Prediction of cognitive abilities at the age of 5 years using developmental follow‐up assessments at the age of 2 and 3 years in very preterm children

Longitudinal research which maps cognitive recovery in the context of conditions involving potential brain compromise, such as very preterm birth, has a number of important goals: (1) informing families and health professionals about the relative risk of persisting problems; (2) identifying residual impairments which may be amenable to treatment; (3) establishing factors that predict poor outcomes, and thus target children for intervention; and (4) contributing to our theoretical understanding of the impact of early brain insult for ongoing development. Despite the importance of such endeavours, there are significant challenges for researchers undertaking this work. As illustrated by Potharst et al.1, children born very preterm provide an excellent opportunity to address all of these issues. As has been demonstrated by many groups, prospective cohorts of infants born very preterm can be readily established at birth, with families eager for monitoring of their child’s development. Further, as preterm birth is relatively common, such cohorts are generally sufficiently large to address complex research questions, and to cope with the unavoidable problem of sample attrition. Potharst et al. take advantage of these opportunities to follow a substantial sample (n=102) over 5 years, and, to comprehensively describe sample characteristics, thus reassuring the reader regarding potential for sample bias. In this context, they discuss issues that plague longitudinal child health research and threaten the potential for translation of this work into clinical practice and optimal child outcomes. In particular, they raise the issues of interest to those working with children born very preterm, but also those in the field of early brain insult more broadly: (1) when to test?; (2) what to use?; and (3) how to interpret findings? Researchers documenting the cognitive consequences of early brain insult continue to debate the question of optimal timing for assessment.2 On the one hand, later assessment, when brain development is more complete, allows the identification of the range of cognitive impairment, capturing deficits that may emerge later in development as more complex skills fail to develop as expected.3 In contrast, early assessment facilitates the prompt detection of impairments and allows for implementation of timely and appropriate treatments. Further, assessment at school entry may provide important information to guide decisions regarding education and special education needs. Potharst et al. contribute to this debate by comparing outcomes across a range of time points but more importantly, reporting on the ability of early assessments to predict later cognitive outcomes. Longitudinal research is difficult in adult populations and the challenges magnify with children. No single test can be administered across infancy, toddlerhood, and childhood.2 Early assessment must limit itself to broad outcomes, with many cognitive skills not sufficiently developed for standard measurement. Researchers must employ different cognitive measures at each time point, limiting the reliability of longitudinal comparisons. Potharst et al. address this issue, highlighting disappointingly low correlations between the same measure across time points and different measures across time points. They also identify significant variability in rates of impairment derived from these measures; for example, at 2 years 35% of children in their sample demonstrated at least one impairment, with a drop to 15% at 3 years, and then an increase to 50% at 5 years. While these differences may be argued to reflect the variability present in timing of skill acquisition in young children, recent studies also document the psychometric weaknesses of cognitive measures in these age groups.4 In considering their findings, Potharst et al. suggest cautious interpretation of results from longitudinal child health studies. A number of technical ‘take-home’ points are discussed for readers to bear in mind when evaluating such research findings. First, cognitive outcomes may look vastly different depending on age at testing. Second, care needs to be taken when interpreting differences in results across time points, where different assessment tools have been employed. Of broader significance, these authors extrapolate their findings to provide direction regarding the benefits of clinical assessment at different ages during childhood and in particular, the greater sensitivity of later assessments for detecting subtle, but functionally significant cognitive impairment.