Just a few posts ago, we introduced the idea of NNT as being an ‘absolute’ measure of how effective a treatment is; that is, the number of folk needed to treat to get one extra good result, compared to something else. This can be used to balance against stuff that might be negative – such as side effects, or costs/resource uses.

An alternative way of measuring effectiveness is to look at the relative efficacy, with measures such as relative risks or odds ratios.

The imaginary trial of super8 vs factor VIII had 5/50 joint replacements in the super8 group, and 10/50 in the fVIII group.

EER = 0.1

CER = 0.2

ARR = 0.2-0.1 = 0.1

NNT = 10

Another way of looking at it is to look at the ‘relative risk’ of a joint replacement in the super8 group, compared to the fVIII group: this is 0.1 compared to 0.2, or half as likely.

RR = EER/CER = 0.1/0.2 = 0.5

The value of using a relative measure, as compared with an absolute one, is that

a) it usually sounds better (half the chance or one-in-ten will benefit … which would you prefer?)

b) for medicines, it usually reflects a fairly similar reduction across all risk strata (if the event rate were much lower in a different group of boys, say only 2/100 having replacements on fVIII, you can estimate the rate if this group has super8 would be 0.5 >> 1/100)

c) it keeps consistency with systematic review/meta-analysis results.

It’s disadvantages are

a) it usually sounds better (half the chance or one-in-ten will benefit … which would you prefer?)

b) it’s harder to use to balance up the risks vs benefits as adverse events are often something that happens ‘every so often ‘ e.g. 1:10,000 having anaphylaxis, or 1:100 developing haemolytic antibodies

c) it usually sounds better (half the chance or one-in-ten will benefit … which would you prefer?)

But when you’re appraised with both ways of expressing an effectiveness, you can get a better way of measuring up how good a treament might be for your patient group and finessing the answers.

– Archi