Using Bayes Method And Risk-Banifit Analysis To Scanning The Advers Drug Reaction Signals

Drug - Safety is so important that any country will put it at first place. A high efficient surveillance system for drug safety is needed to minimize the harm of the advers drugs reactions in an acceptable safety boundary.Before put onto the market, drugs have to pass three phases tests on human body to collect the basic information of safety and effect. Also the tests can get many information of the drug, but that's not enough, because:1,The sample number of the tests are not large enough to detect rare adverse drug reactions;2,The surveillance of the tests are not long enough to detect long delay adverse drug reactions.3,The sampling does not come from the whole population, so adverse drug reactions occurred on special population will be ignored.According to the three reasons above, to collect safety information of the drugs after putting onto the market is important. There're mainly seven methods: Spontaneous Reporting System (SRS), Intensive Hospital Monitoring, Prescription-Event Monitoring (PEM ), Information Linkage System, Drug Utilization Survey, case control study and cohort study. Among these methods, Spontaneous Reporting System (SRS) is used by many country, because it can cover the largest population, detect rare rate ADRs and can save a lot of money.The drug safety information which comes from a lot of methods is large, complex and their quality is questionable. So how to select useful signals from the data base is very important to the ADR surveillance system. This screen process usually have three steps, first uses computer to screen raw signals from the original data base; second experts reviews the raw signals screened by the computer; third reevaluate the safety of the drug reviewed by the experts. Before the first step of the screen process, an arithmetic has to be designed. The arithmetic designed to fit the characteristic of the data, different country may use different arithmetic. To the spontaneous reporting system, the so called disproportional method is effective and widely used, disproportional method is not one arithmetic but a family, which include Proportional ADR Reporting Ratio ( PRR), Reporting Odds Ratio ( ROR ), Bayesian confidence propagation neural network method (BCPNN), Empirical Bayesian Gamma Poisson Shrinker (GPS ) and etc..The disproportional method suppose that:1,because of effect from random events, the trueness of the reports is doubtable.2,If a certain drug causes a certain adverse reaction, this drug-reaction pair has a higher chance to appear in the reports.The disproportional method can check out these pairs if the chance is high enough. Though the disproportional method is simple and effective, it only concerns the mathematics rate of the drug-reaction pairs. The row signals the experts receive only have probability, but not importance. For example: if a patient takes chemotherapeutics has the adverse reaction of sickness and anorexia will be ignored; but if a patient of flu has the adverse reaction of sickness and anorexia will seize the eyes of the doctor.So signal screening should not only concern about the mathematics rate but also the benefit - risk of taking the drug.If the arithmetic concern these two aspects, the signal screening will more efficiency and the whole system will be more efficiency.Benefit - risk analysis is a complex and high demanding job. It concerns not only how bad and how long the adverse reaction is but also should concern about how good the drug to the patient is. The original disease is also takes into account. The analysis method mainly consist description analysis, half- quantitative analysis, and quantitative analysis.Because the database is to simple to make quantitative analysis, we use half-quantitative analysis method to reevaluate the signals screened by the BCPNN arithmetic. In this study we discussed 30 signals selected from 1310 row signals screened by the BCPNN arithmetic.This study downloaded the adverse drug reaction information from the Shanghai ADR surveillance center. After cleaning and coding, we use BCPNN to screen the database and get 1340 row signal. At last we select 30 signals from the 1340 row signals which are highly benefit-risk concerned.This study use Visual Basic in Excel to be the programming language to implement the functions like data cleaning, coding and signal screening.