 |
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
Re: st: Re: applying sampsi: sub-groups
Many thanks to Paul Seed, Carlo Lazzaro, and
Peter Lachenbruch for your astute insights. Yes,
Paul, it is your third example (Different
treatment according to diagnosis) that is what
I'm seeking. It's a fertility intervention with
modest differences in treatment technique, and
all with the same outcome: successful
implantation after IVF. The central question I'm
deliberating is whether the sub-groups need
separate power calculations.
It's very kind of you to have responded.
Michael
Dear all,
Michael McCulloch recently asked about a particular randomised
controlled trial (RCT) he is planing in which
the active treatment is to be given at
several different doses , and Carlo Lazzaro compared it to a published paper
In my experience, multiple planned doses in a RCT
can come about in 3 different ways, requiring different
handling:
1) Dose determination.
Effectively a 4 armed study (placebo, low, medium, high).
Allowance for multiple testing may be needed
(e.g. adjustments to p-values), depending on the
planned analysis, as declared in the protocol
& data analysis plan.
2) Complex intervention (usually unblinded).
Following an initial period on standard dose,
the dose may be raised, lowered, supplemented by
"rescue medication" etc., all according to a predefined plan.
This can be treated as a simple 2-arm RCT, where one of the arms is
a treatment plan, rather than a drug.
Patients are analysed according to their original
randomised treatment, irrespective of later changes.
This is known as the intention to treat principle &
applies in the main analysis of all Stage III RCTs.
Per-protocol analysis, allowing for changes,
may be used in addition, but it should not be
the main analysis.
3) Different treatment according to diagnosis at trial entry.
It is effectively 3 RCTs for 3 different patient groups.
It is therefore essential to determine the diagnostic
category for both active & placebo arms, and in the analysis
to compare only like with like.
For sample size purposes, each of the three
sub-trials needs to be fully powered - so if the answer is
60 subjects per group, complete data on 360 participants
(=6*60) are needed.
However, it might well be that the diagnostic groups require separate
sample size calculations -
e.g. if one is harder to recruit,
if the likely effect of the intervention is different
if the planned primary endpoint is different
or even if the standard deviation of the measure is different between groups.
This appears to be what Michael is dealing with.
However, Carlo draws attention to a fourth option,
(the MOSAIC study) in which "usual care" is determined
by the physicians according to their own preferences.
This strikes me as an awkward design, particularly if there are
substantial differences between the 3 forms of usual care,
and I am not surprised at the criticisms that Carlo implies.
There may of course have been practical reasons why this
was the only way the trial could be organized.
.
Inviato: mercoled� 6 febbraio 2008 23.09
A: Statalist
Oggetto: st: applying sampsi: sub-groups
I am applying sampsi to a project design, and am puzzled by a
question which perhaps one of the list members may have an opinion.
In my study design, subjects are randomized to one of two groups,
treatment and control. Treatment group receives the new intervention,
and control groups receives standard care.
My question is the following: after random allocation, within the
treatment group subjects will be assigned to one of three treatment
doses of the same treatment, depending on a differential diagnosis.
In practice, in estimating sample size requirements, would one
consider the three sub-groups as distinct groups, or aggregate them
all into the treatment group?
Thank you.
- --
Best wishes,
Michael McCulloch
Pine Street Foundation
124 Pine St., San Anselmo, CA 94960-2674
Tel: (415) 407-1357
Date: Thu, 7 Feb 2008 10:37:16 +0100
From: "Carlo Lazzaro" <[email protected]>
Subject: st: R: applying sampsi: sub-groups
Dear Michael,
as far as I am concerned, what you outlined in your post already happened.
In this clinical trial (please, see full reference below), patients were
randomized to moxifloxacin (1st arm) or to one out of three antibiotics (2nd
arm), chosen according to physicians'opinion. Clinical as well as
microbiological results reported by patients enrolled in the 2nd arm of the
trial were pooled and compared with those repored by patients enrolled to
receive moxifloxacin (1st arm).
Wilson R, Allegra L, Huchon G, Izquierdo JL, Jones P, Schaberg T, et al;
MOSAIC Study Group. Short-term and long-term outcomes of moxifloxacin
compared to standard antibiotic treatment in acute exacerbations of chronic
bronchitis. Chest 2004; 125: 953-964.
On the grounds of my experience in the economic evaluation of healthcare
programmes, I would take the liberty to advise you to carefully consider
whether this study design may cast some doubts on the robustness (even from
a statistical point of view)of the results as far as readers (or the
referees) are interested in head-to-head comparisons (ie, new treatment vs
1st standard treatment dose of the drug administered in the control group;
new treatment vs 2nd standard treatment of the drug administered in the
control group; new treatment vs 3rd standard treatment of the drug
administered in the control group).
HTH and Kind Regards,
==========================
Paul T Seed MSc CStat
Senior Lecturer in Medical Statistics
King's College London
Division of Reproduction and Endocrinology
St Thomas' Hospital,
Westminster Bridge Road,
London SE1 7EH
tel (+44) (0) 20 7188 3642
fax (+44) (0) 20 7620 1227
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/