Frequent readers of this blog will know that I have been critical of the promotion of pre-discharge MRI as a universal screening standard for very preterm babies. The positive predictive value of most findings on MRI at term-equivalent age is low, especially when you take into account that most significant findings will already be seen on prior head ultrasound. The PPV in particular of white matter abnormalities is well below 50% in almost all studies. Nevertheless, people like to look, and try to convince themselves that seeing more bad things in the brain of a very preterm baby must be a good thing to do. Hence we get articles like this one : Melbourne L, et al. Clinical impact of term-equivalent magnetic resonance imaging in extremely low-birth-weight infants at a regional NICU. J Perinatol. 2016.

The authors report results from 103 term equivalent cerebral MRIs in infants with birth weight less than 1000g. They report that they found new abnormalities in about half of the MRIs, not seen or suspected on serial head ultrasounds. Where the paper gets weird is that they asked a pediatric neurologist to look at the scans (US and MRI) and predict what the prognosis would be, and they then use this prediction as an outcome variable, even performing a statistical test to show that “simulated prognosis” was worse when you looked at the MRI.

Only 26 of the babies actually had a follow-up assessment, so it’s not really even worth talking about that part, except to say that they note there were a lot of “false positives”. The authors note that based on the MRI 14% of those 26 babies (that is, 3.6 of them(?)) were predicted to have cortical blindness, and none of them did. The one baby with cortical blindness was not predicted by the MRI.

How on earth a pediatric neurologist can give a  prognosis of the degree of developmental delay based on an MRI is beyond me. There is so little correlation between MRI findings for an individual baby and Bayley scores, (or between head ultrasound findings and Bayley scores for that matter) that to predict that an individual will have moderate delay, or mild delay or severe delay is impossible. To use that prediction as the outcome variable for this study is bizarre. Of the very few babies with follow-up there were 9 who had a Bayley (version 2) MDI more than 2 SD below the mean, of whom 4 were predicted from the MRI, and 12% of the babies (3, I presume) were predicted to have that low an MDI and were false positives. Which is pretty useless.

The authors end the abstract with this statement:

TE-MRI detects new abnormalities and impacts developmental prognosis in the extremely low birth weight, which supports its use despite the added financial cost.

and end the entire article with a similar conclusion.

There is no data about how the parents felt about the new abnormal findings of doubtful prognostic interest.

The MRI cost $1600, which is a large additional cost at the end of already costly hospitalizations. They note that this cost “is not negligible and should not be incorporated into practice without measurable advantages to providers, patients and/or families.” I agree with the patients and/or families part of that sentence; I don’t think we should be doing things to babies that cost money and disturb them and their families because of benefit to providers, though. They certainly have not demonstrated any benefit to anyone of routine MRI.

Basically the authors are saying we found more bad things, so that must be a good thing. I would say that they discovered imaging findings of uncertain significance, and low prognostic value, and that might equally be a bad thing.