Comparison Between Simultaneously Recorded Amplitude Integrated Electroencephalogram (Cerebral Function Monitor) and Standard Electroencephalogram in Neonates
Objective: To assess the value and the limitations of amplitude integrated electroencephalogram (EEG) using the cerebral function monitor (CFM) in comparison with standard EEG in neonates who have hypoxic ischemic encephalopathy or were suspected of having convulsions.
Methods. In 36 neonates with a gestational age ≥36 weeks, CFM and simultaneously recorded EEG traces were analyzed off-line and independently classified. CFM background activity: continuous normal voltage; continuous normal voltage, slightly discontinuous (DNV); burst-suppression (BS); continuous extremely low voltage; flat tracing. CFM epileptiform activity: suspected epileptic activity, single seizure (SS), repetitive seizures (RS), status epilepticus (SE). EEG background activity: normal, depressed, low voltage undifferentiated, excessive discontinuity, BS, no activity. Epileptiform activity: interictal unifocal, interictal multifocal, ictal unifocal, ictal multifocal, SE.
Results. A total of 33 traces were suitable for analysis. Interobserver agreement on background activity was reached in 31 cases (κ = 0.92) for CFM and in 27 cases (κ = 0.74) for EEG. There was full agreement on CFM ictal activity (RS, SS, or SE) and EEG ictal activity. A normal CFM (continuous normal voltage) corresponded with a normal or a depressed EEG in 90% of the cases. The positive predictive value for a severely abnormal CFM (BS, continuous extremely low voltage, flat tracing) to correspond with a severely abnormal EEG (excessive discontinuity, BS, low voltage undifferentiated, no activity) was 100% (negative predictive value, 80%; sensitivity, 76%; specificity, 100%). DNV (10) on CFM corresponded either with depressed (6) or excessive discontinuity (4) on EEG. Ictal activity on EEG corresponded with SS, RS, or SE on CFM in 8 cases (sensitivity, 80%; specificity, 100%; positive predictive value, 100%; negative predictive value, 92%).
Conclusion. CFM is a reliable tool for monitoring both background patterns (especially normal and severely abnormal) and ictal activity. Certain focal, low amplitude, and very short periods of seizure discharges can be missed. We recommend using CFM as a monitoring device and performing intermittent standard EEG whenever there is any doubt about the classification of the CFM (ie, DNV pattern or suspected epileptiform activity).
Single channel amplitude integrated electroencephalogram (EEG) (cerebral function monitor [CFM]) is increasingly being used in neonatal intensive care units. The CFM signal gives a continuous, on-line trend recording of cerebral electric activity at the cot side. Although it cannot provide information on cerebral activity outside the pick-up area, background activity and generalized seizure activity can be identified easily.1–4
Several CFM studies have recently been performed in full-term infants with hypoxic-ischemic encephalopathy (HIE). They have shown that prediction of neurodevelopmental outcome is possible as early as 3 to 6 hours after birth. The technique has therefore been chosen for selection of infants for the present multicenter hypothermia trial.5–10
It has been claimed that CFM, on the basis of 1-channel EEG, has a very high concordance with multichannel standard EEG.2,3,8,9,11 For a standard neonatal EEG, 9 electrodes are used to produce 14 bipolar derivations (channels) of EEG together with channels for eye movement, muscle activity, and electrocardiogram. The standard duration of recording is approximately 30 minutes.12,13
For comparing these 2 methods in detail, CFM and EEG traces, recorded simultaneously, were evaluated in a standardized way by 2 neonatologists for the CFM and 2 clinical neurophysiologists for the EEG. The recordings obtained using these 2 methods were subsequently compared.
The aim of this study was to assess the value and the limitations of the CFM, with regard to both the background pattern and the recognition of epileptiform discharges in neonates who have HIE or were suspected of having convulsions.
Between May 1998 and September 2000, 36 consecutive newborn infants were enrolled in the CFM-EEG comparison study. They were admitted to our regional tertiary neonatal intensive care unit because of suspected convulsions and/or moderate to severe HIE. They all had a gestational age of 36 weeks or above. As soon as possible after the infants were admitted, the CFM recording was started by the attending neonatologist as part of routine care in patients who were deemed to be at risk of developing convulsions. When epileptiform activity was clearly recognized on the CFM recording, before or after the EEG was done, antiepileptic drugs (phenobarbitone, midazolam, lidocaine, or clonazepam) were administered.
