Documenta Ophthalmologica98: 247–256, 2000.© 2000Kluwer Academic Publishers. Printed in the Netherlands.

Effect of sleep state on the flash visual evoked potentialA case study

ASHLEY SHEPHERD1, KATHRYN SAUNDERS2 and DAPHNEMcCULLOCH3

1Department of Nursing and Midwifery, University of Stirling, Stirling, FK9 4LA, UK;2School of Biomedical Sciences, University of Ulster, Coleraine, Northern Ireland BT52 1SA,UK; 3Department of Vision Sciences, Glasgow Caledonian University, Cowcaddens Road,Glasgow G4 OBA, UK/I3

Accepted 1 November 1999

Abstract. Controversy exists regarding the influence of sleep state on the flash visual evokedpotential. This study recorded the visual evoked potential in a new-born infant in four differentsleep states; wakefulness, drowsiness, active sleep and quiet sleep over a five hour period. Theinfant’s heart rate, breathing rate and breathing regularity were also recorded. It was clear thatwhen this subject was awake the VEPs recorded differed substantially from those recordedwhen sleeping. Two of the four main components had shorter peak latencies, one componentwas prolonged and one of the peak to trough amplitudes was consistently smaller when alert.This study highlights an important and often overlooked aspect of developmental research thatthe state of the infant may affect developmental measures.

Key words: flash visual evoked potential; neonate, sleep, state

Introduction

Assessment of the visual pathways early in life can be made using visualevoked potentials (VEPs) to flash stimulation. Abnormal flash VEPs are as-sociated with a poor neurological prognosis in both preterm [1] and fullterminfants [2]. Despite a number of studies claiming that sleep state has no in-fluence on flash VEP recordings in young infants [3–6], other authors havereported significant effects. Whyte et al. [7] concluded that the peak latencyof the large negative trough (N3) was unaffected by sleep state but therewas a significant reduction in amplitude of this N3 wave during quiet sleep.These authors also found that the positive component (P2) was particularlysensitive to sleep state and was present in awake and drowsy infants but notin infants who were asleep. Apkarian et al. [8] concluded that sleep state hada significant effect on the morphology, amplitude and peak latencies of theVEP waveform in term infants. As wakefulness increased a decrease in P2and N3 peak latencies and an increase in the P2:N3 amplitude were observed.

248

Apkarian et al. [8] suggested that the reliability of the VEP could be greatlyincreased if VEP measurements were made while the subject is awake. Thisrecommendation may be difficult to adhere to in clinical situations whentesting new-born infants who are asleep for prolonged periods.

There is considerable inter-individual variability among flash VEPs ofhealthy infants. Repeated recording in one infant allows separation of thesystematic influence of sleep state from this inter-individual variability. Thiscase report describes the influence of sleep state on the flash VEP of onefullterm infant over a 5 h period.

Materials and methods

A fullterm female infant born at 39.3 weeks gestation with a birth weight of3120 g was studied 3 days after birth (i.e. post-menstrual age of 39.8 weeks).

Sleep state was assessed during each VEP recording by two researchersusing the behavioural variables of Anders et al. in Table 1 [9]. Sleep statewas categorised into four different states; quiet sleep, active sleep, drowsi-ness or wakefulness by observation of body, facial and eye movements. Onlywhen agreement of the sleep state was reached between both researchers didVEP recording begin. Heart and respiration rates were also measured usinga Synectics microsleep channel recorder which was attached to three self-adhesive ECG electrodes (medicotest blue sensor P-100-S electrodes) andone respiratory sensor (RespSponse pressure sensor) which was positionedacross the infant’s chest. The recorder displayed the time on a small screenbut the measured variables were not displayed until the unit was attached toa computer at the end of the complete series of tests. The software used wasthe Synectics Medical-Multigram, Sleep Apnoea Edition version 5.20W16.

Flash visual evoked potentials (VEPs)

The clinical use and recording procedure for flash VEPs in new-borns hasbeen previously described [1]. Briefly, a hand held strobe light was presented20 cm from the infants’ eyes at a rate of one flash per second with an intens-ity of 160 cd.s/m2 (Grass intensity 16). The strobe, 13 cm in diameter, wascovered in a pebbled diffuser. VEPs were recorded from an occipital electrode(Oz) with a frontal reference electrode (Fz) [10]. The time was noted fromthe Synectics microsleep channel recorder at the start and end of each flashVEP recording. Thirty averages were recorded for each flash VEP trial, butrecording was stopped if the infant changed sleep state. The full recordingsession took five hours during which 40 flash VEP trials were recorded (13recorded when awake, 6 when drowsy, 10 during active sleep and 11 during

249

Table 1. Behavioural variables used to differentiate sleep states [9]

