Electroencephalography and clinical Neurophysiology, 84 (1992) 44-54 © 1992 Elsevier Scientific Publishers Ireland, Ltd. 0168-5597/92/$(15.0(I
44
EVOPOT 91031
The late somatosensory evoked potential in premature and term infants. II. Topography and latency development W a l t K a r n i s k i , L a n c e W y b l e , L o r e t t a L e a s e a n d R. C l i f f o r d Blair Department of Pediatrics, Unil'ersity of South Florida College o[" Medicine, Tampa, FL 33612 (U.S.A.) (Accepted for publication: 3 June 1991 )
Summary The maturation of latency and scalp voltage topography of the simultaneously bilateral somatosensory evoked potential was studied in 53 neurologically intact pre-term and term infants, from 31 to 40 weeks post-menstrual age. Four peaks (N1, PI, N2 and P2) were reliably identified in all infants. The latency of each peak decreased as the infants matured. Each peak had a unique voltage scalp topography that remained stable as infants matured, even though the maps changed in amplitude intensity. N2 was large, easily identifiable with a central peak, and extremely stable in topography, suggesting that it might be used to evaluate the functional status of the somatosensory cortex in pre-term and term infants who are at high risk for developing intracranial hemorrhage leading to abnormalities of tone and delays in motor development.
Key words: SEP; Topography; Premature; Infants
Premature births occur in approximately 1 out of every 10 deliveries. Almost 7% of infants born in the United States weigh less than 2500 g (Wegman 1989) and 1% of all newborns weigh less than 1500 g (Chase 1977). Because of their immaturity, 35-40% of the very low birth weight pre-term infants will experience either intraventricular or intracerebral hemorrhage, with a risk that is indirectly proportional to their gestational age (Ahmann et al. 1980; Levine and Starte 1981; Dolfin et al. 1983; Van De B o r e t al. 1986). Outcome studies indicate that many of these children develop within a normal range, while other children experience mental and motor delays and cerebral palsy (Boyzinski et al. 1984; Krishnamoorthy et al. 1984; Amato et al. 1986). The somatosensory evoked potential (SEP) has become an effective tool to assess the integrity of the peripheral somatosensory system in both adults and children. Within the first 18 msec following a stimulus, several early peaks have been identified in adults, each representing different aspects of the peripheral, farfield response (see Desmedt 1988 for review). The N20 follows the far-field potentials and there is now compelling evidence that this SEP peak represents the arrival of the afferent volley from a median nerve
Correspondence to: Walt Karniski, M.D., Department of Pediatrics, MDC Box 15, University of South Florida College of Medicine, 12901 Bruce B. Downs Blvd., Tampa, FL 33612 (U.S.A.).
stimulus at the contralateral anterior parietal somatosensory cortex (Broughton 1969; Allison et al. 1980; Desmedt and Cheron 1980; Wood et al. 1985). In infants and children, this peak is referred to as N1 and the developmental changes of N1 latency and amplitude have been well described (Desmedt and Manil 1970; Desmedt and Debecker 1972; Cullity et al. 1976; Desmedt et al. 1976; Lager 1982; Hashimoto et al. 1983; Willis et al. 1984; Lauffer and Wenzel 1986; Tomita et al. 1986; Bartel et al. 1987; Zhu et al. 1987; Laureau et al. 1988; Taylor and Fagan 1988; George and Taylor 1991). However, only a few studies have investigated these developmental changes in pre-term infants (Hrbek et al. 1973; Gallai et al. 1986; Klimach and Cooke 1988; Bongers-Schokking et al. 1990). Pre-term infants' SEPs are characterized not only by the N1, but also by a larger amplitude N2 and smaller positive peaks. However, previous studies have focused primarily on the N1. Thus very little information is available about the cortical contribution to the SEP after the initial volley first reaches the cortex. Furthermore, none of these studies have examined the topographical distribution of these peaks across the scalp, since the SEP has been measured at only very few electrodes. As part of a separate study of the principal component topography of the late SEP to right median nerve stimulation in the same infants participating in this study, 3 principal components were identified that corresponded to 3 of the 4 late peaks to be analyzed in this investigation (Karniski 1992).
SEP D E V E L O P M E N T IN P R E M A T U R E I N F A N T S
This study was designed to investigate the systematic maturational changes of the late SEP peaks in pre-term and term infants. It is anticipated that different SEP peaks will have characteristic topographic profiles indicating different neural sources.
Methods
Subjects Fifty-three infants were studied in the Neonatal Intensive Care Unit and Newborn Nursery at Tampa General Hospital. In order to insure that they were relatively free of neurologic disorders, infants were included only if they met all of the following criteria: (1) 1 and 5 rain Apgar scores of at least 4 and 7 respectively, (2) length, weight, and head circumference appropriate for gestational age, (3) absence of major congenital malformations, (4) absence of known neurologic abnormalities or evidence of a seizure disorder, (5) normal cranial ultrasound in infants less than 34 weeks gestation at birth or less than 1500 g birth weight, (6) absence of asymmetric limb lengths, and (7) stable clinical condition without assisted ventilation or oxygen requirements. Parental consent was obtained for all infants. Because of the importance of an accurate assessment of post-menstrual age (PMA) for this study, an infant was included only if the post-menstrual age as assessed by examination (Dubowitz et al. 1970) was within 2 weeks as estimated by history (date of mother's last menstrual period). The PMA at birth ranged from 28 to 40 weeks. Infants were grouped by PMA into 2 week increments (31-32 weeks, 33-34 weeks, 35-36 weeks, 37-38 weeks, 39-40 weeks). Post-menstrual age is defined as the gestational age at birth plus chronological age at the time of testing and thus represents the time that has elapsed since the mother's last menstrual period. Because the time of the last menstrual period usually antedates conception by 2 weeks, the PMA is approximately 2 weeks longer than the time from conception. Thus, the PMA represents a maturational age that is more relevant than chronological age. Characteristics of these infants have been presented in a previous study (Karniski 1992) and will be only briefly summarized. The infants' mean chronological age at the time of testing was 13 days. The groups were generally very similar in race, sex and Apgar scores
Procedures Infants were all asleep at the time of testing. Stimuli consisted of simultaneous bilateral median nerve stimulation at each wrist, with a stimulus duration of 100 msec and a variable frequency of 0.9-1.7/sec. Stimulus intensity was set below sensory threshold, then gradually increased until a slight thumb twitch was observed.
