NNT or LBHH?

I was interested in an apparent inconsistency in the article in Bandolier 21 on the presentation of statistical information about the effects of treatment. Although you conclude that the Number-Needed-to-Treat (NNT) is becoming "the presentation of choice", you suggest earlier in the article that the "Dear Doctor" letter about oral contraceptives would have been more useful if it had expressed the risks associated with them using the "clinically interpretable numbers 1 in 6667 and 1 in 4000".

I agree with you. The NNT is an excellent way of presenting statistical information for those who are considering policies; but I do not think it is as helpful to individual clinicians or recipients of health care as the "Likelihood of Being Helped or Harmed" (LBHH) [1]. Clearly there is scope for some more empirical research here to find out which among the substantial numbers of opinions on these matters - including that expressed above - can be supported by evidence

Iain Chalmers
UK Cochrane Centre, Oxford.

I Chalmers. Applying overviews in meta-analysis at the bedside. Journal of Clinical Epidemiology 1995 48: 67-70.

previous story in this issue