Biotech Values

[terry hallinan]

The numbers racket.

Wouldn't exactly be a shock if KERX's CEO was fudging the figures on its clinical trial. That's pretty much what CEO's do. They are after all paid promoters and Weiss appears to be superb at the job.

There is a more basic problem that few here seem to appreciate. I would be interested if anyone here makes his living as a statistician or has made a deep study of statistics. I noticed once when I was advised by a biostatistician to get a particular book on the application of statistics of clinical trials that the 3 or 4 books in our local library were rather dusty. He actually expected me to buy a book but I didn't see why I had to be the only one who did such an unheard of thing as actually study the theory instead of the mechanics.

"Razor" (Rosario) was a long-retired, aging chess player in Utica who spent much of his retirement playing dozens of correspondence games. Razor was a regular at the chess club when there was a blub. He wasn't a great player, at least over the board, but the challenge was the thing.

Once Razor figured some odds for a local bookie and then was forced to fend off repeated offers he couldn't refuse.

The bookie made his living quoting odds though he was obviously mathematically illiterate.

Way it goes with statistics.

A 95% confidence interval (CI) is the gold standard for trials for the FDA. No reason really except that's the custom.

I was a bit surprised when the FDA proposed that if they allowed drugs to be approved on the basis of a single controlled trial with 95% CI for efficacy, then there was a 1 in 5 chance roughly that the drug was no better than a placebo.

Not too surprised but a little. The bookie could have told them where they went wrong I bet.

One could rationally make an argument that it doesn't matter what the p value is. Volunteers in a clinical trial are not mere digits. The biostatistician makes things fit that don't.

We used to do that.

A stereo photographic madel had to be adjusted to ground truth for charting. Once a control point in an arctic area was picked on a low lying cloud. We made it fit ground truth. Wasn't easy but if you have a computer and a photogrammetrist that is handy with numbers, you can make anything fit. Well, most anything. The cloud was relatively easy.

I wish, I wish upon a cloud that I could make things fit like the good old days but you lose the knack.

Do numbers get better as well as worse?

Yes.

Do promoters change?

No.

Best, Terry