ABSTRACT

Phase III therapeutic trials in oncology are conducted to compare the effectiveness of treatment regimens. In most settings an accepted standard therapy exists, and the motivation is the hope that a new treatment regimen E will prove to be superior to the standard therapy S in some respect, for example survival or response rate. Such trials are called superiority trials. Let ρ(X, Y) denote a parameter that characterizes the difference between outcomes with treatments X and Y. Without loss of generality, we assume that ρ(E, S) is parameterized so that ρ (E, S) 0 if E and S are equally effective and ρ(E, S)0 if E is superior to S. For example, for comparisons of response rates, ρ(E, S) might be rate difference PE – PS or the log odds ratio ln{PE (1 – PS)/[PS(1 – PE)]}, where PX denotes the response rate for arm X S or E. Similarly ρ(E, S) might be the log hazard ratio for S relative to E in comparing survival. In the hypothesis testing context, a superiority trial tests the null hypothesis H0:ρ(E, S)0 against the one-sided alternative HA:ρ(E, S)0.