Bias caused by migration in case-control studies of prenatal risk factors for childhood and adult diseases.