Main

The sick newborn infant is vulnerable to brain injury and impaired cerebral autoregulation is thought to contribute to this (1, 2). Most studies of cerebral autoregulation use the static, classic model of autoregulation (3). This model contains a plateau defining the limits of autoregulation. Subjects are grouped together to produce this model as it is not possible to make sufficient observations of cerebral blood flow (CBF) at different times in one individual. The estimates of CBF included in these studies are results of average values obtained over several minutes, repeated over a period of hours. This model has been questioned recently. Blood pressure (BP) changes have to be induced and this can only be achieved after a significant time interval. This means that it is difficult to control for other variables known to affect CBF, namely hematocrit and Paco2 (4, 5). In addition, it is not ethical to induce large BP changes in sick newborn infants.

The autoregulatory response to changes in BP must take a finite amount of time to occur. In adults, Aaslid et al. (1989) introduced the thigh cuff method to produce sudden drops in arterial blood pressure (ABP) and studied the evolution of the immediate CBF velocity (CBFV) response with Doppler ultrasound (6). Doppler ultrasound is the only noninvasive method that allows sufficient temporal resolution to measure these rapid changes. Both Aaslid et al. (6) and other workers (7) have shown that the CBF velocity response to a sudden change in ABP begins within 2 s and is complete in 10–15 s. The short time frame of dynamic studies minimizes the influence of variables such as Paco2, sleep state, or hematocrit on CBFV.

The thigh cuff method is clearly unsuitable for neonates. We have developed an alternative way of assessing autoregulation without disturbing the infant by using coherent averaging to assess the CBFV response to small spontaneous transient changes in BP. In this model, intact cerebral autoregulation produces a characteristic response consisting of a brief period when CBFV follows ABP but quickly returns to baseline value (Fig. 2 in this paper). An impaired autoregulatory response shows CBFV mirroring the ABP curve closely.

Figure 2
figure 2

Autoregulatory response curves for the four groups of infants studied. These curves were constructed after the ABP and CBFV waveforms from the infants in each group were averaged and normalized with respect to their respective peak values.

Full size image

The commonest clinical manifestation of neurologic illness in the newborn infant is seizures. CBF increases during neonatal seizures and this increase is present even when the clinical component of the seizure is absent or has been abolished with anticonvulsants or therapeutic paralysis (8, 9). Seizures in the newborn are often very subtle or lack an obvious clinical signature, and the true seizure burden may be grossly underestimated (10–12) with long-term developmental consequences (13).

We chose to examine the cerebral autoregulatory response of infants with neurologic illness who were undergoing intensive care treatment. We compared them to a group of infants also undergoing intensive care treatment but without evidence of neurologic illness. Our hypothesis was that sick infants with neurologic illness would have impaired autoregulatory ability, whereas other sick infants without neurologic illness would autoregulate.

METHODS

Over a 2-y period, a group of infants admitted to the Neonatal Intensive Care Unit (NICU) at Kings College Hospital was studied. The study had ethical approval from the hospital and written parental consent was obtained for each infant.

Subjects.

Infants who were admitted to the unit with any of the following conditions, hypoxic ischemic encephalopathy (HIE), meningitis, suffering from drug withdrawal, or with known cerebral pathology were eligible. A group of control infants was also recruited from infants admitted to the NICU with mild respiratory distress syndrome or requiring neonatal surgery.

Recording procedures.

Video-EEG polygraphy was recorded in all infants using the method described previously (8). During the EEG, CBFV was recorded from the middle cerebral artery at intervals, using a semicontinuous system similar to that described by Evans et al. (14) but based on a 486 PC. In short, this consisted of a small continuous wave Doppler ultrasound probe with a total acoustic output of <5 mW and I (spatial peak temporal average) <50 mW cm–2 that was fixed to the skin overlying the middle cerebral artery. BP was measured simultaneously from an existing umbilical or peripheral arterial catheter using a Medex (Medex, Columbus, OH, U.S.A.) MX860 transducer connected to a Mennen (Mennen Medical Corp., Clarence, NY, U.S.A.) Horizon 2000/s or Marquette (GE Medical Systems, Milwaukee, WI, U.S.A.) Solar 8000 monitor. No infant had a catheter sited purely for the purposes of this study.

In all infants studied, simultaneous recordings of CBFV, BP, and EEG (5- to 12-min periods) were obtained at intervals during the recording session. CBFV, BP, and two channels of EEG data (C4-P4, C3-P3) were stored on a digital instrumentation tape recorder (PC-108 M) and transferred to a personal computer.

Information on blood gases, electrolytes, blood glucose, serum bilirubin, sedation, and all drugs, fluids, and anticonvulsants were collected prospectively. A detailed history of pregnancy, labor, delivery, and postnatal condition was also obtained for each infant. The attending neonatologist (J.R.) clinically assessed all infants on the day of the examination and performed a cranial ultrasound scan just before CBFV measurement. All high-risk and locally born preterm control infants had an Amiel-Tison neuromotor assessment (15) and a Griffiths neurodevelopmental assessment (16) at 6 mo to 1 y, performed by a developmental pediatrician. In the remaining infants (five control term and three control preterm infants), follow-up information was obtained from hospital notes and local health visitors

Data analysis.

