Replies to post #147296 on Biotech Values

[DewDiligence]

Actually, it’s not uncommon for randomized cancer trials to allow for crossover following disease progression. Most phase-2 cancer trials are non-randomized, but that’s a separate issue.

[jq1234]

Generally Phase II oncology trials that are SOC+drug vs SOC don't have a crossover. Companies hate crossovers for obvious reasons (clouds any OS advantage), and such a trial without crossover is perfectly ethical in a situation like this where before the trial there was only very modest evidence that the drug had good efficacy.

That's not true. It depends on what primary endpoint is. For PFS primary endpoint, most ph2 randomized oncology trial allow crossover. The reason is very simple, patients are going to have subsequent cancer treatments anyway, it is better to offer placebo arm the study drug than you have no control whatsoever, plus you can further analyze the drug effect after crossover. As of clouding OS advantage, since the primary endpoint isn't OS, that isn't main concern. You still should see OS trend if the drug really works even with crossover. A couple of recent examples: ARQ197 + Tarceva in 2nd line NSCLC, THLD302 + Gemcitabine in 1st line PanC allowed crossover after progression. However, if OS is primary endpoint, then crossover typically is not allowed.

This is certainly very strange, given the trial has been unblinded and has no protocol that I know of which would interfere with placebo arm crossover.

The trial designed ORR as primary endpoint which is why no crossover occurred. It is very odd for them to design ORR as primary endpoint if they really believed the mechanism of action of the drug. By the way, it missed both ORR primary and PFS secondary endpoint.

At the end of the day, it's the OS numbers from this trial that will move the stock in the short/mid-term.

I don't think OS number by itself is key in this case, the key is whether the arms are balanced especially true in NSCLC. With ORR as primary endpoint, not sure they stratified enough factors during randomization to make the arms balanced. Without balance, the OS number doesn't mean much really.