Helen MacDonald

what these might show.

And Carl kindly agreed to look at test number one to nine, which were looked at by this group, and give us some information on whether these antibody tests would give us hope.

Will you tell us, Carl?

So what's that good for then?

So do you have a sense, Carl, of whether those antibody tests, and I suppose you're taking...

the performance characteristics you took there of 70 and 98%, was that from one particular test or was that from a range across all of the different tests?

So we don't actually have that specific test right now, is that?

And to what extent do you think that matters?

There's another puzzling thing about this paper, which is that these tests are numbered one to nine.

And that was hard to get my head around.

And so I called on additional help.

I called John Deeks, who is a professor of biostatistics at the University of Birmingham and happens to be the BMJ's chief statistical advisor.

Because I'd noticed him getting a bit riled up on Twitter about these tests one to nine.

And it turned out he was the perfect person to ask about this problem.

I wish we'd gone to him weeks ago as well, because John is in the middle of doing multiple systematic reviews for Cochrane with a huge international team of people on lots of aspects of testing on PCR tests, on antibody tests, on point of care tests, but also on signs and symptoms tests.

And I asked him just to give us a sense of what the field of diagnostics on COVID-19 is doing and what he thought of this Oxford study.

A problem that I have found looking at some of this research is there just seems to be tens, if not hundreds of different tests available.

How do we begin to work out

what of this vast offering works.

Tell us why it's so mysterious.