Comparative Study Of Different Methods In Dealing With Missing Data In Clinical Trials

In clinical trials,the lack of data is common.In most clinical trials,especially those with a particularly long trial period,there will always be subjects who exit early for a variety of reasons,resulting in incomplete data collection.The existence of missing data will lead to the reduction of valid data,increase the difficulty of statistical analysis,affect the accuracy of statistical data.The incomplete display of the overall information of clinical trials may lead to bias in the statistical process,affect the accuracy of the assessment,and cause the statistical inference to be biased or invalid,thus affecting the final statistical decision.Therefore,it has important research value in how to deal with the missing data.Different clinical trial projects,because the researchers are different in the field of knowledge,even the researcher's personal preferences,often determine their different methods of dealing with data missing,so its corresponding statistical analysis results may also be affected by the researcher's own subjective factors.In this paper,taking psoriasis as an example,the two main indicators of its therapeutic effect,SPGA and PASI,are compared with different missing data processing methods,except for the last observation forward carrying(LOCF),the worst observation forward carrying(WOCF),the forward carrying of baseline observations(BOCF)and No Response Imputation(NRI)in clinical trials.In addition,this paper will also try to introduce a new multiple imputation method pMI(Placebo Multiple Imputation)to replace the traditional multiple imputation,and introduce tipping point as a supplement to the fill method.Through comparison,the following conclusions can be drawn: in this clinical trial of psoriasis,the new multiple imputation method pMI is more effective in dealing with missing data of PASI,and combined with linear model of repeated measurement mixed,the results are more objective and realistic,and use the new multiple imputation method pMI to deal with PASI75 missing data is the more effect.The new multiple imputation method pMI is same as No Response Imputation(NRI)inprocessing SPGA(0,1)missing data.At the same time,combined with Tipping Point analysis,the proportion of successful trials is calculated,so as to make a comprehensive and accurate statistical analysis of clinical trials containing missing data.