Serial EEGs were performed in several infants, but only the first EEG of a patient was used for this study. CFM and EEG were registered simultaneously, and the beginning and the end of the EEG recording were carefully marked on the CFM trace. Both CFM and EEG traces of the simultaneous recordings were analyzed off-line and independently classified by 2 investigators, who were unaware of the clinical condition of the patients and their neurodevelopmental outcome. Informed consent was obtained from all parents.
Amplitude Integrated EEG
For CFM recording the CFM 4640 (Lectromed, Devices Ltd, UK) was used. The CFM records a single-channel EEG from biparietal needle electrodes (P3, P4, reference Fz). A second tracing continuously records the electrode impedance. The filtered signal is rectified, smoothed, and amplitude-integrated before it is written out at slow speed (6 cm/h) at the cot side.1,2,5
• Continuous normal voltage (CNV): continuous activity with voltage 10 to 50 μV
• Continuous normal voltage, slightly discontinuous (DNV): mainly continuous normal voltage with periods of more discontinuous intermittent low voltage (no burst-suppression) activity
• Burst-suppression (BS): discontinuous background pattern; periods of very low voltage (inactivity) intermixed with bursts of higher amplitude
• Continuous extremely low voltage (CLV): continuous background pattern of very low voltage (around or below 5 μV)
• Flat tracing (FT): very low voltage, mainly inactive (isoelectric) tracing with activity below 5 μV
• Sleep-wake cycling (SW): continuous activity with regularly cycling patterns; intervals of continuous activity (awake, active sleep) intermixed with periods of slightly lower voltage. (In a term infant, the broadest bandwidth: minimum, 6–8 μV, maximum, 15–20 μV; narrowest bandwidth: minimum, 6–8 μV; maximum, 9–15 μV.) Because of the short duration of the simultaneous recording, it was not possible to diagnose the presence or absence of the SW pattern. The length of the cycle takes at least 1 hour, with periods of deep sleep lasting 20 to 30 minutes
Epileptiform activity (characteristic pattern looking, with increased amplitude during epileptic seizure activity and lower voltage in interictal period) was classified as follows:
• A single seizure (SS)
• Repetitive seizures (RS): ≥3 discharges during 30-minute period
• Status epilepticus (SE): “sawtooth pattern”
• Suspected epileptiform activity (e): irregular lower edge of the CFM trace with downward bursts
A 19-channel EEG with a duration of 30 minutes was recorded simultaneously with the CFM. Nine EEG electrodes were placed on positions of the International 10–20 system (FP1,2, T3,4, O1,2, C3,4, Cz) to produce 14 bipolar derivations (channels). In addition, 2 electrodes were placed just outside the outer canthus to record horizontal eye movements.
One electrode was placed on the right side of the chin and 1 was placed on the left cheek for the recording of sucking movements. A movement sensor was placed over the area of visible respiration, usually the lower part of the thorax. A second movement sensor was available to record jerking of arms or legs. One channel was used to record electrocardiogram. Conventional 10-mm Ag-AgCl electrodes were fixed to the scalp with tape or with collodium 3%. The effective resistance between skin and electrodes was reduced to <2 kOhm by means of scrubbing with Nupress before fixing the electrodes.
The 19-channel EEGs were recorded with a digital Brainstar system. The EEG signal was sampled digitally at a frequency of 512 Hz. For assessment, a bandwidth of 1 to 70 Hz was used. All EEGs were assessed using a long-distance bipolar montage.
• Depressed: depression of voltage (≥25 μV) with preservation of normal patterns
• Excessive discontinuity: low voltage with absence of normal patterns (interburst interval <6 seconds, interburst low voltage 5–25 μV)
• BS: bursts 1 to 10 seconds, suppression amplitude <5 μV
• Low voltage undifferentiated: 5 to 25 μV during all states
• No activity: <5 μV during all states
• Interictal unifocal
• Interictal multifocal
• Ictal unifocal (activity was considered as ictal when its duration was at least 6 seconds)
• Ictal multifocal
Both EEG and CFM have an event marker. Especially when doing long-term continuous CFM recording, it is very important that the nurses write down, using the event marker, what they are doing to the patient, as some manipulations may change the CFM and could be mistaken for an electrical discharge.
Both CFM and EEG recordings were independently assessed by 2 observers, with expertise for their own technique: 2 neonatologists for the CFM and 2 clinical neurophysiologists for the EEG. The CFM interpreters were not aware of the CFM recording before or after the EEG recording and vice versa. The interobserver agreement for either CFM or EEG was measured using κ statistics. The 2 EEG observers and the 2 CFM observers reached consensus on the interpretations on which they disagreed previously. The consensus recordings for EEG and CFM were compared.