State Eyes Body movement Facial movement Other

Wakefulness Open, blinks, pursuit, Slow, writhing, rapid, jerky, Frown, smile, grimaces or Vocalisations, penile

searching spontaneous startle relaxed, suck erections

Drowsiness Open or closed, blinks, Slow, writhing, Frown, smile, grimaces or Vocalisations, penile

intermediate luster spontaneous startle relaxed, suck erections

Active Sleep Closed Slow, writhing, Frown, smile, grimaces, Vocalisations, penile

spontaneous startle suck erections

Quiet Sleep Closed Spontaneous startle Relaxed, jaw jerks

250

Table 2. Psychological variables recorded during behaviourally defined sleepstates

Sleep state Heart rate (SD) Breathing rate (SD) Breathing reg:irreg

Wakefulness 131.2 (7.5) 38 (7.1) 1:12

Drowsiness 128.6 (6.5) 38.5 (8.5) 2:4

Active Sleep 118.3 (9.0) 41.9 (7.4) 3:7

Quiet Sleep 111.4 (3.8) 44.1 (5.7) 11:0

quiet sleep). Testing was interrupted for cleaning and feeding of the infantand flash VEP recordings were made for a total of 20 min during the 5 h.

Physiological measures

After the full recording session the microsleep channel recorder and all elec-trodes were removed from the infant. The recorder was connected to a per-sonal computer on which the heart and breathing rates could be calculated foreach of the 40 flash VEP recording trial times. These rates were measured forthe full period of the flash VEP trial (i.e. between two and three minutes) anda mean rate was calculated. Breathing regularity was classified as regular orirregular based on the uniformity of the breathing rate voltage signals duringeach flash VEP trial.

Results

Physiological and behavioural sleep state measurements

The relationship between sleep state (as defined by the observers using be-havioural variables described in Table 1) and the physiological variables wascompared using analysis of variance (ANOVA). As wakefulness increased sotoo did the heart rate (p=0.0001). The breathing rate showed a tendency todecrease as wakefulness increased. This proved significant only when com-paring the breathing rate in the awake state with that in the quiet sleep state(Fisher PLSD,p <0.05). Breathing regularity increased significantly witha decrease in wakefulness (Chi squared,p=0.0001). Table 2 describes themean values for heart and breathing rates and the ratios of regular to irregular

251

Figure 1. Examples of a flash VEP recorded in each sleep state in the same infant: (a) awake,(b) drowsy, (c) active sleep, and (d) quiet sleep. (The negative spike at the flash onset is anartefact of the flash stimulus.)

breathing for each sleep state (as defined by behavioural variables in Table1). The sleep state, as defined by the behavioural variables, appeared to re-late accurately to published heart and breathing rate values, suggesting thatour observers definition of sleep state using only behavioural variables wasaccurate [11].

Effect of sleep state on the flash VEP

WaveformThe flash VEP in fullterm neonates consists of a series of two to five peaks[12]. The four most consistent peaks, three positive (P1, P2 and P3) and onenegative (N3) were measured in terms of their peak latencies (measured fromstimulus onset to peak) and amplitude (measured from peak to trough). In thisinfant, clearly recordable VEPs were present for all trials in all sleep states.Typical VEPs in each sleep state are illustrated in Figure 1. A large positivity

252

Table 3. Prevalence and peak latency of Flash VEP components for each sleepstate

Sleep state Peak latencies (ms)

P1 (SD) P2∗∗ (SD) P3∗∗ (SD) N3∗ (SD)

Wakefulness 100.7 (9.9) 185.1 (11.7) 389.2 (42.7) 280.5 (14.1)

(n=13) [6/13] [10/13] [12/13]

Drowsiness 109.3 (16.7) 198.8 (9.1) 361.2 (21.6) 278.3 (15.8)

(n=6) [3/6] [5/6] [6/6]

Active Sleep 100.0 (27.9) 205.2 (7.5) 348.6 (15.9) 296.0 (38.7)

(n=10) [5/10] [7/10] [9/10]

Quiet Sleep 78.0 214.6 (20.1) 308.3 (55.4) 333.0 (71.0)

(n=11) [2/11] [6/11] [10/11]

∗Significant difference between sleep states (p <0.05).∗∗Significant difference between sleep states (p <0.005), present in 100% oftrials.

(P2) at approximately 200 ms was clearly identifiable on all trials. This wasusually followed by a negativity (N3) at about 300 ms. In less than half ofthe trials an early positive component (P1) was present at approximately 100ms. P3 was defined as a positive component following P2. This was presentin most trials but more variable (peak latency 234 to 416 ms). The prevalenceof each component peak is given in Table 3.

