Protective Durability Of The COVID-19 Vaccine Based On Randomized,Double-Blind,Crossover Design Simulation Studies Evaluating Statistical Analysis Methods

With a large number of COVID-19 vaccines at home and abroad have entered phase III clinical trials,some vaccines have been approved for marketing.Confirmed safety and efficacy are the criteria for successful approval of vaccines,but most phase III clinical trials of COVID-19 vaccines evaluate the protective efficacy of vaccine effectiveness in a short period of time.From a scientific point of view,even if the vaccine has been approved for marketing,it is still necessary to follow up the subjects and collect data to evaluate the durability of the vaccine’s protection.Since the vaccine has been proven to be effective,it is unethical for placebo subjects not to be vaccinated,and the operation in the trial is more complicated.Therefore,this article discusses the trial design to study the durability of the COVID-19 vaccine,how to calculate the protective efficacy under the trial design,and how to evaluate the long-term protective efficacy of the COVID-19 vaccine.The paper was simulated under a blinded state-based crossover experimental design.The simulations were performed from two perspectives: calculating the two stage protective efficacy based on the incidence rate from disregarding the time factor,and fitting the protective efficacy change curve from considering the time factor,respectively.The main research contents of the paper are as follows:The chi-square test or Fisher’s exact probability was used to test the difference between groups,and the use of both conditions needs to be judged according to the use of Clopper-Pearson to calculate the confidence interval of the protective efficacy of the two stages,and the lower limit of the confidence interval is used to judge whether the criteria are met.According to the results of the simulated data,the confidence intervals for different incidence rates with the same protective efficacy are different,and therefore,a comprehensive evaluation from the statistical and clinical perspectives is required.The calculation of protective efficacy and 95% confidence intervals for different number of cases and incidence rates at two stages using the Beta distribution model should use smaller prior information and have less impact on the posterior probabilities.Unlike the confidence interval results estimated by Clopper-Pearson when both incidence and protective efficacy are the same,sensitivity analysis can be used to examine the robustness of the statistical analysis by calculating from multiple perspectives.The chi-square test or Fisher’s exact test probability and Beta distribution models calculate the protective efficacy based on the number of cases and incidence rate,without considering the time factor.In the COX model with time-dependent covariates,not only the number of cases and incidence rate are considered,but also the effect of the time of vaccination and other factors on the protective efficacy,and thus the curve of the protective efficacy over time can be fitted.