When can cerebral palsy be prevented? The generation of causal hypotheses by multivariate analysis of a case-control study.

Causal hypotheses for spastic cerebral palsy were sought by comparing a population based sample of 183 cases with 549 matched controls. A time-ordered multivariate analysis was used to distinguish confounders and consequences of disease from possible causes, which could be single factors or sequences of factors. Eighteen factors were identified as having an association with spasticity that did not arise by confounding with other identified factors nor as a consequence of the disease. Nearly half the cases (48.6%) but only 14.4% of controls experienced one or more of these factors, but no one factor was experienced by > 11%, and most by < 5%, of cases. Those factors identified as occurring before labour commenced affected 35% of all cases. The proportion of cases experiencing identified factors and the distribution of those factors between epochs varied with gestation of delivery and with description and severity of impairment. The possible timing of causes in cases without identified factors and the role of preterm birth and poor intrauterine growth are discussed. We conclude that there were many pathways to spastic cerebral palsy many of which could not be identified. Each contributed only a small proportion and many may have been multifactorial. Intrapartum initiation of the aetiological pathway was relatively unimportant, being likely in about 9% of cases, but the majority of pathways commenced predelivery.