The sensitivity, specificity, positive predictive values (PPV), and negative predictive values (NPV) of different items were calculated. After the comparison of EEG and CFM was done, we had access to the CFM before and after the EEG recording.
Assessment of Neurodevelopmental Outcome
The survivors were routinely seen in the follow-up clinic at 3, 9, 18, and 24 months. Assessment of outcome was made using the Griffiths mental developmental scale and items from Amiel-Tison and Grenier and Touwen.15–17 The Alberta Infant Motor Scale was also used in children who were younger than 18 months.18 Cerebral palsy was classified according to the criteria of Hagberg et al.19
The initial diagnoses at admission on the 36 patients who were enrolled in the study were as follows: 26 HIE, 2 parenchymal hemorrhage and convulsions, 1 posterior cerebral artery infarction, 1 severe hypoglycemia and convulsions, and 6 seizures of unknown cause.
Three EEG traces could not be analyzed properly as a result of (movement) artifacts and were excluded from the study. The other 33 traces were analyzed separately for CFM and EEG. Antiepileptic drugs were administered in 29 of the 33 patients before the standard EEG was performed. All of these patients received phenobarbitone because of a clinical suspicion of seizures at the referring hospital and thus before CFM monitoring was started. When epileptic activity (SS, RS, or SE) clearly persisted clinically and/or on the CFM trace, lidocaine and/or midazolam and sometimes clonazepam were added, according to our treatment protocol. CFM and EEG classifications together with medication and outcome are listed in Table 1.
Interobserver agreement for CFM and EEG is listed in Table 2.
The CFM traces were independently analyzed by 2 investigators according to the classification mentioned above. Interobserver agreement on background activity (CNV, DNV, BS, CLV, FT) was reached in 31 cases (κ = 0,92). In the other 2 cases, there was disagreement on whether the background showed CNV or DNV.
Interobserver agreement on epileptiform activity was reached in all but 6 cases (κ = 0,70). In all of these 6 cases, the question was whether there was suspected epileptiform activity (e). There was full agreement on obvious epileptiform activity, RS, SS, or SE.
Ten traces were classified as a CNV pattern, with RS or SS activity in 2 cases. Another 10 traces were classified as a DNV pattern (partly CNV in 2), with RS or e in 6 traces. A BS pattern was seen in 8 cases, with epileptiform activity (SE, RS, SS) in 3 traces. A CLV pattern was seen in 3 cases, with RS in 2 traces. A FT was seen in 2 traces.
Interobserver agreement about background activity was reached in 27 cases (κ = 0.74). In 2 cases, there was disagreement between “depressed” and “normal.” In another 3 cases, there was disagreement between “excessively discontinuous” and “depressed.” In another case, there was disagreement between “excessively discontinuous” and “suppression burst.” Complete interobserver agreement about epileptiform activity existed in all cases (κ = 1).
Two EEG traces were classified as normal, 13 traces were classified as depressed, 11 traces were classified as excessive discontinuity, 2 traces were classified as suppression burst, 3 traces were classified as low voltage undifferentiated, and 2 traces were classified as no activity. In 7 traces, no epileptiform activity was seen. Ten traces were classified as showing ictal activity, either focal or multifocal, and 16 traces were classified as showing interictal multifocal activity.
Comparison Between EEG and CFM
Comparison between EEG and CFM background activity, together with the comparison of epileptiform activity, is listed in Table 2. A normal CFM (CNV) background pattern corresponded with a normal or a depressed EEG background pattern in 9 of 10. Seven of the 10 infants with a normal CFM pattern (CNV) showed a depressed background pattern on EEG. The infants with a DNV pattern on CFM (10) showed a depressed (6) or an excessively discontinuous (4) pattern on EEG. The infants with a BS pattern on CFM (8) showed an excessively discontinuous pattern (6) or a BS (2) pattern on EEG. Overall, 28 of 33 classifications showed a good correspondence between CFM and EEG.
The child with CNV on the CFM and excessive discontinuity on the EEG and the 4 children with DNV on the CFM and excessive discontinuity on the EEG were considered a poor correspondence, with a more severe background pattern on the EEG than expected on the basis of the CFM. The most pathologic CFM patterns (FT, CLV) corresponded with the EEG background patterns (no activity, and low voltage undifferentiated) in all cases. The PPV for a severely abnormal CFM pattern (BS, CLV, FT) to correspond with a severely abnormal EEG (excessive discontinuity, BS, low voltage undifferentiated, no activity) was 100% (NPV, 75%; sensitivity, 72%; specificity, 100%).
