Sample size and power for comparing two or more treatment groups in clinical trials.BMJ 1989; 299 doi: http://dx.doi.org/10.1136/bmj.299.6700.663 (Published 09 September 1989) Cite this as: BMJ 1989;299:663
- S. J. Day,
- D. F. Graham
Methods for determining sample size and power when comparing two groups in clinical trials are widely available. Studies comparing three or more treatments are not uncommon but are more difficult to analyse. A linear nomogram was devised to help calculate the sample size required when comparing up to five parallel groups. It may also be used retrospectively to determine the power of a study of given sample size. In two worked examples the nomogram was efficient. Although the nomogram offers only 5% and 1% significance levels and can be used only for up to five treatment groups, this is sufficient for most researchers.