Development of a Drug Safety Signal Detection Reference Set Using Japanese Safety Information

One of the main objectives of pharmacovigilance activities is to confirm unknown adverse drug reactions (ADRs), and data-mining methods have been developed to detect signals that are candidates for ADRs. Reference sets have been developed to evaluate the performance of the data-mining methods. However, reference sets generated in previous studies are not based on Japanese safety information; therefore, they are not suitable for use in evaluation studies in Japan because some drugs have not been approved or marketed for a long time in Japan. This study aimed to develop a reference set using drug safety information marketed in Japan and to evaluate its performance.

A reference set was developed for 43 drugs and 15 events. For each combination of the selected drug and event, those that were listed as important identified risks in the Japan Risk Management Plan (J-RMP) were set as "positive controls" and those that were not listed as adverse reactions in the package insert were set as "negative controls." In addition, we performed data-mining using Japanese Adverse Drug Event Report database (JADER) and evaluated the results against the reference set to empirically confirm its effectiveness.

The reference set included 127 positive and 386 negative controls. A comparison of the signals obtained from data-mining using JADER with the reference set revealed higher correlations than those in previous studies.

A reference set was developed using the safety information of drugs approved in Japan to promote research on data-mining methods.