Antigenics Reports Phase 3 Results for Oncophage in Kidney Cancer; Study Did Not Meet Endpoints; Number of Events Insufficient for Analysis
The news that Antigenics’ Oncophage failed to meet its primary endpoint won’t come as a surprise to hedge fund managers, very few of whom owned long positions in the company. Indeed, the stock has been a favorite short for some time, with an astounding short ratio (the number of shares of a security that investors have sold short divided by average daily volume of the security over a defined period) over 27 in February according to Yahoo! Finance (short ratios up to 10 aren’t unusual for development-stage biotech companies awaiting Phase 3 data).
However, based on my discussions with fund managers, I believe investor pessimism was fueled more by general skepticism about the therapeutic approach and the competence (or lack thereof) of Antigenics’ management rather than anticipated failures of clinical operations. Yet, it appears from the press release that at least one operational failure contributed to the study outcome: a conflict in numbers between investigator-reported events and events as adjudicated by an independent events committee. As this was an event-driven analysis, such a conflict proved fatal to making a definitive statistical inference. How does this happen? Having no facts other than the press release to guide me, I can only speculate on what went wrong. However, there are only a few possible operational failures that can result in this outcome. Here is the short-list of possible operational breakdowns that can lead to a “substantial” discrepancy between unadjudicated and adjudicated clinical endpoints:
- Poorly defined, ambiguous or poorly communicated (including language barriers) endpoint definitions, resulting in different endpoint interpretations (i.e. among investigators, between investigators and adjudicators, or a combination of both);
- Difference of available information to investigators and adjudicators (i.e. information asymmetry);
- Difference in clinical practices (culture), equipment, or personnel among investigators that could not be overcome by training;
- Insufficiently audited investigators (i.e. lack of oversight);
- Insufficiently trained or incompetent investigators;
- Insufficiently trained or incompetent adjudicators;
- Investigator fraud or related misconduct;
- Adjudicator fraud or related misconduct;
- Difference of clinical judgment without one of the above factors.
Generally speaking, the first five factors are more common than the last four. I’d also add that development managers must be aware that FDA can and does (usually) perform its own endpoint adjudications, particularly when there are important differences between the company’s own unadjudicated and adjudicated data. If I were advising Antigenics, I’d first find out whether the results are substantially better for Oncophage when relying on unadjudicated data only. If the results are substantially better, so much so that they could result in an approval, then it is likely wise to try an NDA (assuming the drug is approvable otherwise) and devote substantial resources to investigating and explaining why there was a discrepancy. It would be far cheaper and faster than doing another clinical trial.
