Anyone else notice how Pyr seems to disappear for weeks when the PPS is stable, then suddenly starts posting multiple posts per day when the PPS starts moving in big ways (on no news)?

Pyr - Your odds make no mathematical sense unless you assume a very low placebo rate (below 5%). We know there are 1290 events at the 80% IA, and we know that the patient years is approx. 26,500 or more based on company statements. A 4.9% placebo rate would be necessary to show that the highest odds are for RRR = 2%, as in your table. Someone recently posted a link to an analyst report that went over a meta-analysis of placebo event rates in diabetic populations (ignoring the high TG risk factor), and found that the 3 "hard MACE" outcomes had a rate of 5.2%, which would increase to 5.4% if you added in stroke and angina, as in the R-It PE. Fourier also had a >5.2% placebo rate, even with substantially fewer diabetics and substantially lower TGs than in R-It. So, your estimates are probably right if you assume the worst case scenario, but much evidence would lead us to think the placebo rate will be at least 5.2%, in which case the 15% RRR for the PE would be hit. Even at the 80% IA, a 15% RRR will meet their p<.02 cutoff for success.

Pyr - So would you assume that applies as an underestimate from the 5.9% placebo rate they proposed in the original design (the "mistake" most studies make)? The revised 5.2% rate that they used as a basis for increasing the sample size is 12% lower than the original design rate, presumably for exactly the reason you point out. If the placebo rate is <5.2%, there will however be problems meeting the 15% RRR for the PE at the 80%.

MrMain - Numbers I ran said that if placebo rate is between 5.2 - 5.9% as expected, there is about an 84% chance (based on what we know about event rates and patient years) that the 15% RRR is hit at the 80% IA. I think this sets the low end of the probability range, and my guess would be closer to 90%.

PPS drop the last 2 days on what was essentially no (bad) news after a period of stability has all the hallmarks of the manipulations of the last 6 months where stock rose dramatically for several days on no news, then would drop precipitously for no obvious reason, and then repeat a couple of weeks later. My guess is that it drops a bit more, then in a week or so suddenly increases for no reason, likely trying to take advantage of people's hopes for a data leak or pending announcement (and maybe cycle once more before the 80% IA announcement occurs). I am just ignoring daily PPS - won't make any difference once positive results come out.

You were right - they were just slow to acknowledge. Still possible its lower placebo AND V rate, but slowing of placebo rate as sicker patients were added later in the study makes no sense. I choose to think it is the more parsimonious explanation, i.e., mostly V effects.

True, but "weeks rather than months" could be anywhere from 2 - 7 weeks. If closer to the latter, that could be meaningful. Every 4 weeks accrues approx 600 patient years - 1290 events with more patient years = lower composite rate = placebo rate can be lower and still show the desired RRR for V.

Examples:
For 1290 events/26,000 PYs with placebo rate = 5.3, RRR = 12.8
For 1290 events/27,000 PYs with placebo rate = 5.3, RRR = 19.7

MM - Could in theory be due to reduced placebo rate. However, the sicker patients (higher TG cutoff) were enrolled later in the study, and we would therefore expect increasing placebo event rates later in the study because of this. Taken in context of these sicker patients now being enrolled longer in the study, the fact that overall event rate was "on track as expected" at the 60% IA, but now is a bit slower than expected, suggests that the active intervention (V) is more likely responsible for the delay. I do not consider any delay a bad thing (other than longer to wait for the outcomes).

PPS drop today in funny (almost 5% right now). The company announced missing earnings estimate by .01 during a quarter known historically to be worse than the other 3 each year, prescription numbers that are exceeding what they expected, and approval to proceed with a short clinical trial (like Marine) in a potential 150,000,000+ patient population (in China). Possible further dilution noted in the earnings report seemed at least partially to be in context of needing to expand hiring and promotion to address expanded market after R-It success, which would not really be a bad thing. Everything else is on track. No surprises. What's not to like? I suspect the PPS will quickly bounce back up to where it has been - just people taking advantage of the situation.

GG - Thanks for posting. I thought the following was interesting:

JL - Thanks - agreed. Given the median RRR value in these scenarios, the data say approaching 30% RRR is certainly plausible.

Zum - Good question. AVI can address this much better than I can. As long as dropouts are relatively similar across groups, it wouldn't bias the pattern of results. However, even if evenly distributed, if we had 20% dropouts that would be like running the study with a sample of 6,400, which I believe would make it substantially harder to show significance for a given event rate. The other issue is how an intent-to-treat analysis (where dropouts are counted in analyses as members of the group they were originally assigned to) would play into this, especially if dropouts were highly selective to one group. My understanding is that this is an important issue when patients in the active treatment arm drop out due to side effects - they are counted as having received treatment but did not actually get it, which reduces the effect size for the treatment. From everything I have read, this does not seem likely to be a big problem for V, given its absence of major side effects. Maybe AVI can weigh in on the power effects of various drop out rates in more detail.

