Minimising errors in clinical studies by proper selection of sample size.

In the planning of a clinical trial, the statistical aspects of the choice of sample size should be considered. The meaning of statistical terms such as Type I and Type II errors, significance and power are explained. Formulae are given for calculating the required sample size for parallel groups' designs (two independent samples) and matched pair or cross-over designs for both quantitative and qualitative responses. Samples size tables are given with instructions for their use and there are four worked examples. The sample size for a trial depends on the specification of the levels of significance and power and the size of the difference the trial is designed to detect, and these can be somewhat arbitrary. Furthermore, it is often necessary to estimate the standard deviation on the basis of limited information. Thus, the sample size required for a trial cannot be determined precisely, and there can be no exactly right or wrong sample size. However, sensible consideration of sample size should ensure that the trial is not so small as to make the results trivial, nor so large as to be unethical or a waste of resources.