Abstract:
|
Phase II clinical trials are used to study possible efficacy of a new treatment and whether further research is warranted. The dominant statistical theory for these trials is Neyman-Pearson (NP) decision theory. With NP methods, investigators must conclude that the treatment is potentially effective and worthy of further research, or ineffective and unworthy of further research. Routinely, this decision is based on a single endpoint: response rate. Trials are designed to control erroneous conclusions in identically repeated trials. Exact error rates are estimated but unrealistic due to practical conduct and one may accept a hypothesis which is less credible than its competing hypothesis. Likelihood methods are proposed for phase II clinical trial design and reporting. Data analysis can occur at anytime, not just after an exact number of patients are accrued, providing more flexibility and shortening trial duration. One can calculate a measure of evidential strength, unavailable using NP methods, which depends only on observed results and not on trials which are not carried out. All endpoints, not just response rate, can be objectively analysed by investigators within each trial.
|