Implementation of Signal Detection Methods in Pharmacovigilance – A Case for their Application to Safety Data from Developing Countries
Background: Pharmaceutical companies, regulatory agencies, or other government agencies conduct safety signal detection as one of the ways for mitigating the possible risks due to medical products. The safety signal is defined as information that arises from one or multiple sources (including observations or experiments), suggesting a new, potentially causal association, or a new aspect of a known association between an intervention like administration of a medicine and an event or set of related events, either adverse or beneficial, that is judged to be of sufficient likelihood to justify verificatory action. Definition of signal in this paper is restricted for pharmacovigilance to adverse events. Signal detection as a pharmacovigilance activity is highly recommended in the United States of America and it is mandatory in Europe. However, at the moment, the same cannot be said for any African country. Findings: Traditional techniques for signal detection work well when the adverse drug reaction reports are few and manageable to enable manual review. The data mining algorithms for signal detection perform well with large and/or pooled drug safety monitoring reports. The methods discussed are proportional reporting ratio (PRR), the Report Odds Ratio (ROR), the Bayesian Confidence Propagation Neural Network (BCPNN) and the Multi-item Gamma Poisson Shrinker (MGPS). These sophisticated methods are routinely used in developed countries at the moment. Conclusion: Data mining algorithms for signal detection need to be adopted in developing countries because the increased use of pharmaceutical products has led to exponential increase in safety reports and datasets.