SErENdipITy - study of electrophysiological biomarkers (electroencephalogram and heart rate variability) in neonatal seIzures and encephalopathy

Thumbnail Image
Pavel, Andreea
Journal Title
Journal ISSN
Volume Title
University College Cork
Published Version
Research Projects
Organizational Units
Journal Issue
Background In recent years, the use of conventional electroencephalographic (EEG) monitoring in newborns has increased and it is now recognised worldwide to be the gold standard monitoring for neonatal seizure detection, as well as a reliable biomarker for brain injury and long term prognosis in this population. Similarly, heart rate variability (HRV) is a non-invasive monitoring which provides information regarding the autonomic nervous system function, and has also been explored as a biomarker for brain function and prognosis in newborns. These physiological monitoring have been proven useful, especially in newborns with encephalopathy following hypoxic ischaemic injury. Despite the increased use of therapeutic hypothermia (TH) and the decrease in adverse outcomes following hypoxic ischaemic encephalopathy (HIE), the incidence of HIE is 1.6 per 1000 live births in high-income countries. Early identification of newborns with HIE which might benefit from TH is still challenging in clinical practice. HIE is also the main cause of seizures in newborns. Due to the unique physiological properties of the neonatal brain, this is the age period with the highest risk of seizures, with an incidence of 1-5/1000 live births. Recent evidence showed that seizures themselves might add to the degree of brain injury regardless of the background pathology, and a high seizure burden was associated with worse long-term neurodevelopmental outcome. The majority of neonatal seizures are subclinical (electrographic only) or have very subtle clinical manifestations, thus the importance of early biomarkers to identify newborns which will develop seizures, as well as the importance of prolonged EEG monitoring for seizure detection. Aims The aims of this thesis were to explore the value of physiological biomarkers, such as early clinical parameters, EEG and HRV features for early prediction of neonatal HIE severity and seizures, as well as neonatal seizure detection and evolution. Using early EEG and HRV analysis I aim to develop prediction models for seizure occurrence and HIE severity. In addition, I plan to assess the clinical value of a neonatal seizure detection algorithm (ANSeR algorithm). By analysing newborns with HIE and electrographic seizures, I aim to describe the temporal evolution of seizures during TH and to evaluate if seizure burden and intensity (power) correlates with HIE severity and long-term outcome. Furthermore, I aim to assess if the time to treatment of the first electrographic seizure in newborns with HIE had an impact on subsequent seizure burden and outcome. Methods To achieve my goals, I have analysed infants recruited from two European multicentre cohort studies, across eight European tertiary neonatal intensive care units, between January 2011 and February 2017 ( Identifier: NCT02160171 and NCT02431780). The studies included infants born at 36 to 44 weeks corrected gestational age, requiring prolonged EEG monitoring for being at high risk of developing seizures or being suspected of having seizures. All infants had continuous EEG monitoring using disposable electrodes according to the 10:20 electrode placement system for neonates (F3, F4, C3, C4, Cz, T3, T4, O1/P3 and O2/P4), with simultaneous electrocardiography (ECG) monitoring. The clinical diagnosis of HIE was established by the teams at each recruiting site based on signs of perinatal asphyxia and encephalopathy on neurological examination (modified Sarnat score within 24 hours of age) and retrospectively corroborated with abnormalities suggestive of HIE on EEG and brain MRI. The clinical grade of HIE was based on the most severe score of modified Sarnat score. The encephalographic grade of HIE was established by visually analysing the EEG background using a system described previously by our group: 0-normal EEG background, 1-mild abnormalities, 2-moderate abnormalities, 3-major abnormalities and 4-inactive EEG background. The quantitative EEG analysis was performed using the NEURAL software package, extracting a set of features for power, discontinuity, spectral distribution and inter-hemispheric connectivity. The HRV analysis was performed using an in-house software (HRV Analysis, Beta Version 1.12, ©University College Cork 2008-2012) which automatically identified R-peaks on ECG recording. The RR interval was generated as the time difference between each R peak. HRV was expressed in time, frequency and complexity features. Electrographic seizures were defined as at least one EEG channel with sudden, repetitive and evolving waveforms for minimum 10 seconds. For all infants, electrographic seizures were identified by two neurophysiology experts in neonatal EEG. Seizure characteristics and seizure quantification were calculated based on these expert annotations. Results The two studies included 504 newborns, out of which 266 newborns had a diagnosis of HIE (3 newborns with HIE following postnatal collapse). Machine learning models were developed for early