Clinical Trials

❘❙❚■ Chapter 9 | Sample Size and PowerFigure 2. The relationship between the number of patients required per treatment group and the smallestanticipated treatment effect, with a 5% significance level and 80% power.50% –α = 5%β = 20%Anticipated smallest treatment effect (δ)40% –30% –20% –10% –n = 25n = 50n = 100n = 5000% –|0%|10%|20%|30%|40%|50%|60%|70%Response rate in the control group (π 1)|80%|90%|100%them to choose a sample size that is too small to offer a reasonable chance ofavoiding a false-negative error (a Type II error).Not all negative trials are due to insufficient power. In some cases it might be thatthe event rate in the control group was lower than expected or that there wereconfounding factors, such as changes to routine treatment methods during theduration of the study. A branch of medical statistics known as meta-analysiscombines the results from many such small studies to try to estimate a true meaneffect more closely. If this analysis shows that the new treatment has a favorablebenefit, then this should be verified by performing a larger, definitive RCT. However,one must always take into consideration the outlay of resources required to realizethe potential benefit, and even then, large RCTs might produce unexpected results.86