Post-randomisation exclusions: the intention to treat principle and excluding patients from analysisBMJ 2002; 325 doi: https://doi.org/10.1136/bmj.325.7365.652 (Published 21 September 2002) Cite this as: BMJ 2002;325:652
All rapid responses
Rapid responses are electronic comments to the editor. They enable our users to debate issues raised in articles published on bmj.com. A rapid response is first posted online. If you need the URL (web address) of an individual response, simply click on the response headline and copy the URL from the browser window. A proportion of responses will, after editing, be published online and in the print journal as letters, which are indexed in PubMed. Rapid responses are not indexed in PubMed and they are not journal articles. The BMJ reserves the right to remove responses which are being wilfully misrepresented as published articles or when it is brought to our attention that a response spreads misinformation.
From March 2022, the word limit for rapid responses will be 600 words not including references and author details. We will no longer post responses that exceed this limit.
The word limit for letters selected from posted responses remains 300 words.
In an interesting article on post-randomisation exclusions from
intention to treat analysis (ITT) in randomised controlled trials,
Fergusson et al discuss the issue of false inclusions, ie. subjects
found after randomisation not to satisfy the entry criteria. This is a
particular problem in population based screening intervention trials, and
we have encountered the problem in two studies. In the Breast Screening
Programme, GP Prior Notification Lists (PNLs) are often used as the basis
for randomisation. Screening Offices issue invitations from the PNLs,
which list women deemed eligible for screening. This register is
constantly being updated because of new information sent from the FHSA,
GPs and other sources, so it is inevitable that further ineligibility will
be detected after invitation. We describe two RCTs 2 3 in which
information about ineligibility only came to light after randomisation,
causing subsequent problems with analysis.
In the first RCT, investigating the effectiveness of a home visit
intervention on the uptake of breast screening 2 among Asian women, a
study population of 527 women were randomised from PNLs. Subsequent
updated information from the FHSA , before invitation, found 29 women to
be ineligible: intervention group (17), control group (12). These women
were not therefore sent an invitation by the Office, were non-participants
in the screening programme and so were excluded from the analysis. During
the home visits, a further 62 women were discovered to be ineligible (34
not resident at address, 28 returned to Asia/visitor only). These women
had been found because of the intervention and so could not be excluded
from the analysis, which was conducted on an ITT basis for 498 women.
In the second example, the effect on attendance for breast screening
was investigated according to whether the offer of a cervical smear test
was included along with the breast screening invitation (1065 vs 1066
women) 3 . After invitation the Screening Office was notified that 219
(10%) women were ineligible for breast screening (100 vs 119 in each
group), because they were either not resident at the address (183), had
died (9), been screened recently (5) or were too ill to attend (22). The
women were false inclusions resulting in a missing outcome measure. It was
assumed that these reasons for ineligibility were unrelated to the
invitation for screening; they were evenly distributed between the two
groups. Since a main objective was to assess detriment to breast screening
uptake when invitations for cervical screening were sent in advance, these
219 women were excluded from the analysis, because this approach resulted
in the largest difference in attendance between the groups.
These RCTs illustrate the difficulties inherent in designing
intervention trials to fit in with the organisation of large screening
programmes. Our studies add to the debate on appropriate analysis
Dr Tanya Hoare
Dr Gill Lancaster
Authors: No competing interests.
Dr Tanya Hoare
Dept Health Care Studies,
Manchester Metropolitan University,
Manchester M13 OJA
Dr. Gill Lancaster
Lecturer in Medical Statistics
Medical Statistics Unit,
Division of Statistics and OR,
Department of Mathematical Sciences,
M & O Building,
University of Liverpool
1. Fergusson D, Aaron S, Guyatt G, Hebert P (2002) Post-randomisation
exclusions: the intention to treat principle and excluding patients from
analysis BMJ 325 652-4
2. Hoare T, Thomas C, Biggs A, Booth M, Bradley S, Friedman E.
(1994) Can the uptake of breast screening by Asian women be increased? A
randomised controlled trial of a linkworker intervention. J Pub Health Med
3. Lancaster G, Elton P (1992) Does the offer of cervical screening
with breast screening encourage older women to have a cervical smear test?
J Epid & Comm Health 46 523-7.
Competing interests: No competing interests