Intended for healthcare professionals

Rapid response to:

Papers

Neonatal examination and screening trial (NEST): a randomised, controlled, switchback trial of alternative policies for low risk infantsCommentary: “Switchback” allocationdangerous bends ahead!

BMJ 1999; 318 doi: https://doi.org/10.1136/bmj.318.7184.627 (Published 06 March 1999) Cite this as: BMJ 1999;318:627

Rapid Response:

Analysis by individual or by cluster?

Dear Sir

Analysis by individual or by cluster?

The trial reported by Glazener and colleagues (1) is of interest not
only for its substantive content but also on methodological grounds. The
Commentary by Deeks (2) alludes to a particular issue, that of the unit of
randomisation. Neonates were not randomised individually but in clusters
(wards). Yet the analysis ignores any potential clustering effect. Why
did the authors choose this approach? Adjusting for clustering is often
complicated and may lead to a loss of power. It would therefore be
helpful if the authors would explain whether there is a simple way of
being able to identify situations in which cluster randomised trials can
be analysed as if they were individually randomised trials without
jeopardising their validity. This would have implications for others
undertaking such trials in the future.

Yours sincerely

Diana Elbourne

(1) Glazener C, Ramsay C, Campbell MK, Booth P, Duffty P, Lloyd D,
McDonald A, Reid A. Neonatal Examination and Screening Trial (NEST): a
randomised controlled switchback trial of alternative polices of low risk
infants. BMJ 1999; 318: 627-631
(2) Deeks J. Commentary: "Switchback" allocation - dangerous bends ahead!
BMJ 1999; 318: 631-2

Competing interests: No competing interests

25 March 1999
Diana R Elbourne
Senior Lecturer
Medical Statistics Unit, London School of Hygiene and Tropical Medicine