Amarin Corp Plc (AMRN)

[AVII77]

... if that was what was agreed with FDA as part of the SPA.

Yes, if that is what was agreed, that is what should happen.

I just don't believe that was part of the agreement.

The protocol says (emphasis mine):

"the interim analysis will occur after at least 967 primary efficacy endpoint events have been observed".

I think the FDA would encourage that the most complete and accurate data set be used for the interim analysis. That is in the patients and public's interest.

The Lan DeMets alpha spending approach allows this to be implemented. Whatever the number of events end up to be, they can calculate the exact p value required for that interim (and how it affects the final p value).

I misspoke when I suggested they may have intentionally over-ran events in order to make it easier to stop. That was a foolish thing to suggest. We now know the real reason: they were modifying statistical aspects of the protocol. And that takes time. (and it must be completed before the DB is unlocked).

Regarding currancy of data, Fleming says:

"A third opportunity for the DMC to have a positive influence on data quality arises during its conduct of formal interim data analyses. Through careful review of study reports, the DMC should attempt to determine whether proper standards for data completeness and accuracy have been implemented. One such standard will relate to the recency of follow-up available on all patients. It is desirable that the database be as current as possible, since interim analyses of the available data could lead to early termination of the trial. Some amount of time, however, must be allowed for editing and correcting the database, performing the interim analyses and transmitting the interim reports to the DMC members for review prior to the meeting. Experience from the perspective of both the statistical data center and the DMC member suggests that it is reasonable to expect the DMC to receive complete and accurate major endpoint information that is current for each study participant to within about 2 months of the date of the DMC meeting. This standard (i.e., allowing a lag of no more than 2 months for each study participant) would minimize the risk that the number of outcomes that would have occurred by the time of the DMC meeting but would not be available to the DMC would
substantially alter conclusions from those data presented to the committee."

And that was from a decade ago. electronic CRF's, IMO, would likely make it possible to achieve even less lag time.

Fleming goes on to note:

"Even if these procedures and this timeline are met for ensuring accurate and complete data through the date of data freeze, it is still likely in many trials that important information will have evolved on the trial’s primary endpoint(s) between the date of data freeze and the date of the DMC meeting. For example, some endpoints may have occurred prior to the date of data freeze but were not adjudicated until after the date the database was locked; other endpoints may have occurred after the date of data freeze and may have been subsequently reported.

Thus, in many trials it has been standard practice for the statistical center to take a ‘snapshot’ of the primary endpoints in the database approximately 7–14 days before the date of the DMC meeting. Some investigators urge that all available primary endpoints should be included in the interim analyses even though not all are fully adjudicated (Pocock and Furberg, 2001; Wittes, 2000). Even when the primary analysis focuses on adjudicated endpoints, a separate brief summary could be generated that would provide a more current (though less cleaned or validated) update of the primary endpoint analysis."