Read on Twitter
Today’s Pfizer trial announcement is exciting, but it’s still just a press release.
As a trial methodologist I want more details, but I am cautiously optimistic. This is a step in the right direction even if it isn’t the end of the marathon.
As a trial methodologist I want more details, but I am cautiously optimistic. This is a step in the right direction even if it isn’t the end of the marathon.
One question in particular I would like answered: the press release says the denominator is trial participants who received two doses.
Does this mean that the >90% number is an attempt to estimate a per-protocol effect? If so, how have they done it?
Does this mean that the >90% number is an attempt to estimate a per-protocol effect? If so, how have they done it?
Using the more common naive per-protocol analysis approach will likely result in a poor estimate of effectiveness. And the resulting wrong estimate could be an over-estimate or an under-estimate.
Similarly, loss to follow-up could potentially have inflated their estimate if it occurred differentially between trial arms and I’d like to know how much drop out there was & whether and how they’ve adjusted for it.
Another thing I noticed in the press release: events were only counted if they occurred more than 7 (or maybe 14? i forget the exact number) days after vaccination. How do they handle the competing event of COVID within the grace period? And how many of those events were there?
Those are just my questions about how well the vaccine actually worked in the trial population. I also want to see data on who participated so I can assess how well the results of the trial might reflect what we expect in terms of how well the vaccine will work in other people.
And then there are questions about how well we can expect this vaccine to scale — it requires a cold chain, which means it has to be kept at cold temperatures at all times. That makes it very hard to distribute in rural areas or areas with irregular or interrupted electricity.