EEG records from all infants were examined and background activity was graded according to postconceptional age. EEGs were graded according to the scheme outline in Table 1. The diagnosis of an electrographic seizure required the evolution of sudden, repetitive, evolving stereotyped waveforms with a definite beginning, middle, and end (17). Infants were then divided into four groups according to postconceptional age and status (i.e. controls or high-risk infants) (Table 2). The CBFV and ABP signals from all records were low pass filtered at 20 Hz with a zero-phase Butterworth digital filter and the beginning and end of each cardiac cycle was detected using the foot of the BP pulse. Systolic, mean, and diastolic values of CBFV and BP were extracted for each cardiac cycle and their temporal sequence was interpolated with a third-order polynomial to produce a uniform time series with time intervals of 0.2 s.

Table 1 Summary of background EEG activity grades used for all infants
Full size table
Table 2 Four groups of infants studied and number of transients averaged in each group
Full size table

The beat-to-beat mean CBFV time series was then filtered with a zero-phase Butterworth digital filter at 0.4 Hz to remove any oscillations of the signal due to respiratory motion. Points in the ABP signals where spontaneous transient peaks occurred with corresponding increases in the CBFV signal were detected visually. Transients were only accepted if they were at least 6 s apart and their relative amplitude (peak to foot) was ≤2% of the baseline value (Fig. 1). The position of the maximum derivative of each BP peak was used as the point of synchronism for coherent averaging. Data periods for both the ABP and CBFV were synchronized about the transient so that each period contained 8 s before the transient and 12 s after. Each data period was then normalized with respect to the baseline value for that period and a coherent average formed from these periods for each group mentioned above. In high-risk infants, data periods were then subdivided into those with and without electrographic seizure activity.

Figure 1
figure 1

The beat-to-beat mean CBFV time series and ABP trace over 240 s from one infant. Points in the ABP signal where spontaneous transient peaks occurred with corresponding increases in the CBFV signal were detected visually (as indicated by arrows).

Full size image

Statistical analysis.

For each group of neonates a number of data periods of ABP and corresponding CBFV about a point of synchronization were collected as described above. Before analysis, ABP and CBFV waveforms were normalized with respect to their respective peak values (taken to be 1 s after the point of synchronization). A linear regression analysis was performed on each transient for a period from 1 s to 12 s after the point of synchronization (approximate peak in CBFV), i.e. the region where the response dependent on the current state of autoregulation would be seen. Distributions of ABP and CBFV gradients for each of the six groups were obtained. An unpaired t test was then performed between the transient CBFV gradient distributions obtained from the high-risk infants (term and preterm) during periods of seizure activity and periods without seizure activity. For both the term and preterm infants there was no significant difference (p > 0.05) in the means of the distributions obtained in the presence or absence of seizure activity. This justified pooling the transient CBFV gradients (for term and preterm infants) obtained during seizure and nonseizure periods. A one-way between-groups ANOVA test was performed on the CBFV gradient distributions for the resulting four groups to see if a significant (p < 0.05) difference between the means of the gradient distributions existed. Duncan and Scheff 233 tests (18) were performed on the CBFV gradient distributions to specify which of the means was significantly different. All calculations were done using the Statistica v5.1 software package (Statistica for Windows 1997, Release 5.1, StatSoft Inc.).

RESULTS

Twenty-five infants were studied. Thirteen infants with a gestational age range of 25–41 wk had seizures. Twelve control infants with a gestational age range of 26–42 wk were also studied. Two infants with a gestational age of more than 36 and a half weeks of gestation were included in the term control group. Clinical details of all infants are summarized in Table 3. No parent refused to allow their child to be studied and all infants tolerated the procedure very well.

Table 3 Clinical, EEG, and outcome details of all infants studied RDS, Respiratory distress syndrome; HIE, hypoxic ischaemic encephalopathy; CAM, congenital adenomatoid malformation; IUGR, intrauterine growth retardation; DH, diaphragmatic hernia; HC, haemachromatosis; CA, conceptional age; BW, birth weight.
Full size table

As outlined earlier, four groups were then formed according to gestation and status. Table 2 lists the number of transients used to compute the coherent average per group.

Cerebral autoregulatory response curves were plotted for the four groups and these are displayed in Figure 2. High-risk term and preterm infants had a passive autoregulatory response. Preterm control infants also failed to show intact cerebral autoregulation. Term control infants, however, showed an intact cerebral autoregulatory response.

