The number needed to treat (NNT) is an increasingly popular way of presenting the effects of treatment. However, the NNT varies markedly depending on the baseline risk of patients, the outcome considered, and the clinical setting. Furthermore geographic and secular trends make the NNT unstable between places and over time. Particular caution is needed in deriving the NNT Keywords: meta-analysis; hypertension; randomised controlled trials Eight hundred and thirty-three. This is the number of mildly hypertensive people who must be treated with antihypertensives for a year to avoid one stroke 1 and is widely quoted as an argument for not bothering to detect and treat hypertensive people. Why has the number needed to treat (NNT) become so popular recently? Is it because it can be used to highlight the workload and cost implications of treatments? Is it because the trade off between benefits and risks of treatment is made more obvious? Or is it just the novelty value of a new(ish) method of expressing the effects of treatment? Despite the promotion of NNT as easier for clinicians to understand, when presented with a range of measures of treatment effects, most found NNTs no easier to understand or explain to others than relative risk. 2
Describing the effects of treatmentThe effects of treatment are conventionally expressed in relative terms-the ratio of the event rate in the treatment group divided by the event rate in the control group. Ratios greater than one imply that treatment is harmful, less than one, that treatment is beneficial. This hides the fact that a small relative benefit applied to patients at very high risk will generate more lives saved (and non-fatal events avoided) than the same relative benefit applied to much lower risk patients. However, the relative treatment effect may be dependent on the baseline level of risk in the control group. For example, surgery for carotid endarterectomy is effective for high risk tight stenosis but not for low risk smaller degrees of stenosis. 3 It cannot be assumed that treatCorrespondence: Shah Ebrahim, Department of Social Medicine, University of Bristol, Canynge Hall, Whiteladies Road, Bristol BS8 2PR, UK Received 18 May 1999; accepted 22 May 1999 from pooled absolute risk differences in meta-analysis. In general, the NNT should be calculated by applying the relative risk reduction on treatment estimated by trials or meta-analysis to populations of specified absolute high, average or low risk to illustrate a range of possible NNTs. ment effects are constant across a wide range of baseline levels of risk.The absolute effects of treatment take account of baseline level of risk and are simply the difference between intervention and control group rates. From a clinical and public health standpoint it is useful to have some idea of the amount of effort required to avoid one adverse event. This is given by the reciprocal of the absolute risk difference-the number needed to treat. 4 The absolute effect and the NNT will vary-often greatly-according to the baseline leve...