Update of my take on RI -
In recent months I took another hard look at the available RI-related evidence, thinking about what to expect at the first interim analysis (IA) and checking for anything important we enthusiastic longs have missed. The bad news is that there are too many imprecisely known variables to give a certain or precise estimate of RRR, but the good news is we do have quite a lot of information that IMO points in a very promising direction (but there’s never a guarantee). If it were too obvious, we wouldn’t have had the excellent buying opportunity we’ve had, and still have IMO.
A few months ago I estimated odds of interim stop for efficacy at about 60-70% despite what I believe is prudent under-promising by Amarin management, a tad below JL and HDG at 70% and 75%. I cited, among other concerns, the performance requirements for several unknown secondary endpoints and promising but statistically underpowered and therefore imprecise empirical evidence from prior studies (and their subgroups) in support of the blockbuster hypothesis that is suggested by medical theory and which seems likely for several reasons. (I was encouraged to hear the words “potential blockbuster” from management recently.) I recently posted I raised my odds of early stop for efficacy to 70-75%. If not stopped early, I still expect strong performance later. I remain very long.
I see several new positive signs. I was encouraged to hear at the CC that the secondary endpoints must look good for interim stop but not necessarily statistically significant, so we won’t get hung up by one secondary p value missing by a smidgen. I think that was one more in a series of important good decisions by the Amarin team. There were also several smaller but encouraging studies I learned about from other posters here that seemed to consistently suggest RRR in the 70% range (either CVD trial or otherwise correlating EPA level with CVD risk). I view the 53% JELIS subgroup result as plus or minus about 20 or 25%, which means probably at least 30ish for that group. Seeing several other high empirical results, albeit small and imprecise, helps bolster my confidence the 53% wasn’t a fluke.
Most of us know JL’s line of reasoning that using V to raise low RI EPA levels to about the JELIS active arm level suggests very high RRR for RI (perhaps 50% or more). I agree with that, but considered two potential limiting factors.
The first is that, due to limited RI trial duration, especially at interim, slower modes of action of EPA might be reflected more fully by the JELIS active arm (many of whom were life-long fish eaters) than in RI. That concern was reduced somewhat IMO by some recent smaller studies showing substantial benefits within several months (incuding CHERRY), but I still assume modest performance loss from that, decreasing over time.
The second concern stems from the lower statin dose for JELIS. Following JL’s point that much of the statin benefit derives from its anti-inflammatory properties (see JUPITER), one could argue that the additional statin in both arms of RI could mitigate some of the inflammation, leaving less for V to mitigate and reducing RRR for RI due to diminishing returns of anti-inflammatory action. This concern is reduced for me by the fact that JELIS showed that raising even an already quite high EPA level has substantial benefits that persist across subgroups, and a recent post referenced a paper that suggested some JELIS subjects probably would have benefitted from even more EPA. I think what left a very viable market niche for V is that statin doses have been limited by their adverse side effects and based on LDL instead of being set higher to offset inflammation to the optimal degree. If memory serves, Barry Sears suggested EPA doses of 5 or 10 grams daily. I doubt the CVD benefits of additional EPA have topped out at 4 grams, even on top of high dose statins.
Finally, placebo rate P for RI is the primary unknown factor that, with known event rates, will determine performance. I have always felt P was higher than many assume, and in extensively double checking that and my other beliefs against available evidence, I reached some conclusions with about 80-90% confidence.
I believe P for incoming enrollees increased substantially over time as recruitment standards evolved, with the minimum trig level increase in May 2013 followed by substantial additional tightening of recruitment criteria (not just trigs) as reported by two posters here and confirmed by reported or inferred event rate changes. (Kiwi, and another poster who said he was rejected because his trig level over 250 wasn’t high enough.)
I believe P started fairly close to the reported minimum initial expected value of 5.9%. The designers of RI were aware, pre-trial, of the P-lowering issues that prompted the precautionary FDA request for additional subjects. Those issues were summarized in a paper we thoroughly discussed here months ago, with the punchline that most of the problems were due to too many low risk subjects. RI has lots of high risk subjects.
I did a detailed analysis of slides 72-73 of the Amarin adcom presentation (graphs of actual and projected enrollment and events vs time as of October 2013) and other available data. Despite being substantially imprecise and incorrect in several respects, these charts contain some extractable real data. (For example, the Christmas 2012 break in enrollment is clearly visible.) This led to some interesting conclusions.
Q1 of 2013 apparently had a statistically aberrant low number of events, which likely caused moderate concern about whether the placebo rate was lower than expected (a high profile issue then, especially to the FDA) and helped motivate the increased minimum trig level for enrollees. (Amarin commented to that effect at adcom. They later said they already had enough subjects with trigs 150-200. Both statements are true.) The event rate for Q2 of 2013 was significantly higher than Q1, back in line with the average rate around that time. The change in enrollment criteria couldn’t possibly have had that much effect that soon, so the low Q1 2013 event rate was a statistical aberration (not too surprising given the relatively low number of patient years involved), not indicative of low P for those enrolled prior to May 2013 (about 5/8ths of enrollees).
So, there’s no reason to believe P was ever lower than expected, and P was increased probably at least about 0.2% by the announced trig level increase in May 2013 (e.g., from about 5.9% to about 6.1%). It seems quite clear from subsequent acceleration of the event rate (confirmed by a couple anecdotal reports) that P continued increasing as enrollment standards continued evolving to higher risk (additional increases in trig level and other risk factors).
P varies for various time intervals. We are now most interested in the interval ending at the imminently expected data lock. P will continue increasing moderately but significantly as higher-risk later enrollees contribute an increasing fraction of the total patient years, in addition to age-related P increase.
Furthermore, since the benefits from V’s various modes of action ramp up at various speeds (perhaps from days for blood thinning to several months for plaque stabilization), there’s a startup period of reduced benefit, which becomes an increasingly smaller fraction of the trial over time. So, with the bar lowered at the second IA and even more at final, there are multiple reasons to be optimistic about eventual success if RI continues this fall.
I’ve focused recently on the yes or no question of whether RI is likely to be stopped for efficacy at the 60% IA. I have somewhat arbitrarily picked 28% RRR as likely sufficient, allowing for some imperfect luck with primary and secondary endpoints. I think RRR is pretty likely higher than that, possibly a lot higher. I could be wrong.
I know a lot of us will take some off the table if there’s a big gap-up. That raises questions about what price trajectory to expect for various levels of success (what level, and when the near-term peak might be expected). Good news often gets sold very quickly, but, I think Amarin is far enough off a lot of people’s radar that it may gradually build for days or weeks, as some other stocks have done. I’m interested in opinions about that, and welcome comments on my remarks.
Market Data 
Markets 
AI