45
The intensity was then reduced until the twitch was barely visible. Two blocks of 300 stimuli each were presented to each infant, 15 min apart, then the 2 blocks were averaged together. Single trials with voltage in excess of a pre-set level for each infant were automatically excluded from subsequent averaging. Sixteen pediatric gold electrodes (5 mm in diameter) were used with a linked ears reference at the following sites: Fpl, Fp2, F3, F4, F7, F8, C3, C4, T3, T4, T5, T6, P3, P4, O1 and 02. A Cadwell 8400 was used for signal generation and data acquisition. Impedance was maintained below 5000 gL The sweep extended from 50 msec prior to the stimulus to 483 msec after the stimulus. Voltage was sampled at a frequency of 960 Hz. A bandpass of 1-100 Hz was used. Rationale for this methodology can be found in Karniski (1992). Latency measures were initially made on wave forms with this bandpass, but they were usually digitally filtered with a high-cut Blackman filter at 45 Hz (Blackman and Tukey 1958) to help determine the exact latencies of the slower late peaks.
Data analysis" Latency was measured at each peak (N1, P1, N2 and P2) for each block for each infant at 6 electrodes (F3, F4, C3, C4, P3, P4). These were the only electrodes at which consistently identified peaks were observed. Other electrodes were used for the generation of topographical maps only. The N2 was generally large and easy to identify in most central electrodes, so that it was used as a landmark. The P1 was identified as the first positive peak occurring before the N2 and then the N1 was identified as the first negative peak before P1. P2 was usually a broad positive peak that more slowly returned to baseline after the N2. Occasionally two negative peaks occurred prior to the P1. In these cases, the latency of the largest was identified as the N1. When a peak could not be seen, the peak latency was not measured. Test-retest reliability was estimated by Pearson product-moment correlation by comparing the latencies for each peak on the first block of stimuli against the latencies on the second block for electrodes F3, F4, C3, C4, P3 and P4. To study the effects of latency maturation at different topographic sites, two 3-way, repeated measures ANOVAs were performed with the following main effects: post-menstrual age (5 groups), peak (N1, P1, N2 and P2), left-right position (left (F3, C3, P3)/right (F4, C4, P4)) for the first A N O V A and post-menstrual age, peak, and electrode (anterior (F3, F4)/central (C3, C4)/posterior (P3, P4)) for the second. The dependent variable was peak latency. All interaction effects were tested as well. Voltage topography was also analyzed with two 3-way repeated measures ANOVAs. Because the stimuli were
46
simultaneously bilateral and the anticipated changes in topography were expected to be in an anterior/posterior direction, homologous left and right electrode voltages were averaged. Both ANOVAs used the same conditions and levels as in the second analysis above (post-menstrual age, peak, and electrode (anterior (F3, F4)/central (C3, C4)/posterior (P3, P4)). For the first ANOVA, the dependent variable was the baseline adjusted voltage measured at the peak latency determined for each subject at each electrode minus the average voltage across the 50 msec pre-stimulus baseline. The dependent variable for the second ANOVA was the scaled baseline adjusted voltage, produced by subtracting from each subject's voltage the minimum voltage for each level of each condition across all subjects and electrodes and dividing by the range (McCarthy and Wood 1985). Thus a condition by electrode interaction for the unscaled voltages that persisted for the scaled voltages would indicate a change in topography. On the other hand, the absence of a significant interaction in the scaled analysis would indicate that an initial interaction for the unscaled voltages was caused by a modulation of voltage, proportionate across the scalp, without a difference in topography (McCarthy and Wood 1985). Topographical voltage maps were generated for the voltages at each electrode using the average reference (Offner 1950; Lehmann 1987), calculated by subtracting the mean of all electrodes from each electrode. The maps were then scaled by a slightly different method than that used for the statistical analysis. Voltages were divided by the root-mean-square (rms) or global field power (GFP). The GFP is the spatial standard deviation of all electrodes for a given map (Lehmann and Skrandies 1980) and is less sensitive than the range to noise in a single electrode. Scaling average reference data by the GFP is equivalent to generating spatial Z scores. Wave forms were averaged across subjects within the 5 PMA groups, then topographical maps were generated for the average voltage of a time segment around the peak latency. The peak latency was taken as the average of the latencies at the frontal, central and parietal electrodes, specific for each PMA group and for each peak. The time segment for N1 was from 5 msec before the N1 peak to 5 msec after the peak, for P1 the segment was peak plus and minus 10 msec and for N2 and P2 the segment was peak plus and minus 20 msec.
Results
Wave form description The wave forms averaged by group (Fig. 1) revealed 4 major peaks (N1, P1, N2 and P2). Each peak de-
W. K A R N I S K I E T A L .
creased in latency with increasing maturation. The NI was bi-peaked in some subjects and this is seen in the averages for the less mature infants. The amplitude of the P1 increased with increasing age, while the amplitude of the N2 decreased. There also appeared to be a slow, late, third negative wave clearly seen at 400 msec in the 39-4(} week infants, while only the first half of this peak could be seen near the end of the sweep in the less mature infants. Latencies or voltages of this peak were not measured and were not included in subsequent analyses because the sweep would not allow visualization of the peak for all infants.
