Results

The total population of infants born at >31 weeks gestation in our hospital between 1987 and 1993 was 44 518. Over the same period, 100 babies of 32 weeks gestation or above (0.22%) were investigated for neonatal seizures. A clear cause for the seizures was elicited in 85 (85%); the relative prevalence of causes are shown in table 1. Hypoxic-ischaemic encephalopathy was the most common and neonatal cerebral infarction the second most common diagnosis. The 12 cases of neonatal cerebral infarction accounted for 12% of the total number of infants with seizures and 14% of those with an established diagnosis.

All of the cases were at least of 37 weeks gestation. The characteristics of the 12 individual infants with neonatal cerebral infarction are shown in table 2. Eleven presented with seizures on the postnatal wards and were admitted to the nursery for investigation; one was admitted because of respiratory distress and subsequently had seizures. Convulsions were focal and clonic in eight infants, generalised and clonic in three, and the remaining infant had subtle seizures.

All babies were alert and responsive between episodes and were not encephalopathic. Onset of seizures was between 12 and 76 hours, with a median of 37 hours. This may reflect short postnatal confinements, because babies who had a first convulsion after discharge would not have been identified in this study. Computed tomography between days 2 and 8 of life was abnormal in all 12 infants, showing areas of low attenuation (table 3). Most lesions were peripheral, and in only one case was involvement of the basal ganglia reported. In 11 infants the computed tomogram abnormalities were unilateral. Case 8 was included, although there was a possible area of decreased attenuation in the right temporal lobe as well as the predominant left frontal lesion. Cerebral ultrasound scans showed unequivocal unilateral echodensities in only one infant when first scanned, and repeat scans showed asymmetrical echodensities in two others. Figs 1A-C show the abnormal computed tomogram and cerebral ultrasound scan from case 10. Figure 2shows the abnormal computed tomogram from case 1. Electroencephalographic (EEG) recordings were carried out on 11 cases, and were frankly abnormal in five (42%), showing focal abnormalities in four. In most cases EEG and imaging results corresponded to each other and to the clinical features, but in two cases there was a discrepancy. Case 9 presented with left sided focal seizures: the computed tomogram showed an infarction in left middle cerebral artery territory, and the EEG was normal. Case 12 presented with left sided focal seizures and the EEG was consistent with asymmetrical and predominantly right sided abnormalities. However, the computed tomogram showed left cerebral infarction. In most cases seizures were easy to control, and in eight (67%) there was a prompt response to an intravenous loading dose of 20 mg/kg phenobarbitone.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Computed tomogram on day 4 (A); coronal and parasagittal ultrasound scans on day 3 (B) (C) from case 10. All show left frontal neonatal cerebral infarction.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Computed tomogram on day 3(A) from case 1 showing left parietal   neonatal cerebral infarction. Serial ultrasonography (B), (C) revealed no changes   in this case.

Perinatal risk factors for the 12 cases and the 24 controls matched for gestation, sex, and approximate date of birth, are compared in table 4. There were no significant differences between cases and controls with respect to maternal age, pregnancy complications, cardiotocographic abnormalities in labour, mode of delivery or umbilical cord acidosis. Cases had a lower Apgar score at 1 minute, and were more likely to require assisted ventilation for resuscitation (OR 7, 95% CI 1.04 −53.5), but by 5 minutes Apgar scores ranged between 8 and 10 and were no different from those of controls. There were no differences between cases and controls in relation to birthweight, platelet count, haematocrit or plasma biochemistry on admission to the neonatal unit.

Table 3 shows the neurodevelopmental outcome of the 12 cases of neonatal cerebral infarction. Latest follow up data were recorded in the hospital notes at 11-60 (median 27) months of age—in 11 cases by consultants and in one case by a paediatric lecturer. Three children had been discharged from hospital follow up at 11, 12, and 14 months and attempts were made to determine their subsequent neurological outcome. The family of the child who was discharged with no neurological abnormalities at 11 months (case 9) had moved to America and could not be traced. Cases 8 and 11 (last seen in hospital at 14 and 12 months) had been regularly assessed by health visitors and their general practitioners and had no overt neurological impairment at 41 and 33 months of age. None of the infants had a persisting seizure disorder, none was causing any concern with respect to hearing or vision, and only one had a major motor deficit. Case 6 had a contralateral spastic hemiparesis but was developing normally in other respects. There was no evidence of cognitive deficits in any of the cases of neonatal cerebral infarction, although no formal testing was carried out. Parents were counselled in the postnatal period about the possibility of motor deficits in the future, but no specific interventions such as physiotherapy were initiated. The hemiparetic infant started treatment when physical signs became evident.

