Neonatal risk factors for cerebral palsy in very preterm babies: case-control study

Selection of subjects

All the selected babies were singletons of less than 32 completed weeks of gestation who survived to hospital discharge, born to mothers resident in Oxfordshire and West Berkshire between 1984 and 1990. Multiple births were excluded from this study as current evidence suggests that the risk factors for cerebral palsy in this group may differ from those in singleton births.11

Gestational age for all groups was estimated by using a combination of menstrual dates and an ultrasound scan performed before 20 weeks' gestation. The scan date was preferred if the menstrual date was uncertain or there was a discrepancy of more than 14 days between the menstrual date and the scan estimate.

Cases–Fifty nine children with cerebral palsy were identified from the Oxford region register of early childhood impairments.12 The definition of cerebral palsy used by the register is that of a permanent impairment of voluntary movement or posture presumed to be due to permanent damage to the immature brain. The register includes children of mothers who were resident within the former Oxford health region at the time of delivery. Multiple sources of ascertainment are used to compile the register, and the condition of the children is determined at the age of 3 and 5 years.

Controls–A total of 474 babies who survived to discharge and did not develop cerebral palsy were identified from two sources to ensure maximum ascertainment: the hospital admission registers and the Oxford record linkage study. On the basis of an audit of preterm birth at the John Radcliffe Hospital (unpublished data) we estimated that an adverse neonatal factor would be present in 25-30% of controls. We predicted that a study population of 59 cases of cerebral palsy with four controls for each case would be large enough to detect for each neonatal factor an odds ratio of 2.5 with 80% power and an α level of 0.05. We wished, therefore, to select approximately half of the potential controls, and 235 controls were thus randomly selected. Controls were selected from the entire geographical population of very preterm babies (<32 completed weeks' gestation) but were not matched with the cases for gestational age. An unmatched case-control study design allowed us to investigate the relation between gestational age and cerebral palsy among very preterm babies.

Data sources

The neonatal notes of babies included in the study were reviewed by a researcher unaware of the children's outcomes. A detailed dataset was completed, encompassing 52 variables including characteristics at birth; cardiovascular, respiratory, systemic, and metabolic complications; neurological sequelae; and cerebral ultrasound findings. Patent ductus arteriosus (clinical diagnosis supported by cardiac ultrasonography, requiring indomethacin or surgical ligation), hypotension (mean blood pressure <30 mm Hg on at least two occasions), blood transfusion for either anaemia or hypotension, prolonged mechanical ventilation (duration of at least seven days), pneumothorax (diagnosis confirmed by chest x ray, requiring insertion of chest drain), and sepsis (clinical diagnosis confirmed with microbiology, requiring antimicrobial therapy) were of special interest. Details of diagnosis, onset, duration, and management were recorded. Birth trauma referred to severe bruising or x ray evidence of a fracture.

Ultrasound data were included if at least two scans were available, the first recorded during the first week of life and the second as near as possible to six weeks after birth. This approach was likely to identify lesions developing in both the early and late neonatal periods. In fact most babies had daily scans for the first week and weekly scans thereafter, with additional scans if clinically indicated. Ultrasound scanners (Advanced Technical Laboratories) used were the ATL 300C until 1988 and the UM4 thereafter with 7.5 MHZ transducer heads. The findings were described by using a classification modified from a data sheet used in a neonatal trial (OSIRIS).13 The right and left cerebral hemispheres were described separately in terms of germinal layer or intraventricular haemorrhage, ventricular dilatation, parenchyma echodensity, and parenchyma cysts. Moderate ventricular dilatation was assigned where the ventricular index was above the 97th centile and hydrocephalus was assigned when the dilatation was more than 4 mm above the 97th centile, using a centile chart from an unpublished study by M Levene et al. Parenchyma cyst was an umbrella term used for any parenchymal echolucency suggesting a cavity. For the purposes of this study parenchymal echodensities and echolucencies were grouped together and termed parenchyma damage.