For each VEP trial the presence or absence of each component peak andthe total number of peaks was noted. There were no significant differences inthe prevalence of any individual peak with sleep state (Chi square,p >0.05).However, the complexity of the flash VEP as measured by the total num-ber of component peaks increased with alertness (Kendall rank correllation,Tau=−0.27,p=0.01).

Peak latenciesTable 3 also details the peak latencies of the major components with respectto sleep state. The latencies of the flash VEP tended to be earlier when thesubject was awake and later when asleep. Specifically, both the P2 and N3components showed significantly shorter peak latencies with an increase inwakefulness (ANOVA,p <0.005). Post hoc testing demonstrated that the P2peak latency in the alert state was shorter than in the light and quiet sleepstates (Fisher PLSD,p <0.01). The N3 latency proved to be longer in quiet

253

Table 4. Amplitudes of Flash VEP components with respect to sleep state

Sleep state Amplitudes (µV)

N0:P1 (SD) P1:N1 (SD) N1:P2∗ (SD) P2:N3 (SD)

Wakefulness 2.7 (1.6) 5.5 (2.0) 5.9 (2.8) 19.7 (9.1)

(n=13)

Drowsiness 5.8 8.9 (7.7) 12.75 (12.8) 22.4 (9.4)

(n=6)

Active Sleep 5.0 (2.9) 6.3 (2.4) 10.4 (5.2) 14.5 (5.7)

(n=10)

Quiet Sleep 3.2 9.8 15.9 (7.5) 18.6 (16.9)

(n=11)

∗Significant difference between sleep states (p=0.01).

sleep than in alert or drowsy states (Fisher PLSD,p=0.02). The P3 componentshowed a trend in the opposite direction with the shortest latency for quietsleep (ANOVA,p <0.01). Thus, when awake, the flash VEP of this infantcharacteristically had an earlier P2 (mean 185 ms) followed by N3 around 100ms later and a P3 at 389 ms. As wakefulness decreased, P2 and N3 becamelater and the P3 became earlier so that the flash VEP tended to consist of adouble component P2:P3 followed by N3 (see examples in Figure 1). The P1component, which was present in 40% of the flash VEPs recorded, demon-strated no significant change with respect to sleep state (ANOVA,p >0.05).

AmplitudesTable 4 details the amplitudes of the major components present with respectto sleep state. The amplitude of P2 measured from the preceding trough wasfound to differ significantly with sleep state (ANOVA,p=0.01). Its amplitudewas significantly smaller when alert compared with quiet sleep (Fisher PLSD,p <0.01). Thus, the general trend was towards an increase in N1:P2 amp-litude as wakefulness decreased. No difference in the amplitude of the othercomponents was found (ANOVA,p >0.05). The total amplitude (sum ofpeak to trough measurements) also showed no differences among sleep states.

254

Discussion

In the term new-born, there are two well-defined sleep states: active andquiet sleep. Wakefulness and drowsy states may also be recognised [13].Ideally careful observation of body movements along with measurement ofphysiological variables should be used to determine state. However, Werneret al. [13] suggest that if only one variable is assessed then the assessment ofbody movement should be chosen as it is the most reliable indicator of state.

Heart rates and breathing regularity appropriate for each state have beennoted by a number of authors [7–9, 11, 14]. Briefly, they all describe reg-ular breathing and decreased heart rate for quiet sleep with more irregularbreathing and increasing heart rate as wakefulness increases. These relation-ships between heart rate and breathing regularity with sleep state have beenconfirmed in the present study. This finding suggests that using behaviouralmeasures only, an accurate assessment of sleep state can be achieved.

It is clear from the present study that sleep state did affect flash VEPpeak latencies and amplitudes of the new born infant under study. The flashVEP recorded from the infant when awake had smaller N1:P2 amplitudesand shorter peak latencies compared to flash VEPs recorded when sleeping.No difference was noted in the presence or absence of any particular VEPcomponent with respect to sleep state but the total number of peaks presentincreased with respect to increased wakefulness. Flash VEPs recorded whenthis infant was drowsy had similar variability to those recorded in other states.However, in a cross-sectional study of term infants (each tested once) P2 wasextremely variable in the drowsy state [15]. It would therefore appear moreadvantageous to record flash VEPs in one of the other three states to avoidrecording extremely variable VEPs.

The results of this study are in agreement with the findings of Apkarian etal. [8] with respect to latency but differ in amplitude results. Apkarian et al.[8] found a significantly larger P2:N3 amplitude in more wakeful infants. Nosuch difference was noted in the present study although the earlier componentamplitude N1:P2 decreased with wakefulness. This difference may be a char-acteristic of this individual infant. There may also be differences in scoringP2-N3 amplitude, especially in alert infants who tend to have a double peakprior to N3. Alternatively, the difference found in Apkarian’s group wouldbe explained if those infants with larger P2:N3 components had a tendencyto be more wakeful. Repeated measurement on a group of infants would berequired to distinguish between these alternatives.

