Grading of images in a clinical trial

We developed a computer model to compare the power in clinical trials that use either of two grading schemes for objectively recorded dependent variables. The example we use is an eye lens photo. The first grading system consists of side by side comparison of the before and after treatment photos; the second involves separate grading of the two photos. This modelling experience suggests that separate grading may often yield more power than side by side comparison, and offers other advantages as well. For applications in which grading error is considerably smaller in the side by side method, power may be markedly better. Most importantly, strong prejudice in favour of one or the other approach in all situations seems unwarranted.