Design and analysis of two-period studies of potentially disease-modifying treatments

A number of two-period designs have been introduced in the neurological clinical trials literature to evaluate the effects of treatment on progressive diseases, using names such as withdrawal, active-extension, randomized withdrawal, randomized start, and staggered-start designs. After parallel groups complete the first period, treatment is either initiated or withdrawn in the second period in a subset of subjects. The purpose of the second period is to provide evidence of whether any effect of treatment observed during the first period is long-term ("disease-modifying") or short-term and transitory ("symptomatic"). A four-arm version, which we term a complete two-period design, has active/active, active/placebo, placebo/active and placebo/placebo respective treatment assignments in the two periods. We provide statistical models for these designs, describe some of the appropriate analyses, and investigate the relative efficiencies of various allocations and special cases. We describe extensions to full and partial factorial versions of such designs which permit efficient and simultaneous evaluation of disease-modifying and symptomatic effects of two or more treatments, along with possible interactions. Advantages and limitations of the various designs are discussed.