When you’re thinking about applying the results of a clinical trial, its’ often difficult to get a meaningful handle on the balance that should be made between the beneficial and adverse effects of a treatment. If the medicine gives pain relief from your laparotomy to 1 extra patient in every three that take it (NNT=3), but makes 1 in 10 significantly very very constipated (NNH = 10) then is it worth using?
There’s a coarse way to approach this: say ‘yes, of course it is’ and give the drug. Then there’s a purely qualitative way: ‘how would you like this drug that might be good at making your pain go away but might make you very sick?’. And then there’s another way which seeks to quantify the differences by asking “how much worse is realy bad constipation than being in pain?” Then take this value – say it’s twice as bad – and use it to adjust the estimates like this:
NNT for good thing vs. NNH for bad thing / relative importance
3 vs. 10 / 2
which is the same as or NNT 3 vs. NNH 5 ….
meaning it would be rational for this person to take the medicine as the chance of benefit (1 in 3) outweighs the adjusted chance of adverse effects (1 in 5). This method of adjustment of NNTs adds a way of individualising, in a rational rather than purely emotional way, the potential benefits and harms of therapies and further hones the use of raw ‘evidence’ in clinical practice.