Amarin Corp Plc (AMRN)

[AVII77]

I thought I remembered Steve stating that one enrollee could have more than one event.

Yes, absolutely people can have more than one event. Only the first will count in the Primary composite. Subsequent events may count in other endpoints.

Somehow I had overlooked that and do not remember that detail being brought up at any of the CCs.

If you look at the SAVOR trial (the one where I took the screen caps of the FDA review docs) and read through the actual paper presenting the results in the NEJM, you'll see they never actually mention it is the "First Occurrence" they are capturing for their primary. This detail is only spelled out in the
Protocol.

Don't take this the wrong way but I don't think most physicians really care (or should care) about such nuances of trial design. Their education is not focused on biostatistics but rather treating patients. But for modeling it is an important nuance to understand.

I do not understand the rationale for this proviso in the trial design.

It's really an issue of the math. Please have a look at this paper. The authors provide some methods to analyze data in the way most probably think it is actually analyzed. Techniques such as the ones they discuss have not gained wide acceptance. (When you get to page 4 of the paper you might appreciate why these nuances are not covered in medical school - personally, I'd prefer my doctors focus their training on other things).

I am unaware of a single trial where the total number of CVD events does not equal the the number total up in the list of individual events. Is this the case here and how is that done?

I'm not sure I follow your question.

Sometimes it is obvious and sometimes not.

Take the AIM-HIGH trial for example. Below you will see the CV deaths that were "first events" are different from all the CV deaths in the trial. Here they offer a footnote to help the reader.



In contrast, consider the PARADIGM-HF trial (recently stopped early at an interim). It takes some reading between the lines to see that only the deaths that are first occurrences are included in the composite primary endpoint.

It seems what would follow from your discussion is the risk percentage in the placebo group must then relate to first time MACE occurrence. Which would mean if this is correct that the numbers should be about the same.

I don't put much faith in the trialists accurately predicting the event rate for a trial. In that SAVOR trial, the actual event rate was almost double the rate they predicted. In contrast, for the ACCORD-LIPID trial they went through incredibly detailed gyrations to predict the event rate (for powering purposes) and got pretty close; 2.2 vs 2.71 IIRC (though in the paper they say they nailed it, the calc's in the protocol tell a different story).

would appreciate any further information on how this works.

It's fairly straightforward. For modeling, once a patient has his first event remove him from the "at risk" population. So, for example, suppose your model shows we are at full enrollment (8000) and 900 events. The population at risk for having a first event is 7100.