Scenario: We are looking at a 2 arm parallel group trial of two weight loss products, A and B. The trial is sample sized to detect a difference of 5kg in the mean loss of bodyweight between the two groups with the assumption that the mean reduction in Group A is 10kg and the mean loss in group B is 5kg. With a power level of 90% and a significance level of 5% it is calculated that 672 patients are required and with an expected drop-out/non-evaluable rate of 10% a total of 746 patients will be recruited. The data are assumed to be normally distributed with a standard deviation of 20.

Question: If the expected difference was reduced from 5kg to 4kg corresponding to mean reductions of 10kg and 6kg what would the effect be on the sample size?