Skip navigation

Is It Worth Doing?

Bandolier has panic attacks over the loose use of the word effective. Numbers-needed-to-treat (NNT [1]) is one way to define effectiveness. NNTs are easier to interpret than absolute and relative rate reduction. Two examples may help. Using flu vaccination the reduction in relative risk was 0.42. This translates to an NNT of 23 compared with not vaccinating. One case of flu was prevented for every 23 patients vaccinated. In a powerful review of matters cardiovascular [2] the AIRE study of angiotensin converting enzyme inhibitors after myocardial infarction [3] is cited as "one of the most important clinical trials in cardiovascular medicine to be published for many years". 2006 patients with clinical evidence of heart failure were randomised to ramipril or placebo (minimum follow up six months). There were 170 deaths in the 1014 patients randomised to ramipril compared with 222 deaths in the 992 patients randomised to placebo. This relative rate reduction of 27% translates to an NNT of 18. One death was prevented for every 18 patients treated.

Both examples are of preventative interventions. It is important to appreciate that NNT is by definition context-dependent. It is flu vaccination versus no vaccination and ACE inhibitor versus placebo. The NNT would change if comparisons were different, for instance one flu vaccine versus another. For a common disorder a cheap preventative intervention with minimal adverse effects and an NNT of 50 may well be worth doing. If there are substantial adverse effects then it may not be worth doing. For a serious condition then a risk of significant adverse effects may be worth taking. NNT makes it easy to quantify this clinical decision-making, because NNT for adverse effects can be calculated as well as NNT for effectiveness. NNT for drug-related study withdrawal in the AIRE study was 18 for ramipril compared with placebo. Like Bandolier , patients may find NNT easier to understand. "By taking this medicine you will reduce the chance of dying from your heart problem by xx; the chance of your not being able to tolerate the drug is yy" . The next twist is to use NNTs for treatment as well as prevention.

We have a simple Excel template for these calculations. We will happily provide this over our imminent Internet connection hoping that numerically literate readers will improve on our crude efforts.

1. Sackett DL, Haynes RB, Guyatt GH, Tugwell P. Clinical Epidemiology: a basic science for clinical medicine. 2nd edition. Boston: Little, Brown; 1991.

2. McMurray J, Rankin A. Cardiology - I: Treatment of myocardial infarction, unstable angina, and angina pectoris. British Medical Journal 1994 309:1343-50.

3. Acute Infarction Ramipril Efficacy (AIRE) study investigators. Effect of ramipril on mortality and morbidity of survivors of acute myocardial infarction with clinical evidence of heart failure. Lancet 1993 342:821-8.

next story in this issue