Raf - I am not smart enough to do that.

JL - To your point, I repost the following you may have missed:

Tas - The posts were conflating 2 things. 1000 extra patients were added before the study started to address possible 5.2% rather than 5.9% event rate in the placebo group. Separately, the minimum TG requirement was increased mid-stream in R-It, which led to more higher risk subjects being added to the study later on. This should change (increase) the overall event rate in the late phase of the study but it is hard to predict by how much.

Kiwi - I can't really estimate anything for the 100% completion because we have no clear idea on how many PYs we will have then. Wish I could though.

Isaeed - I don't have time to actually run those for the 60% IA, but assuming results were good at that time, the 60% no stop may have had more to do with the secondaries and the desire to see a clear trend over time. We have much more data now.

Isaeed - Patient years in the month prior to them announcing the 80% event was listed in an investor presentation as ">26,000 patient years." We don't know if it was exactly this, but since they accrue a max of about 660 PYs per month, it pretty much has to be between 26,000 - 27,000. Exact dates in your items a-c do not affect these calculations (aren't necessary to know). The range of PYs I used in analyses captures the impact different PY values would have on the RRR, and it does make a difference. We don't need to know the treatment group event rate for these calculations - just the overall event rate (events/PYs) and what we assume the placebo rate will be. I have posted before that with the number of events available at the 80% IA, an RRR of 15% would meet their PE criterion for a stop (using simple stats rather than kaplan-meier approach for which we do not have enough info).

In case you (or anyone) is interested, if I restrict these models to those in which there is a placebo event rate between 5.2 - 5.9 (inclusive), the range in their original design and subsequent modification, and use the 26,000 - 27,000 patient years as above, 83% of the 72 scenarios meet the 15% RRR cutoff, with a median RRR of 24.5%. Assuming they are right in their placebo estimates, we should be good.

RRR likelihood given what we know for those interested - We know the number of MACE events at the 80% interim is likely between 1280 and 1300, and that patient years are between 26,000 and 27,000. I ran some numbers using these high/low estimates to derive overall event rate, and calculated the RRR given a range of placebo event rate assumptions, from 4.9% to 6.0% (testing a range around 5.4%). Even with this relatively broad range of possibilities, we hit the targeted RRR from the original design (15%) in 64% of the 108 scenarios tested. In all scenarios, the critical placebo rate for reaching the target RRR for the PE is between 5.2 - 5.4%. The highest possible RRR in these models is 41.98%, with a median of 21.3% RRR. Each 0.1% increase in placebo rate increases the RRR by about 3%. Of the 108 scenarios tested, 24 (22%) have an RRR < 10%. This 64% likelihood of meeting the PE target of 15% RRR should provide fairly conservative odds (includes scenarios with a lower than expected placebo rate) of obtaining significant results at the 80% IA.

FSU - Potent systemic anti-inflammatory effects is the reason if your TGs are not elevated.

JL - I do read all your posts, closely. I recall you many times arguing for increasing divergence with time, and I always assumed that would mean the overall event rate would decrease over time. Clearly not. Probably combination of acceleration in placebo, deceleration in V, accelerating events towards the end due to change to higher TG cutoff in midstream, and fewer events possible as the sample size decreases (those with events can no longer accumulate more). Bottom line - its complicated. I still agree with your main argument that there should be some increased divergence over time, but its just not easy to detect from the overall event rate numbers.

So nobody really knows anything

Agree that there is a lot we don't know exactly. My overall point is that we are not seeing a clear reduction in event rates over time, which we might expect if the Placebo event rate was stable over time and V was reducing events. Maybe the Placebo group is increasing in event rate over time (aging?), and V is reducing some simultaneously with net effect of 0 on overall event rate. In either case, assuming the higher end of the composite event rate estimate based on what we now know (1290/26,500 = approx 4.9) and the placebo event rate is at least 5.3 or so, the RRR should meet the intended target (15%+). If we are overestimating the current composite event rate, and it is closer to 4.7-4.8 due to more patient years(1290/27,000), then that only improves the RRR. Also, for each increase of 0.1 for the placebo event rate, RRR increases by about 3%.

