Article Text

Download PDFPDF

Is survival and neurodevelopmental impairment at 2 years of age the gold standard outcome for neonatal studies?
  1. Neil Marlow
  1. Correspondence to Professor Neil Marlow, Neonatal Medicine, UCL EGA Institute for Women's Health, 74 Huntley Street, London WC1E 6AU, UK; n.marlow{at}

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Designing perinatal trials is a continuing challenge. In the process, the choice of primary outcome is one of the critical decisions and one that will determine the necessary trial size and ultimately the success or failure. The primary outcome should be directly causally relevant to the intervention under study and the difference sought must be clinically relevant. Alongside the test of efficacy, there is also a need to ensure that a treatment is safe. Few trials are now designed without a measure of developmental outcome at 18–24 months as a primary or coprimary outcome.1 This is a complex outcome, being a composite usually of death and four to five domains of developmental impairment, based on value judgments as to severity of impairment in the areas of motor, cognitive, hearing and visual functions, and sometimes communication. I would like to address a range of issues with the use of this as a composite outcome and the extent to which we should rely on it to show benefit when we are assessing research based evidence about interventions.

This paper discusses the relevance of 2-year outcomes in several trials and develops some ideas as to how we might consider 2-year outcomes, in terms of their interpretation and in the implementation of trial data into practice.

Robustness of 2-year outcomes

Recently I reviewed the challenges in measuring 2 year outcomes in research practice.1 They pose real issues in quality control, diagnostic accuracy and interpretation of the measures themselves in practice. Their predictive value for an individual is relatively poor, except for the most severe impairments.2 Most studies use a combined severe and moderate outcome classification similar to that published by British Association of Perinatal Medicine.3 We can improve the predictive accuracy by reducing noise in the measure from variation in assessment technique or by statistically correcting …

View Full Text

Linked Articles

  • Fantoms
    Martin Ward Platt