Longitudinal Data Analysis And Model Comparison Based On Type Ⅱ Diabetes Clinical Trial

Clinical trial is critical process for new drug research and its target is to evaluate the new drug efficacy and safety.Clinical trial is characterized with a limited number of patients,using statistical analysis method while conforming to clinical meaningful results,to generate the treatment guidance to have judgment for overall patients with similar situation.This paper stands on a new type II diabetes clinical trial with inadequate glycemic control and metformin monotherapy.This paper follows the general statistical principles and uses common longitudinal data statistical methods to analyze the efficacy and safety for collected data in clinical trial,such as two independent sample t-test,fisher exact test,ANCOVA model,LOGISTIC model,constrained longitudinal data analysis model and so on.Then this paper will start to compare the ANCOVA model with constrained longitudinal data analysis model from different aspects including variance,bias,coverage of confidence interval,type I error rate and power.Simulated database will be created to further investigate the advantages and disadvantages of these two models.As CLDA uses full likelihood approach which includes the baseline value as well as post-baseline value into model while ANCOVA may omit the missing data and their corresponding records.CLDA is superior to ANCOVA model as observed data are fully used.