Not sure if this was ever posted, but I went back to info available Feb-March 2016 when the 60% interim event hit. At the 60% interim, based on the 20,000 patient years reported in a press release the month prior to the 60% event occurring, our overall event rate had to be between 4.6% - 4.8% (lower is better for V). By comparison, at the 80% interim, we are at an overall event rate of 4.7 - 4.96 (depending on whether we are at 26,000 or 27,000 patient years). So, event rate seems to be increasing some over time, possibly reflecting the sicker patients enrolled later in the study (increase in qualifying TG level). Not sure this supports the idea of increasing V effects over time, unless we are seeing a correspondingly larger acceleration in the Placebo group events to more than offset decreasing rates in the V group (meaning it has increased over time in the Placebo group from the expected 5.2 or 5.9).

Hmmm, hadn't considered that possibility.

BB - I think you are being overly pessimistic. If there is money to be made, investors will buy the stock, period. It would be totally irrational to refuse to buy the stock after R-It success simply because it had been going nowhere prior to R-It success - they are not connected in any way. Right now, there's just no money to be made on the stock due to the "dead period" it's in. Just imagine all the press R-It will get if it is successful. Investors will be clamoring to make money off of buyout speculation, if nothing else.

BB- Check the chart for the last year. The PPS made its big move up on July 28- August 3, which was approx 6 weeks before the 60% interim announcement (Sept. 12). If we are on the same schedule, PPS should start to rise early-mid June in anticipation of 80% interim.

BB - The facts (stable PPS) do support your argument so far, but I suspect that is only because we are 3 months away from knowing anything about R-It results, at the earliest. Let's see what the PPS is in mid-July. I am betting it is higher than $3.14 by then. I think we are neglecting the fact that most PPS changes in the past year were driven by manipulation, shorts, day-trading, etc., as well as 60% stop expectations (last August). They were not due to any fundamental aspect of the stock. The Ad Com failure obviously is why this stock is not at $20/share now, but that has no bearing on what the PPS will be if R-It is successful.

Great - UCSF is one of the top medical schools in the country, so sounds like you are being treated by the best. I have not met a fat cardiologist either. They must know something...

Thanks Kiwi - Great article. Regarding:

Controversial editorial says CV disease is due to inflammation, not clogged arteries. Implies that anti-inflammatory treatments would be a better mechanisms based approach for CV treatment. Wonder how that could be accomplished?

http://www.cnn.com/2017/04/25/health/saturated-fat-arteries-study/index.html

Kiwi - Higher risk + shorter time in study may be a wash IMO.

Kiwi - Thanks. I think you often take the position of being the "devil's advocate," but in a reasonable way, which I appreciate. You always raise good questions and make me think.

JL - Makes sense. "Well-trained patients" are ideal.

Agreed - I do not plan on looking into this further. It is beyond my pay grade.

Raf - I have tried to go to the literature to get information on what placebo rate we might expect. It is difficult to draw firm conclusions based on a cursory review due to differences in study entry criteria, MACE definitions, how data are reported, etc. If you look at Fig 1 in a meta-analysis in the Lancet (2010, vol 376, p. 1670), trials of higher vs. lower intensity statins show annual event rates of 3.8 - 13.1% in lower dose statins, and 3.6 - 11.3% in high dose statins (but all different study populations). Diabetes (approx 50% of the R-it sample) would increase this. Also presumably increasing risk compared to other trials would be pre-selecting for high TGs and patients having combined High TG/low HDL (approx 50% of sample). As someone else mentioned (you?), we have to have faith that the people that helped them design the trial thought these issues completely through and were accurate in their estimates of a placebo rate above 5.2%.

Kiwi - I suspect you are correct. As has been argued by you and others, maybe this means going to the end will have better results. There really are just too many moving pieces and unknowns to really make good sense of all this.

Raf - I agree with you that the observed overall event rate even at 80% has to be right around 4.9% even with a fudge factor. Assuming this as the composite rate, here is the RRR for various placebo rates by my calculations:

PL Rate = RRR =
5.0 4%
5.1 8%
5.2 12%
5.3 15%
5.4 19%
5.5 22%
5.6 25%
5.7 28%
5.8 31%
5.9 34%
6.0 37%

Assuming placebo rate is at least 5.3%, we are golden. If not, I have to believe that diabetics, High TG/Low HDL-C, Male, etc. subgroups will show a higher RRR than overall and still be useful for expanding the market via 1st amendment promotions.

Kiwi - No - slower enrollment just means they had trouble getting qualified patients who agreed to participate. Trial enrollment never go as quickly as one would hope. This would not have anything to do with event rates.

Raf - The FDA does not reward novelty, and I am willing to bet AMRN would be the first to propose a drug for an indication based on EPA/AA ratio. May be scientifically justified but would be corporate suicide.