Jonathan,
I agree that the co. may just be being thorough (sorry if I didn't say that here, there has also been a discussion of this going on the VD Discussion thread exchange2000.com. Afterall, previous delays by the company have been for that very reason. The problem I have with that reasoning is that in this case they had X amount of data that should have been in a format that they designed, they had Y people processing it, therefore they should have been able to predict (and schedule the panel) better than that. A one week overrun I'd buy that more easily; over a month begins to make me think that something unexpected has occurred or that management is really poor at timeline development.
If its the latter problem with timeline planning, then IMO we should expect similar problems with the NDA timeline.
If its the former and something unexpected in the data, then what is it? Once again, I think the key issue is was the "adjudication panel," which the company describes as reviewing clinical evaluations, part of the original process? If not, then IMO there may be a problem with clinical judgement on one of the endpoints. Since the secondaries are all pretty cut and dry (length of hospitalization, days in ICU, antibiotic usage) then IMHO such a problem would have to fall with judgment of the primary endpoint (serious infection). IF (big IF), this is the case, then any number of things could cause the concern. I can think of two right away:
1) Because this is a trial revolving around post-surgical infection, clinicians involved may have been more likely to notice even minor infection, leading to elevated occurence in both the drug and control groups. This would be a problem, because if you have a historical baseline and your control deviates strongly from that baseline, the FDA statisticians start to get a little worried.
2) Because the protocol specifically calls for antibiotic usage, the clinical sites may have been more diligent in their use of antibiotics and both groups may have fallen below the baseline. IMO, this might be an even worse problem because if the placebo group has a low enough incidence then statistical significance may not be achieved with this sample size, EVEN IF THE DRUG WORKS. I'm not sure what a panel could do to resolve this, so I consider this the less likely of the two.
This is all conjecture and worst case scenario, but IMO both of these results would be apparent before unblinding of the results. It is probable that there is no problem with the data, so we return to Rosemary's dissatisfaction and my concerns about future timelines.
Never apologize for being a contrarian. If there aren't two sides to discussions we'll never learn anything from them.
biowa |
