The Impact Of Drug Assignment In SRS On Signal Detection Of Adverse Drug Reactions

Background and PurposePharmacovigilance is an important part of the drug safety, whose history and developmentis closely connected with the development of drug safety monitoring. It has experienced along process from realize the importance of drug safety to establish a complete scientificmonitoring system. Many developed countries have established mature adverse reactionmonitoring systems at the beginning of the21st century. Adverse Drug reaction (ADR)analysis is the core content of Adverse Drug Reaction monitoring, and it used to detectwhether the occurrence and development of adverse reactions is related with any drug, anddegree of the relationship. The term ‘data mining’ refers to the use of computerisedalgorithms to discover signals in large SRS databases. Spontaneous Reporting Systemdatabases are the key data source currently used for drug safety data mining. However,there are no standards for the use of signal detection methods in routine pharmacovigilanceat present. Some safety databases such as our country maintained by pharmceuticalmanufacturers typically classify each drug cited in a report as ‘suspect’ or ‘concomitant’,and others will be defined as three categories: suspected drugs, concomitant drugs andinteracting drugs. Obviously, the value in a particular22cell will be different dependingon the choice to only count reports where a drug is coded as suspect (S only) versus allreports containing the drug irrespective of coding as suspect or concomitant (S+C).Presently, there is no reason to believe that either strategy is superior and further researchon the utility of the different approaches is required. This study was to analyze the impactof drug properties selection on ADR signal detection by the simulation and actual data.MethodsDrug-ADR monitoring data was simulated with Monte Carlo simulation method bySAS9.3, which was based on the characteristics of real data in Shanghai SpontaneousReport System. We compared the results of databases with or without concomitant drugsby sensitivity, signal quantity and intensity of common signal detection methods, includingROR, PRR, MHRA and IC. And the effect in real data was validated.ResultsIn simulation experiment, the sensitivity and specificity of all the detection methods have improved after removing the concomitant drugs. And the sensitivity was higher if theconcomitant drug has larger proportion in database.A total of24292reports were submitted to Shanghai SRS in2009. Concomitant drugsaccounted for13.36%of all drugs, and14.52%of Drug-ADR combinations wereconcomitant drug-ADR combination. If only analyzed suspect drugs, the number ofsuspicious signal detected by ROR, PRR, MHRA and IC were decreased23.80%,22.66%,13.37%and23.80%, respectively. The inclusion of both suspected and concomitantmedicines led to a substantial decrease in the strength of most indicators for intersectionsignals, in our study, the strength of suspicious signal detected by ROR, PRR and IC wereincreased65.97%,61.06%and51.77%, respectively. More than90%of the frequency forthe signal that lost after removing concomitant drugs were≤3.ConclusionsIn simulation experiment, the excursion of concomitant medicines led to a substantialincrease in the sensitivity and positive predictive value of the common signal detectionmethods. The experience of our study, using different signal detection methods, is thatcomputed relative statistic are slightly higher with ‘S only’ compared with ‘S+C’and thata greater number of drug-event pairs meet or exceed a numerical ‘signal’ threshold with ‘S+C’compared with ‘S only’for actual data. We should consider the impact of concomitantdrugs on ADRs signal dection when it has a large proportion in a database.