Testing a drug compound for efficacy increasingly involves simultaneously testing for efficacy in the entire patient population (broad efficacy), and efficacy in one or more pre-specified subgroups. Two scenarios leading to such testing are as follows. The first is that the mechanism of action of the compound might make it more beneficial to a subgroup of the patients. This is often true with cancer drugs, with recent examples such as crizotinib (Xalkori) for non-small cell lung cancer patients with ALK translocation, and vemurafenib (Zelboraf) for skin cancer patients with BRAF mutation. The second scenario is efficacy of a compound may be affected by polymorphism in genes in its metabolic and transport pathway. Hotly debated is whether polymorphism in CYP2C19 affects efficacy of clopidogrel (Plavix) and Tamoxifen, for example.
When a drug is approved for use by the entire patient population, there may be concern that this efficacy is driven by extreme efficacy in a subgroup only. If a drug is approved for use in a subgroup, due to the practice of off-label use, some assessment of its efficacy for patients in the complementary subgroup is also desirable.
Our research shows Partition testing for broad efficacy and efficacy in pre-specified subgroups readily recognizes logical relationships among the parameters in the null hypotheses being tested, and readily inverts to simultaneous confidence bounds on efficacy in broad population and in the subgroups. If these confidence bounds show efficacy in a subgroup is of real concern, then an appropriate statement can be included on the label.