| In randomized controlled clinical trials, subjects are allocated to treatment andcontrol or placebo group randomly, for receiving the treatment of the trial drug orplacebo independently, following up subjects and documenting the baselinecharacteristics, laboratory measurement and time-dependent covariates simutaniously.At the end, intention-to-treat (ITT) or per-protocol approach are applied to evaluate effectsize of trial drug.Different to general clinical trials, cancer clinical trials should take long period andat high risk of failure, experts prefer to implement the interim analysis. When the trialdrug show better effects in the interim analysis, the trial will go ahead and some subjectsmay transform from control to treatment group or drop out. This transform design iscalled one way crossover design. One way crossover design is usually used in cancer clinical trials. For one waycrossover survival data, traditional analysis approaches such as intention-to-treat andpre-protocol are simple and possible lead to bias. To address this limitation, Robins andRotnitzky presented an innovative strategy named Inverse Probability of CensoringWeighted (IPCW) Method, and successfully imployed to adjust the bias coming from thesubjects’ crossover much more efficiently.IPCW method is to calculate weight based on the baseline and lab measurement ofeach interval and then use it to reweight the visit time of each patient who crossed over.So the weighted time is defined as the overall survival of patients who crossed over, thenCox proportional hazards model may be applied to compare the overall survival betweenthe two groups. Now this method has been applied as the supplyment of traditionalmethods in many clinical trials in Europe.In this article, the detecting power and typeⅠerror under different conditions ofIPCW method compared with traditional methods such as ITT method and Censoredmethod were evaluated by Monte Carlo simulation. In addition, application conditions ofIPCW method were also studied. The main works are as follows.Firstly, the power of IPCW, ITT and Censored three methods were compared underdifferent median survival time ratio between treatment and control groups.The results ofthe simulation show that the power of IPCW method is better than ITT approach underdifferent hazard ratio. When the trail drug is much better than control, the ITT performswell. On this condition, ITT is also a plausable choice considering its easily calculation.When the effect of two groups is similar, IPCW method will be better than ITT method,and it can save sample size and shorten study period. So when the median survival timeof two groups is similar, IPCW method is prefered.The type Ⅰerror of three methods were also compared when median survival timeof two groups are equal (i.e. HR=1.0).the typeⅠerror of the two approaches fluctuatesnear0.05. The power of Censored method is highest, but its type Ⅰerror is highest tooand it often overestimate the effect of trail drug. So IPCW method performs well and isoptimal method for one way crossover survival data. Secondly, the comparison of power and typeⅠerror between IPCW method andtraditional methods were implemented while cross-over time changed. The results of thesimulation show that the sooner subjects cross over, the higher power of the IPCWmethod is, and the power of IPCW method is better than ITT approach.The results of the simulation also show that the typeⅠerror become higher underthe situation that control subjects crossed over before median survival time. But on thecontrary, when the event happened latter than median survival time, the typeⅠerrorcontrolled well. Indeed, the lack of the information about visit time has greatly hamperedthe effort of IPCW method. As a result, it can’t evaluate the efficacy of drugs properly.In conclusion, when the patients crossed over after followed up for a long time,IPCW method works better. |
PDF Full Text Request