Biotech Values

[iwfal]

AGIX - Summary of CC after announcement of Diabetes Trial:

a) The liver enzyme elevations typically occurred early in the treatment. This would be important in that close monitoring for a month or two might be feasible (whereas close monitoring for years definitely is not). But no literature about severe liver issues ever talks about this as the solution - presumably because the cliff analogy is correct. Patients are completely fine until they walk off the cliff and the only chance of rescue is the one step they make on the crumbly edge - not a big window for rescue.

b) Said that ‘with increasing HbA1C at baseline there was progressively greater cardiac risk reduction for all three cardiac endpoints’. That is a strong seller. I'd put out a p value for that trend and if it is below 0.01 it is worth something.

c) Improvement in HbA1C was substantially better in the patients with higher baseline HbA1C - gave example that in patients with baseline HbAC1>9 the improvement in HbA1C was nearly 2x that of patients with baseline <7.0. Note that this is entirely predictable since it is common is most (all?) diabetes drugs.

d) Improvement in HbA1C was seen regardless of other diabetes medications being used.

e) They expect to start partner discussions later this year.

f) ANDES Trial:

1) 1200 patients at 3 doses (plus placebo presumably). Highest dose is that used in ARISE.

2) 6 Month regiment with HbA1C measured at 24 weeks.

3) Protocol same as other diabetes trials as published at FDA - Scott actually hinted that there was a particular trial that served as the primary model.

4) Expect average (median? mean?) baseline HbA1C to be 9 to 9.5 (vs 7.2 for ARISE). So bigger improvement will be seen as per point number c above. Note that to first order this will NOT improve the power of the trial because the standard deviation will also increase. But it will improve the marketability of the drug to be able to tout a bigger improvement.

5) Entry criteria is 2 HbA1C readings above 7.5.

6) Trial locations are North Am, South Africa, Eastern Europe and India. Warning! This is where there is a really big problem. Running a trial in Eastern Europe is catastrophically dangerous. Docs in Eastern Europe just don't know how to run a trial. Lax entry criteria. Corruption. Poor compliance. Ex-communist states just don't understand the rigor needed in medicine (my family was significantly involved in trying to fix pieces of this in the early 90's - but the recent results in trials indicates there is a long long way to go). How might this sabotage the trial? Cheating to get past the entry criteria - go off drugs for a few weeks beforehand and then go back on. This will increase the 'improvement' seen in placebo and the resulting variance seen will kill the stats. Non-compliance - this is a known and significant problem even in the US. Diabetes trials work hard to get only patients with a history of complying with medical instructions. I'd bet 100:1 we will hear of a 'surprising' number of non-compliant patients in ANDES. Other issues will also appear in Eastern Europe. The only exception might be if they use Quintiles or the like (other companies with a history of successfully run trials in Eastern Europe) - but then they lose most of the reason for using Eastern Europe (the $ savings).

7) They will have an interim look at the data in mid 2008. Note that here Scott said something silly - that there should be no alpha penalty if there was no intent to change trial design at the interim. This is patently false - I'll pay you 5:1 if in 100 coin flips you allow me to calculate p value and I win if it ever passes under 0.05. But note that this is more of a pet peeve than a real one since Scott acknowledged that he will be paying an alpha penalty for the interim look.

8) Expect all patients to be enrolled by end of 2007. Given biotech that means they won't be enrolled until end of Q1 2008 at earliest. Just my cynical opinion.

9) They expect to release not just top line results with the interim peak but such things as 'fasting glucose and HbA1C'.