Database Construction And Signal Detection Of Antibacterial Adverse Drug Reactions Based On MedDRA

In order to improve China’s drug regulatory capacity and level,simultaneously enhance the international competitiveness of China’s pharmaceutical industry,the National Medical Products Administration(NMPA)clearly stated that MedDRA will be fully applied to report of Adverse Drug events(ADE)from July 1,2022.Since the application of China’s Spontaneous Reporting System(SRS),ADE has been reported with the use of World Health Organization Adverse Reaction Terminology(WHO-ART)as standard.Compared with WHO-ART,the number of MedDRA terms is greatly increased,and the effect on signal generation is unknown under the original signal detection conditions.The U.S.Food and Drug Administration’s Adverse Event Reporting System(FAERS)is relatively mature and efficient in the application of MedDRA standard.Before the comprehensive application of MedDRA,it is undoubtedly of great practical significance for China to carry out signal detection based on SRS,which is the mature application of M1 guiding principle.The task of drug ADR monitoring is to perform function operation on the ADEs collected based on SRS by using disproportionate algorithm.The drug-ADE combination exceeding a certain threshold is preliminarily defined as a signal.Then signal evaluation will be done to find new suspicious ADRs.During the signal evaluation step,experts are required to compare the obtained signals with the safety information in package inserts one by one to judge whether the signals are known ADRs.ADR database collects and standardizes all ADR information in drug instructions.Using ADR database in signal evaluation process can greatly reduce the workload of expert signal comparison and improve work efficiency.Since China has not established a standardized ADR database,experts still need to manually judge whether ADR is known according to the safety information of package inserts in the existing signal evaluation step.With the continuous increase of ADE reports,China’s drug regulatory ability is expected to be in line with the world,therefor the construction of ADR database based on MedDRA standards has become a practical problem that needs to be solved urgently.The number of ADE reports of antibacterials has always been in the first place among all kinds of drugs,which is the focus of drug ADR monitoring domestic and overseas.Therefore,this study focused on antibacterials and carried out a series of relevant studies centering on MedDRA based on FAERS and our country’s SRS,providing theoretical reference data for the successful application of M1 guidelines in ADR monitoring of drugs in China.In view of the above issues,our study mainly research on the following aspects:In the first part of this paper,we took FAERS,which has mature application of MedDRA standard,as the model.Through signal detection of carbapenems with the strongest antibacterial activity and significantly increased consumption,the necessity and feasibility of constructing MedDRA based ADR database were summarized and discussed.We downloaded the most recently available FAERS data from January 1,2015 to December 31,2018 through the FDA website.Four kinds of carbapenems approved by FDA were choosen as the targeted antibacterials.We calculated the 95%confidence intervals of PRR,ROR,MHRA and information component(IC)of BCPNN to detect signals.A signal was defined when it satisfied the criteria that 3 or more targeted drugADE pairs,a PRR was at least 2,chi-suared was at least 4,the lower limit of the 95%confidence intervals of PRR and ROR was larger than 1,and that of IC was larger than 0.In addition,when the signal was in the SIDER and ADReCS,it was defined as a known ADR.We performed descriptive analysis on the information of whole classes of carbapenems and the known ADRs.Results showed that conforming to the trend of the ADE reports data,the annual number of carbapenem-related signal has kept increasing over the years,which was 229,266,310 and 409 respectively.Cumulative detected signals accounted for 36.4%of all carbapenem associated ADEs pair.There were 123 known ADRs,which accounted for 15.8%.Among ADEs resulting from carbapenems,serious ADEs were reported in 41.2%(N=8313).A potential safety signal for carbapenems(PRR 1.43)was detected for serious ADEs.PRRs for serious ADEs were 1.14 for imipenem.The drug ineffectiveness(10.5%)was the most frequently reported ADE related to carbapenems.The results indicates that Carbapenems are more likely to be associated with severity and death compared to the rest of β-lactam antibacterials.Carbapenem resistance is alarming.In addition,we find that there are relatively authoritative standardized ADR databases in foreign countries,which provide effective and convenient resources for FAERS signal detection.At present,there is no relatively reliable ADR database based on international standard in China.China can learn from foreign experience and construct ADR database based on MedDRA in advance to lay a good foundation for the operation of M1 guideline.Based on the above inspirations,we analyzed the resource conditions for constructing the MedDRA based ADR database of antibacterials in China.We found that most of the ADR information of antibacterials in China was poorly standardized and lacked of clinical trial data.The relatively comprehensive ADR database of antibacterials could not be constructed only by the resources of antibacterial package inserts in China.In view of all antibacterials used in China were imported or domestic generic