Latency reliability and regression Each of the 4 peaks could be identified in both blocks in most subjects (Table I). All 4 of the peaks were more easily identified in the central electrodes. Each peak could be identified in at least 1 of the 2 blocks in the central electrodes for all of the infants. Peak latencies were consistent between the 2 blocks across subjects, more for the early than the late peaks (Table II). The peak latencies tended to be more reproducible at the frontal and central electrodes for N1 and P1, and at the central and parietal electrodes for N2 and P2. Linear, quadratic and cubic regressions were used to evaluate the relationship between PMA and the latencies of each peak. Because there was little improvement in the R z values with quadratic or cubic regression, only the estimated latencies from linear regression will be reported here. N 1 (lat) = 168 - 3.44 x P M A ( R 2 = 0 . 4 9 , P < 0 . 0 0 0 5 ) P l (lat) - 2 6 1 - 4 . 5 5 × N20at)=345
PMA (R 2 = 0.38, P < 0.0005)
4.55xPMA(R
2=0.30,P 4.00, P < 0.015 in all cases). Thus each peak was topographically distinct from each other peak. Except for some minor, systematic changes noted above, the topography changed little across PMA groups with maturation. However, there was a voltage modulation effect since the unscaled Electrode x Group interaction was significant ( F (8, 8 6 ) = 2.21, P = 0.03), while the scaled interaction was not ( F (8, 86) = 1.57, P = 0.15). This modulation effect would not be visible in the scaled maps (Fig. 4), but can be appreciated in Fig. 5 and in the wave forms (Fig. 1) as reflected in an increased N1 and P1 amplitude and a decrease in N2 amplitude with maturation in the central electrodes.
50
W. KARNISKI ET AL.
N
1
P
1
N2
31-32
l.O
luU
33-34
35-36
0
37-38 -1 .0
pu
39-40
SEP DEVELOPMENT IN PREMATURE INFANTS
51
AMPLI (IuV)TDEI~ U -
Discussion P2
£y~-o
N2
!/ P1 o
e, , \
-1 -2
\,,j,
~
"=
N1
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
Electrodes F
C
P
F
C
P
F C
P
F
C
P
F
C
P
Weeks 31-32
33-34
35-36
37-38
39-40
Fig. 5. Topographical distribution of voltage for the bilateral median nerve SEP. The average homologous unsealed voltages for the average of the left and right electrodes at frontal (F), central (C) and parietal (P) locations are graphed for each PMA group and for each peak. The "tilting" up of the line for the N1 is due to the posterior movement of the negative maximum. The P1 is maximal frontally in the less mature infants, but becomes central as they mature. The N2 becomes markedly reduced in amplitude as the infants mature, while maintaining the same topography. Finally, the P2 reveals a consistent central topography, with the exception of the 37-38 week infants. These findings are consistent with the development of the topography seen in the topographical maps in Fig. 4.
Although the 3-way interaction (PMA group x Peak x Electrode) was not significant for the unscaled voltages (F (24, 137) = 1.39, P = 0.12) it was for the scaled voltages (F (24, 137)= 2.08, P = 0.005). This interaction is present because the topography shows only a modulation across groups but changes topographically across peaks.
The brain of the 25 week fetus is lissencephalic, or almost completely smooth with little evidence of fissures or gyri. The sylvian fissure is evident in the third month post conception but the central sulcus is barely visible until 6 months (Bailey and Miller 1927; Patten 1968; Crelin 1974). Over the next 15 weeks, the fetal brain progresses through extensive structural, microscopic and functional changes. The central sulcus deepens and the pre-central gyrus and post-central gyrus (containing the motor and somatosensory cortices respectively) become visible and fold in. As the association areas of the frontal lobe increase in size, the central sulcus moves more posterior. Myelin begins to develop around neuronal axons, neuronal cells differentiate, migrate to the surface of the cortex and begin to establish multiple cell-to-cell connections (Herschkowitz 1988; Mrzljak et al. 1988). Thus it is not surprising that the somatosensory evoked potential also changes dramatically over the critical 9 weeks from 31 weeks PMA to full-term. This study analyzed the developmental changes of 4 peaks in the bilateral median nerve SEP. These peaks arise when the somatosensory stimulus reaches the cortex and thereafter and most likely reflect intracortical activity. These 4 peaks were reliably reproduced over short periods of time. As pre-term infants matured, the peak latencies became shorter and latency profiles changed. Each peak produced different voltage topographies and these landscapes were stable in shape as the infant matured but changed in magnitude. These 4 peaks are probably generated by at least 4 distinct sources and this is supported by a number of findings in this study. First, the latencies of the 4 peaks mature at different rates, as indicated by the different slopes of the regression lines. Second, this is confirmed by the ANOVA showing a significant interaction of the latency between Peak and PMA group. Third, the significant Peak x Electrode interaction for latency as well as unscaled and scaled voltages indicate that the different peaks evolve in time and space in distinct fashion. Fourth, separate paired ANOVAs were performed that revealed a Peak × Electrode interaction for both unscaled and scaled voltages, indicating unique scalp topographies for each of the possible pairs of
Fig. 4. Scaled topographical maps of the 4 peaks of the bilateral median nerve SEP for each PMA group. Topographical maps were generated for each of the 4 peaks for each PMA group utilizing the average voltage over a time segment surrounding the average peak latency. All maps were created Using the voltage from 16 electrodes with the average reference. Midline electrode values were interpolated. The maps were scaled by dividing the average reference voltages by the root-mean-square of the map voltages or global field power. The time segment for N1 was from 5 msec before the average N1 peak to 5 msec after the peak, for P1 the segment was peak minus and peak plus 10 msec and for N2 and P2 the segment was peak minus and peak plus 20 msec.
