The Dose-Finding Method For Single-Agent And Two-agent Combination In Phase ? Clinical Trials

Clinical trials have always been one of the hot topics in biomedical research,which have played a critical role in the development of a novel drug.Clinical trials can generally be classified into four sequential phases.Phase I clinical trials mainly focus on the safety and toxicity profile of the investigational compound,which study the dose-finding problem of a novel drug and find the maximum tolerated dose of the drug or the maximum tolerated dose combination of the drug combinations.Phase II clinical trials will be undertaken to examine the efficacious activities based on a short-term efficacy endpoint.If the test drug shows promising anti-disease effects,the study will then move forward to a large-scale phase ? trial for confirmative evaluation of the drug's efficacy.If the new drug has successfully undergone extensive testing through phase ?,? and ? clinical trials,it will be filed to the regulatory authority,for approval of widespread use in the general patient population.After the drug becomes available on the market,phase ? trials may be initiated to keep drugs'efficacy,toxicity,and rare side effects under long-term surveillance.Phase I clinical trials is an important stage in clinical trials research and has a huge impact on the following stages.The method widely used in the phase ? clinical trials is Bayesian adaptive dose-finding algorithm,which is based on a dose-toxicity parametric model,and using Bayesian method to estimate model parameters and determine the maximum tolerated dose.However,in practical applications,the dose-toxicity relationship model is usually unknown,it may be unreasonable to model the dose-toxicity relationship by using parametric model in phase ? clinical trials.Therefore,to compensate for the above deficiency,this dissertation study the dose-finding problem for single-agent or two-agent combinations systematically based on the nonparametric model.The main contents of this dissertation include:1.The traditional continual reassessment method assume that the dose-toxicity relationship is a parametric model with increasing monotonically on doses.This assumption is unreasonable in many practical problems.Therefore,this dissertation propose the nonparametric Bayesian continual reassessment method,which relax the parametric assumption imposed on dose-toxicity relationship by using Dirichlet process to approximate dose-toxicity relationship.The proposed method is used to study the dose-finding for single-agent from Bayesian nonparametric perspective.A hybrid algorithm combining the Gibbs sampler and the adaptive rejection Metropolis sampling algorithm is developed to estimate the dose-toxicity curve,and a two-stage nonparametric Bayesian adaptive design procedure is presented to estimate the maximum tolerated dose.The novel nonparametric Bayesian continuous reassessment method does not depend on any parametric model.Therefore,it is more robust than the traditional method under the model misspecified.The simulation results show that nonparametric Bayesian continuous reassessment method behaves better than classical continuous reassessment method.An example analysis shows that the method is effective.2.In dose-finding trials for two-agent combinations,parametric Bayesian method mainly based on Copula regression,column-wise continual reassessment method and partial order continual reassessment method.As we all know,frequency method has better interpretability in statistical modeling,but it has not been widely used in studying dose-finding for two-agent combinations.In order to deal with the dose-finding problem for two-agent combinations from frequency perspective,maximum likelihood method is used to estimate the maximum tolerated dose based on the above three parametric models.This dissertation adopts the two-stage design to study identifiability of model parameters under limited sample size.simulation studies show that the frequency method is also effective in dose-finding for two-agent combinations.3.The traditional partial order continual reassessment method specify a parametric model for all possible sequences of dose combinations.Then the classical continual reassessment method is used to deal with the dose-finding problem for two-argent combinations.However,parametric modeling of the dose-toxicity curve may be problematical when there is little prior information on the shape of the dose-toxicity curve.Therefore,this dissertation propose the nonparametric Bayesian partial order continual reassessment method,which relax the parametric assumption imposed on dose-toxicity relationship by using Dirichlet process to approximate dose-toxicity relationship of every possible sequence,respectively.The proposed method is used to study the dose-finding for two-agent from Bayesian nonparametric perspective.A hybrid algorithm combining the Gibbs sampler and the independent doubly adaptive rejection Metropolis sampling algorithm is developed to obtain the robust Bayesian estimation of toxicity probability for dose combination,and a two-stage nonparametric Bayesian adaptive design procedure is presented to estimate the maximum tolerated dose.The proposed novel nonparametric Bayesian partial order continuous reassessment method does not depend on any parametric model.Therefore,it is more robust than the traditional method under the model misspecified.The simulation results show that nonparametric Bayesian partial order continuous reassessment method behaves better than partial order continuous reassessment method.An example analysis shows that the new method is effective.