Sample sizes for cancer trials where Health Related Quality of Life is the primary outcome

Health Related Quality of Life (HRQoL) instruments are increasingly important in evaluating health care, especially in cancer trials. When planning a trial, one essential step is the calculation of a sample size, which will allow a reasonable chance (power) of detecting a pre-specified difference (effect size) at a given level of statistical significance. It is almost mandatory to include this calculation in research protocols. Many researchers quote means and standard deviations to determine effect sizes, and assume the data will have a Normal distribution to calculate their required sample size. We have investigated the distribution of scores for two commonly used HRQoL instruments completed by lung cancer patients, and have established that scores do not have the Normal distribution form. We demonstrate that an assumption of Normality can lead to unrealistically sized studies. Our recommendation is to use a technique that is based on the fact that the HRQoL data are ordinal and makes minimal but realistic assumptions.