52
peaks. And finally, the distinct topographies as revealed by the topographic maps tend to confirm these findings. However, it is not possible to draw conclusions about generator source from the topographical maps produced in response to a bilateral stimulus, because activation of a unilateral somatosensory generator can produce both contralateral and ipsilateral effects on the field measured at the scalp surface. In addition, it is possible that simultaneous bilateral stimuli will interact to produce an effect on the wave forms and scalp topography that is different from what would be expected from the simple sum of the effects from two unilateral stimuli. To test this hypothesis, the sum of two separate unilateral SEPs to left and right median nerve stimuli were compared to bilateral median nerve evoked potentials. Only minimal differences were found and these differences occurred after 150 msec. The SEP produced by a right median nerve stimulus was studied with principal component analysis in the same infants studied here (Karniski 1992). Three factors emerged from this study, accounting for approximately three-fourths of the variance. The first factor (N!/P1) had a left posterior minimum and a left frontal-central maximum. The second factor (N2) was characterized by a consistent left central minimum, with a maximum that appeared to rotate in space as the infants matured. A third factor (N3) appeared to represent the first evidence of ipsilateral cortical activity and this factor's topography remained stable with maturation. The unilateral N1 in this study appears to be analogous to the N20 in adults and it produces a contralateral posterior minimum and contralateral frontalcentral maximum which indexes the arrival of the peripheral volley at the primary somatosensory receiving area. It is generally well accepted that the N20 is generated by a tangentially oriented dipole in the posterior bank of the central fissure with the negative end directed posteriorly and the positive end anteriorly (Broughton 1969; Allison et al. 1980; Desmedt and Cheron 1980; Wood et al. 1985). The estimated latencies identified for the N1 in this study were similar to those found in Klimach and Cooke (1988) as indicated by the similarity between their regression equation (N1 (lat)= 162-3.05 x PMA) and the equation found in this study (N1 (lat) = 1 6 8 - 3.44 × PMA). However, no other study has systematically analyzed the peaks occurring in the SEP after N1. In premature infants the N1 is followed by a P1, and in adults the N20 is followed by an overlapping P22. While the P22 was initially thought to represent the positive end of the tangentially oriented N20 dipole, a separate, radially oriented dipole in the pre-central gyrus is the probable source of the P22 (Desmedt and
W. KARNISKI ET AL.
Cheron 1981; Maugui~re et al. 1983; Desmedt and Bourguet 1985; Desmedt et al. 1987; Allison et al. 1989). In this study, the P1 corresponds to the P22 found in adults. Following P1, a large N2 is clearly evident and the most prominent feature of the late SEP. It has a central topography that remains very stable as the infants mature. The N2 is probably analogous to the N26 described by Tsuji et al. (1988) or the N30 described by Desmedt and colleagues (Desmedt and Bourguet 1985; Desmedt et al. 1987). They have proposed a source in the supplementary motor area for the N2, while Allison et al. (1989) propose a radial dipole located in the post-central gyrus. The P2 maintains a posterior orientation throughout maturation, but the shape of this topography is more variable than the previous peaks. Unfortunately, longitudinal or cross-sectional developmental studies of the N2 and P2 from infancy to adulthood are not available that would allow one to conclusively determine the analogs of these two peaks in the adult SEP. Furthermore, very little is known about the generators of the late peaks of the adult SEP. This study indicates that there are different topographies for each of the identified peaks and while there are systematic changes with maturation, there is a relative stability of topography with maturation. This topographic stability allows for a constant metric of brain function, regardless of the maturational level of the infant. The N2 is the most stable peak topographically. Because it occurs after the peripheral volley first reaches the somatosensory cortex, it probably represents intracortical connections. Since it is so easily identifiable, even in the most premature infants in this study, and since the topography is so stable as the infant matures, the topography of the N2 might be a more fruitful, future area of study in neonates with ischemic or hemorrhagic insult. The authors wish to extend their appreciation to Richard Weibley, M.D., for identification of many of the infants eligible for this study, to Rodney Vanderploeg, Ph.D., for his assistance in the review of this paper and to Roger Daniels for assistance in the Neonatal ICU. A special thanks is extended to Dietrich Lehmann, M.D., who supported the first author's sabbatical and stimulated the thinking that lead to some of the techniques presented in this paper. This research was funded in part by the State of Florida, Department of Education, Florida Diagnostic and Learning Resource System and Children's Medical Services, State of Florida.
References Ahmann, P.A., Lazzara, A., Dykes, F.D. et al. Intraventricular hemorrhage in the high risk pre-term neonate: incidence and outcome. Ann. Neurol., 1980, 7: 118-124. Allison, T., Goff, W.R., Williamson, P.D. and Van Gilder, J.C. On the neural origin of early components of the somatosensory
SEP D E V E L O P M E N T IN P R E M A T U R E INFANTS evoked potential. In: J.E. Desmedt led.), Clinical Uses of Cerebra[, Brainstem and Spinal Somatosensory Evoked Potentials. Prog. Clin. Neurophysiol., Vol. 7. Karger, Basel, 19811: 51-68. Allison, T., McCarthy, G., Wood, C.C., Darcey, T.M., Spencer, D.D. and Williamson, P.D. H u m a n cortical potentials ew)ked by stimulation of the median nerve. I. Cytoarchitectonic areas generating short-latency activity. J. Neurophysiol., 1989, 62: 694-710. Amato, M., tlowald, H. and Von Muralt, G. Neurological prognosis of high-risk prc-term infants with peri-intraventricular hemorrhage and ventricular dilatation. Eur. Neurol., 1986, 25: 241-247. Bailey, F.R and Miller, A.M. Text-book of Embryology. William Wood and Co., New York, 1927. Barrel, P., Conradie, J., Robinson, E., Prinsloo, J. and Becket, P. The relationship between median nerve somatosensory evoked potential [atencies and age and growth parameters in young children. Electroenceph. clin. Neurophysio[., 1987, 68: 180-186. Blackman. R.B. and Tukey, J.W. The M e a s u r e m e n t of Power Spectra. Dover Publ., New York, 1958. Bongers-Schokking, J.J., Coltm, E.J., Hoogland, R.A., Van den Brande, J.L. and De Groot, C.J. Somatosensory ew)ked potentials in term and preterm infants in relation to postconceptual age and birth weight. Neuropediatrics, 19911, 21: 32-36. Boyzinski, M.E.A., Nelson, M.N., Rosati-Skertich, C., Genaze, D., O'Donnell, K. and Naughton, P. Two year longitudinal follow-up of premature infants weighing
44
EVOPOT 91031
The late somatosensory evoked potential in premature and term infants. II. Topography and latency development W a l t K a r n i s k i , L a n c e W y b l e , L o r e t t a L e a s e a n d R. C l i f f o r d Blair Department of Pediatrics, Unil'ersity of South Florida College o[" Medicine, Tampa, FL 33612 (U.S.A.) (Accepted for publication: 3 June 1991 )
Summary The maturation of latency and scalp voltage topography of the simultaneously bilateral somatosensory evoked potential was studied in 53 neurologically intact pre-term and term infants, from 31 to 40 weeks post-menstrual age. Four peaks (N1, PI, N2 and P2) were reliably identified in all infants. The latency of each peak decreased as the infants matured. Each peak had a unique voltage scalp topography that remained stable as infants matured, even though the maps changed in amplitude intensity. N2 was large, easily identifiable with a central peak, and extremely stable in topography, suggesting that it might be used to evaluate the functional status of the somatosensory cortex in pre-term and term infants who are at high risk for developing intracranial hemorrhage leading to abnormalities of tone and delays in motor development.
