Brett Manley, and a group of the CAP investigators, including yours truly, have just published an article about the long term cognitive testing results of the trial subjects. (Manley BJ, Roberts RS, Doyle LW, Schmidt B, Anderson PJ, Barrington KJ, Bohm B, Golan A, van Wassenaer-Leemhuis AG, Davis PG et al: Social Variables Predict Gains in Cognitive Scores across the Preschool Years in Children with Birth Weights 500 to 1250 Grams. The Journal of pediatrics 2015). The 5 year outcomes of the CAP trial, if you remember, were interesting in showing that only 18% of the babies who had a Bayley 2 score at 18 months corrected age which had been below 70 still had cognitive testing scores (on the WISC-R) less than 70 at 5 years of age. The babies who had the lowest scores on the Bayley had, overall, the greatest “improvement” in their scores at 5 years of age.

You may also remember that the primary outcome was no longer different between the caffeine and control groups at 5 years, although there were some differences between the groups, which still favoured the caffeine exposed infants. The average IQ score of the whole group of infants at 5 years of age was 98.9, almost identical to the population standardised mean.

We thought that the changes in scores, and the lack of difference between the groups when they reached 5 years, was probably evidence of the variable influences of environmental factors, which become much more important as babies age, so we used those social variables that we had collected in the CAP data set to see if they were associated with the changes in cognitive scores.

Obviously the tests that were used were not the same at 2 years and at 5 years, a developmental evaluation at 2 years, and a more formal test of “intelligence” at 5 years, but, with that proviso, we were able to show that much of the difference in test scores was explained by the contribution of a number of social variables, which actually all had additive effects. The most important social variables were paternal education, maternal education and parental employment. Having two parents in the home was important on individual analysis, but dropped out when combined with the other factors in the model.

Our data confirmed that developmental testing at 18-20 months is very poorly predictive for cognitive abilities at 5 years, among very preterm infants. Which is entirely consistent with other published data.

One important implication of this is that we need to be very circumspect making treatment decisions or therapeutic choices based on predictions of 18-20 month developmental outcomes.

It also strongly suggests that we should focus programs to try and improve the long term function of very preterm infants on those with more social limitations. Consistent with this suggestion is the recognition among several results of trials of early intervention programs that benefits are difficult to demonstrate, except among those with more social disadvantages. Presumably, families with more resources (of all kinds) may well already be able to provide the kind of environment which aids the very preterm infant to achieve their potential, those with more difficulties need more assistance to end up in the same place.

I think in the future we should find ways to screen families in neonatal follow-up to identify those that will most benefit from intervention aimed at improving outcomes, the families most likely to benefit are likely to be those who have more limited resources.