Antenatal and intrapartum data had been recorded from the obstetric notes of the mothers; these data were available from our earlier study.8 Factors included in the logistic regression model as potentially important confounders were antepartum haemorrhage, maternal infection, chorioamnionitis, prolonged rupture of membranes, pre-eclampsia, and the mode of delivery.

Ethical approval

The approval of the Oxfordshire and West Berkshire ethics committees was obtained before the start of the study.

Statistical methods

The odds ratio associated with a given factor estimates the risk of cerebral palsy given the factor relative to the risk of cerebral palsy without the factor. The 95% confidence intervals for crude odds ratios were calculated with the programme CIA.14 The odds ratios with adjustment for potential confounders were calculated by logistic regression, using the statistical package for the social sciences.15 As the study populations were unmatched for gestational age, in the first instance an odds ratio was calculated for each neonatal factor with adjustment for gestational age. Biologically plausible interactions of neonatal factors were investigated by entering variables into a logistic regression model in a forward conditional fashion. Only variables with a strong association with cerebral palsy remained significant (P<0.05) independent of the other variables. We included these factors in further regression models, looking for relations between neonatal factors and cerebral palsy independent of the presence of adverse antenatal factors and the delivery mode. There were no more than six variables in a logistic regression model at any one time. Trends in the rates of survival and cerebral palsy among survivors by gestational age were tested by χ² tests for trend.

Neonatal factors

An Apgar score of ≤3 at 5 minutes was significantly associated with an increased risk of cerebral palsy (odds ratio 5.3 (1.4 to 21) (table 1). Otherwise the two groups did not differ significantly in characteristics of the babies or condition at birth.

Cardiovascular and respiratory complications were common among these babies: 232 (79%) had at least one complication. Patent ductus arteriosus, hypotension, transfusion, prolonged ventilation, and pneumothorax were associated with cerebral palsy after gestational age was adjusted for (table 2). On forward conditional logistic regression of the cardiovascular and respiratory factors described, transfusion and pneumothorax were independently associated with cerebral palsy (odds ratios 2.2 (1.1 to 4.7) and 4.8 (2.2 to 10.8) respectively).

Sepsis, total parenteral nutrition, and hyponatraemia were associated with an increased risk of cerebral palsy (table 3). The numbers with hyponatraemia were small, however, and this association could be a chance finding. The association with sepsis was independent of other systemic, cardiovascular, or respiratory complications (odds ratio 3.3 (1.6 to 6.8)). The sequence of antenatal infection and neonatal sepsis was strongly associated with cerebral palsy, but this occurred in only a few subjects (table 4).

Neonatal seizures occurred in 28 (9.6%) babies and were associated with a highly significant increased risk of cerebral palsy (table 5). Cerebral ultrasound scans were available for a total of 239 (82%) babies, with a similar proportion for cases and controls. Isolated intraventricular haemorrhage was not associated with an increased risk of cerebral palsy, but there was a strong association between cerebral palsy and parenchymal lesions and ventricular dilatation. Retinopathy of prematurity (all grades) occurred more frequently among cases than controls, but the difference was not significant at the 5% level.

Neonatal factors controlled for antenatal and intrapartum events

Patent ductus arteriosus, hypotension, transfusion, prolonged ventilation, pneumothorax, sepsis, hyponatraemia, and total parenteral nutrition were associated with an increased risk of cerebral palsy after adjustment by logistic regression for gestational age, antenatal complications, and the mode of delivery (table 6). The only antenatal factors of importance in the logistic regression model were chorioamnionitis, any maternal infection, and mode of delivery.

Neonatal complications

Several cardiovascular, respiratory, and systemic factors investigated in this study of very preterm babies were associated with an increased risk of cerebral palsy. In earlier studies, hypotension, transfusion, and patent ductus arteriosus have been associated with periventricular leukomalacia,16 17 18 19 an ultrasound finding which predicts later handicap (especially cerebral palsy) more accurately than any other antecedent.20 21 22 23 24 25 26 27 28 29 Pneumothorax and prolonged ventilation have been associated with both periventricular leukomalacia17 19 30 and cerebral palsy.31 32 The findings in this study are consistent with these observations, supporting the hypothesis that cardiovascular and respiratory disturbances have a role in the aetiology of cerebral ischaemia in very preterm babies.

