Computer software that can calculate confidence intervals is now availableBMJ 1995; 310 doi: https://doi.org/10.1136/bmj.310.6989.1269b (Published 13 May 1995) Cite this as: BMJ 1995;310:1269
- Iain E Buchan
EDITOR,—Richard J Cook and David L Sackett usefully show applications of the number needed to treat statistic for clinical decision making.1 Confidence intervals are an important aid to meaningful inference from this statistic.2 Mention in the paper of the construction of confidence intervals for the number needed to treat is limited to their direct inverse relation with confidence intervals for the absolute risk reduction. In statistical terms, the absolute risk reduction is analogous to the comparison of unpaired proportions. A widely published formula for approximate confidence intervals for the difference between unpaired proportions is often used to calculate confidence intervals for absolute risk reduction and thence number needed to treat.3 With large numbers, such as those quoted by Cook and Sackett, this approximation is acceptable, but with smaller numbers it is unreliable.4
This problem can be overcome with computer software that uses robust iterative methods to construct these confidence intervals. Arcus Pro-Stat version 3.23 provides “near exact” confidence intervals for relative risk, relative risk reduction, absolute risk reduction, number needed to treat, and other statistics used in clinical epidemiology (Medical Computing, Aughton, Lancashire). Any reader who needs computer software for these calculations can obtain a royalty free copy of a limited version of Arcus Pro-Stat for IBM compatible computers by contacting me.
Log in using your username and password
Log in through your institution
Register for a free trial to thebmj.com to receive unlimited access to all content on thebmj.com for 14 days.
Sign up for a free trial