We welcome the comments of Shukinami et al but it is important to realise the basic theoretical and practical difference between a risk factor and prediction of illness in the premorbid phase. When exposures are common (as are many obstetric complications) but incidence ratios of the illness are not high and outcomes fairly rare (as is schizophrenia), prediction of future disease is difficult.

Abnormal gestational age may or may not be a subtle risk factor for schizophrenia. This has been analysed in a recent meta-analysis (Reference Cannon, Jones and MurrayCannon et al, 2002) of eight prospective population-based studies of the association between obstetric complications and schizophrenia. Gestational age under 37 weeks was weakly associated with schizophrenia (odds ratio = 1.22, 95% CI 0.90-11.65). The results within the Northern Finland 1966 Birth Cohort were similar (Reference Jones, Rantakallio and HartikainenJones et al, 1998).

Our review mentioned abnormal foetal growth and development as a potential risk factor for schizophrenia, as did Cannon et al, but the predictive power of abnormal foetal growth is weak as it is a rather common phenomenon. Prediction in this situation is not easy at the population level. Our aim was to describe the best known risk factors for schizophrenia, which is why we did not conduct a detailed analysis of gestational age.

The references included in the letter of Shukinami et al suggest that the association of gestational age with other mental disorders may be stronger than for schizophrenia.