Carl Heneghan

So there are two things when you measure blood pressure.

One is the machine.

and the machine has to be accurate but that can have some tolerance and actually the recognized standard for a blood pressure machine is about plus or minus five millimeters of mercury if it's more than that it doesn't get recommended so that's your analytical variability and the thing about that is a systematic error in effect it will go up or down but in one way

And then you've got your blood pressure, which is your biological variability, which can vary hugely by plus or minus 20 millimetres of mercury.

And in some people, even more, depending on what activities and what you're doing.

So one of the things with blood pressure to get a true measure is we take blood.

more measures in effect and you have to increase the number of measures to get an accurate measure over a period of time and use the average blood pressure so that's the bit about the serial measures but i think this is really helpful where you often see people and take the cholesterol the cholesterol is a good example total cholesterol they call it the reference change value is about 20 percent

How much does it have to change by for you to say this is a significantly different value?

So you've got a total cholesterol of five and you come back and say, I've got my new super diet and now I'm down to 4.6.

You go, that could actually just form within your biological variability and the analytical variability.

It has to go beyond 20%.

So it has to come down.

No, I haven't.

But I tend to go, oh, that's quite interesting.

But what's interesting about that is to know you shouldn't be changing treatments because the opposite can happen.

Somebody's cholesterol is 5 and it goes up by a certain amount, 10% to 5.5.

And you should go, hold on a minute, that's normal variability.

What the issue is, is a tendency to then pile in with a higher dose, which actually might not get you much reduction compared to the standard dose.

Yes.

And so what you really should do is say, hold on a minute, let's repeat the measure and get some serial measures to understand what the average is and whether this is just all noise or is it actually a signal?