The same issue was brought up by one of us in response to the article on
gabapentin for diabetic neuropathy in JAMA. They do it all the time for pain
scales. The solution seems to be to before the study, survey patients and
ask them what would a clinically significant improvement be on any sort of
nominal scale and then report the results as the number of people that
achieved that outcome in each group (or the number that failed to achieve
improvement if you want) and then get your benefit or risks differences
calculated and your NNTs or NNPs.
Thanks
Victor Montori
-----Original Message-----
From: [log in to unmask] [mailto:[log in to unmask]]
Sent: Tuesday, November 16, 1999 5:23 PM
To: [log in to unmask]
Subject: Re: witch incendary
I read the paper that started this off with critical interest
(Arch Intern Med. 1999;159:2273-2278) as indeed I have followed the
subsequent entertaining correspondence. I wonder if others have
also been concerned with what appears to be a fundamentally flawed
methodology slipping past the reviewers, presumably because of the
sincerity of the authors' approach to the subject matter.
The double-blind study seems set up immaculately, but the problem
comes with the method used for recording the outcome data. The
authors describe a new scale which they claim to be a continuous
variable, so they then treat it arithmetically and record mean
scores to compare the two groups. This scale does not appear to have
been validated and comprised six stages from needing antianginal
drugs to death. I think that although each of these stages is
related to the other items, it is not in a linear way. Each number
allocated within the scale is nominal and so should not be treated
as if it had quantitative value even though each stage is clearly
worse that its predecessor. Surely the correct way to handle this
data would be to record for each group the number of
patients achieving each of the outcomes described and then perhaps
calculate NNP to avoid - say - death or cardiac arrest.
What do others think?
Oliver Samuel fax: +44(0)20-7419 4622
24 Lancaster Grove London NW3 4PB phone:+44(0)20-7419 4624
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|