A new publication and an editorial address some of the issues around whether we can predict outcomes in preterm infants.
This was a large cohort study from the NICHD network, the usual multi-multi-multi-author paper, 34 authors wrote this paper apparently, Hintz SR et al: Neuroimaging and Neurodevelopmental Outcome in Extremely Preterm Infants. Pediatrics 2014. There were 480 infants of less than 28 weeks who were enrolled in the SUPPORT trial (from 16 of the 20 SUPPORT centers), they had early head ultrasounds, late head ultrasound near term, and MRIs close to term. Then followed up with standardized neurodevelopmental testing. 10% had major anomalies on early ultrasound, 6% on late ultrasound. 20% had MRI white matter injury defined as moderate or severe, and 16% had cerebellar lesions seen on MRI.
The authors showed that if you have imaging brain abnormalities you are more likely to have adverse outcomes. There are many p-values less than 0.001.
Which is hardly surprising.
They show that early head ultrasounds are of little value when you take into account clinical factors and late findings. They also show that
a substantial proportion of children with adverse late Cerebral US or MRI findings in our cohort did not have severe adverse outcomes at 18 to 22 months, emphasizing that neuroimaging must not be used in isolation to predict outcomes.
To me this is the most important finding of this study.
But first lets look at early ultrasound findings: of those without a major finding on a 4 to 14 day ultrasound, there were 6% with “neurodevelopmental impairment”; of those with a serious hemorrhage or cystic PVL there were 28%. If we make that a ‘glass half-full’ statement, that means that if you have serious early ultrasound findings there is a 72% probability that you will not have “impairment”.
Late imaging findings were more strongly associated with “impairment”, but only the 18 babies who had severe MRI white matter injury were really any different in terms of outcomes (the babies with moderate abnormalities don’t seem much different to those with ‘normal MRI’), and among those babies, half had serious cerebral palsy and 22% had Bayley 3 cognitive scores below 70, (in other words, 78% of infants with severe white matter abnormality on late MRI did not have low cognitive scores on Bayley 3 testing).
The positive predictive value of cerbellar lesions was also very poor, 10% of infants with cerebellar lesions had moderate to severe cerebral palsy, and 15% had cognitive scores below 70.
Late head ultrasounds, (scored as abnormal in the presence of a shunt, moderate to severe ventricular enlargement, cystic PVL or a porencephalic cyst), were to my eyes, just about as good (or bad) as an MRI; with PPVs ranging from 23 to 50%.
What does all this mean in clinical practice? I think we should, as the authors of this study state “re-evaluate the practice of early head imaging” in very preterm infants. I think the data also suggest that we should “re-evaluate the practice of late head imaging”! We should continue to examine closely why we are doing such studies, what we want to get out of them, and how we inform parents, both for prior consent and after we have the result.
An editorial accompanying the study asks many of the same questions, as Eric Eichenwald states, “we need to understand the potential impact of of our predictive uncertainty on the parents of these vulnerable infants. …neuroimaging techniques.. cannot yet be used to determine follow-up strategies or target interventions after discharge and thus may be of little or no benefit to many parents.” (and he references the article by Annie and me!)
I think an objective evaluation of the usefulness of these tests for predicting outcomes in individual babies would lead to the inescapable conclusion that they are close to useless. Sometimes worse than useless, as they give false reassurance, or create false anxieties.
Both early and late head imaging fail to satisfy any of the criteria for a useful screening test. We should consider limiting early head ultrsound to use for detection of treatable lesions, and later imaging for research purposes only, with extensive informed consent to ensure that parents are well aware of the limitations of both MRI and near term head ultrasound.