During the course of this study, it was evident that lengthy recordingsessions would be required to record flash VEPs in all sleep states. This islargely due to the extremely limited periods of wakefulness in fullterm new-

255

borns who are reported to sleep for an average of 14 h per day during theirfirst month of life [16]. This study of one infant has highlighted two mainissues with regard to the clinical use of flash VEPs. Firstly, a modest in-crease in recording time would allow the examiners to avoid recording duringdrowsiness but it is unlikely to facilitate VEP testing in all sleep states. Hence,the utilisation of VEP recordings made during either sleep or wakefulness isreasonable if comparative normative data are available. Secondly, it highlightsthe need for sleep state to be considered when reporting the flash VEP. Basedon this case study, it appears that behavioural observation may be an accuratemethod of determining sleep state in neonates if physiological measurementis not available.

Acknowledgements

We gratefully acknowledge the financial support of the Disability and Con-tinuing Healthcare Research Committee of the Chief Scientists Office, Scot-tish Office Health Department (Grant No. K/RED4/C265). We also thankMrs Jackie O’Neil, Respiratory function laboratory, Yorkhill Hospital forassistance with monitoring and measurement of physiological data.

References

1. Shepherd AJ, Saunders KJ, McCulloch DL, Dutton GN. The prognostic value of theflash visual evoked potential in preterm infants. Developmental Medicine and ChildNeurology 1999; 41: 9–15.

2. Whyte HE, Taylor M, Menzies R, Chin KC, MacMillan LJ. Prognostic utility of visualevoked potentials in term asphyxiated neonates. Pediatric Neurology 1986; 2: 220–223.

3. Ferriss GS, Davis GD, Dorson M, Hackett ER. Changes in the latency and form of thephotically induced averaged evoked response in human infants. ElectroencephalographyClinical Neurophysiology 1967; 22: 305–312.

4. Ellingson RJ. Variability of the visual evoked response in human newborns. Electroen-cephalography Clinical Neurophysiology 1970; 29: 10–19.

5. Barnet A, Friedman S, Weiss I, Ohlrich E, Shanks B, Lodge A. VEP developmentin infancy and early childhood. A longitudinal study. Electroencephalography ClinicalNeurophysiology 1980; 49: 476–489.

6. Baitch L, Srebro R. Binocular interactions in sleeping and awake infants. InvestigativeOphthalmology and Vision Science 1990; Suppl 31: 251.

7. Whyte HE, Pearce JM, Taylor MJ. Changes in the VEP in preterm neonates witharousal states, as assessed by EEG monitoring. Electroencephalography Clinical Neuro-physiology 1987; 68: 223–225.

8. Apkarian P, Mirmiran M, Tissen R. Effects of behavioural state on visual processing inneonates. Neuropediatrics 1991; 22: 85–91.

256

9. Anders T. In: Werner S, Stockard J, Bickford R, eds. Atlas of Neonatal Electroenceph-alography. New York: Raven Press, 1971.

10. Harding GFA, Odom JV, Spileers W, Spekreijse H. Standard for VEPs 1995. VisionResearch 1996; 36: 3567–3572.

11. Harper RM, Hoppenbrouwers T, Sterman MB, McGinty DJ, Hodgman J. Polygraphicstudies of normal infants during the first six months of life. I Heart rate and variabilityas a function of state. Pediatric Research 1976; 10: 945–951.

12. Taylor M. Evoked Potentials. In: Barber C, Taylor M, eds. Evoked potentials, ReviewNo. 4. England: IEPS Pubs, 1991.

13. Werner SS, Stockard JE, Bickford RG. Atlas of neonatology. New York: Raven Press,1997.

14. Mercuri E, Siebenthal K, Tutuncuoglu S, Guzzeta F, Caser P. The effect of behaviouralstates on visual evoked responses in preterm and full-term newborns. Neuropediatrics1995; 26: 211–213.

15. Shepherd AJ. Development of the visual evoked potential in high and low risk preterminfants. PhD Thesis, Glasgow Caledonian University, Glasgow, 1998.

16. Ardura J, Anders J, Aldana J, Revilla MA. Development of sleep–wakefulness rhythmin premature babies. Acta Pediatrica 1995; 84: 484–489.

Address for correspondence:A. Shepherd, Department of Nursing and Midwifery, Universityof Stirling, Stirling FK9 4LA, UKPhone: 01 786 466 334; Fax: 01 786 466 344; E-mail: Ashley.Shepherd@stir.ac.uk