Abstract:
|
Demonstrating that A is efficacious uses a two group design: therapy (a) and placebo (1), with efficacy estimated by a - 1. When another therapy, B, has known efficacy ethics suggests that the placebo group is unethical in these patients. A 22 factorial design {1,a,b,ab} would provide 2 efficacy estimates a - 1 and b(a - 1) which are pooled provided these therapies do not interact. Eliminating the placebo {a,b,ab} gives one estimate b(a - 1) which estimates A's efficacy provided the two therapies do not interact, a judgement that has to be made from therapy theory. These ideas extend to a higher number of 2 level factorial designs 2k (k = 3) where the placebo group is dropped there are 2k -1 estimates of the efficacy of the new therapy that can be pooled together provided the new therapy does not interact with any of the other therapies, although the latter can interact with each other. These designs permit efficacy estimation when a placebo group is unethical. Time will tell whether these designs will be tried, and the evidence becomes acceptable to the scientific community.
|