Key words: SEP; Topography; Premature; Infants
Premature births occur in approximately 1 out of every 10 deliveries. Almost 7% of infants born in the United States weigh less than 2500 g (Wegman 1989) and 1% of all newborns weigh less than 1500 g (Chase 1977). Because of their immaturity, 35-40% of the very low birth weight pre-term infants will experience either intraventricular or intracerebral hemorrhage, with a risk that is indirectly proportional to their gestational age (Ahmann et al. 1980; Levine and Starte 1981; Dolfin et al. 1983; Van De B o r e t al. 1986). Outcome studies indicate that many of these children develop within a normal range, while other children experience mental and motor delays and cerebral palsy (Boyzinski et al. 1984; Krishnamoorthy et al. 1984; Amato et al. 1986). The somatosensory evoked potential (SEP) has become an effective tool to assess the integrity of the peripheral somatosensory system in both adults and children. Within the first 18 msec following a stimulus, several early peaks have been identified in adults, each representing different aspects of the peripheral, farfield response (see Desmedt 1988 for review). The N20 follows the far-field potentials and there is now compelling evidence that this SEP peak represents the arrival of the afferent volley from a median nerve
Correspondence to: Walt Karniski, M.D., Department of Pediatrics, MDC Box 15, University of South Florida College of Medicine, 12901 Bruce B. Downs Blvd., Tampa, FL 33612 (U.S.A.).
stimulus at the contralateral anterior parietal somatosensory cortex (Broughton 1969; Allison et al. 1980; Desmedt and Cheron 1980; Wood et al. 1985). In infants and children, this peak is referred to as N1 and the developmental changes of N1 latency and amplitude have been well described (Desmedt and Manil 1970; Desmedt and Debecker 1972; Cullity et al. 1976; Desmedt et al. 1976; Lager 1982; Hashimoto et al. 1983; Willis et al. 1984; Lauffer and Wenzel 1986; Tomita et al. 1986; Bartel et al. 1987; Zhu et al. 1987; Laureau et al. 1988; Taylor and Fagan 1988; George and Taylor 1991). However, only a few studies have investigated these developmental changes in pre-term infants (Hrbek et al. 1973; Gallai et al. 1986; Klimach and Cooke 1988; Bongers-Schokking et al. 1990). Pre-term infants' SEPs are characterized not only by the N1, but also by a larger amplitude N2 and smaller positive peaks. However, previous studies have focused primarily on the N1. Thus very little information is available about the cortical contribution to the SEP after the initial volley first reaches the cortex. Furthermore, none of these studies have examined the topographical distribution of these peaks across the scalp, since the SEP has been measured at only very few electrodes. As part of a separate study of the principal component topography of the late SEP to right median nerve stimulation in the same infants participating in this study, 3 principal components were identified that corresponded to 3 of the 4 late peaks to be analyzed in this investigation (Karniski 1992).
SEP D E V E L O P M E N T IN P R E M A T U R E I N F A N T S
This study was designed to investigate the systematic maturational changes of the late SEP peaks in pre-term and term infants. It is anticipated that different SEP peaks will have characteristic topographic profiles indicating different neural sources.
Methods
Subjects Fifty-three infants were studied in the Neonatal Intensive Care Unit and Newborn Nursery at Tampa General Hospital. In order to insure that they were relatively free of neurologic disorders, infants were included only if they met all of the following criteria: (1) 1 and 5 rain Apgar scores of at least 4 and 7 respectively, (2) length, weight, and head circumference appropriate for gestational age, (3) absence of major congenital malformations, (4) absence of known neurologic abnormalities or evidence of a seizure disorder, (5) normal cranial ultrasound in infants less than 34 weeks gestation at birth or less than 1500 g birth weight, (6) absence of asymmetric limb lengths, and (7) stable clinical condition without assisted ventilation or oxygen requirements. Parental consent was obtained for all infants. Because of the importance of an accurate assessment of post-menstrual age (PMA) for this study, an infant was included only if the post-menstrual age as assessed by examination (Dubowitz et al. 1970) was within 2 weeks as estimated by history (date of mother's last menstrual period). The PMA at birth ranged from 28 to 40 weeks. Infants were grouped by PMA into 2 week increments (31-32 weeks, 33-34 weeks, 35-36 weeks, 37-38 weeks, 39-40 weeks). Post-menstrual age is defined as the gestational age at birth plus chronological age at the time of testing and thus represents the time that has elapsed since the mother's last menstrual period. Because the time of the last menstrual period usually antedates conception by 2 weeks, the PMA is approximately 2 weeks longer than the time from conception. Thus, the PMA represents a maturational age that is more relevant than chronological age. Characteristics of these infants have been presented in a previous study (Karniski 1992) and will be only briefly summarized. The infants' mean chronological age at the time of testing was 13 days. The groups were generally very similar in race, sex and Apgar scores
Procedures Infants were all asleep at the time of testing. Stimuli consisted of simultaneous bilateral median nerve stimulation at each wrist, with a stimulus duration of 100 msec and a variable frequency of 0.9-1.7/sec. Stimulus intensity was set below sensory threshold, then gradually increased until a slight thumb twitch was observed.
