Carl Heneghan

And if you've got them all, you're high quality.

And if they're not there, you're low quality.

And I think bias is just a much more interesting issue and subject, and it's not been studied as much as some of the other issues like the analytical aspects of trials, because I think it's a very interesting widespread issue, and it's much more than just blinding.

And I'll give you an example.

We've just been publishing our catalogue of bias, and we've got about 65 biases on there already.

And I think when you look at a trial, it's about the overall assessment of all the biases that are contained.

And blinding sometimes is incredibly important.

And I'll give you one of the examples, for instance, was the trials of knee arthroscopy, where they used a sham procedure.

So they actually put the patients in the control group to sleep, give them a little nick, and said, you've had an operation similar to the intervention group.

And that showed no difference.

The trials with the absence of that actually did show there was a difference in the intervention group with the surgery.

Whereas sometimes I think, particularly non-drug interventions, it doesn't matter.

And I suspect in certain times in drug interventions, you can be using a pragmatic design where you're trying to match what actually happens in clinical practice.

It's important.

to see blinding in the context of all the biases.

Yeah, and I think understanding biases and the implications on the effect, because really what you're saying, given this effect, do we believe it or not?

To what extent is it biased by issues within the trial design or the population we've used or the measures we've used?

And I think what's happening here is we're starting to think bias is much more complex, is really a skill in assessing it.

And I totally agree.

I think a section which sets out and says, to what extent is this result uncertain, would be much more useful than what we see now is, oh, the overall assessment of the quality of evidence was moderate.