A second hypothesis concerns infection, and several studies have shown associations between neonatal sepsis and both periventricular leukomalacia17 18 19 and cerebral palsy.32 Our results support this hypothesis as neonatal sepsis and cerebral palsy were strongly associated even after other potentially confounding neonatal complications were adjusted for.

These findings suggest a role for several neonatal complications in the aetiology of cerebral palsy in preterm babies. The difficulty in interpreting these findings, however, lies in determining which neonatal factors are causes of cerebral palsy and which are consequences of earlier disturbances in the antenatal and intrapartum periods and already part of the outcome. Some neonatal factors, such as transfusion, may be markers of severity of neonatal illness or may be the consequence of a disabling cerebral haemorrhage. Our previous study of antenatal and intrapartum risk factors for cerebral palsy in very preterm babies found a strong association between maternal infection and, in particular, chorioamnionitis and an increased risk of cerebral palsy. Maternal infection occurred, however, in only 37% of cases and 17% of controls and is likely to explain only a proportion of cases of preterm cerebral palsy. It is possible, therefore, that the origins of cerebral palsy lie in the neonatal period for a large proportion of very preterm babies. In addition, because of the design of a case-control study it is not possible to predict the timing of cerebral damage in relation to the insult and it is possible that the ischaemia associated with chorioamnionitis is not manifest until the neonatal period and may occur only if the baby suffers an additional further insult in the neonatal period. Our finding that the sequence of maternal infection followed by neonatal sepsis was strongly associated with cerebral palsy lends some strength to the theory of a continuum of insults in the pathogenesis of preterm cerebral palsy. However, this sequence of events affected only a small proportion of the study population.

Timing of brain injury

The question of the timing of cerebral damage was addressed in a study of the ultrasound findings and clinical events of preterm babies with cerebral palsy.4 Although antenatal complications were common, only a small proportion of babies (2/18) had evidence of antenatal cerebral damage; most had evidence of parenchymal damage of neonatal onset. These observations suggest a role for neonatal complications in the pathogenesis of preterm cerebral palsy.

Grether et al,9 in a recent study of prenatal and perinatal factors and cerebral palsy in very low birthweight infants, considered the interaction between prenatal and neonatal events and found chorioamnionitis to be associated with cerebral palsy only when seizures occurred in the neonatal period. With our previous study we had a full dataset of antenatal, intrapartum, and neonatal factors and could examine the independent associations between neonatal complications and cerebral palsy by controlling for the presence of antenatal and intrapartum factors. Transfusion, prolonged ventilation, pneumothorax, and sepsis had strong associations with cerebral palsy, adding further support to the hypothesis that cardiorespiratory disturbances and infection contribute to the aetiology of cerebral palsy in at least a proportion of cases and, more particularly, that this contribution can be independent of antenatal and intrapartum disturbances.

Cerebral lesions

As in previous studies of periventricular leukomalacia and cerebral palsy, we found strong associations between neonatal seizures, ultrasonically diagnosed parenchymal damage and ventricular dilatation, and preterm cerebral palsy. It is possible, of course, that cerebral ultrasound lesions arise incidentally as a result of severe physiological disturbances which in themselves cause cerebral palsy. This is unlikely, though, because ultrasound findings are so much more predictive of cerebral palsy than are cardiorespiratory complications. The statistical power of this study was limited for assessment of the complex interaction of these neurological factors and antenatal, intrapartum and neonatal events, and the results of multivariate analyses could be misleading. These interrelationships could be evaluated, however, with combined data from multiple sources.

In conclusion, we suspect that cerebral palsy has multiple risk factors, both causes and modifiers, but that a proportion of cases of cerebral palsy among very preterm singletons have their origins in the neonatal period. It would seem, therefore, that a major reduction in cerebral palsy among very preterm babies will arise only from an integrated approach throughout the antenatal, intrapartum, and neonatal periods to the management of any baby at risk. The possibility that new neonatal interventions may lead to a reduction or an increase in the frequency of cerebral palsy among very preterm babies can be tested by well designed randomised controlled trials.