Discussion

Among our babies, neonatal cerebral infarction is the second commonest cause of seizures in neonates over 31 weeks gestation. Our finding that 12% of all neonatal seizures in full term infants were due to neonatal cerebral infarction is similar to previously reported rates of 14%6 and 17.5%.5 Over the period of data collection, there was an overall incidence of 11/44 518 (0.025%) in our population. This is very similar to the reported incidence of 0.02% by Perlman.8 Our hospital population is not from a strictly defined geographical area and includes antenatal referrals but this would not have had a major influence on the prevalence calculation. We omitted the one case of neonatal cerebral infarction from outside our health district, and the effect of antenatal referrals on the denominator is small. This study and others have probably significantly underestimated the prevalence of neonatal cerebral infarction, because the diagnosis arises through imaging (usually computed tomography) which is only undertaken if there are seizures or other overt neurological signs. Many of the babies are well, without any evidence of neurological disturbance between seizures, so cases are likely to remain undetected if seizures do not occur, and may be missed or ignored if seizures are brief or subtle. Furthermore, the tendency towards short postnatal stays will lead to symptoms occurring at home and presentation to paediatricians and paediatric neurologists rather than neonatologists.

The varied and sometimes subtle symptomatology raise the possibility that some cases may not be detected in the perinatal period. If this is the case, neonatal cerebral infarction may be responsible for some cases of hemiplegic cerebral palsy in babies reported to have completely unremarkable perinatal histories. It is sometimes assumed that hemiplegic signs in infancy, which may be associated with contralateral computed tomography changes, are likely to be antenatal in origin unless there is a clearly documented history of perinatal insults. The computed tomograms in our cases showed acute changes rather than established atrophy, and the transient seizure disorders suggested an acute neurological event. Cerebral infarction can therefore occur in the few days before or after birth in healthy infants without overt underlying pathology.

This was not a prospective study, and the imaging and investigations were performed according to clinical need, rather than as part of a research protocol. Because the progress following initial discharge from hospital was usually benign, there was rarely an indication for a follow up computed tomogram. Sequential changes on magnetic resonance imaging (MRI) following neonatal cerebral infarction have recently been closely documented by Mercuri et al.4 MRI was not available to us during most of this period, and latterly we were reluctant to sedate or anaesthetise infants who were developing normally or who had a limited and predictable neurological deficit. We are therefore open to the criticism that the original computed tomography findings may have been misinterpreted, the diagnosis of cerebral infarction erroneous, and the good prognosis therefore much less surprising. However, all of our cases presented with clinical seizures, all the computed tomograms were reported by consultant neuroradiologists, and table 3 shows that there was ancillary evidence of a focal cerebral lesion in many cases. A good outcome has been reported by others7 9 10 following neonatal cerebral infarct, and there has even been a recent suggestion that MRI shows cerebral “regeneration” following infarction to explain this.11 Follow up images are, of course, not available to clinicians advising parents in the neonatal period. Our finding of a good medium term prognosis following presentation with seizures and a computed tomographic diagnosis of neonatal cerebral infarction is therefore clinically relevant, even considering the constraints of a retrospective review and lack of follow up imaging.

Although periventricular and subcortical infarction may be visualised very well by cerebral ultrasonography, we found that scanning with 5 and 7.5 MHz transducers was often unhelpful in the diagnosis of unilateral neonatal cerebral infarction. Ultrasound imaging may be normal even when interpreted in the light of information from a computed tomogram (fig 2). Others have suggested, like us, that ultrasonography has limitations in the diagnosis of neonatal cerebral infarction.5-12 However, Mercuri et al 4 found that 11 of 14 cases of neonatal cerebral infarction detected by MRI were also evident on ultrasonography. Koelfen et al 13 found that all six infants with computed tomographic evidence of neonatal cerebral infarction had abnormal ultrasound scans, and Perlman et al 6used ultrasonography as a primary investigation and only proceeded to MRI if ultrasound scanning suggested neonatal cerebral infarction. Allen and Riviello14 reported that “we feel confident in US scanning to the point of not doing a CT scan in the face of negative serial US scans”. Our data do not support this view.

