Read on Twitter
Since I’m on the topic of the possible limitations of COVID vaccine trial design..
1/ https://twitter.com/drecwalsh/status/1283577605686472705
1/ https://twitter.com/drecwalsh/status/1283577605686472705
And clearly exorcised enough to demand Gantt charts from people after 8pm on a Wednesday..
2/
2/
Probably worth us breaking down the likely limitations of these upcoming Ph3 studies as well, right?
3/
3/
First off - key to mention - these Ph3 studies are difficult in the best of times. Difficult to design and execute. Much harder when the hopes of the world are pinned on them. So nothing in this strand is a critique..
4/
4/
Ok so what’s the approach:
30K volunteers
- great, big trials means greater numbers of participants with all demographic backgrounds. Important to understand safety in everyone not just the young (18-55)and healthy who are typically in Ph1/2 trials..
5/
30K volunteers
- great, big trials means greater numbers of participants with all demographic backgrounds. Important to understand safety in everyone not just the young (18-55)and healthy who are typically in Ph1/2 trials..
5/
Also:
Placebo controlled
- 1 trial lists a 1:1 randomization stratified across demographics. E.g. this means that randomly getting a bunch of young folks on treatment & older folks on the placebo won’t happen (which is great given how that could confound interpretation)
6/
Placebo controlled
- 1 trial lists a 1:1 randomization stratified across demographics. E.g. this means that randomly getting a bunch of young folks on treatment & older folks on the placebo won’t happen (which is great given how that could confound interpretation)
6/
Further:
Masked trial
- quadruple masking in one case such that Participant, Care Provider, Investigator, and Outcomes Assessor don’t know what group the volunteer is in...
*but*
7/
Masked trial
- quadruple masking in one case such that Participant, Care Provider, Investigator, and Outcomes Assessor don’t know what group the volunteer is in...
*but*
7/
Given the data from earlier stage trials:
- https://www.medrxiv.org/content/10.1101/2020.06.30.20142570v1.full.pdf
- https://www.nejm.org/doi/full/10.1056/NEJMoa2022483
- https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(20)31208-3/fulltext
- and still looking for the initial data on this one (if anyone has it) https://www.bloomberg.com/news/features/2020-07-15/oxford-s-covid-19-vaccine-is-the-coronavirus-front-runner
8/
- https://www.medrxiv.org/content/10.1101/2020.06.30.20142570v1.full.pdf
- https://www.nejm.org/doi/full/10.1056/NEJMoa2022483
- https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(20)31208-3/fulltext
- and still looking for the initial data on this one (if anyone has it) https://www.bloomberg.com/news/features/2020-07-15/oxford-s-covid-19-vaccine-is-the-coronavirus-front-runner
8/
It would seem that a majority of volunteers (across platform technologies) has a “local” and/or systemic reaction to the vaccination. And that these reactions are not seen as frequently in placebo controls (although not every trial used these).
9/
9/
What this means is that even with excellent masking in the study, volunteers will be experimentally “unblinded” due to these reactions.
10/
10/
And given this, they may alter their behaviors after injection accordingly (I sure would go out less if I felt nothing after two shots of a “treatment” in a 1:1 randomizes vaccine trial.. I would presume I was in the placebo group).
11/
11/
This means, unfortunately, that while comparisons for objectively measured endpoints (post injection clinical signs/adverse reactions, antibody generation, post-infection clinical consequences) could still be meaning meaningfully compared between treatment &placebo arms...
12/
12/
The rate of infections may be very hard to compare meaningfully if the behavioral risk for exposure is different between groups due to the unintentional physiologic unblinding.
13/
13/
Now in a case of a 100% effective vaccine this is no problem. Because you will see 0 cases in the treated group and some number even among the compliant social distancers in the placebo.
However this is frankly unlikely based on what we know today about the virus.
14/
However this is frankly unlikely based on what we know today about the virus.
14/
Previously discussed, challenge trials which volunteers sign up for a vaccine followed by an intentional dosing with COVID could help to fill the gap of information here (at least for the demographics covered in the study).. even if they won’t “accelerate” the process...
15/
15/
But another analysis we will likely all be doing in 6 months is a comparison between infection rates in treatment arms of different vaccine candidate trials.
My sense is that with @FDACBER Peter Marks involved this will be as possible insofar that it is tactically possible.
16/
My sense is that with @FDACBER Peter Marks involved this will be as possible insofar that it is tactically possible.
16/
However each trial will likely be run in a different geography with different underlying prevalence and different exposure risks. So this may be harder than we anticipate particularly among the patient populations that need the vaccines most (65+ or underlying disease).
17/
17/
It would be great if ahead of these trials reading out, we could have a real statistical “think” about how powered they will be for these populations and what differences in behavioral exposure avoidance we can tolerate between groups and still hit the 50% efficacy target...
18/
18/
Laid out by the FDA
https://endpts.com/peter-marks-on-covid-19-vaccine-efficacy-euas-and-challenge-trials/
End/
https://endpts.com/peter-marks-on-covid-19-vaccine-efficacy-euas-and-challenge-trials/
End/
