PGNX, does Azedra really achieved primary endpoint?
Progenics released long expected Azedra pivotal trial data, the data is disappointing or "concerning" which is exactly the situation as Mr. Baker described in the earnings CC in May 6th, 2015. It technically met the SPA with a surprisingly poor performance in the new cohort of patients. More "concerning", during the CC, Mr. Baker repeated ask people to disregard the difference between the two phases, which is completely opposing his previous position.
Should FDA follow his suggestion?
Absolutely Not. Before combining the two cohorts to evaluate the overall performance as whole, the two stages MUST be separately evaluated to ensure comparable performance. This is because the two stages used products from two different manufacturers. The manufacture and distribution are major challenges to the quality and consistent performance of radiopharmaceuticals. Given Iodine-131 has a half-life of 8 days. One day change in shelf will change the relative activity of the product. This is the reason why Progenics took a whole year to work with a new manufacturer (CPDC) to make the product. Notably, the new manufacturer improved (changed?) the procedure to "manufacture small batches of radiopharmaceuticals in a cost-effective manner". Whatever intention of the change, for an radiopharmaceutical product, there is no way FDA will ignore the possible difference from different manufacturers, especially with modified procedures. In fact, in all previous presentations and conferences, Mr. Baker repeated stating that the new phase will enroll "a few more" patients and intended to see the drug is as active in Progenics as in MIP. This is clear that the comparison between these two stages is not only meaningful but essential.
What really is the difference? Let's crack some numbers
As for primary endpoint,
In MIP, 13 responders out of 41 patients, or 32%
In PGNX, it is 4 out 27 or 15%, less than half.
Though, combined, it turned to be 17 out of 68, exactly hitting the 25% bar.
As for secondary endpoint of ORR,
In MIP, 12 PR out of 38, or 32%
(Note: later 2016 presentations cited 14 out of 34, which seems adapted from long term followup study and not match the total patient numbers in combined analysis, anyway this will only make the new cohort looks worse.)
In PGNX, it is 3 out of 26, or 12%, the new cohort has less than 40% the response rate. (even in the two doses subpopulation, the PGNX product is only half of MIP activity.)
The real "concerning" is that
In MIP, 34 received 2 doses and 7 received 1 dose.
In PGNX, it is 16 for 2 doses and 11 for 1 dose.
This is a statistical significant difference (p=0.048)
In the CC, Mr. Baker explained it using sick patients and toxicity. When Johnathan asked Mr. Baker to clarify further whether the patients population are different between PGNX and MIP stages (indicating elder/sicker patients), Mr. Baker refused to confirm it (because it is not the reason?!) but asked him to disregard the difference between the cohorts. To me it is clear that substantially higher toxicity weighs in the poor performance in PGNX newly manufactured product.
There is another interesting disclosure during the CC, that there is another primary endpoint cutoff of lower CI95>10%.
At first look, it does not make any sense.
You see, the CI95 for proportion is fixed once you have n=58 and 25% (or whatever fixed n and target %). Actually the Lower CI95>10% is No-Bar when n>41, assuming this endpoint is indeed applying to the combined population. Even if all 17 new patients fail (then it would be the original 13 responder out of total 58) will still give the lower CI95> 13.6%.
But if you consider how FDA would reinstate the SPA and establish an quantitative measurement to make sure the newly manufactured drug can achieve a minimal meaningful activity in PGNX phase, they might ask, in addition to the combined population achieving the 25% activity, they will want to see the drug to achieve the low CI95>10% in PGNX cohort. This would be 4 in 16 (25%!), 5 in 22 (22%) or 6 in 28 (21%) etc. This way the regulator can ensure a minimal trustable 10% activity within the finally approved manufacture product. You cannot expect more accommodating the regulator could be, regardless how desperate the condition is. And this is the real reason, I guess, why Progenics want to enrollment "a few more" patients.
Unfortunately, the data failed to meet this low bar. Given 4 out of 27, you have lower CI95 as 5.9%. So in real population, the result can only ensure 6% patient may benefit from the treatment on the price of high toxicity of 40% discontinuation.
What regulator would approve such a product? and
Did Progenics really achieved primary endpoint, even technically?