Cerebral palsy (CP) is a disorder of movement and posture that results from a non-progressive disturbance to the developing brain. The condition encompasses a variety of aetiological pathways, pathogenic mechanisms, and clinical manifestations. Magnetic resonance imaging (MRI) provides an in vivo view of brain structure and plays a pivotal role in identification of the most likely pathogenic mechanism. Particular patterns of abnormality on neuroimaging can provide clues as to the nature, timing, and severity of the cerebral injury. Consequently, investigation of potentially preventable antecedent factors is likely to be most informative when performed in subgroups based on neuroimaging patterns. Clinically defined subgroups are less useful for this purpose as different types of adverse event may result in the same clinical pattern of CP and a similar aetiology may produce variable outcomes.
Three systematic reviews have amalgamated imaging data and determined the proportion of abnormal scans in children with CP; however, few of the included studies used population samples or included all CP subtypes.[2-4] More recently, CP registries have adopted comparable classification systems for imaging findings in CP. This has enabled comparison of imaging findings between contemporary population cohorts and has resulted in a small number of papers reporting associations between imaging patterns, birth gestation, and clinical variables.[5-8]
In this study, we aimed to describe the distribution of MRI patterns in a large population-based sample of children with CP and to examine associations between neuroimaging patterns, and antenatal and perinatal factors typically recorded by CP registries. The strength of evidence for causality was also considered, particularly with respect to strength of association, consistency of findings between studies, and biological plausibility. We built on previous work by using a larger, geographically defined cohort, including a range of antenatal and perinatal variables, in addition to birth gestation, and through the use of population comparison data. Our intention was that this work would set the scene for ongoing research, in which MRI characteristics, relationships with clinical outcomes, and causal pathways could be explored in greater depth.
- Top of page
These Australian data pertain to the largest known population cohort of children with CP and classified neuroimaging. Brain MRI findings in 593 children born between 1999 and 2006 were classified as normal or as exhibiting one of five patterns of abnormality that broadly reflect the pathogenesis of the cerebral abnormality. The use of a population cohort allowed assessment of sampling bias, and the decision to assess each child's imaging records without access to information on past history and clinical characteristics reduced the opportunity for the radiologists to be influenced by known associations between MRI findings and clinical information. The size of the cohort enabled exploration of associations between each MRI pattern and multiple antenatal and perinatal variables, whereas previous studies have reported associations only with clinical and birth gestation variables.
The diagnosis of CP is established clinically through the identification of signs and symptoms indicating motor impairment of cerebral origin. Brain imaging may be performed to rule out the possibility of alternative conditions that do not fall under the CP umbrella, and to provide information about the most likely pathogenic mechanism. Stratification of MRI patterns is, therefore, an ideal starting point for the exploration of causal pathways. An important finding of this study was that risk factors for CP that were already identified, including parity, plurality, birth gestation, place of birth, and condition at birth, had differentially stronger associations with particular MRI patterns. Although only a relatively small number of variables were included, there is a strong possibility that future research may identify groups of variables that share common causal pathways.
