The title of this post I stole from the title of a newly published article (Mann PC, Woodrum DE, Wilfond BS: Fuzzy images: Ethical implications of using routine neuroimaging in premature neonates to predict neurologic outcomes. The Journal of pediatrics 2013, 163(2):587-592).

The commentary is a critique of the common practice of performing routine head ultrasounds in the first few days of life to predict long term outcomes in very preterm babies.

As I have noted here before, there is very little good data to support the practice. Head ultrasounds are poorly predictive of neurological or developmental problems, and are of no proven value for prediction of serious impairments.

In addition the large majority of studies that have compared head ultrasound results with outcomes have examined the children much too early to make good long term predictions, and (you might be able to guess what I am going to say next) a low 2 year score on the Bayley MDI scale is not an impairment! There is a recent meta-analysis of all the data that the authors could find comparing Bayley scores, mostly performed around 24 months, to later outcomes. They showed that the MDI correlated poorly with later testing, and that variation in MDI explained only 37% of the variation in IQ.

The commentary refers to one of the few follow-up studies of a largish cohort of very preterm babies who had intracerebral hemorrhages (referred to in the original article as periventricular hemorrhagic infarction, and in this commentary as grade 4 IVH) who examined the children at school age. That study from Roze and colleagues in Groningen in the Netherlands, including the extremely productive Arie Bos, showed that although many of the children had a diagnosis of cerebral palsy (16 of the 21 children) there were only 3 who had a GMFCS of 3 or worse. Meaning that the remainder were able to walk. Only 2 of the 21 had a cognitive outcome worse than 2SD below average. These 21 children were from an initial number of 38 infants who had periventricular hemorrhagic infarction, the remainder we are told ‘died in the neonatal period’. Now what that might well include are a number of infants who had withdrawal of life support because of the head ultrasound findings. It is possible that the infants with the worse appearance on head ultrasound were among that group, so we can’t in any of the studies really know what is the prognosis of all babies with this type of lesion, because there are some non-survivors in all of the studies, and it is rarely made clear if life support was withdrawn. In fact the group from Groningen have published a previous report which seems to include the infants in their newer paper, with a follow up to 24 months of age. In that paper they note that there were only 2 infants who died as a result of withdrawal of life support, so it is not such an issue for this cohort, but in some cohorts it could be much more important.

Despite that limitation there are now several studies that show very little effect of intracranial bleeds on outcomes of very preterm babies; including for example another paper by Roze and Bos, a study which followed 106 very preterm babies to school age, and showed ‘The children with cerebral lesions (grade III intraventricular haemorrhage and periventricular haemorrhagic infarction) had similar scores to those without cerebral lesions on total IQ and the Movement ABC’. Also interestingly ‘IQ scores did not correlate with gestational age’ the babies who were born at 24 and 25 weeks had scores not different to more mature babies.

To go back to the commentary by Mann, their introductory section ends with this sentence

We encourage neonatal practitioners to reconsider whether the perceived screening benefits are valid and the prediction of NDI definitive.

I think they present plenty of reasons to make that sentence much stronger: ‘screening head ultrasounds are of no proven value as predictors of important neurological or cognitive impairments, and it is therefore ethically questionable to use them for decisions regarding life sustaining interventions.  They should be performed only to screen for treatable lesions such as post-hemorrhagic ventricular dilatation.’