drugs.The standardization of package inserts of the brand name antibacterial were better than that of domestic generic antibacterials in China.At the same time,there was no significant difference in the ADR information of antibacterial drugs among ethnic groups.Therefore,we constructed the ADR database of antibacterial drugs based on MedDRA with the ADR resources from antibacterial drug instructions in China and that of foreign resources.The specific steps were as follow:①We extracted ADR information associated with antibacterials from two kinds of sources,including Chinese antimicrobial package inserts from "catalog set",“administration asssistant" and "drug intelligence data",as well as foreign resources from SIDER,ADReCS and Drugs.com website.Bi-LSTMCRF model was used for ADR entity recognition of Chinese antimicrobial inserts and free text downloaded from Drugs.com.②The standardization was according to the key points of MedDRA term selection.ADR terms were mapped to the LLT level of MedDRA(20.0)term set in Chinese and English respectively.③we constructed the entity relationship model of antibacterial and ADR.④Data was stored in MySQL database,and Vue(Element)was used as database management tool to realize database visualization.In addition,we performed the descriptive analysis on the safety profile of antibacterials and bulit the platform based on the constructed database.The results showed that the F values of ADR recognition in Chinese and English were 92.43%and 82.21%,respectively.A MedDRA based ADR database with a total of 15676 drug-ADR pairs,covering 140 antibacterials in 10 categories,including tetracycline,aminoylol,β-lactam,sulfanilamide and quinolone,and 1793 types of associated ADRs was established.The results indicates that Bi-LSTM-CRF model can realize the automatic extraction of ADR in Chinese antibacterial inserts.The constructed ADR database based on MedDRA can provide reference for signal evaluation and provide a standard reference set for the study of safety monitoring methods.In the third part of this study,the MedDRA based ADR database constructed in Part 2 was taken as the "gold standard" reference set to further verify its significance of constructing it,as well as the signal detection capability and applicability of different signal detection methods under the new standard.The framework of the three parts is shown in Figure A of the preface.We collected data from January 1,2016 to December 31,2018 from SRS of Shaanxi Province.Antibacterial were choosen as the targeted antibacterials.We calculated the 95%confidence intervals of PRR,ROR,and information component(IC)of BCPNN and MHRA to detect signals.Signal detection rate,precision rate,recall rate and F value were selected as the evaluation indexes of signal detection ability.We regarded the constructed MedDRA based antimicrobial database as the "gold standard".ADEs in the"gold standard" was defined as known ADRs,and the proportion of known ADRs reported in SRS was calculated.The detection ability of different detection methods under different time background.The results showed that a total of 33825 ADE reports associated with 4221 pairs of antibacterials-ADEs were submitted to the SRS of Shaanxi Province from January 1,2015 to December 31,2018.Among them,1692 pairs were known ADRs,accounting for 40.1%.Under the same period of backgroud,SDR、R°、and P° of different detection methods showed statistically significant differences.ROR and PRR were significantly higher than IC and MHRA in SDR、R° and F°index.IC and MHRA were significantly higher than ROR and PRR in P° index.With the shortening of time interval,ROR and PRR detection methods showed an increasing trend in the four detection indexes.The results of the four methods showed an upward trend.The SDR of IC method was between 0.073 and 0.105.The results show that the proportion of known ADR is high.Reporting ADE based one MedDRA decrease the rate of signal detection.By shortening the time background selection,the accuracy of all detection methods and the comprehensive detection ability of ROR and PRR detection methods can be effectively improved.However,the IC method demonstrates too conservative which may make the signal undetectable.It is necessary to make further study on the adaptability of different detection methods under the new standard.To sum up,this study focuses on MedDRA standard in the context of the upcoming implementation of M1.We find that the standard ADR database can provide great convenience for signal evaluation by signal detection of FAERS.Combining with the fact that there is no standardized ADR database in China,the antibacterials ADR database based on MedDRA is constructed by integrating domestic and foreign antibacterials ADR resources.Finally,the data of China was used to verify that signal detection based on MedDRA standard would dilute signal detection,and at least 40%of detected signals in signal evaluation could be directly judged according to ADR database.The main innovation points of this study are as follow:①Constructing an antibacterial ADR database based MedDRA with a relatively standardized procedure which provided technical reference for the construction of ADR database of other drugs;Meanwhile,a website for ADR retrieval and analysis of antibacterial is established based on the database which can provide guidance for clinical medication.②Different detection methods and background selection are discussed by real world data from SRS based on MedDRA,so as to provide theoretical support for further comprehensive implementation of M1 guidelines.