The Research About Two Statistic Issues In Interim Analysis Of Clinical Trials

In clinical trial,interim analyses are often performed before the completion of the trial,in which the intention is to possibly terminate the trial early.International Conference on Harmonization(ICH)E9 guideline provides the definition of interim analysis,that is any analysis intended to compare treatment arms with respect to efficacy or safety at any time prior to formal completion of a trial.To facilitate the application of the interim analysis and early termination,E9 guideline indicates that the goal of such an interim analysis is to stop the trial early if the superiority of the treatment under study is clearly establishes,if the demonstration of a relevant treatment difference has become unlikely,if unacceptable adverse effects are apparent,or if sample size re-estimation is needed.Early termination after interim analysis for safety reason or weak treatment effect is acceptable widely.And it is also recommended that any decision of terminating a trial is a very complex process in which various elements should be considered,especially the actual treatment effect.Therefore,whether the treatment effect could be estimated correctly is the key point.Otherwise,regarding the application of sample size re-estimation during the interim analysis,regulatory authorities certainly favor blinded sample size reassessment because it better preserves study integrity,but due to the technical issue and the special character of the survival data,the use of blinded sample size re-estimation in this context has rarely been reported.This research focused on the above two issues,evaluated whether the treatment effect could be estimated correctly in early termination stage and tried to derive the blinded sample size re-estimatiomethod for survival data.The first partwas to examine the treatment effect estimation during interim analysis through formula derivation and simulation methods.The inequalities were derived for the relationship between the estimated effects and the actual effects for normal and binary endpoints.Several simulations were also provided to support our findings.The derived inequalities show that the difference between two statistical significant sample means/rates is greater than that between two pre-defined population means/rates.The simulation results show the ratios of the average of the estimated mean/rate differences in the statistical significant early termination stage and the pre-defined effect size are always greater than I even if the type I error rate is controlled by Peto,O’Brien-Fleming or Pocock method.Otherwise,the stage-wise ordering method was applied to estimate efficacy for normal endpoints,the simulation results showed that the adjusted estimations were closed to the unadjusted values in the early stage interim analyses,and the adjusted effects were only appeared in the later stages of the multiple interim analysis design,but the underestimated risk was also appeared at the same time.Therefore,the actual population benefits may be smaller than the estimation when a clinical trial early stopped for benefits because of significant interim analysis.So,it is inadvisable to early terminate the confirmatory trial as well as exploratory trial.The second part tried to apply the EM algorithm to the blinded sample size re-estimation of survival data.Based on the denity function of the exponential and Weibull distribution,expectation-maximization(EM)algorithm of the hazard ratio was derived,and several simulation studies were used to verify its application.Our studies show that the stability of the EM estimation results directly correlated with the sample size,the convergence of the EM algorithm is impacted by the initial values but is acceptable,and the balanced design produces the best estimates.The average estimation results are acceptable,estimation means are closed to pre-defined parameters,but the single values are not stable which presents bimodal distribution.The derived EM method is still limited,but results provide useful information to steer the practitioners in this field from repeating the same endeavor.