Replies to post #3362 on Dendreon Corp fka DNDNQ

[Thurly]

<< first step would be to persuade the FDA to give them a free look at 9902B >>

Elan and Wyeth have a situation where they take peeks at their AAB-001 PII trial in progress through a neutral third party. Neither Elan nor Wyeth are allowed to see any of the data. The third company reports out the results of the peeks allowing Elan and Wyeth to see whether they have found the optimum dose/frequency of AAB-001 to determine when they should move on to a PIII trial.

I don't know if there is a penalty for the peeks in this case since, while the data is unblinded, the companies are still, essentially, blind.

Anybody know if a neutral third party looked at the 9902B data whether or not the FDA would allow them a free look?

Thurly

[p3analyze]

I have given this some thoughts. We can forget about alpha-penalty-free interim look. Any interim analysis, planned or unplanned would incur some alpha penalty.

What I think should be negotiated is that we advance the interim analysis from the current 180 (or whatever it was the number that's greater than 160) deaths to as soon as data become available. The alpha spending algorithm can remain the same, but with less events, the alpha allocated to this advanced interim analysis will be smaller, making it more difficult ot achieve statistical significance, yet in the meantime incurs less alpha penalty to the final overall survival analysis.

The point is that the power of analysis at the currently planned 180-event analysis will not be 80%, more like 40 to 50%, which is why DM inquired whether Gold will consider to increase alpha allocation or wait for more events, but either action would increase the power for interim analysis but also increase hurdle for final analysis if it doesn't pan out, and that Gold said they are evaluating it. So bottom-line, if the current game plan does not guarantee success, advancing it earlier may not be so bad.

Given FDA has already protected the big money interest by this approvable letter (I know Dew is going to say I am hallucinating again - but the approvable letter convinced me that the system is as corrupt here in this country), and that by delaying the approval, the purpose of protecting trial integrity has been served, and Fleming and Hussain's argument will not have a stand (as long as by interim analysis most if not all enrolled patients become symptomatic). Then what FDA really need to show is that, based on interim analysis of 9902b, there is trend if not statistically significant overall survival benefit. If you are concerned that the sample size from 9901 and 02a were too small, then interim data from 9902b has provided further support to the already Ad-com voted "substantial evidence".

I know this would not appease the die-hard statistician, but at least this approach will put FDA in a much better situation.