Children born earlier than 32 to 34 weeks' gestation are at a high risk of perinatally acquired intraventricular haemorrhage/periventricular haemorrhagic venous infarction and periventricular leukomalacia. Consequently, it was not surprising that WMI was found to be strongly associated with preterm delivery, very low birthweight, and multiple birth. Nor was it surprising that WMI was associated with low Apgar scores and the need for intubation, particularly in the least mature neonates. The strong association with birth at a tertiary hospital suggests that the increased risk of WMI was known before delivery in at least some cases. Interestingly, although WMI was observed in 71% of children born preterm, this pattern was also seen in 32% of children born at term. Our figure of 32% was the same as that reported from a population study of term-born singletons with CP, but higher than that reported by other groups.[5, 6, 8] At this stage, little is known about WMI in term-born children, apart from the fact that clinical outcomes appear to differ from the outcomes of children with WMI who were born preterm.[15, 16]
GMI was seen on MRI in 14% of all included children. Compared with all births in Victoria over the same time period, children with GMI were more likely to have signs suggestive of severe perinatal compromise, including low Apgar scores, need for resuscitation, and neonatal seizures. More children with GMI than with any other MRI pattern had 5-minute Apgar scores below 7 or below 4, experienced delay in establishing spontaneous breathing, experienced neonatal seizures, or were admitted to the intensive care or special needs nursery. Most children with GMI were born at term, had a birthweight over 2500g, and were a normal weight for gestational age. They were no more likely than children without CP to be born in a tertiary hospital, suggesting that the risk of brain injury was not recognized early enough to allow in utero transfer. Taken together, these associations suggest a perinatal timing for the brain injury, although additional perinatal data would assist in confirming a diagnosis of acute encephalopathy. Researchers from Utrecht and Hammersmith previously reported that more than 90% of infants with encephalopathy had evidence of perinatally acquired lesions on MRI and a very low rate of established antenatal brain injury or other confounding diagnoses. Notwithstanding the fact that antenatal factors may increase susceptibility to, or intolerance of, stress in labour, current evidence supports the view that antenatally established cerebral damage in infants with an acute encephalopathy is not common.
Nine per cent of the Victorian cohort had MRI evidence of a focal vascular insult. This MRI pattern was seen relatively more frequently in male children. An explanation for the poorer outcomes in males has been suggested by animal studies performed in the context of experimental stroke, in which females demonstrated reduced sensitivity to the effects of induced ischaemia compared with males.
Children with focal vascular insults were the most likely to be born to primiparous mothers. The finding of an association between primiparity and focal vascular insults supports the results of earlier studies of children with perinatal arterial ischaemic stroke. It has been suggested, however, that risk factors such as prolonged second stage of labour and pregnancy-induced hypertension/pre-eclampsia may be confounders in the association between perinatal stroke and primiparity. A 2005 case–control study from California identified chorioamnionitis, infertility, pre-eclampsia, and prolonged second stage of labour as independent risk factors for perinatal stroke, whereas the effect of primiparity lost statistical significance on multivariable analysis.
The focal vascular group accounted for the highest proportion of children who were SGA. A 2004 study from the Californian group found that intrauterine growth restriction was an independent risk factor for perinatal arterial stroke in neonatal infants who developed long-term motor impairment, but their rate of SGA of 13% was substantially lower than the 25% in our study. In terms of the condition after delivery of children with focal vascular insults, our findings are consistent with the observation made by Rutherford et al. that delivery of infants following perinatal stroke can sometimes be difficult and some resuscitation may be required, but the infant usually recovers sufficiently that nursery admission is not required, even though many of these infants present with isolated seizures. In our study, 36% of infants had neonatal seizures, a figure comparable to the 39% reported from the Californian cohort that included only children with motor impairment.
Ten per cent of children in this Victorian cohort had a brain malformation on MRI. Malformations were the only MRI pattern that was not relatively more common in males. Furthermore, multiparity (three or more previous births) was a strong risk factor for malformations, whereas there was no strong evidence for associations between multiparity and the other MRI patterns. Term birth and a low frequency of perinatal difficulties were characteristics consistent with other reports. In this study, children with malformations were those least likely to have had neonatal seizures and to require admission to the NICU or SCN.
The study has some limitations. First, we believe that our estimate of the proportion of WMI in CP, although higher than most other estimates from developed countries,[5, 8] is likely to be an underestimate. In keeping with the findings of an earlier pilot study from Victoria and a study from Quebec, preterm birth was associated with a reduced likelihood that MRI was available, most likely because in many children born preterm the diagnosis of WMI was established through serial cranial ultrasonography during the neonatal period. Furthermore, a proportion of those with ‘normal’ imaging may have had diffuse WMI that was undetectable using current imaging methods and qualitative assessment. Second, only a small number of variables were studied; in future studies we hope to increase the number of variables studied as potential risk factors for CP.