JL – Thanks for your reply and kind words. I read virtually all of your posts, although I can’t always do it as promptly as I’d like. Sometimes it takes me a while to think things through.
Those not interested in detailed speculation can skip to my bottom line below, or to the next post.
A key question is why RI continued at the 60% interim analysis (IA60) when there are so many good reasons to expect high V efficacy (possibly very high). Several factors may have contributed: a startup period while benefits ramp up; a straggler SE; the bar set extra high to defend against the FDA; DMC unpredictability; etc. For reasons we discussed, the startup effect contributes but does not seem a primary driver. Your medical insights agree with the math on that.
To clarify, by “startup period,” I mean the delay before true RRR reaches it’s full value (not the estimate). It’s the delay before the active arm turns the corner and starts to separate from the placebo arm on an ideal events chart. RRR is the difference in slopes of the two arms, which resembles a derivative of a derivative. Differentiation amplifies noise, and roughness in the JELIS placebo arm indicates substantial noise that gets double-amplified, so it’s very hard to reliably quantify startup period from real-world data IMO.
Note that a very long, severe startup period (years) would imply that estimated efficacy should increase substantially over time. But, with moderate startup, measured efficacy for the various endpoints should approach the true efficacy values from either above or below. The odds of a straggler substantially low SE estimate go down over time, as do the odds of a substantially high one. So it’s unclear more time would necessarily increase measured efficacy. (Statistical significance should improve, but it seems likely to be ample already.)
A long time ago you explained the importance of the high background EPA level for the JELIS trial. As someone mentioned again recently, their 19% RRR represents the benefit of only about half the RI dose, the other half was in both arms, and JELIS really suggests RRR for RI around 50% or more (arguably much more for some key subgroups. You also pointed out Japanese genetics were ruled out as the cause.) The strong negative correlation of EPA level and CVD risk in both JELIS arms, and other empirical results, suggest high V efficacy, as does the growing relevant medical theory you and others have discussed.
We must change our views to match evolving facts, and be careful not to succumb to cult-like blind loyalty, but I have looked carefully and found no major flaw in the strong case for high V efficacy, and my best explanation for the continue at IA60 is a combination of the likelihood of a straggler SE as discussed in my prior post, a very high bar set by Amarin for several reasons (including understandably wanting a bullet-proof defense against the FDA) and moderate startup delay. It seems unlikely to me we missed an efficacy stop at IA60 by very much, probably for a SE, not the PE. While continuing at IA60 makes it less likely there’s extremely high RRR (which would probably have dragged all SEs along, even with bad luck), it does not preclude high RRR (in the mid-30s, for example, my current best guess).
Bottom Line: I think the case for high V efficacy remains strong, and RI continued at IA60 most likely because Amarin defensively set the bar high and a laggard SE missed. I expect high efficacy (my guess around mid 30s and a nearly 50-50 chance of efficacy stop at IA80). Of course, as always, no guarantees, there are still lots of missing pieces for this puzzle.
News 
Market Data 
Discover 
AI
