UK trial of extracorporeal membrane oxygenation gave biased estimate of efficacy

BMJ 1999; 318 doi: https://doi.org/10.1136/bmj.318.7185.738a (Published 13 March 1999) Cite this as: BMJ 1999;318:738
  1. Frank Shann, Director of intensive care. (shannf{at}cryptic.rch.unime0lb.edu.au)
  1. Royal Children's Hospital, Melbourne, Victoria, 3052, Australia

    EDITOR—Roberts et al say that extracorporeal membrane oxygenation lowers neonatal mortality at an acceptable cost.1 Their analysis, however, is based on the United Kingdom trial of extracorporeal membrane oxygenation, in which babies were randomised to receive either conventional mechanical ventilation in one of 55 neonatal units or extracorporeal membrane oxygenation in one of only five intensive care units.2 The five units providing extracorporeal membrane oxygenation had substantially more facilities than many of the 55 neonatal units.

    Fewer children died in the group given extracorporeal membrane oxygenation, but we cannot be sure that this difference was because of the extracorporeal membrane oxygenation. Mortality may have been lower in this group because of better care in the units that provided extracorporeal membrane oxygenation: there is substantial evidence that centralisation of intensive care services for children reduces mortality.3 Indeed, the relative risk of 0.55 in the trial of extracorporeal membrane oxygenation is strikingly similar to the odds ratio of 0.48 in our comparison of paediatric intensive care in Victoria (centralised) with that in Trent (decentralised).3

    There were good practical and ethical reasons for the design of the United Kingdom study of extracorporeal membrane oxygenation. The decision to leave control babies at the referring hospital (rather than sending all babies to the centres for extracorporeal membrane oxygenation) means, however, that we do not know whether the observed difference in mortality was due to extracorporeal membrane oxygenation or to other differences in management.


    1. 1.
    2. 2.
    3. 3.
    View Abstract

    Log in

    Log in through your institution


    * For online subscription