A BAYESIAN NEURAL-NETWORK METHOD FOR ADVERSE DRUG REACTION SIGNAL GENERATION

Objective: The database of adverse drug reactions (ADRs) held by the U ppsala Monitoring Centre on behalf of the 47 countries of the World He alth Organization (WHO) Collaborating Programme for International Drug Monitoring contains nearly two million reports. It is the largest dat abase of this sort in the world, and about 35 000 new reports are adde d quarterly. The task of trying to find new drug-ADR signals has been carried out by an expert panel: but with such a large volume of materi al the task is daunting. We have developed a flexible, automated proce dure to find new signals with known probability difference from the ba ckground data. Method: Data mining, using various computational approa ches, has been applied in a variety of disciplines. A Bayesian confide nce propagation neural network (BCPNN) has been developed which can ma nage large data sets, is robust in handling incomplete data, and may b e used with complex variables. Using information theory, such a tool i s ideal for finding drug-ADR combinations with other variables, which are highly associated compared to the generality of the stored data, o r a section of the stored data. The method is transparent for easy che cking and flexible for different kinds of search. Results: Using the B CPNN, some time scan examples are given which show the power of the te chnique to find signals early (captopril-coughing) and to avoid false positives where a common drug and ADRs occur in the database (digoxin- acne; digoxin-rash). A routine application of the BCPNN to a quarterly update is also tested, showing that 1004 suspected drug-ADR combinati ons reached the 97.5% confidence level of difference from :he generali ty. Of these, 307 were potentially serious ADRs, and of these 53 relat ed to new drugs. Twelve of the latter were not recorded in the CD edit ions of The physician's Desk Reference or Martindale's Extra Pharmacop oea and did not appear in Reactions Weekly on-line. Conclusion: The re sults indicate that the BCPNN can be used in the detection of signific ant signals from the data set of the WHO Programme on International Dr ug Monitoring.The BCPNN will be an extremely useful adjunct to the exp ert assessment of very large numbers of spontaneously reported ADRs.