Adjusting for baseline risk: Number needed to treat
Adapted with permission from Chatellier et al, 1996[1]
Background
The number needed to treat (NNT) to avoid a single additional adverse outcome is a meaningful way of expressing the benefit of an active treatment over a control. It can be used both to summarise the results of a therapeutic trial or series of trials and to help medical decision making about an individual patient.
If the absolute risk of adverse outcomes in a therapeutic trial is ARC in the control group and ART in the treatment group, then the absolute risk reduction (ARR) is defined as (ARC (− ART). The NNT is defined as the inverse of the ARR:
NNT = 1/(ARC − ART)
Since the relative risk reduction (RRR) is defined as (ARC − ART)/ARC, it follows that NNT, RRR, and ARC are related by their definitions in the following way:
NNT × RRR × ARC = 1
This relationship can be used to estimate the likely benefits of a treatment in populations with different levels of baseline risk (that is different levels of ARC). This allows extrapolation of the results of a trial or meta-analysis to people with different baseline risks. Ideally, there should be experimental evidence of the RRR in each population. However, in many trials, subgroup analyses show that the RRR is approximately constant in groups of patients with different characteristics. Cook and Sackett therefore proposed that decisions about individual patients could be made by using the NNT calculated from the RRR measured in trials and the baseline risk in the absence of treatment estimated for the individual patient.[2] The method may not apply to periods of time different to that studied in the original trials.
Using the nomogram
The nomogram allows the NNT to be found directly without any calculation: a straight line should be drawn from the point corresponding to the estimated absolute risk for the patient on the left hand scale to the point corresponding to the relative risk reduction stated in a trial or meta-analysis on the central scale. The intercept of this line with the right hand scale gives the NNT. By taking the upper and lower limits of the confidence interval of the RRR, the upper and lower limits of the NNT can be estimated.
- Case Studies
- Editor in Chief of the Norwegian Electronic Health Library discusses why Clinical Evidence was an essential purchase for their users
- Norwegian Electronic Health Library
- Norwegian Training Report - September 2007