Innovative Statistical Methods for Early Phase Clinical Trial

Early phase clinical trials provide some of the first indications of the efficacy of the experimental therapy at treating a disease. Often, the endpoint chosen to evaluate the treatment efficacy requires a lengthy follow-up time on each patient, which can prolong the length of these small studies. In early phase clinical trials, two-stage designs are commonly used for ethical reasons. They provide an early stopping rule that prevents researchers from exposing patients to a potentially inferior therapy. In the commonly used Simon's two-stage design, the lengthy follow-up time on each patient also causes a suspension in patient accrual at the end of the trial's first stage. We must wait to observe patient outcomes before making a decision to continue the trial to the second stage or stop the trial early. This suspension in patient accrual is disruptive and can cause the trial to lose momentum. The motivation for this work is to improve the efficiency of trial design where the efficacy determination can only be made after a long follow-up time on each patient.;We propose a two-stage design for a single-arm clinical trial with an early stopping rule for futility. Unlike Simon's design, our proposed design employs different criteria to assess early stopping and efficacy. The early stopping rule is based on a criteria that can be determined more quickly than that for efficacy, which helps to reduce the suspension in patient accrual at the end of the first stage. These separate criteria are nested in the sense that efficacy is a special case of, but usually not identical to, the early stopping criteria. This design readily allows for planning in terms of statistical significance, power, and expected sample size. This proposed method is illustrated with a Phase II design comparing patients treated for lung cancer with a novel drug combination to a historical control. In this example, the early stopping rule is based on the number of patients who exhibit progression-free survival at 2 months post treatment follow-up. Efficacy is judged by the number of patients who have progression-free survival at 6 months.;Curtailed sampling rules are introduced into this clinical trial design to allow the study to end as soon as a predefined statistical endpoint is reached. This helps to reduce the required number of patient enrollees and the expected duration of the trial. We have derived the distribution of the sample size of each stage of the trial under curtailed sampling, which describes the number of Bernoulli trials needed to observe a certain number of successes or a certain number of failures. Statistical properties of this distribution are provided, including the probability mass function, moment generating function, and likelihood function. A single-stage design with curtailed sampling rules is introduced as a simple example of the practical application of this distribution in the planning of curtailed clinical trials. This work will simplify the process of proposing and planning new designs with curtailed sampling going forward.;We also have developed an R package which provides tools to assist in the planning of this two-stage nested clinical trial design as well as the single-stage design with curtailed sampling. Visualization tools are provided to aid in selecting a design and functions are provided to compute the power, significance, expected sample size, and additional statistical properties of a chosen design.;The proposed two-stage nested design has the potential to substantially reduce the duration of these types of clinical trials, which is a significant factor in lowering clinical trial costs. Additionally, alleviating the length of the problematic suspension in patient accrual is another benefit of utilizing the nested design. This research provides the theoretical statistical framework and practical tools to assist in the planning of these designs.