Detection of cluster adverse drug events in the spontaneous reporting system of China using a disproportionality filter algorithm.

What Is Known And Objective: Cluster adverse drug events (CADEs) are multiple ADEs with similar clinical manifestations involving the same drug, manufactured by the same company, that occur within a short time period. The disproportionality filter algorithm (DFA) is a promising tool for the identification of historical clusters related to ADEs. The Chinese spontaneous adverse drug reaction reporting system (SRS) may serve as an important database for the detection of CADEs. The objective of this study was to evaluate the usefulness of DFA as an approach to identify CADEs using SRS.

Methods: Suspected adverse drug reaction (ADR) reports collected by the Chinese SRS in 2014-2015 were examined to identify potential CADEs. The reports were divided into 48 15-day subsets. Disproportionate excess reporting of ADEs for drugs from specific companies may be a signal for CADEs. The clusters were categorized as 'confirmed', 'potential', 'unlikely', 'indecisive' or 'information-loss' ADEs when evaluated by report units. Furthermore, early warning information in 2014-2015 from the Chinese early warning system (EWS) classified as 'concern', 'monitoring', 'ignorance' or 'rest' was compared with DFA to explore the applicability of the novel algorithm in Chinese SRS.

Results And Discussion: In total, 2294 CADEs, comprising of 380 confirmed, 1753 potential, 15 unlikely and 59 indecisive clusters, were generated; 87 clusters were missing additional information. There were 263 'significant' clusters with DFA, but only 26 'significant' clusters in EWS. The sensitivity of DFA was 88.5%, but the specificity and positive predictive value were low.

What Is New And Conclusion: Spontaneous adverse drug reaction reporting system in China may be a potential database for the identification of CADEs engaging the DFA. This could supplement the EWS of CADEs in China. The DFA may be of value in detecting CADEs with high sensitivity, although expert screening is required given the low specificity and positive predictive value.