RE: [NMusers] Reporting Modeling Results

Hi Bruce, Steve,

If the subjects had been removed immediately after consent, this would
be fine. According to the manuscript, however, the subjects were in
fact randomized (all 4 into the novel arm) and therefore should really
be included in the analysis, no matter how convincing the reasons for
exclusion sound. This may seem harsh, but these are the rules for RCTs
and, as far as I know, they are the same in academia and industry.

Just say "no adverse events occurred in the 4 subjects excluded from the
data analysis" and I will put a note in the margin that this is not a
concern. If you can't tell me this, for whatever reason, I place a
question mark. I suspect the world will continue to spin. Even in the
extreme scenario that all 4 excluded subjects had adverse events, it's
still a promising trend in favour of the new regimen. I was surprised
that a relatively simple dose adjustment, based on very cheap
covariates, appeared to prevent 7 out of 8 adverse events on enoxaparin.
It's an extraordinary breakthrough, if you have preserved efficacy
(which can't really be addressed in a small trial).

I think the contribution of academia to patient care is both huge and
invaluable,and I have learnt a great deal from your personal
publications, presentations and critical comments on this forum. In
fact, I was taught to always ask about the impact of excluded subjects
while undergoing my academic training. I will continue to do so, even
if it may occasionally cause offense.

Best regards, James



James G Wright PhD
Scientist
Wright Dose Ltd
Tel: 44 (0) 772 5636914
www.wright-dose.com


-----Original Message-----
From: Bruce Green [mailto:greenb_at_pharmacy.uq.edu.au]
Sent: 30 October 2007 02:42
To: 'James G Wright'; 'nmusers'
Subject: RE: [NMusers] Reporting Modeling Results


James,

They were not removed prior to analysis, they were excluded immediately
after consent. Following consent it became apparent that they were not
eligible. Here is an example to help you follow the workings of a
hospital. A subject is recruited in the ER and given a dose of UFH which
was not charted immediately by the ER physician. The subject was
consented, then the UFH was written up by the physician prior to
transfer to a medical ward. Now they are ineligible, have never had the
drug being tested and need to be excluded according to the protocol. As
such, you do not know what events occurred because they were not in
trial. So, they were not "retrospectively excluded treated subjects" as
they did not have the treatment.

Bruce


-----Original Message-----
From: owner-nmusers_at_globomaxnm.com [mailto:owner-nmusers_at_globomaxnm.com]
On Behalf Of James G Wright
Sent: Tuesday, October 30, 2007 11:37 AM
To: 'nmusers'
Subject: RE: [NMusers] Reporting Modeling Results

Steve,

Retrospectively excluding treated subjects from an RCT is a basic
methodological error (no matter who does it). Doing so casts very
fundamental doubts on the reported results. However, if these 4
exclusions didn't in fact bias the results of your study, then your
conclusions may still be correct.

Did any adverse events occur in the 4 excluded subjects in the novel
treatment arm? This is all I wish to know.

Best regards, James

James G Wright PhD
Scientist
Wright Dose Ltd
Tel: 44 (0) 772 5636914
www.wright-dose.com


-----Original Message-----
From: owner-nmusers_at_globomaxnm.com [mailto:owner-nmusers_at_globomaxnm.com]
On Behalf Of Stephen Duffull
Sent: 29 October 2007 20:21
To: 'James G Wright'; 'nmusers'
Subject: RE: [NMusers] Reporting Modeling Results


James

I had hoped we would move on :-)

> As you are holding up this enoxaparin work as a generalizable example,

It is just a single example - there are many others.

> I think your bold claims merit at least a token
> challenge. When I review academic studies, the first thing I check for

> is patients who are excluded from the data analysis.

I am absolutely astounded that you consider academic studies in a more
critical light than industry driven studies. Shouldn't all studies be
taken on their merit - or are you suggesting that academic studies are
naturally flawed in some way?

> Did excluding these 4 subjects alter the results of the trial?

Simply the patients were removed before analysis. They did not meet
protocol requirements (e.g. some received unfractionated heparin).

It is my best belief that if we continued to recruit patients that we
would see the same signal as the trial found. The trial was adequately
powered so we are not expecting bias inherent in underpowered studies
(compare to the underpowered APPROVe study wrt CVS observations).

In addition, the purpose of this part of the thread, for me, was to show
that a) studies do arise in academic settings that improve patient care
[to respond to Mark's comments] and b) that simplifying the dosing on
the drug label to make the drug "easier to use" doesn't necessarily
result in better patient outcomes.

Regards

Steve
--
Professor Stephen Duffull
Chair of Clinical Pharmacy
School of Pharmacy
University of Otago
PO Box 913 Dunedin
New Zealand
E: stephen.duffull_at_otago.ac.nz
P: +64 3 479 5044
F: +64 3 479 7034

Design software: www.winpopt.com
 


Received on Tue Oct 30 2007 - 05:52:35 EDT