“NOT DISAPPOINTED”

This short video by Bronx-born novelist Jake Lamar is just . . . well, lovely.  No shouting, no ominous music – or soaring music, for that matter – just quiet common sense.

Post it on your Facebook page?  Retweet me?

(Retweet, Suh?  Never!)

(Sorry; new to Twitter, still amused by the puns.  But seriously: watch the clip.)

VOLUNTEER

Looking for something great to do?  On serve.gov, enter your zip code and areas of interest and up come the nearest volunteer opportunities.  Very neat.

IF YOU OWN CBRX

Almost a year ago, I posted:

Guru thinks it could be $5 within a year, so I bought a bunch yesterday at $2.55.  Guru is often but not always right.  So only with money you can truly afford to lose.

Two months later:

With CBRX closing at $4.10 in after-hours trading, Guru writes:  “The data they presented look great.”  PROCHIEVE, he thinks, will cut health care costs for some premature births.  And it “reduced the incidence of respiratory distress syndrome by 60% – a huge finding, as this is a major source of morbidity and mortality among pre-term babies.”

So despite the allure of a 60% two-month gain, I’m in no rush to sell.

I was – with the stock having dipped to $1.10 Tuesday – an idiot.

Guru:  “It’s a good buy again!  Tuesday, the FDA released its analysis of CBRX’s Phase III data for Prochieve to prevent pre-term births.  In the documents, the FDA acknowledges that in April 2010 (before the study was unblinded), they met with CBRX and agreed on the statistical analysis plan.  The study was then unblinded and the study met its primary endpoint with a p value of 0.02 using the pre-defined statistical test agreed to with the FDA in April 2010.  (Less than 0.05 is the threshold for success.)  Using a different statistical test that was also pre-defined at the April 2010 meeting and deemed a ‘secondary analysis,’ the data were also a success (just barely) at p = 0.044.  The FDA acknowledged all of this yesterday. . . . However, yesterday, the FDA went on to do a secondary analysis, comparing the US patients (about 50%) and the non-US (about 50%) and again, using the predefined primary statistical test, the p was 0.033.  However, the FDA’s statistical reviewer said that given how the study  turned out, there were not enough people in various strata to make this test the best one to use.  (Perhaps this point is correct, but the FDA very well knows that the rules of the game must be specified in advance and that we can’t change the rules once the game is over.)  Instead, the statistical reviewer decided that the second statistical test was the better one and produced a p value of 0.056 – just missing technical statistical significance.  The FDA’s report then went on to break down the study by regions within the US and by different countries and to show that there was a disparity in how the drug worked.  In one of the four regions of the US, the placebo actually worked better than the drug.  In one of the four hospitals in India, the placebo worked better than the drug.  In some countries, the drug worked much better than in other countries. . . . The problem with this analysis is that the study was not designed for it to be done:  You can’t start looking through small subgroups and still make valid conclusions.  In any given hospital, only a few patients participate in the study and there may be factors NOT affected by Prochieve that overwhelm the effects of Prochieve’s efficacy.  It is true that the drug appeared to perform better outside the US than inside the US, but the centers outside the US were widespread (India, Ukraine, Chile, South Africa, Czech Republic), so there is no obvious systematic factor that would bias in favor of Prochieve.  The women in the US, on average, were older and much heavier (both independent risk factors for pre-term birth), which may have biased against efficacy, though overall, in the US, the treated patients always did numerically better than the placebo patients.  The FDA had no safety concerns. . . . I have seen this situation several times:  The FDA agrees on an analysis plan in advance with the company, then the study is performed and then an FDA reviewer does additional non-planned analyses and questions are raised.  For example, in HGSI’s Benlysta trials for lupus, an analysis of the US population alone also showed no evidence of statistical significance.  In DYAX’s pivotal trial, which succeeded, the FDA reviewer pointed out that the product worked from 0 to 2 hours, but not from 2-4 hours (where the primary endpoint was 0 to 4 hours).  In ACOR’s phase III trials, the agreed-upon analysis was a responder analysis, but the FDA chose to do an analysis of averages and reported no statistical significance.  In all three cases (and there are others), the FDA panel voted ‘yes’ and the FDA approved the products in the end. . . . In my experience, the panel always votes what is the primary prospective data, not the post-hoc subgroup analysis.  In the case of CBRX, the primary prospective data met its endpoint.  I expect the panel to vote ‘yes.’  I also expect that in the end, the FDA will give Prochieve approval.  If so, the stock should rally back close to 3/share.

☞  Which will be good, if it happens, but I was still an idiot not to sell with a two-month 60% gain.

 

Comments are closed.