The NNT – number needed to treat – is thought by many to be a central core of EBMing. (I’d argue that the numbers in EBM are only a small part of the equation – and that understanding the question and the biases inherent in the study design are greater – but I’d still agree it’s important.)
So – what’s the NNT?
The number needed to treat follows the same PICO format as the study is arises from. It is the number of patients needed to treated with one intervention, comared with the other, to prevent (or produce) one ‘extra’ outcome.
Consider an imaginary trial of super8 [i] vs. normal factor VIII [c] for preventing joint replacements at 10 years [o] in 8-14yrs old boys with haemophilia [p].
There are 5 / 50 replacements in the super 8 group, and 10/50 in the fVIII group.
This can also be expressed as super8 ‘experimental event rate’ (EER) of 0.1, and a fVIII ‘control event rate’ (CER) of 0.2
The absolute difference in rates (absolute rate reduction) is 0.2 – 0.1 (= 0.1)
This means for every hundred patients treated with super8, 10 of the, would not have their joint replaced who would have had if they had been treated with fVIII.
Alternatively, for every 10 patients treated with super8, one would not have their joint replaced who would have had if they had been treated with fVIII — the ‘number needed to treat’ is 10.
(The shortcut to this is 1/ARR = NNT)
Now, the reduction can also be expressed as a relative proportion … but that’s for another post.