The pathology of neonatal cerebral infarction is unknown, and few necropsy data are available because the babies rarely die. Some have suggested an embolic or thrombotic origin; these reports have been reviewed by Mannino and Trauner.1 Ment et al 3 reported a high incidence of perinatal asphyxia and proposed an hypoxic-ischaemic origin, while others1 15 have implicated birth trauma. Some infarctions clearly correspond to the vascular distribution of a major cerebral artery—usually the middle cerebral—but this is not always the case. Because of the similarities in presentation, clinical course, and outcome, we included neonatal cerebral infarction of various sites and did not confine this report to so-called “middle cerebral artery” infarctions. We did not see the unexplained predominance of left sided lesions reported by other authors.6-8

Most of the published reports of neonatal cerebral infarction concern very small numbers of patients. Lien et al 5have looked at risk factors for early onset neonatal seizures of any cause, but there are no previous published data comparing perinatal risk factors for neonatal cerebral infarction with controls. The case control data presented here concern a very small number of cases, and the analysis clearly has a very low power. As neonatal cerebral infarction is likely to have a multifactorial origin, it is not surprising that our study sheds very little light on possible aetiology. Resuscitation by assisted ventilation, either by “bag-and-mask” or endotracheal intubation, was more common in cases than in controls. In two cases endotracheal intubation was performed in the context of meconium stained liquor, and the umbilical cord artery pH was above 7.0 in all cases where it was measured. The postnatal course also suggests that the cases were not significantly asphyxiated as none had a low 5 minute Apgar score or a clinical course consistent with hypoxic-ischaemic encephalopathy. We used babies who had been admitted to the neonatal unit as controls because we wanted to compare haematological and biochemical results with cases. We felt it would be inappropriate to include as controls infants who had been admitted to the unit because of concerns about prolonged resuscitation, metabolic acidosis, or clinical suspicion of hypoxic-ischaemic encephalopathy. Our selection of “non-asphyxiated” controls will therefore have tended to increase any apparent excess of asphyxial markers in the cases of neonatal cerebral infarction.

Koelfen et al 13 have reported a small series with a high incidence of assisted delivery, but this was not seen in our study. We have no evidence of cocaine abuse in any of the mothers of our cases, and none of the 12 was a twin. Polycythaemia has been implicated as a cause of neonatal cerebral infarction.16 Neonatal texts vary widely in their recommended threshold for dilutional partial exchange transfusion in asymptomatic polycythaemia, and one justification for an aggressive approach is the risk of “cerebral sludging.” Interestingly, in our study neonatal cerebral infarction cases did not have a significantly higher mean haematocrit than controls. One baby with neonatal cerebral infarction had an initial haematocrit of 70 and received a dilutional exchange transfusion, but the rest all had an initial haematocrit of 60 or less.

The very low incidence of hemiplegic cerebral palsy at follow up was surprising. The parents were counselled that unilateral motor deficit of some degree was the most likely outcome. However, the previously published reports (summarised in table 5) show a wide range of outcomes.

de Vries and Levene12 summarised the data from published series and suggested that over 50% of infants with neonatal cerebral infarction are entirely normal at the age of 12-18 months, which is consistent with the data in table 5. However, our study shows significant motor deficit in one of 12 cases. One case was lost to follow up at 11 months but the rest were assessed at 20 months or more and judged to be normal, apart from one child with an obvious hemiparesis. This study confirms previous suggestions that cognitive development is unlikely to be seriously impaired following neonatal cerebral infarction. All of our cases, including the child with hemiparesis, were thought to have normal cognitive and sensory development when last seen, although they were not formally tested. Wulfeck et al reported that half of their cases showed a degree of language delay, although there was no significant delay in global cognitive function.18 Our study cannot exclude subtle cognitive or language deficits. We have established that 11 of 12 of these children are functioning without apparent disadvantage in the medium term, but problems may, of course, become apparent at school age. None has a persisting seizure disorder at present, although late recurrence of seizures after an eight year seizure free interval has been documented by Sran and Baumann.10

This seven year review of newborn infants admitted with seizures to a neonatal unit showed unilateral neonatal cerebral infarction to occur with a frequency of about 1 in 4000, and to be the second most common identifiable cause of seizures in infants of 32 weeks gestation and above. In our cases, the aetiology was usually unclear, and a small case-control study did not reveal any major risk factors for neonatal cerebral infarction. The prognosis in our patients was more benign than has previously been reported.