08-21-2020, 08:51 PM
Here's their pre-publication paper: https://www.medrxiv.org/content/10.1101/...20176651v1
Note that they are planning an event-driven study (at least for an interval analysis). With 30,000 participants, they'd be powered to hit a 50% efficacy, with a lower bound of the confidence interval of 30% efficacy, (which is the FDA's requirement as best I can tell from the guidance) with 75 COVID-19 infections. If they get 50 events in the placebo group and 25 in the treatment group, the 95% efficacy confidence interval is 31 - 81%, and of course, that is a reduction of 50%.
To get that many infections, they would need about 0.17% of participants to become infected, so that is reasonable.
Ironically, the worse the outbreak, the faster we may have a vaccine. If the outbreak slows down (generally a good thing), it might delay the development of a vaccine.
BC
Note that they are planning an event-driven study (at least for an interval analysis). With 30,000 participants, they'd be powered to hit a 50% efficacy, with a lower bound of the confidence interval of 30% efficacy, (which is the FDA's requirement as best I can tell from the guidance) with 75 COVID-19 infections. If they get 50 events in the placebo group and 25 in the treatment group, the 95% efficacy confidence interval is 31 - 81%, and of course, that is a reduction of 50%.
To get that many infections, they would need about 0.17% of participants to become infected, so that is reasonable.
Ironically, the worse the outbreak, the faster we may have a vaccine. If the outbreak slows down (generally a good thing), it might delay the development of a vaccine.
BC