Saturday, June 16, 2007

Pagaclone open-label release misleading??

Here is what Indevus wrote about the new open-label results for Pagaclone:

The open label data demonstrated an improvement in nearly all efficacy measures compared to baseline that was at least double the magnitude seen in the initial 8 weeks of double-blind treatment. During the 8-week, double-blind portion of the study, patients who were randomized to placebo experienced a mean reduction in the percentage of syllables stuttered of 5%. After three months in the open label portion of the study, these same patients experienced a 31% mean reduction. These open label gains lagged slightly the gains made by patients originally randomized to 8 weeks of active pagoclone (5 months total exposure). For those patients randomized to pagoclone in the double- blind phase, the mean reduction in percent syllables stuttered was 18%, while in the open label phase, the mean reduction was 40%.

Similarly, data for the Clinician's Global Impression of Improvement (CGI- I) showed considerable gains during the three months of open label treatment........
Superficially good results. But I am a bit suspicious due to this relevant quote later in the press release:
More than 70% of the 119 patients who entered the open label phase of the EXPRESS trial were included in the three month open label data being presented at NCDEU.
In effect, 30% of the people decided to discontinue the use of Pagaclone. I would not be surprised that only the people who had a positive effect continued, so it is not surprising at all that there was at "least double the magnitude seen in the initial 8 weeks of double-blind treatment"? And most importantly, should you not actually evaluate all the patients after the 3-month open label, even those that did not take part to see the overall efficacy of Pagaclone?? Their statement of improvements is a bit misleading, because they only mention much later that only 70% took the open-label invitation up.

12 comments:

WILLME said...

This is an example of selectivity bias which could be handled if they used a regression model to analyze their results (see Maddala, G.S., Limited Dependent and Qualitative Variables in Econometrics, Cambridge University Press, 1983, pp. 278-283). Basically, the problem needs to be reformulated as a switching regression model.

Also what does a 31% mean reduction mean? Did a portion of the study group not stutter at all while the rest continued to stutter with their normal frequencies, resulting in a 31% mean reduction? Or are we talking about something in between?

Ora said...

Tom -

First, a minor correction to your statement “should you not actually evaluate all the patients after the 8-week open label”. The recent results represent the first three months of the open label period, not 8 weeks.

I agree that the “at least double the magnitude” statement is slightly misleading for the reasons that you suggest. But this shouldn't obscure the fact that even in the worst case, the overall positive results remain positive.

Let’s adjust the study in the way that you suggest: include all of the patients, rather than only the 70% who entered the open label phase but did not complete it. Let’s make the extreme assumption that all of the 30% who dropped out were not helped at all, i.e., that they had a 0% reduction in syllables stuttered. Using a weighted average, that would reduce the overall mean reduction in syllables stuttered from 31% to 22% (placebo group) and from 40% to 28% (pagoclone group).

The adjusted numbers now fall short of the “at least double the magnitude” standard for the pagoclone group, since mean reduction in syllables stuttered now improves from 18% to only 28% (adjusted number) and not from 18% to 40% (original number).

But we should still recognize that even 28% is a significant improvement.

(To rule out a potentially confounding factor, we know from another Indevus press release in May 2006 that “approximately 90% of patients continued into the open-label phase” (http://phx.corporate-ir.net/phoenix.zhtml?c=92810&p=irol-newsArticle&ID=860169&highlight= ) So there’s not much drop-out there. And there’s no particular reason to think that there would be much difference between the pagoclone group and the placebo group in the dropout rate at this point.)

Ora said...

For some reason the link in my previous post was incomplete. Let me try again:

Full link is: http://phx.corporate-ir.net/phoenix.zhtml?c=92810&p=irol-newsArticle&ID=860169&highlight=

Or if that still doesn't work, use my original link, and add this at the end (after the p=):
irol-newsArticle&ID=860169&highlight=

Tom Weidig said...

Ora,

I corrected to three months. Sorry.

I think one needs to do what willme proposes to do it properly.

>>> Using a weighted average, that would reduce the overall mean reduction in syllables stuttered from 31% to 22% (placebo group) and from 40% to 28% (pagoclone group).

>>> But we should still recognize that even 28% is a significant improvement.

Yes, but I would rather emphasise that the 22% in the placebo group is an even more amaying effect!!

Maybe I am a bit pedantic, but now you need to redo the ttest, and it should be less significant because the difference between placebo and non placebo went down from 9% to 6%.

Tom Weidig said...

As far as I understand the 31% reduction is the average reduction of each patient from before trial to after trial. The patients have various degree of dysfluency. For example, a severe stutterer might reduce from 30% stuttered syllables to 15% and he reduces bz 50%. A mild stutterer might reduce from 4% to 2% which is also a 50% reduction. By the way 2% is considered to be normal speech.

Ora said...

Tom - "Yes, but I would rather emphasise that the 22% in the placebo group is an even more amazing effect!!".

To avoid misunderstanding... do we have the same interpretation of "placebo group" and "pagoclone group"? This "placebo group" is the group that took placebo for 8 weeks followed by pagoclone for 3 months. Their result was 31% reduction (observed) or 22% (using my worst-case adjustment). This is compared with the pagoclone group, which took pagoclone for 8 weeks followed by 3 months. Their result was 38% reduction (observed) or 28% (using my adjustment).

So the "placebo group" took pagoclone for 3 months, while the "pagoclone group" took it for 5 months.

If you're saying that even 22% reduction after 3 months of pagoclone is pretty amazing, I guess I agree, although 28% reduction after 5 months is even better.

Hugo said...

A relevant question doing in this case would be: Which the normal average of withdrawals in pharmacological tests? If the average is around 30%, then the reason of the withdrawals must be due to more general factors, like the repulse to medicines and the fear evoked by the risk of dependence and side effects.

Tom Weidig said...

Sorry Ora, I got it wrong. Of course, both took Pagaclone in the open-label phase.

Anonymous said...

I am one of those study subjects and there is a definite improvement in quality of life. THe lower dosage studies did not help much, but with the increased dosages there is a signifcant difference.

Anonymous said...

I am stuttering since my childhood. I took recently a medicine against anxiety and my stutterer diminished considerably, even my friends realized it. But unfortunately, I must stop, because there is dependency with that product (prazepam, named Lysanxia in France).But I think that for me and for several stammerers, the chronic anxiety is a big cause. And I wait with impatience to try Pagaclone.

Patrick Thomas said...

Hey, im actually in Atlanta and I know this is really late like 2 years late but im actually going to be in a medical research thing on pagaclone, im a moderate stutterer and its going to be a 8 to 9 month study and afterward i have the choice to continue taking it with a free prescription but yeah, i'll let you guys know how that goes!
-Patrick

UrbanMermaid said...

Hi there, after reading so much about Pagaclone I was wondering if we could create some kind of petition to request for it to be released early or at least as soon as possible. Stuttering IS a life altering condition and it's about time people recognized that. If this was a cancer drug it probably would've been out right after Phase II... but since it's for stuttering it's like no one is in a rush to release it. As far as I know it's currently in Phase III which can take another 2 yrs. I wish we could do something to show people how important this is for us instead of sitting around waiting for it to happen.