Antibe Therapeutics Inc(Pre-Merger) ATBPF

RE:RE:RE:RE:RE:RE:RE:Thank you to the objective people.

[Actuarial]

If you do read ATE-346 study (the British one) in details, you would realize that they have tried their best to group these people uniformaly by age, gender, etc., before their trial . Though I agree that there must be other facotrs impact the trial, hardly it be uniformly (i.e., it would introduce more within variance, which would bring P-value up), agree?

Many drugs with p-value less than 0.05 in PII but they eventually failed in PIII. This doesn't mean p-value useless. Actually, it is due to FDA (or industry wide) statistical significant standard being too loose. The arbitary 0.05 is not the right standard. There have been many argument around this topic. In this ATE-346 case, P-value < 0.001, is truly statistical significant = the drug works. Actually, it works even better than it was expected. It might be something else derived from naproxen+H2S but beyond naproxen.

qwerty22 wrote: So you know that in that study the patients were told if they were recieving drug or placebo? Thats never been revealed. Double-blinding can be broken in many ways either thru the patient,or the person taking the results or the people doing the final analysis or any combo of them. Just one other error the drug was measured against naproxen not placebo.

We dont really need any more lessons on stats. The thing the stats guard against is the null hypothesis, that the two group are the same with a reasonable degree of chance. The very low p value suggests the results difference from the two groups did not arise from chance, it doesnt tell us what was the cause of that difference though. It may be the action of the drug, it may be that not being double-blinded impacted the result in some way, it may be that some other factor (say patient age, weight, disease progression) was not well controlled in this small sample. When we get this trials results we wont just get stat significance on pain relief we'll get it on many other characteristics of the patient group which will help to rule out different possible causes for the difference and hopefully leave us with just the action of the drug as the explanation for the differences.

I think the point you're failing to understand is a very low p-value only rules out the possibility the difference arose thru chance, it does not, in anyway, prove what factor did cause the difference. Many factors need to be ruled out to leave just the action of the drug as the cause of the difference. These include observer bias and the placebo effect from the trial being double-blinded. It also comes from other factors being well controlled (with their own stats tests) to show they werent the source of the difference, like the ones I mention above. Leaving the action of the drug as the source of the difference alone is the aim of this trial. It is a fact that many drugs that appear to have statistically significant difference lose that difference when tested in larger groups that happens because its very difficult to control all possible factors in a small sample, this larger group size should help with that.

Actuarial wrote: Double-blinded is important because it can differentiate true medicine efficacy from placebo (psychological relief). Some times, physicians give patients placebo for cetain symptoms and so often it works!

The trial was small but its p-value delivers so important information. In order to calculate P-value, or paired student t-test, we need two key measures, between variance and within variance. Between variance is average pain relief, that's 7 points. This number is the larger the better. Within variance is to measure consistency of the paired group pain level improvements. This number is the smaller the better. FDA requires P-value be smaller than 0.05 (i.e.,95% confident level) and ATE-346 P-value is less than 0.001. What it means? It means 7 points pain relief is not something placebo can do. And, the pain relief is so consistent across the 12 people. And therefore, though the sample size is small, it is a robust result.

The other purpose of PII is to identify best dose for PIII, and this is so obvious. If 100mg works, you won't take 250mg. But, we don't know it before this repeating trial.

qwerty22 wrote: There's no logical connection in your argument. Double-blinded is intended to mitigate observer bias, a stats test result does not prove observer bias did not give rise to result any more than it can prove the result came from the action of the drug, it just protects against the result arising from random chance once everything else has been well-controlled against. Not everything has been well controlled in this small trial.

But I'm not arguing the trial wont be a success, there are a number of lines of reasoning to lead you to think this drug is an active NSAID beyond the stats in that small trial. This larger trial meets the requirement from pharma execs to see efficacy in "hundreds" of patients (in a well controlled double-blinded trial). Its also going to give us the added info you talk about.

Actuarial wrote: Two reasons for repeating the test: 1. it was not a "double blind" test; 2. it was "same" dose test. PII is to confirm two things, one is efficacy (vs. placebo), the other is to determin best dose for PIII.

To be honest, for item1, I don't worry too much about it cause its P-value is < 0.001, which is far more than satisfy (0.05 FDA standard). While the sample size is small at 12, it also requires much more stable and consitent results to reach a p-value < 0.001. Paired student t-test indirectly adjust calculation for P-value with respect to sample size.

qwerty22 wrote: Of coutse 12 is too small thats why they are essentially repeating that experiment again now with larger numbers. Its not just statistical tests they need to satisfy.

Actuarial wrote: If you do look for facts, the fact is that 12 people show average pain relief score 7 with P-value < 0.01. I see why you would challenge 12 being too small because obviously you don't understand how P-value was calculated. Google "paired student t-test" and you will understand it is a fair test for both small sample or large sample.