With your numbers,
Proportion of controls suffering an event = 8/1255 = 0.006375
Proportion of Rezulin-treated patients suffering an event = 28/1255 =
0.022311
Relative risk = 3.5 (1.63 to 7.51)
Absolute risk increase = 0.0159 (0.00702 to 0.0263)
Number needed to harm 62 (38 to 142)
So...statistically significant...
The argument was that they did not see that proportion of problems in
widespread use (and remember that a lab abnormality (i.e. 3x elevation in
liver enzymes) could be monitored and detected as the drug company wanted
but the people suffering from fatal liver failure appeared to escape
monitoring. Furthermore, there was never an answer to the question of which
patients were at higher risk for this complication so you couldn't really
use it without fear of killing a liver and its carrier...
-----Original Message-----
From: [log in to unmask]
[mailto:[log in to unmask]]
Sent: Tuesday, March 13, 2001 9:46 AM
To: [log in to unmask]
Subject: Re: Qoute from Warner-Lambert about Rezulin
I am not a statistician so perhaps Steve Simon could give us some further
commentary based on the following.
According to information on the W-L web site, (
http://www.warner-lambert.com/Rezulin/henney.html)
'During all clinical studies in North America (N=2510 patients),
a total of 20 RezulinŽ-treated patients were withdrawn from treatment
because of liver function test abnormalities.'
and
'During controlled clinical trials, 2.2% of RezulinŽ-treated patients
had reversible elevations in AST or ALT greater than 3 times the
upper
limit of normal, compared with 0.6% of patients receiving placebo.'
On this limited information we might assume that approx 1255 patients
received Rezulin
and 1255 received placebo.
If so, then 28/1255 in the active group would have suffered raised liver
enzymes
and 8/1255 in the placebo group would have suffered raised liver
enzymes
Intuitively this looks to be a real difference. Statistically I'm not so
sure.
On the basis that they may not be 'statistically different' then the
company
might be justified in saying they are 'comparable'.
Turning the question around, would we be confident that a treatment which
produced a positive response in 28/1255 patients is really superior to a
placebo
response rate of 8/1250.
Whatever the answer to these questions, on the basis of 'first do no harm'
should we
anyway use different evidence standards when assessing adverse reactions?
Jonathan
-----------------------------------
Jonathan Harris
Pharmaceutical Adviser
NHS Executive (Trent)
Sheffield, UK.
e-mail [log in to unmask]
----------------------------------
-----Original Message-----
From: Simon, Steve, PhD [mailto:[log in to unmask]] Sent: Monday, March
12, 2001 10:40 AM
To: [log in to unmask] Subject: Qoute from
Warner-Lambert about Rezulin
There was an interesting article about Rezulin in Sunday's paper. I'm
sure most of you are familiar with this, but this drug was taken off
the market recently in the United States because of problems with side
effects like liver failure. There is some question about whether the
Food and Drug Administration approved the drug too hastily.
There was one quote in particular that caught my eye. During the
approval process, the company noted that 2.2% of the patients
receiving Rezulin had liver problems compared to 0.6% of the placebo
group. The company characterized these two rates as "comparable".
In defending that characterization, Randall Whitcomb (vice president
for diabetes research) testified as follows:
"'Comparable' is, is, you know, is an interesting word. Is 2.2 percent
different than 0.6 percent? ... I think you could look at 2.2 and 0.6
and say that those are similar numbers, you know, when you look at
this now. I mean, 'similar' is a -- is a very broad term ... I don't
think that these numbers are, are all that different."
Anyone care to comment on this quote? Are 2.2 and 0.6 percent similar
or comparable numbers in this context?
Steve Simon, [log in to unmask], Standard Disclaimer.
STATS: STeve's Attempt to Teach Statistics. http://www.cmh.edu/stats
|