Restriction lift date: 2022-05-30
Electroencephalography of premature infants
Lloyd, Rhodri O.
University College Cork
Background and Objectives: The early prediction of neurodevelopmental outcome in very preterm infants remains challenging. An objective tool with the potential to provide useful information about preterm brain health is the electroencephalogram (EEG) but current knowledge remains incomplete in infants of this age group. The ability to record continuous conventional EEG is usually overlooked due to the ease of application and maintenance of the amplitude-integrated EEG (aEEG). The aEEG is routinely used to identify seizures, assess background EEG and predict outcome, despite the fact that it has considerable limitations for preterm infants in particular. Research using conventional multichannel EEG in preterm infants is ongoing but studies tend to be of short duration and at varying periods post-birth. To progress and achieve future information about the predictive ability of EEG for neurodevelopmental outcome in preterm infants, more in-depth analysis is required. In this thesis, I aim to progress current knowledge in very preterm infants <32 weeks gestational age (GA) by investigating the ability of EEG to assess neurological wellbeing and to predict neurodevelopmental outcome at 2 years. Furthermore, I aim to investigate and described the frequency and characteristics of electrographic seizures during the early postnatal period in very preterm infants, and compare this to the existing literature. In addition, I aim to develop a standardised scheme for assessing both the normal and abnormal EEG features of preterm infants according to post-menstrual age. Finally, I aim to investigate the EEG of preterm twins and assess EEG concordance between monochorionic-diamniotic (MCDA) and dichorionic-diamniotic (DCDA) twins. Methods: Two cohorts of preterm infants <32weeks GA were recruited from 2009-2014 (cohort 1; 2009 – 2011 and cohort 2; 2013 – 2014). All infants had continuous conventional video-EEG monitoring. The EEGs from cohort 1 were recorded as soon as possible after birth, while the EEGs from cohort 2 were recorded within the first 12 hours of age, continued for approximately 72 hours with further short follow-up recordings at 32 weeks corrected gestational age and at pre discharge. EEG was graded as normal (normal or mildly abnormal) and abnormal (moderately abnormal or severely abnormal). Clinical demographics, clinical risk scores and details of the clinical course in the Neonatal Intensive Care Unit (NICU) were also collected. Neurodevelopmental outcome was assessed at 2 years of age via the Bayley Scales of Infant Development-III (Bayley-III). We used cohort 1 to develop a multimodal model for the prediction of neurodevelopmental outcome. This model incorporated simultaneous multi-channel electroencephalography (EEG), peripheral oxygen saturation (SpO2), and heart rate (HR) recordings. One-hour epochs of EEG, HR and SpO2 were then extracted at 12 and 24 hours of age from each recording. EEG grades were combined with GA and quantitative features of HR and SpO2 in a logistic regression model to predict outcome. Clinical status was also incorporated into the model to predict neurodevelopmental outcome. EEGs from both cohorts were used to examine seizures in very preterm infants. The entire video-EEG recording for each infant was reviewed and all electrographic seizures were visually identified, annotated, and analysed. Quantitative descriptors of the temporal evolution of seizures were calculated including total seizure burden, mean seizure duration, and maximum seizure burden. For each seizure, the onset location, morphology and evolution were described. We used cohort 1 to develop a standardised EEG assessment scheme for preterm infants. Initially a comprehensive literature review was performed by two electroencephalographers (EP & RL ) to identify existing descriptions and definitions of both normal and abnormal EEG features of preterm infants. This was followed by development and testing phases of a new standardised EEG assessment scheme. Two neonatal EEG experts, not involved in the development phase of the study then evaluated the scheme using random 2-hour EEG epochs from 24 infants <32 weeks GA. Where disagreements were found between both experts, the features where further checked and modified. Finally, both experts used the scheme to independently evaluate 2-hour EEG epochs from 12 additional infants <37 weeks GA. Percentage of agreement between observers were calculated. In the penultimate study, using infants from cohort 2, the concordance between continuous video-EEG recordings in preterm twin pairs was examined. EEGs commenced almost synchronously in twin pairs and continued until the infants were approximately 72 hours of age, while additional, shorter recordings at 32 weeks and post discharge were also recorded in unison. Visual EEG interpretation was assessed using standardised criteria from the previous chapter. Correlations were estimated within twin pairs and compared to age-matched singletons. Additionally, quantitative, mathematical EEG features were extracted and generated to represent EEG power, discontinuity, and symmetry. While controlling for GA, intra-class correlations (ICC) estimated similarities within twins. For the final study, EEGs from cohort 2 at 3 time-points over the neonatal course was used. EEGs were reviewed and scored by two electroencephalographers (EP & RL) based on the newly developed standardised EEG assessment scheme, which considered normal and abnormal activity. Bayley-III assessed neurodevelopmental outcome at 2 years corrected age. Results: Data from forty-three infants, from cohort 1, were used to