Research On Identification Method Of Treatment Effect Heterogeneity Based On Random Search

The treatment method is tailored according to the patient’s specific physical condition,usually called personalized or stratified medicine,and its development will bring revolutionary and unprecedented changes to drug development.The new diagnostic tools will match the patient’s characteristics,disease pathophysiology and/or factors affecting drug metabolism with safe and effective drugs to treat their diseases.Due to the existence of individual heterogeneity,different patients often respond differently to the same treatment,and it is difficult to develop treatments that can benefit everyone.Therefore,the use of targeted therapies is increasing.An important goal of precision medicine is to identify the subgroups of patients whose treatment effect is much higher(or lower)than the average treatment effect of a certain treatment method,and adopt corresponding effective treatment methods for these patient subgroups to achieve the best treatment effect.In order to solve the problem of identifying subgroups of patients who may have a positive therapeutic effect in randomized trials,the main research contents of this paper include:(1)This thesis proposed a new hypothesis: it is believed that when the same treatment is used for samples with the same or similar characteristic variables,the treatment effects of these samples are the same.And use the multivariate normal distribution,binomial distribution and other mathematical distributions to randomly generate simulated clinical data based on linear regression model and tree model for the evaluation of subsequent algorithms.(2)The PRLSA algorithm is developed based on the principle of simulated annealing algorithm,and the simulated clinical data generated was used to verify the PRLSA algorithm and the eight existing algorithms for identifying the heterogeneity of treatment effects(Findit,Interaction Trees(IT),SIDES,MOB),PRIM,seq BT,SIDES,VT),and evaluate the specificity,sensitivity and kappa coefficient of each of the above algorithms.The evaluation results show that the overall performance of PRLSA is better than the existing eight algorithms.(3)In the R language environment,use the Shiny expansion package to build a treatment effect heterogeneity analysis platform.This platform is interactive and has a friendly user interface,making it more convenient for medical staff or other researchers without programming foundation.This platform integrates eight subgroup recognition algorithms,and optimizes these algorithms in parallel.Using this analysis platform can greatly increase the running speed,reduce the running time,and evaluate and compare multiple algorithms.