Hi Ann,
I would need to say that the intention-to-treat principle has been violated.
The intention-to-treat principle states the
1) all patients are included in the analysis; and
2) patients are analysed according to the original randomised group.
The groups are formed by randomisation and so should (on average) be
balanced for known (and unknown) factors, if you violate that randomisation
by excluding patients because they were non-starters then the analysis
cannot be an intention-to-treat. The false-starters (those later found to be
non-eligible) can be excluded provided that the same screening process was
applied equally in the two groups and that they are a result of the trial
and could not happen in reality. For example, if you are comparing a
surgical intervention to a drug intervention and those in the surgical
intervention were found to be non-operable then they should not be excluded.
Unless you provide more information I can only be very general, but
generally any post-randomisation exclusions violate the intention-to-treat
principle. How you would deal with the drop-outs, false-starters and
non-starters is a different question.
kind regards
allan
----------------------------------------------
Dr Allan Clark
Lecturer in Medical Statistics
School of Medicine, Health Policy and Practice
University of East Anglia
-----Original Message-----
From: A UK-based worldwide e-mail broadcast system mailing list
To: [log in to unmask]
Sent: 11/11/04 15:01
Subject: intention to treat
Hi
In a recently completed study x patients were randomised, prior to any
treatment, y patients either withdrew or were found to be non eligible
under
clinical consideration- we therefore had z patients at baseline who
underwent the treatment and were followed up at two subsequent time
periods,
and at both time periods there were some additional drop outs
My understanding of intention to treat, is that those patients who were
randomised and received the treatment (called z patients above) were
included in the analysis under their original treatment allocation
irrespective of the actual treatment received
My problem is the patients who were randomised but never actually took
part
in the study, I do not have data relating to their characteristics e.g
age
at the time of the randomisation and have therefore not been included in
any
of the analysis
I do not consider , by ignoring in the analysis these original
withdrawals
that the intention to treat has been violated , but would appreciate
comments / thoughts
Ann E Fitzmaurice
Medical Statistician
University of Aberdeen
Department of Obstetrics & Gynaecology
Aberdeen Maternity Hospital
Cornhill Road, Aberdeen. AB25 2ZL
Tel: +44 (0) 1224 553876
Fax: +44 (0) 1224 553708
e-mail: <mailto:[log in to unmask]> [log in to unmask]
<http://www.abdn.ac.uk/dugaldbairdcentre/>
http://www.abdn.ac.uk/dugaldbairdcentre/
|