Hacker News new | past | comments | ask | show | jobs | submit login

Looking at the original paper, the author does the analysis looking at breakthrough drug trial design, but doesn't really talk about what it means.

Of the 46 drugs approved under breakthrough designation, 25 were for oncology. The FDA often allows for a lot of flexibility with trial sign in oncology due to the high unmet need.

The author mentions a lack of trials being double-blinded, having a placebo or active comparator and clinical outcome vs. surrogate marker.

All of this is quite common outside of the breakthrough drug designation. Nothing new here. Double-blinding a trial isn't as big a deal when what you are measuring is not impact by observer bias (tumor size). In addition, if you're measuring a surrogate marker (that is an accepted proxy for clinical outcomes), you don't need a comparator arm.

I don't think any of these findings should be all that surprising or concerning.




> if you're measuring a surrogate marker (that is an accepted proxy for clinical outcomes), you don't need a comparator arm

Why is this the case? Wouldn’t there be concerns that a study population might be different from the general population and thus require an internal control?


Here is a good example of FDA guidance on the issue.[1]

FDA acknowledges that FVIII levels can serve as a proxy for clinical outcome (reduced bleeding episodes) in hemophilia A.

Our understanding of what is "normal" in FVIII blood levels is such that a single arm, surrogate biomarker study is sufficient for approval.

[1]https://www.fda.gov/downloads/BiologicsBloodVaccines/Guidanc...


Could you help by pointing out where they talk about not obtaining baseline/control data? After a quick glance, they only seem to mention using a within-subject design.


Sorry, misread your reply. Yes, they do an internal baseline/control, but don't have a control arm (either placebo or active comparator) in the trial.

That was the focus of the JAMA article.


>"Double-blinding a trial isn't as big a deal when what you are measuring is not impact by observer bias (tumor size)."

Have you ever tried to measure something like this and analyze the data? There are so many ways to skew the results...

EDIT:

That's not even to mention basic stuff like handling the treated rats more carefully so they are less stressed, etc.




Applications are open for YC Winter 2022

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: