Innovation Pharmaceuticals Inc (IPIX)

[PlentyParanoid]

Because I, too, have a pain in my big toe I feel entitled to answer.

In p2a trial prurisol 200 mg, sPGA =3 group recorded sPGA 0/1 shift 2 response percentage of 31.6. Historically sPGA 0/1 shift 2 percentages are about 10 % lower than PASI 75 percentages (sPGA 0/1 ~ 0.9 times PASI 75). Therefore, for phase 2b trial, one could reasonably expect PASI 75 response around 35 % if efficacy at 300 or 400 mg dose is what was recorded for 200 mg in p2a. This should be good enough for approval if placebo response stays below 14 %. Historically only about 1 out of 6 reported placebo responses have been above 10 % and 1 out 20 above 14 %. To me the chances look good.

Why one can think that p2a trial results are unlikely to be coincidences:
1. sPGA 0/1 shift 2 response was accompanied with sPGA 0 response percentage of 10.5 % which corresponds, historically, to PASI 75 and sPGA 0/1 response being in 30 to 60 % range. We have consistency of results in p2a, which is rare with placebo or chance.

2. Highest placebo sPGA 0 response ever reported is less than 2%. Highest sPGA 0/1 reponse reported for placebo is about 18 %. Either we have a record coincidence at our hands or prurisol p2a trial did show real efficacy. We can debate for ever about the REAL magnitude of the response. Did IPIX get lucky with prurisol response? If you think so, then you should also consider that 13.3 % placebo response is quite unlucky occurrence.

About small trials and p values:
As any reputable statistician knows p value scales inversely with trial size. What matters in small trials is the ratio of responses achieved, not the p value. Large trials are for showing statistical significance, small trials are for promising trend (ratio) detection. Prurisol 200 mg vs placebo response ratio reported in phase 2a is sufficient for statistical significance in a two arm trial having about 100 subjects. In prurisol p2b 300 mg dose arm and placebo arm are about 80 subjects, 400 mg arm is about 30 subjects.

This link is to very readable blog by a respected statistician. I recommend reading, at least, Statistic’s Dirty Little Secret and Why do we compute p-values?