Moderna vaccine efficacy is simply astonishing.

Amazing.

In addition to 94% overall effectiveness, there was this gem from Moderna's release today about their phase 3 trials.

Of the 30 people who came down with severe COVID in their trial...30 were from the control group and zero from the experimental group. While the number is small, that suggests that protection from severe COVID is at least as strong as the 94% overall, and possibly 100%.

The number in the trial isn't important. It only determines how long the trial has to run in order to catch a decent number of infections.

They're looking for a vaccine that's at least 50% effective, meaning there would be twice as many infections in the placebo group. The numbers we're seeing are so overwhelmingly lopsided that even at these very low numbers it's statistically certain that the vaccines are far better than that. Roll a die 100 times and count the numbers of 1s and 2s you get. You can roll for the rest of your life and you'll never get 5 out of 100.

It wasn't certain at the outset how long it would take to get conclusive results. A better vaccine would give a clearer signal earlier, and a bigger outbreak would result in a larger sample sooner. The two coming together means that we get a vaccine a month or three sooner than we might have expected, and I'm very happy for that.

Trying to phrase it another way. If you want to be certain that your vaccine is at least 50% effective, i.e. 33% of the cases will be in the vaccine group and 67% from the placebo, you could wait until you have 10,000 sick people and then verify that fewer than, say, 2,900 were in the vaccine group. (Ballparking numbers for illustration, not precise.) At just 1000 sick people, having a 220/780 split might be enough. At 100, a 15/85 might be enough.

Getting 5/95 or so is such an overwhelming result that even with a very small sample, it's abundantly clear the vaccine has cleared the 50% hurdle with room to spare. It can't differentiate reliably between 95% effective and 90% effective, but it can sure tell that it's more than 50.

The one bias I'm aware of - and it's a huge one - tends to make the vaccines look less effective.

Because the vaccine tends to cause reactions in a large number of patients, patients who get a reaction have a strong suspicion that they're immunized, and likely immune. That could lead to more risky behaviors in the experimental group.

But the simple act of randomly assigning patients to groups when the patient and the person don't know what they're giving, seems like a well-understood system. Literally all they know is what # was on the vial they administered to patient X.