AGIX - A summary of my review of the blended ARISE results vs previous trials, primarily IDEAL and PROVE-IT – and note that all opinions in this post are based upon this data (ignoring the fairly promising ph i/ii data). At a first glance the blended ARISE results look very promising indeed. BUT this needs to be tempered by the following facts (the tear down):
Tear down (due to confounders):
a) Standard of Care change - Between IDEAL/PROVE-IT and ARISE there has been the introduction Drug Eluting Stents (DES) and rapid replacement of Bare Metal Stents (BMS) by DES. In the first 9 to 12 months after stent placement DES results in a significantly smaller revascularization rate (without much change on MI or death rate). This change in practice means the overall IDEAL/PROVE-IT curves are not remotely comparable to ARISE.
b) Trial location - ARISE is being conducted in the US/Canada/South Africa, while IDEAL was almost exclusively a Scandinavian trial. The MI treatment paradigm in Scandanavia/Europe is substantially different from the treatment paradigm in the US. Again this makes the IDEAL/PROVE-IT curves incomparable to the ARISE curve.
c) Patient illness – IDEAL and PROVE-IT took everyone except those who effectively had too high a cholesterol level (PROVE-IT had a total cholesterol limit, while IDEAL wouldn’t allow anyone who ‘needed’ more than a minimal dose of statins.) whereas ARISE had no cholesterol exclusion AND all patients had to be in a high risk group (e.g. diabetic). Again this makes the curves incomparable.
The problem - Both a and b are individually big enough (in the 40-50% range) that it means that the overall ARISE curves are within the ‘error bands’ of IDEAL/PROVE-IT.
The corrections:
a) Standard of Care change –
1) The acute phase (the first 12 months) - the MACE rate is highly dominated by the Target Vessel Revascularization rates and those are much lower with DES than BMS. The problem is that the various DES v BMS trials give significantly different improvement factors (ranging from 50% to 90%) AND we don’t know how many ARISE patients got DES (probably between 30 and 70 pct) AND we don’t know how much discount to apply because many ARISE patients are out beyond 9 months (probably another 10% uncertainty in MACE rates). The mapping of PROVE-IT to PROVE-IT enrolled over the same period as ARISE would be a 40% discount +/-20%.
2) Post Acute phase (from 12 months on) – On balance the DES seems to have significantly less benefit during this period. In fact it is probable that DES actually performs somewhat worse than BMS (within(!) this period) – the Revasc rebound effect. The mapping of PROVE-IT to PROVE-IT enrolled over the same period as ARISE would be 0% discount +/- 10%.
b) Trial location – The US has a reputation as being substantially more aggressively interventionalist in regards to heart disease. You can see this in papers such as Geographical Differences in the Rates of Angiographic Restenosis and Ischemia-Drive Target Vessel Revascularization After Percutaneous Coronary Interventions by Singh, M et al or in the difference in PCI rates in IDEAL (a European trial) vs PROVE-IT (a US trial) (PROVE-IT had 70% PCI, IDEAL had 20%). The odd thing about this is that it results in a counter intuitive effect in the expected ARISE curve (given the fact that ARISE is largely American) relative to IDEAL. The amount of aggressiveness is probably 70% +/-20% more interventional for revasc.
1) Acute phase – Because US patients are more likely to have immediate PCI they are less likely to have restenosis during the first 12 months than an IDEAL patient. But compared to a PROVE-IT patient ARISE will have more intervention since they have Canadian and South African patients – assume restenosis rates for ARISE go up 20% plus or minus 10% for ARISE relative to the corrected PROVE-IT. I.e. MACE up 10% +/- 5% (given ARISE era restenosis is about 50% of MACE in acute phase).
2) Post acute phase – Assume 20% less revasc +/-10% (10% less MACE +/-5%) due to Canadian/South African patients.
c) Patient Enrollment – IDEAL and PROVE-IT actually enrolled healthier than average MI patients since they excluded those with high or uncontrollable cholesterol. ARISE not only does not exclude such patients but requires that every patient be in a high risk category. The most important of these factors is that ARISE has 20% more diabetics (who have >50% more risk of MACE) and has about 50 pct higher LDL than PROVE-IT (and the commonly used metric in cardiology is 1% additional LDL is 1% additional risk – but I am uncomfortable taking all of that credit). In toto ARISE patients probably have at least 30% greater MACE rate +/- 15%. Note that in this case I have assumed there is no difference between acute phase and post acute phase.
Results:
a) Nominal - The adjusted-per-the-above-corrections PROVE-IT High Dose Lipitor arm is 72% pct above ARISE in acute phase and 84% above ARISE in the post acute phase. This corresponds to about an HR of 2.5 or 0.40 depending on how it is calculated and this would be substantially better than even the highest dose of Lipitor.
b) Worst case – Stacking all of the worst cases the adjusted-per-the-above-corrections PROVE-IT High Dose Lipitor arm is actually 7% below ARISE in the acute phase and 39% above ARISE in the post acute phase. I.e. the error bounds actually encompass ARISE actuals for acute phase but do not for post Acute phase. Thus it is likely that PER THE STATED ASSUMPTIONS(!!!!) AGI-1067 works in post acute phase. The issue is that the post acute phase, by itself, doesn’t have enough events to cause stat sig. So the interesting question is how likely is it that it has HR=1.78 in post acute phase and HR<1.10 in acute phase?
Liens and Notes:
a) Clearly if any of my above assumptions is substantively wrong then the analysis can be substantively wrong. GIGO.
b) Although the use of Niacin is still very small compared to the overall statin use rate, it has gone up substantially since IDEAL and PROVE-IT (due to very compelling, for small trials, data). If that usage has been preferentially concentrated in trial-type patients the too-low blended ARISE data could be the result of Niacin usage instead of AGI-1067.
c) The uncertainty rates given here are rough. But wherever there is a doubt I’ve been conservative. Of course this still doesn’t account for an OOOPPPPSSS
d) Interesting check is to run the same kind of adjustment for IDEAL vs PROVE-IT and see how close the two come to each other. The results are that for both the acute and post acute phase the nominal adjusted PROVE-IT is about 20-30% different than IDEAL, but IDEAL is within the error bounds of the adjusted PROVE-IT.
Market Data 
Markets 
AI