45
The intensity was then reduced until the twitch was barely visible. Two blocks of 300 stimuli each were presented to each infant, 15 min apart, then the 2 blocks were averaged together. Single trials with voltage in excess of a pre-set level for each infant were automatically excluded from subsequent averaging. Sixteen pediatric gold electrodes (5 mm in diameter) were used with a linked ears reference at the following sites: Fpl, Fp2, F3, F4, F7, F8, C3, C4, T3, T4, T5, T6, P3, P4, O1 and 02. A Cadwell 8400 was used for signal generation and data acquisition. Impedance was maintained below 5000 gL The sweep extended from 50 msec prior to the stimulus to 483 msec after the stimulus. Voltage was sampled at a frequency of 960 Hz. A bandpass of 1-100 Hz was used. Rationale for this methodology can be found in Karniski (1992). Latency measures were initially made on wave forms with this bandpass, but they were usually digitally filtered with a high-cut Blackman filter at 45 Hz (Blackman and Tukey 1958) to help determine the exact latencies of the slower late peaks.
Data analysis" Latency was measured at each peak (N1, P1, N2 and P2) for each block for each infant at 6 electrodes (F3, F4, C3, C4, P3, P4). These were the only electrodes at which consistently identified peaks were observed. Other electrodes were used for the generation of topographical maps only. The N2 was generally large and easy to identify in most central electrodes, so that it was used as a landmark. The P1 was identified as the first positive peak occurring before the N2 and then the N1 was identified as the first negative peak before P1. P2 was usually a broad positive peak that more slowly returned to baseline after the N2. Occasionally two negative peaks occurred prior to the P1. In these cases, the latency of the largest was identified as the N1. When a peak could not be seen, the peak latency was not measured. Test-retest reliability was estimated by Pearson product-moment correlation by comparing the latencies for each peak on the first block of stimuli against the latencies on the second block for electrodes F3, F4, C3, C4, P3 and P4. To study the effects of latency maturation at different topographic sites, two 3-way, repeated measures ANOVAs were performed with the following main effects: post-menstrual age (5 groups), peak (N1, P1, N2 and P2), left-right position (left (F3, C3, P3)/right (F4, C4, P4)) for the first A N O V A and post-menstrual age, peak, and electrode (anterior (F3, F4)/central (C3, C4)/posterior (P3, P4)) for the second. The dependent variable was peak latency. All interaction effects were tested as well. Voltage topography was also analyzed with two 3-way repeated measures ANOVAs. Because the stimuli were
46
simultaneously bilateral and the anticipated changes in topography were expected to be in an anterior/posterior direction, homologous left and right electrode voltages were averaged. Both ANOVAs used the same conditions and levels as in the second analysis above (post-menstrual age, peak, and electrode (anterior (F3, F4)/central (C3, C4)/posterior (P3, P4)). For the first ANOVA, the dependent variable was the baseline adjusted voltage measured at the peak latency determined for each subject at each electrode minus the average voltage across the 50 msec pre-stimulus baseline. The dependent variable for the second ANOVA was the scaled baseline adjusted voltage, produced by subtracting from each subject's voltage the minimum voltage for each level of each condition across all subjects and electrodes and dividing by the range (McCarthy and Wood 1985). Thus a condition by electrode interaction for the unscaled voltages that persisted for the scaled voltages would indicate a change in topography. On the other hand, the absence of a significant interaction in the scaled analysis would indicate that an initial interaction for the unscaled voltages was caused by a modulation of voltage, proportionate across the scalp, without a difference in topography (McCarthy and Wood 1985). Topographical voltage maps were generated for the voltages at each electrode using the average reference (Offner 1950; Lehmann 1987), calculated by subtracting the mean of all electrodes from each electrode. The maps were then scaled by a slightly different method than that used for the statistical analysis. Voltages were divided by the root-mean-square (rms) or global field power (GFP). The GFP is the spatial standard deviation of all electrodes for a given map (Lehmann and Skrandies 1980) and is less sensitive than the range to noise in a single electrode. Scaling average reference data by the GFP is equivalent to generating spatial Z scores. Wave forms were averaged across subjects within the 5 PMA groups, then topographical maps were generated for the average voltage of a time segment around the peak latency. The peak latency was taken as the average of the latencies at the frontal, central and parietal electrodes, specific for each PMA group and for each peak. The time segment for N1 was from 5 msec before the N1 peak to 5 msec after the peak, for P1 the segment was peak plus and minus 10 msec and for N2 and P2 the segment was peak plus and minus 20 msec.
Results
Wave form description The wave forms averaged by group (Fig. 1) revealed 4 major peaks (N1, P1, N2 and P2). Each peak de-
W. K A R N I S K I E T A L .
creased in latency with increasing maturation. The NI was bi-peaked in some subjects and this is seen in the averages for the less mature infants. The amplitude of the P1 increased with increasing age, while the amplitude of the N2 decreased. There also appeared to be a slow, late, third negative wave clearly seen at 400 msec in the 39-4(} week infants, while only the first half of this peak could be seen near the end of the sweep in the less mature infants. Latencies or voltages of this peak were not measured and were not included in subsequent analyses because the sweep would not allow visualization of the peak for all infants.