Ten infants showed ictal activity on the standard EEG recording. This corresponded with ictal activity (SS, RS, or SE) on the CFM recording in 8 cases (sensitivity, 80%; specificity, 100%; PPV, 100%; NPV, 92%). Sixteen infants showed interictal multifocal epileptiform activity on the standard EEG recording. This corresponded with e on the CFM recording in only 4 cases (sensitivity, 25%; specificity, 94%; PPV, 80%; NPV, 57%). Seven infants showed no epileptiform activity on the standard EEG recording. This corresponded with no epileptiform activity on the CFM recording in all cases.
In 2 cases, without clear ictal activity on CFM (RS, SS, or SE), ictal discharges were seen on the standard EEG. In 1 case, epileptiform activity was suspected on CFM. This patient had short periods (6–10 seconds) of multifocal seizure discharges with low amplitude and was on antiepileptic medication. In the other case with a posterior cerebral artery infarction, there were focal (occipital) low-amplitude epileptic discharges. Phenobarbitone was administered to this patient because of a clearly recognizable RS pattern on the CFM recording 2 hours before the standard EEG was done.
The number of discharges identified on the CFM (RS) was lower on the CFM traces than on the standard EEG. When the ictal activity on EEG remained restricted to the occipital area, this was not recognized on the CFM. When the ictal activity on EEG remained restricted to the frontopolar area, this was not always picked up by the CFM. Very short episodes of ictal activity (<10 seconds) or very low-amplitude seizure discharges were missed as well. When ictal activity was missed on CFM during the period of simultaneous recording, it was picked up somewhere on the CFM trace in all patients before or after the standard EEG was done (Figs 1 and 2).
Although EEG is widely used, CFM is more easily applicable and available, especially during nighttime, and is especially suitable for continuous monitoring.5 In this study, interobserver agreement for both CFM and EEG was obtained in the majority of the traces. Classification of EEG background patterns was composed of 6 categories, and classification of CFM was composed of 5 categories. It would have been easier to have the same number of categories for both techniques. As the EEG is a much more sophisticated technique, more subcategories have previously been described by Stockard-Pope et al,12 Holmes and Lombroso,13 and Ortibus et al.14 As these subcategories are usually used in the literature on neonatal EEG, we did not want to change this and therefore also have used these 6 categories. The same applies for the CFM; 5 categories have been used in previous papers (Helström-Westas and colleagues3,5,6).
Agreement on background activity seemed to be more readily obtained for the CFM (κ = 0.92) than for the EEG (κ = 0.74). In contrast with previous reports,20 we did not have any disagreement in classifying a DNV pattern versus a BS pattern on CFM. The only disagreement we had was on DNV versus CNV in 2 patients. Agreement on epileptiform activity seemed more difficult for the CFM (κ = 0.70), especially for suspected epileptiform activity (e), than for the EEG (κ = 1.0)
There is still a lot of discussion about the validity of CFM in comparison with the EEG.20 We found that the PPV of severely abnormal CFM background pattern (FT, CLV, BS) was 100% for severely abnormal EEG (excessive discontinuity, burst suppression, low voltage undifferentiated, no activity). A CNV pattern on CFM corresponded well with a normal or depressed background pattern on standard EEG. Although only 2 standard EEGs were completely normal, most patients with a depressed pattern on standard EEG were normal or showed mild global delay on early follow-up. One infant (patient 10) with a CNV pattern and a subsequent normal outcome showed an excessively discontinuous pattern on standard EEG. Looking at this CFM recording during the first 10 hours of life, it already showed a SW pattern within the first 3 hours. During the standard EEG (5 hours after birth), it seemed that this infant was partly recorded during quiet sleep. This could explain the excessively discontinuous pattern on the standard EEG. One infant with CNV shows severe cognitive impairment and has epilepsy of unknown cause.
A DNV (10) pattern on CFM seemed to be the most difficult pattern to classify. This corresponded either with a depressed (6) or an excessively discontinuous (4) pattern on standard EEG. The 4 children who had excessive discontinuity, which was worse than expected on the basis of the CFM, had a normal early neurodevelopmental outcome. In this group of patients, suspicion of epileptiform activity (e) was also difficult to assess. In these children, it is especially important to obtain a standard EEG during office hours to get more reliable and detailed information.