prediction of newborns with seizures in HIE, by analysing a cohort of 164 newborns with EEG monitoring before 12 hours of age. The best predictive models included both clinical parameters and EEG features: clinical and qualitative-EEG model (MCC (95% CI) 0.470 (0.336 to 0.602), AUC (95% CI) 0.721 (0.681 to 0.813)) and clinical and quantitative-EEG model (MCC (95% CI) 0.513 (0.376 to 0.645), AUC (95% CI) 0.746 (0.700 to 0.833)). A randomised controlled trial evaluating the real-time performance of a seizure detection algorithm (ANSeR algorithm) showed a higher detection of seizure hours in the algorithm group compared to the non-algorithm group (absolute difference (95% CI): 20.8% (3.6% to 37.1%)). Seizure hours detection between the two groups was even greater at weekends (Saturday-Sunday vs Monday-Friday), difference (95% CI): 16.6% (0.1% to 32.3%). Another study assessed the impact of the time to treatment of the first electrographic seizure on subsequent seizure burden showed significantly lower seizure burden and less seizures were noted in infants treated with anti-seizure medication (ASM) within 1 hour from seizure onset (p value=0.029 and 0.035, respectively). A similar trend was noted in the subgroup of infants who had a diagnosis of HIE (n=42). Analysis of newborns with HIE requiring TH with EEG monitoring throughout the rewarming phase, showed that newborns with seizures during active cooling and rewarming had a significantly higher seizure burden compared with newborns with seizures during active cooling exclusively (median (IQR) 167(54-275) vs 69(22-104) minutes, p=0.003). Seizure burden peaked at approximately 24 hours in both study groups and had a secondary seizure peak at 85 hours of age for the group of newborns with seizures during active cooling and rewarming. Newborns with seizures during active cooling and rewarming had a significantly higher risk of abnormal outcome compared to infants without seizures (OR(95% CI):4.62(1.40 to 15.24), p=0.012). In another study all electrographic seizures from 64 newborns with HIE and 2-year neurodevelopmental outcome were analysed. Infants in the severe HIE group had a higher seizure period, with more frequent seizures but less intense (lower mean seizure power), compared to infants in the moderate HIE group. Similar characteristics were associated with abnormal outcome at two years. Early HRV analysis was assessed to predict the EEG grade in neonatal HIE within the first 12 hours of life in 120 newborns. Performance of the HRV model was AUROC 0.837 (95% CI: 0.759-0.914), however performance of the HRV and clinical model combined had an AUROC of 0.895 (95% CI: 0.832-0.958). Conclusion Early qualitative and quantitative-EEG features alone and in combination with early clinical information can reliably predict infants that will later develop seizures in HIE, hours before seizure onset. The quantitative-EEG model proved as reliable as the analysis of a neonatal neurophysiologist expert (qualitative-EEG analysis) in predicting the likelihood of seizures, which could be also used to individualise the neurophysiology review frequency of the continuous EEG monitoring. The findings of the neonatal seizure detection algorithm trial validated in real time the performance of the ANSeR seizure detection algorithm by demonstrating its’ usefulness as a support tool for clinicians, especially during weekends when a limited number of health care professionals are available on site. The study of ASM timing for the first electrographic seizure showed that inappropriate treatment remains a concern in clinical practice, and that early anti-seizure treatment was associate with lower total seizure burden. Current findings suggest that treatment of neonatal seizures might be time-critical. The study of seizure evolution in HIE showed that one third of infants with HIE undergoing TH continued to have seizures after the completion of active cooling, increasing the overall seizure burden which might have an impact on long-term outcomes. Supporting the current guidelines recommendations, there is a clear need for continuous EEG monitoring during active cooling, rewarming and beyond when seizures persist. The seizure analysis in HIE showed that seizures were more frequent and were less intense in severe HIE compared to moderate HIE, and in newborns with adverse outcome compared to newborns with normal outcome at two years. This may have implications for seizure identification as low power seizures are usually harder to detect, especially using aEEG monitoring, which might have an impact on anti-seizure treatment and subsequently on long-term outcome. The study of early HRV analysis showed that HRV and clinical model had a good prediction of encephalopathy HIE severity in the early newborn period and may be a very useful additional tool for neonatologists who are often faced with challenging decisions about TH, especially where EEG monitoring is not available or feasible.
Neonatal encephalopathy , Seizures , EEG , Electroencephalography , Heart rate variability , HRV
Pavel, A. M. 2023. SErENdipITy - study of electrophysiological biomarkers (electroencephalogram and heart rate variability) in neonatal seIzures and encephalopathy. PhD Thesis, University College Cork.
Link to publisher’s version