Latency reliability and regression Each of the 4 peaks could be identified in both blocks in most subjects (Table I). All 4 of the peaks were more easily identified in the central electrodes. Each peak could be identified in at least 1 of the 2 blocks in the central electrodes for all of the infants. Peak latencies were consistent between the 2 blocks across subjects, more for the early than the late peaks (Table II). The peak latencies tended to be more reproducible at the frontal and central electrodes for N1 and P1, and at the central and parietal electrodes for N2 and P2. Linear, quadratic and cubic regressions were used to evaluate the relationship between PMA and the latencies of each peak. Because there was little improvement in the R z values with quadratic or cubic regression, only the estimated latencies from linear regression will be reported here. N 1 (lat) = 168 - 3.44 x P M A ( R 2 = 0 . 4 9 , P < 0 . 0 0 0 5 ) P l (lat) - 2 6 1 - 4 . 5 5 × N20at)=345
PMA (R 2 = 0.38, P < 0.0005)
4.55xPMA(R
2=0.30,P 4.00, P < 0.015 in all cases). Thus each peak was topographically distinct from each other peak. Except for some minor, systematic changes noted above, the topography changed little across PMA groups with maturation. However, there was a voltage modulation effect since the unscaled Electrode x Group interaction was significant ( F (8, 8 6 ) = 2.21, P = 0.03), while the scaled interaction was not ( F (8, 86) = 1.57, P = 0.15). This modulation effect would not be visible in the scaled maps (Fig. 4), but can be appreciated in Fig. 5 and in the wave forms (Fig. 1) as reflected in an increased N1 and P1 amplitude and a decrease in N2 amplitude with maturation in the central electrodes.
50
W. KARNISKI ET AL.
N
1
P
1
N2
31-32
l.O
luU
33-34
35-36
0
37-38 -1 .0
pu
39-40
SEP DEVELOPMENT IN PREMATURE INFANTS
51
AMPLI (IuV)TDEI~ U -
Discussion P2
£y~-o
N2
!/ P1 o
e, , \
-1 -2
\,,j,
~
"=
N1
I
I
I
I
I
I
I
I
I
I
I
I
I
I
I
Electrodes F
C
P
F
C
P
F C
P
F
C
P
F
C
P
Weeks 31-32
33-34
35-36
37-38
39-40
Fig. 5. Topographical distribution of voltage for the bilateral median nerve SEP. The average homologous unsealed voltages for the average of the left and right electrodes at frontal (F), central (C) and parietal (P) locations are graphed for each PMA group and for each peak. The "tilting" up of the line for the N1 is due to the posterior movement of the negative maximum. The P1 is maximal frontally in the less mature infants, but becomes central as they mature. The N2 becomes markedly reduced in amplitude as the infants mature, while maintaining the same topography. Finally, the P2 reveals a consistent central topography, with the exception of the 37-38 week infants. These findings are consistent with the development of the topography seen in the topographical maps in Fig. 4.
Although the 3-way interaction (PMA group x Peak x Electrode) was not significant for the unscaled voltages (F (24, 137) = 1.39, P = 0.12) it was for the scaled voltages (F (24, 137)= 2.08, P = 0.005). This interaction is present because the topography shows only a modulation across groups but changes topographically across peaks.
The brain of the 25 week fetus is lissencephalic, or almost completely smooth with little evidence of fissures or gyri. The sylvian fissure is evident in the third month post conception but the central sulcus is barely visible until 6 months (Bailey and Miller 1927; Patten 1968; Crelin 1974). Over the next 15 weeks, the fetal brain progresses through extensive structural, microscopic and functional changes. The central sulcus deepens and the pre-central gyrus and post-central gyrus (containing the motor and somatosensory cortices respectively) become visible and fold in. As the association areas of the frontal lobe increase in size, the central sulcus moves more posterior. Myelin begins to develop around neuronal axons, neuronal cells differentiate, migrate to the surface of the cortex and begin to establish multiple cell-to-cell connections (Herschkowitz 1988; Mrzljak et al. 1988). Thus it is not surprising that the somatosensory evoked potential also changes dramatically over the critical 9 weeks from 31 weeks PMA to full-term. This study analyzed the developmental changes of 4 peaks in the bilateral median nerve SEP. These peaks arise when the somatosensory stimulus reaches the cortex and thereafter and most likely reflect intracortical activity. These 4 peaks were reliably reproduced over short periods of time. As pre-term infants matured, the peak latencies became shorter and latency profiles changed. Each peak produced different voltage topographies and these landscapes were stable in shape as the infant matured but changed in magnitude. These 4 peaks are probably generated by at least 4 distinct sources and this is supported by a number of findings in this study. First, the latencies of the 4 peaks mature at different rates, as indicated by the different slopes of the regression lines. Second, this is confirmed by the ANOVA showing a significant interaction of the latency between Peak and PMA group. Third, the significant Peak x Electrode interaction for latency as well as unscaled and scaled voltages indicate that the different peaks evolve in time and space in distinct fashion. Fourth, separate paired ANOVAs were performed that revealed a Peak × Electrode interaction for both unscaled and scaled voltages, indicating unique scalp topographies for each of the possible pairs of
Fig. 4. Scaled topographical maps of the 4 peaks of the bilateral median nerve SEP for each PMA group. Topographical maps were generated for each of the 4 peaks for each PMA group utilizing the average voltage over a time segment surrounding the average peak latency. All maps were created Using the voltage from 16 electrodes with the average reference. Midline electrode values were interpolated. The maps were scaled by dividing the average reference voltages by the root-mean-square of the map voltages or global field power. The time segment for N1 was from 5 msec before the average N1 peak to 5 msec after the peak, for P1 the segment was peak minus and peak plus 10 msec and for N2 and P2 the segment was peak minus and peak plus 20 msec.