Most patients had already received antiepileptic drugs before the first standard EEG was performed and thus could have influenced background activity. In the standard EEG, more subtle changes as a result of these drugs may have been picked up, with more undifferentiated changes on CFM.
Correlation of ictal activity on EEG and ictal activity (SS, RS, SE) on CFM was 80%. In 2 traces, the ictal activity was missed, and both patients had already received antiepileptic drugs for ictal activity, seen on the CFM, before the EEG was recorded.
One would expect that very short, low-voltage, and certain focal ictal activity would be missed on CFM. This was clearly the case: occipital focal, low-amplitude, or very short epileptic discharges were missed on the CFM. However, there were no false-positive cases of ictal activity on CFM. When ictal activity was missed on CFM during the period of simultaneous recording, it was recognized somewhere on the CFM trace before or after the standard EEG was done in all patients. This seemed to be attributable to more generalized and longer ictal activity at other moments in the same patient. Thus, monitoring for long periods of time with CFM, in contrast with usually short (30 minutes) periods of standard EEG recording, compensates for not detecting all ictal discharges.
On the CFM, suspicion of epileptiform activity without any overt epileptiform discharge visible was the least reliable item. Interictal multifocal epileptiform activity (16 cases) on EEG corresponded with suspected epileptiform activity (e) in only 4 cases. Fourteen of these patients had received antiepileptic medication because of clinical convulsions before the CFM recording was started (4 cases) or obvious ictal activity (SS, RS, SE) on CFM (10 cases) before the EEG was recorded.
We have previously published data concerning predictive value of the CFM with regard to neurodevelopmental outcome. In these studies,7,8 only children with HIE were studied and all children had CFM recorded at either 3 or 6 hours. As the infants in this study were more heterogeneous with regard to the underlying pathology and the recordings were performed between days 1 and 10, predictive value for neurodevelopmental outcome was not calculated but outcome data are provided in Table 3. The main aim of the study was the comparison of 2 techniques conducted simultaneously.
CFM seemed to be a reliable tool for monitoring both background pattern (especially normal and extremely abnormal) and ictal activity. Because of long periods of registration, the CFM is especially useful to evaluate changes in background pattern over time and detect the occurrence of seizures. One should realize that certain focal, low-amplitude, and very short periods of seizure discharges visible with EEG can be missed with CFM. We recommend using CFM as a monitoring device and performing intermittent standard EEG whenever there is any doubt about the classification of the CFM (DNV pattern or suspected epileptiform activity).
- ↵Prior PF, Maynard DE. Monitoring Cerebral Function. Longterm Recordings of Cerebral Electric Activity and Evoked Potentials. Amsterdam, the Netherlands: Elsevier;1986
- ↵Bjerre I, Hellström-Westas L, Rosen I, Svenningsen NW. Monitoring of cerebral function after asphyxia in infancy. Arch Dis Child.1983;58 :997– 1002
- ↵Azzopardi D, Roberton NJ, Cowan FM, Rutherford MA, Rampling M, Edwards AD. Pilot study of treatment with whole body hypothermia for neonatal encephalopathy. Pediatrics.2000;106 :684– 694
- ↵Al Naqeeb N, Edwards AD, Cowan FM, Azzopardi D. Assessment of neonatal encephalopathy by amplitude-integrated electroencephalography. Pediatrics.1999;103 :1263– 1271
- ↵Murdoch-Eaton D, Toet MC, Livingston J, Smith I, Levene M. Evaluation of the cerebrotrac 2500 for monitoring of cerebral function in neonatal intensive care. Neuropaediatrics.1994;5 :122– 128
- ↵Stockard-Pope JE, Werner SS, Brickford RG. Atlas of Neonatal Electroencephalography. New York, NY: Raven Press;1992
- ↵Griffiths R. The Abilities of Babies: A Study in Mental Measurement. Amersham, United Kingdom: Association of Research in Infant and Child Development;1976
- Amiel-Tison C, Grenier A. Evaluation Neurologique du Nouveau-né et du Nourisson. Paris, France: Masson;1980
- ↵Touwen BLC. Examination of the Child With Minor Neurological Dysfunction. Clinics in Developmental Medicine No 71. London, United Kingdom: SIMP/Heinemann;1979
- ↵Piper MC, Darrah J. Motor Assessment of the Developing Infant. Philadelphia, PA: WB Saunders Company;1994
- Copyright © 2002 by the American Academy of Pediatrics