Some Contributions Of Statistical Methods For Oncology Clinical Trials

There are many challenging statistical problems in oncology clinical trials that need to be solved urgently.In this thesis we mainly focus on some problems in phase I and phase II clinical trials.The landscape has recently changed with the emergence of molecularly targeted a-gents and immunotherapies.These new the rapeutic agents appear more likely to induce multiple low-or moderate-grade toxicities rather than dose-limiting toxicities.To opti-mize the dose of these novel agents,it is important to account for the grade of the toxicity during dose finding.In Chapter 2,we propose a practical phase I trial design that accounts for toxicity grades under a unified framework as a generalization of the Bayesian optimal interval(gBOIN)design.Compared to the existing designs that account for toxicity grade,the biggest advantage of the gBOIN design is its practical simplicity and transparency.Implementation of the gBOIN design is as simple as that of the 3+3 design.The decision of dose transition for the gBOIN design only involves a simple comparison of the sample mean of the endpoint with two prespecified dose escalation and de-escalation boundaries that are derived based on a unified theory of the exponential family of distributions to minimize the decision error of dose assignment.We show that gBOIN has the desirable finite property of coherence and large-sample property of consistency.Numerical studies show that the gBOIN design yields good performance that is comparable or superior to that of some existing,more complicated model-based designs.In Chapter 3,we consider the problem of finding maximum tolerated dose with tar-geted dose-limiting toxicity rate in drug combination trials.We propose a novel Bayesian adaptive design which features adaptive local modelling and weighted learning along the sequential search path.The,method is robust in the sense that neither pre-specifications of marginal toxicity probabilities nor subjective priors for model parameters are required Extensive simulation studies show that the proposed method is comparable to some lead-ing methods in terms of correct selection and superior to them when the priors are mis-specified.The proposed model can be extended to the case of trials with more than two drugs.In Chapter 4,we propose a model-based design for dose finding in phase I oncology trials.It is based on a Bayesian stochastic approximation for quantile estimation.The design works for both single drug and combination drug with arbitrary number of agents The method features robustness in the sense that neither prespecification of marginal toxicity probabilities nor subjective priors for model parameters are required.We demon-strate the superiority to some commonly used methods by simulationIn clinical trials with time-to-event outcomes,it is of interest to predict when a prespecified number of events can be reached.Interim analysis is conducted to estimate the underlying survival function.When another correlated time-to-event endpoint is available,both outcome variables can be used to improve estimation efficiency.In Chapter 5,we propose to use the convolution of two time-to-event variables to estimate the survival function of interest.Propositions and examples are provided based on exponential models that accommodate possible change points.We further propose a new estimation equation about the expected time that exploits the relationship of two endpoints.Simulations and the analysis of real data show that the proposed methods with bivariate information yield significant improvement in prediction over that of the univariate method.