52
peaks. And finally, the distinct topographies as revealed by the topographic maps tend to confirm these findings. However, it is not possible to draw conclusions about generator source from the topographical maps produced in response to a bilateral stimulus, because activation of a unilateral somatosensory generator can produce both contralateral and ipsilateral effects on the field measured at the scalp surface. In addition, it is possible that simultaneous bilateral stimuli will interact to produce an effect on the wave forms and scalp topography that is different from what would be expected from the simple sum of the effects from two unilateral stimuli. To test this hypothesis, the sum of two separate unilateral SEPs to left and right median nerve stimuli were compared to bilateral median nerve evoked potentials. Only minimal differences were found and these differences occurred after 150 msec. The SEP produced by a right median nerve stimulus was studied with principal component analysis in the same infants studied here (Karniski 1992). Three factors emerged from this study, accounting for approximately three-fourths of the variance. The first factor (N!/P1) had a left posterior minimum and a left frontal-central maximum. The second factor (N2) was characterized by a consistent left central minimum, with a maximum that appeared to rotate in space as the infants matured. A third factor (N3) appeared to represent the first evidence of ipsilateral cortical activity and this factor's topography remained stable with maturation. The unilateral N1 in this study appears to be analogous to the N20 in adults and it produces a contralateral posterior minimum and contralateral frontalcentral maximum which indexes the arrival of the peripheral volley at the primary somatosensory receiving area. It is generally well accepted that the N20 is generated by a tangentially oriented dipole in the posterior bank of the central fissure with the negative end directed posteriorly and the positive end anteriorly (Broughton 1969; Allison et al. 1980; Desmedt and Cheron 1980; Wood et al. 1985). The estimated latencies identified for the N1 in this study were similar to those found in Klimach and Cooke (1988) as indicated by the similarity between their regression equation (N1 (lat)= 162-3.05 x PMA) and the equation found in this study (N1 (lat) = 1 6 8 - 3.44 × PMA). However, no other study has systematically analyzed the peaks occurring in the SEP after N1. In premature infants the N1 is followed by a P1, and in adults the N20 is followed by an overlapping P22. While the P22 was initially thought to represent the positive end of the tangentially oriented N20 dipole, a separate, radially oriented dipole in the pre-central gyrus is the probable source of the P22 (Desmedt and
W. KARNISKI ET AL.
Cheron 1981; Maugui~re et al. 1983; Desmedt and Bourguet 1985; Desmedt et al. 1987; Allison et al. 1989). In this study, the P1 corresponds to the P22 found in adults. Following P1, a large N2 is clearly evident and the most prominent feature of the late SEP. It has a central topography that remains very stable as the infants mature. The N2 is probably analogous to the N26 described by Tsuji et al. (1988) or the N30 described by Desmedt and colleagues (Desmedt and Bourguet 1985; Desmedt et al. 1987). They have proposed a source in the supplementary motor area for the N2, while Allison et al. (1989) propose a radial dipole located in the post-central gyrus. The P2 maintains a posterior orientation throughout maturation, but the shape of this topography is more variable than the previous peaks. Unfortunately, longitudinal or cross-sectional developmental studies of the N2 and P2 from infancy to adulthood are not available that would allow one to conclusively determine the analogs of these two peaks in the adult SEP. Furthermore, very little is known about the generators of the late peaks of the adult SEP. This study indicates that there are different topographies for each of the identified peaks and while there are systematic changes with maturation, there is a relative stability of topography with maturation. This topographic stability allows for a constant metric of brain function, regardless of the maturational level of the infant. The N2 is the most stable peak topographically. Because it occurs after the peripheral volley first reaches the somatosensory cortex, it probably represents intracortical connections. Since it is so easily identifiable, even in the most premature infants in this study, and since the topography is so stable as the infant matures, the topography of the N2 might be a more fruitful, future area of study in neonates with ischemic or hemorrhagic insult. The authors wish to extend their appreciation to Richard Weibley, M.D., for identification of many of the infants eligible for this study, to Rodney Vanderploeg, Ph.D., for his assistance in the review of this paper and to Roger Daniels for assistance in the Neonatal ICU. A special thanks is extended to Dietrich Lehmann, M.D., who supported the first author's sabbatical and stimulated the thinking that lead to some of the techniques presented in this paper. This research was funded in part by the State of Florida, Department of Education, Florida Diagnostic and Learning Resource System and Children's Medical Services, State of Florida.
References Ahmann, P.A., Lazzara, A., Dykes, F.D. et al. Intraventricular hemorrhage in the high risk pre-term neonate: incidence and outcome. Ann. Neurol., 1980, 7: 118-124. Allison, T., Goff, W.R., Williamson, P.D. and Van Gilder, J.C. On the neural origin of early components of the somatosensory
SEP D E V E L O P M E N T IN P R E M A T U R E INFANTS evoked potential. In: J.E. Desmedt led.), Clinical Uses of Cerebra[, Brainstem and Spinal Somatosensory Evoked Potentials. Prog. Clin. Neurophysiol., Vol. 7. Karger, Basel, 19811: 51-68. Allison, T., McCarthy, G., Wood, C.C., Darcey, T.M., Spencer, D.D. and Williamson, P.D. H u m a n cortical potentials ew)ked by stimulation of the median nerve. I. Cytoarchitectonic areas generating short-latency activity. J. Neurophysiol., 1989, 62: 694-710. Amato, M., tlowald, H. and Von Muralt, G. Neurological prognosis of high-risk prc-term infants with peri-intraventricular hemorrhage and ventricular dilatation. Eur. Neurol., 1986, 25: 241-247. Bailey, F.R and Miller, A.M. Text-book of Embryology. William Wood and Co., New York, 1927. Barrel, P., Conradie, J., Robinson, E., Prinsloo, J. and Becket, P. The relationship between median nerve somatosensory evoked potential [atencies and age and growth parameters in young children. Electroenceph. clin. Neurophysio[., 1987, 68: 180-186. Blackman. R.B. and Tukey, J.W. The M e a s u r e m e n t of Power Spectra. Dover Publ., New York, 1958. Bongers-Schokking, J.J., Coltm, E.J., Hoogland, R.A., Van den Brande, J.L. and De Groot, C.J. Somatosensory ew)ked potentials in term and preterm infants in relation to postconceptual age and birth weight. Neuropediatrics, 19911, 21: 32-36. Boyzinski, M.E.A., Nelson, M.N., Rosati-Skertich, C., Genaze, D., O'Donnell, K. and Naughton, P. Two year longitudinal follow-up of premature infants weighing
