Predicting outcomes, but not before birth

One very important article that I failed to blog about last year when it came out is this one (Ambalavanan N, Carlo WA, Tyson JE, Langer JC, Walsh MC, Parikh NA, Das A, Van Meurs KP, Shankaran S, Stoll BJ et al: Outcome trajectories in extremely preterm infants. Pediatrics 2012.) It refers to an outcome prediction tool that can be applied after birth, that takes into account many different prognostic factors.

I think this is on the right lines. Prediction of outcomes before birth is so limited, so uncertain and so imprecise, except at the very extremes, that its value is limited. The only thing we can say with much certainty before birth is that if the baby survives their quality of life will be highly acceptable to them and their families, nearly 100% of the time.

So can we predict after birth, as time goes on, how a baby will turn out?

That is what this publication was all about. They took the data from the NRN of the NICHD (if you haven’t been into neonatology for long, that is the neonatal research network of the NIH institute for children) and constructed predictive models at 4 critical times. In the delivery Room, at 7 days, at 28 days of age and at 36 weeks post-menstrual age.  Those times were presumably chosen based on the structure of the NRN database.

At each time, by entering a number of different clinical factors you can calculate the probability that the baby will die, based on the relative impact of those individual factors, in previous babies in the database, on survival. Also a calculation of ‘neurodevelopmental impairment’, NDI, can be made and a combined likelihood of survival without ‘NDI’.

I think as far as predicting death is concerned, if the calculation gives a very high likelihood of death, then that might be an appropriate indication that it is time to mention the prognosis to the parents.

As for ‘NDI’, well I am going to get on my hobby-horse one more time (I am sure it won’t be the last):

A Bayley mental development index (MDI) at 18 to 22 months corrected age below 70 is not an impairment, and I wish the NRN would stop using this term. Most of the babies classified as ‘NDI’ are so classified because of a poor Bayley result. The majority of ex-preterm infants with a Bayley 2 are not cognitively impaired in the long term, as determined by IQ testing at early school age. The proportion of those with a Bayley (version 2) MDI <70 who are cognitively impaired, when tested later, varies between 20 and 33%. I think keeping parents informed about prognosis, and having on-going discussions with them is vitally important; but I really don’t know how to use the information about ‘NDI’, I think it would be a huge mistake to tell a parent that there is X risk of neurodevelopmental impairment, when much of that risk is due to developmental delay which will likely substantially improve with time.

There is an on-line calculator that you can use, but beware of the calculation of NDI, not only for the reasons I have given above, but also because it varies a lot in other populations. A publication from the Melbourne group in 2012 compared their results to what the previous NICHD calculator, (designed to be applied around birth and therefore only including sex, birth weight, steroids, multiple birth and gestation) came up with. The Melbourne groups results for mortality were similar but for developmental delay their figures were substantially lower. That may be in part because of the use of 24 month Bayleys by the Melbourne group, rather than 18 month score from the NICHD. The Melbourne group have already shown that there is an improvement in scores between 18 and 24 months (of around 4 points on average). As a result the 24 month scores are a bit more predictive of long term outcomes than the 18 month scores, but if you look at the graphs in that paper, you can see that they still are not very good as predictions. But most importantly, different populations, with different backgrounds may have differing long term outcomes, either because Australian babies are just much tougher, or for a whole host of other reasons.

Also to illustrate my point, in the Melbourne paper, at 24 months there were 4 babies who had ‘severe delay’ (which is an appropriate terminology) that is, a Bayley 2 MDI score below 50. Only one of these infants was severely impaired on IQ testing at 8 years. Even very, very low Bayley scores, and even at 24 months, are very poorly predictive of long-term impairments. We should be extremely careful about using outcome data based on early developmental testing for counseling or decision making.

I think we need confirmation and expansion of this kind of prognostic information, from other networks if possible, to refine the prognoses, especially for mortality, and if at all possible, for really important truly long term outcomes.

About Keith Barrington

I am a neonatologist and clinical researcher at Sainte Justine University Health Center in Montréal
This entry was posted in Neonatal Research and tagged , . Bookmark the permalink.

1 Response to Predicting outcomes, but not before birth

  1. N. Ambalavanan MD says:

    Excellent points. Of course, many centers have now moved on to the Bayley III, which seems to identify a different subset of infants. Many studies evaluating long-term cognitive outcomes unfortunately (to date) do not have prospectively-collected data on what may be important predictors (parental IQ, early educational environment). As a related issue, even school-age IQ is not very good at predicting adult outcomes. However, developing models using late outcomes is not optimal, as clinical practices and outcomes probably change over time as well (outcomes of 24-28 week infants from the 1970s is probably different from that in the 2000s). We still have a long way to go!

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.