Re: Interim looks in clinical trials
>I continue to find it odd that anyone thinks a trial failed when any interim look fails to lead to early unblinding. [Internal unblinding by the safety monitoring board for the interim looks should not be confused with unblinding to the public.]<
It’s a matter of degree.
If the p-value allocated to an interim look is minuscule—as is supposedly the case for GTOP—then missing the trigger for full-fledged unblinding is not bearish to any material degree.
If, on the other hand, a sizable amount of p-value is allocated to an interim look—as was the case, for instance, with BIOM’s Theratope trial a few years ago—then missing the unblinding trigger is indeed bearish. (You’ll recall that BIOM sold off precipitously after the first interim miss.)
This is no more than simple arithmetic. Missing an interim look with a sizable p-value allocation reduces the likelihood that the drug has truly compelling efficacy. Not only does this lower the probability of hitting the final endpoint (or subsequent interim endpoints), it may also call into question the commercial viability of the drug.
These days, most sponsors are smart enough to avert the kind of mistake that BIOM made: either they don’t schedule interim looks at all, or they allocate only tiny amounts of p-value to them.
--
An alternative that is rarely adopted is the “sequential analysis” design employed by YMI in the pivotal Tesmilifene trial. This is fundamentally different from the conventional trial design because each interim looks may cause the trial to be halted with a failed outcome as well as a successful outcome, and the trial continues until the matter is decided one way or the other.
Market Data 
Markets 
AI