Before performing linear regression analysis it was necessary to remove three transients (out of a total 282) from further analysis as they displayed extreme values of either ABP or CBFV gradients with respect to others in the data set (one transient each from infants no. 12, 13, and 15). Figure 3 shows the means and 95% confidence intervals of the CBFV gradient distributions for each of the four groups. The mean of the distribution of CBFV gradients for term control infants is significantly less than that of the other groups. This observation supports the conclusion drawn from the analysis of the morphology of the CBFV coherent averages, namely that term control infants exhibit an intact autoregulatory response.

Figure 3
figure 3

Graph showing means and 95% confidence intervals of CBFV gradient distributions of the four groups studied. CT, control term infants; CP, control preterm infants; HRT, high-risk term infants; HRP, high-risk preterm infants.

Full size image

The ANOVA test of F (3.49) and p (0.0008) showed a significant difference between the means existed. The Duncan test revealed that the mean of the term control gradient distributions was significantly different from the means of the three other distributions (p = 0.000007, 0.00157, 0.000453, respectively). The more conservative Scheff 233 test showed that the term control distribution mean also differed significantly from the other means, (p = 0.00165, 0.0396, 0.0050, respectively).

EEG findings and neurodevelopmental outcomes for infants in each group are also given in Table 3.

DISCUSSION

This study provides new evidence on the cerebral autoregulatory status of infants undergoing neonatal intensive care. Cerebral autoregulation is difficult to measure in neonates as it generally involves some disturbance to the infant to induce a blood pressure change (e.g. tilting (19) or deflating thigh cuffs (6). The method used in this study, devised by Panerai et al. (7), involves no disturbance to the neonate and relies entirely on the measurement of the CBFV changes evoked in response to spontaneous BP peaks. Linear regression analysis of the individual transients offers a new way of looking at the neonatal cerebral autoregulatory response and lends itself well to statistical analysis. Although this method, based on Doppler ultrasound, has the advantages of rapid temporal resolution using a noninvasive technique, the result is inevitably an indirect measure of cerebral autoregulation that will be difficult to validate in the newborn.

Using the coherent averaging technique and linear regression analysis of the response, we have shown that cerebral autoregulation is intact in term infants undergoing intensive care who are neurologically healthy. In comparison, neurologically healthy preterm infants undergoing intensive care have an impaired autoregulatory response. Both of these groups of infants were sufficiently unwell to require intensive therapy and all infants in each group required ventilation and an indwelling arterial line. Infants in both “control” groups had a normal EEG for postconceptional age at the time of study. The only difference between these two groups was postconceptional age. The finding of impaired autoregulation in neurologically healthy preterm infants was first described by Lou in 1979 using 133 Xe clearance to measure autoregulation (20) and later described by Panerai et al. (7) using dynamic measurements. Preterm infants are particularly susceptible to hemorrhagic and ischemic brain damage that may, in part, be due to loss of cerebral autoregulation. Other workers (21) have more recently suggested that preterm infants can preserve CBF even at low levels of systemic blood pressure, and inferred that autoregulation is intact in this group. However, in this study the static model was used to measure autoregulation, and data from a large number of infants was pooled.

The preterm infants used in the present study were tested within 2–3 d of birth and had no evidence of neurologic injury at the time and a normal cranial ultrasound scan. Two infants did, however, develop cerebral palsy. They developed neurologic complications 1–2 wk after the time of these studies. The remaining preterm infants and all term infants had a normal outcome.

Cerebral autoregulation was absent in all term and preterm infants who were neurologically impaired and had seizures. Autoregulation was absent both during the seizures themselves and between seizures (interictal period). During seizures the EEG showed clear electrographic seizure activity in all infants. In the interictal period, the EEG was always abnormal (Table 3). Infants with severely abnormal EEGs had the poorest outcome. Most infants in the high-risk term group of infants had seizures secondary to HIE. Only two had seizures due to other causes, which makes it impossible for us to extrapolate our results to all term infants with seizures. The finding of impaired autoregulation in infants with HIE has already been described, particularly when associated with a severely abnormal EEG (22) and our findings confirm this. The number of high-risk preterm infants with seizures studied was very small (3) and it would be speculative to comment on the implications of an impaired autoregulatory response in this group, particularly as it seemed that autoregulation was not present in preterm control infants. We found it difficult to obtain high numbers of preterm infants with seizures who already had an indwelling arterial catheter for blood pressure measurement.

In conclusion, this study has provided new evidence about dynamic cerebral autoregulation in infants undergoing neonatal intensive care. Term infants requiring intensive care, but without neurologic impairment, had an intact autoregulatory response. Preterm infants and all infants with neurologic impairment and seizures had an absent autoregulatory response. Further work using this technique might allow identification of the aspects of intensive care management that contribute to autoregulatory impairment, e.g. repeated estimates of the autoregulatory response could reveal the influence of drugs, BP, or sepsis on presence or absence of autoregulation. This could lead to therapeutic manipulations aimed at minimizing the duration of cerebral autoregulatory impairment, or in assisting recovery.