interval was computed from the EEG derivations C3–A2 and O2–A1. Mean SAT% values between sleep stages were compared.

Results

All babies exhibited all sleep stages, however the sleep was remarkably fragmentary in infants due to their respiratory issues. The EEG index, SAT% showed temporal behavior that strikingly well compared with the sleep stage fluctuations in the hypnogram. In the statistical analysis we found significant differences in all recordings between the deep (quiet) sleep and the REM sleep.

Conclusion

Genuine sleep states exist in the early preterm babies, and changes in sleep stages are reflected in the EEG activity in a way that can be readily measured by assessing fluctuation of the automatically detected, EEG based index, the SAT%.

Significance

The findings open a possibility to construct automated analysis or monitoring of sleep wake cyclicity into brain monitors in neonatal intensive care unit.

Introduction

Early preterm birth is associated with a significant risk for neurodevelopmental compromises during later life, and an increasing evidence suggests much of this risk is likely to arise from subtle insults affecting brain during earliest prematurity. It has therefore become increasingly evident that an optimal care of brain in the neonatal intensive care unit (NICU) would need continuous monitoring of neurological well-being. Building a physiologically reasoned paradigm for early preterm brain monitoring is challenged by the fact that non-invasive monitors, such as EEG, can only measure activity within cortex and subplate, while the brain insults leading to adverse outcomes typically affect deeper brain areas as well (de Kieviet et al., 2012).

One of the early signs of neurologically stable baby, and consequently of a favorable outcome, is baby’s ability to maintain fluctuation of vigilance stages, often referred to as sleep–wake cycling (SWC; Weisman et al., 2011). Maintenance of proper SWC requires functional integrity of widely distributed brain networks (Villablanca, 2004, Karlsson et al., 2005), as well as a sufficiently stable clinical status (Hayes et al., 2007, Olischar et al., 2007, Weisman et al., 2011). These raise an intriguing possibility that monitoring SWC could provide a valuable surrogate measure of brain’s wealth in the early preterm babies during their critical first few weeks of life. Continuous monitoring of SWC is, however, challenged by the fact that conventional EEG reading does only recognize sleep states after about 30th week of conceptional age (CA; Grigg-Dammberger et al., 2007). This is, however, several weeks after the time window when the preterm babies are most vulnerable and would need most brain surveillance. It is intriguing in this context, that some recent studies using continuous EEG monitoring with aEEG trends (amplitude integrated EEG; Hellström-Westas et al., 2006) have suggested, that visualizations of selectively compressed EEG might reveal fluctuation (often reported as “cycling”) between presumably vigilance stages long before it becomes recognized in the conventional EEG reading (Hellström-Westas et al., 1991, Olischar et al., 2007, Wikström et al., 2012)

Prospects of developing a physiologically reasoned SWC monitor for early preterm babies is currently challenged by the shortage of two crucial pieces of information: First, the presence of genuine sleep stages as defined by behavioral measures (i.e. ocular and muscle activity) should be established in the very early preterm babies. Second, provided that such sleep stages do take place, one should also show that they can be measured from the cortical (EEG) activity in a way that is apt to implementation into currently used brain monitors. It is notable in this context, that preterm EEG is known to fluctuate between background state with extended intervals between activity bouts (trace discontinue), and another background state with more continuous EEG activity (André et al., 2010). Indeed, these two modes of activity may be readily seen as potential candidates to reflect sleep stages. In the present study, we reasoned that the main difference between these background modes is the cumulative amount of activity bouts, which we call Spontaneous Activity Transients (SAT; also called by other names, see Table 1 in Vanhatalo and Kaila, 2010). Consequently, it could be possible to measure difference between these background states by following the proportion of EEG signal covered by SAT events, which we call SAT%. Such quantitation became possible with the introduction of an optimized and validated SAT detector for the EEG of young preterm babies (Palmu et al., 2010a, Palmu et al., 2010b)

The present study was designed to respond to both of the above challenges. First, we employed the gold standard of sleep studies, polysomnography (PSG), to study the occurrence of genuine sleep stages in the early preterm babies. Second, we employed a recently developed measure of early cortical activity, coverage of Spontaneous Activity Transients, (SAT%; Palmu et al., 2010a, Palmu et al., 2010b) to assess how procession of baby’s sleep across different stages could be assessed from an index derived from the raw EEG signal that is available in every routine brain monitors.