The Simultaneous Statistical Inference Of Crossover Design With Two Binary Endpoints In Clinical Trials

Crossover design is often used in clinical medical trials of chronic diseases with little change in the short term.Compared with parallel design and paired design,crossover design comprehensively applies self-comparison and inter-group comparison,which can not only save the number of subjects and test cost,but also exclude the influence of sequential effect.In this paper,we study the difference hypothesis test of crossover design with two binary endpoints,and then explore the non-inferiority test.At first,this paper introduces the research background and development status of crossover design.There has been a lot of discussion and research on crossover design with only one endpoint.However,there are only a few relevant explorations and literatures on the crossover design with multi-endpoints.Therefore,this paper focuses on the study of the crossover design with two binary endpoints,expect to establish a reasonable hypothesis,use appropriate test statistics,and evaluate the therapeutic effect of the drug.The main research contents of this paper are listed as below:(1)starting from the difference hypothesis test(also known as the unequal test),this paper defines that only when there is no significant difference in the therapeutic effect of two endpoints can the two drugs be considered to have no significant difference in therapeutic effect.The union-intersection method was used to test the difference of the two drugs'therapeutic effects on the two endpoints,and calculated score statistics,then compared with the Mainland-Gart statistics.The power and the probability of making the typelerror of the two test statistics were simulated with data.From the results of data simulation,it can be seen that the test effect of score statistics is slightly better than Mainland-Gart statistics(2)studying the non-inferiority hypothesis test of crossover design with two binary endpoints.In this paper,it is defined that only the therapeutic effect of the test drug is not inferior to the standard drug on both two endpoints can the test drug is not inferior to the standard drug.Intersection-union method was used,score statistics and Mainland-Gart statistics were also used for the test,and approximate power formula was calculated to estimate the sample size.The results of data simulation showed that the test power of score statistics is slightly better than Mainland-Gart statistics,while the sample size determination of Mainland-Gart test is better.