develop a multimodal model for the prediction of 2-year neurodevelopmental outcome. Twenty-seven infants had good outcomes and 16 had poor outcomes or died. While performance of the model was similar to a clinical course score graded at discharge, with an area under the receiver operator characteristic (AUC) of 0.83 (95% confidence interval, CI: 0.69 – 0.95) for the physiological model vs 0.79 (0.66 – 0.90) (p=0.633) for the clinical course score, the model was able to predict 2-year outcome days after birth. Although the differences failed to reach statistical significance, the model did have a larger AUC compared to the individual physiological features, highlighting the potential value of multimodal monitoring during the transitional period. After visually analysing 6,932 hours of EEGs from 120 preterm infants from both cohorts, we identified that 6 infants (5%, 95% CI: 1.9% to 10.6%) had electrographic seizures in the first 3 days. Median (interquartile range, IQR) total seizure burden, mean seizure duration, and maximum seizure burden were 40.3 (5.0, 117.5) minutes, 49.6 (43.4, 76.6) seconds and 10.8 (1.6, 20.2) minutes/hour respectively. Seizure burden was highest in two infants with significant abnormalities on neuroimaging. In the final analysis of the EEG assessment scheme, good percentage agreements were obtained from all patients and EEG feature categories. Median agreements of between 80% and 100% were identified from the 4 categories. No difference was found in agreement rates between the normal and abnormal features (p = 0.959), neither between the younger preterm groups (<30 weeks GA) and the older preterm group (>30 weeks GA), (p = 0.249). The twins study saw the recruitment of 10 twin pairs, four monochorionic diamniotic (MCDA) and six dichorionic diamniotic (DCDA) pairs, and 10 age-matched singleton pairs. For the MCDA twins, 17/22 mathematical EEG features had significant (>0.6; p<0.05) ICCs at one or more time-points, compared to 2/22 features for DCDA twins and 0/22 features for singleton pairs. For the MCDA twins, all 10 features of discontinuity and all four features of symmetry were significant at one or more time point. Three features of the MCDA twins (spectral power at 3 – 8 Hz, skewness at 3 – 15 Hz, and kurtosis at 3 – 15 Hz) had significant ICCs over the course of all three time-points. No features for the DCDA group or control singleton pairs had significant ICCs over all three time-points. For the final study, 57 infants were included to establish whether serial multichannel video- EEG has a role in predicting 2-year outcome. From the 57 infants included, 40 had good outcome and 16 had poor outcome or died. All three serial EEGs were individually predictive of abnormal outcome, with AUCs of 0.68 (95% CI: 0.55 – 0.80); 0.84 (0.73 – 0.94); and 0.91 (0.83 – 1), (p<0.001). Comparatively, the predictive value (AUC) for a poor clinical course was 0.68 (0.54 – 0.80), while the presence of Intraventricular Haemorrhage (IVH) grade III/IV or cystic Periventricular Leucomalacia (cPVL) was 0.58 (0.41 – 0.75), (p=0.342). Conclusions: This research has utilised continuous conventional video-EEG of very preterm infants during the early postnatal period to improve understanding of early brain function and its relationship with future neurodevelopmental outcome. I have shown that quantitative analysis of multimodel preterm physiological signals, has the potential to predict mortality or delayed neurodevelopment at 2 years of age. Further studies with increased numbers are required to confirm the observed results. I identified that electrographic seizures are infrequent within the first few days of birth in very preterm infants and that we report a smaller seizure frequency than previous studies in similar cohorts. Seizures in this population are difficult to detect accurately without continuous multichannel EEG monitoring. This is the first study to use continuous, long duration, video-EEG monitoring to qualitatively and quantitatively describe electrographic seizures in preterm infants <32 weeks during the early postnatal period. In addition, for the first time, I have developed and described a standard EEG assessment scheme specifically for very preterm infants. When implemented, this showed good interobserver agreement. This can provide important information to NICU staff about normal or abnormal brain activity, maturation and neuromonitoring during critical care. I report the first study to investigate the EEG of very preterm twins during the early postnatal period. Preterm twin EEG similarities are subtle and difficult to identify visually, however this is clearly evident through quantitative analysis. MCDA twins showed stronger EEG concordance across all time-points, thus confirming a strong genetic influence on preterm EEG activity at this early stage of development. Finally, in a prospective study investigating EEG for the prediction of neurodevelopmental outcome, I have shown using serial multichannel EEG recordings that the pre-discharge EEG was the best predictor of 2-year outcome. This thesis has progressed the state of the art in preterm EEG, paving the way for further conventional EEG studies that use a standardised EEG assessment scheme. This study has also shown that seizure in preterm infants are infrequent in the early postnatal period and that there is high EEG concordance between some twin pairs. An EEG pre-discharge may be the best predictor of 2-year neurodevelopmental outcome. The assessment scheme was developed and its ability to predict 2-year outcome should now be validated in a large scale multicentre study.
EEG , Preterm infants
Lloyd, R. O. 2020. Electroencephalography of premature infants. PhD Thesis, University College Cork.