Microcomputer-Assisted Bayesian Differential Diagnosis of Severe Adverse Reactions to New Drugs: A 4-Year Experience

Abstract

Severe idiosyncratic adverse drug reactions (IADRs) are an important source of morbidity and mortality. These reactions are usually detected by assessing a series of single case reports. The Bayesian Adverse Reaction Diagnostic Instrument (BARDI) can be used for the differential diagnosis of these adverse events. BARDI calculates the posterior probability in favor of a specific drug versus other drug or nondrug etiologies of an adverse event based on background (eg, epidemiologic) and case information (eg, time of onset). Studies showed that BARDI discriminates between multiple drug and nondrug etiologies. However, its widespread use has been limited by the complexity and time required for calculations. We expanded a previously developed prototype program (MacBARDI) designed for assessing neutropenia with a Macintosh SE microcomputer to also calculate the PsP in favor of drug causation for examples of neurotoxicity, hypersensitivity, and hepatotoxicity. We will describe the application of MacBARDI to two IADRs: (1) nine cases of Guillain-Barré Syndrome (GBS) associated with zimeldine, and (2) nine cases of neutropenia occurring during the administration of mexiletine (MXL). The program decreased the time required to complete computations, facilitated the incorporation of new information, and allowed us to model the impact of changes in the value of various BARDI factors. These examples illustrate that MacBARDI makes this method more readily available for postmarketing surveillance studies of severe idiosyncratic reactions to new drugs.