A simple method for exploring adverse drug events in patients with different primary diseases using spontaneous...

Patient background (e.g. age, sex, and primary disease) is an important factor to consider when monitoring adverse drug events (ADEs) for the purpose of pharmacovigilance. However, in disproportionality methods, when additional factors are considered, the number of combinations that have to be computed increases, and it becomes very difficult to explore the whole spontaneous reporting system (SRS). Since the signals need to be detected quickly in pharmacovigilance, a simple exploration method is required. Although association rule mining (AR) is commonly used for the analysis of large data, its application to pharmacovigilance is rare and there are almost no studies comparing AR with conventional signal detection methods.

In this study, in order to establish a simple method to explore ADEs in patients with kidney or liver injury as a background disease, the AR and proportional reporting ratio (PRR) signal detection methods were compared. We used oral medicine SRS data from the Japanese Adverse Drug Event Report database (JADER), and used AR as the proposed search method and PRR as the conventional method for comparison. "Rule count ≥ 3", "min lift value > 1", and "min conviction value > 1" were used as the AR detection criteria, and the PRR detection criteria were "Rule count ≥3", "PRR ≥ 2", and "χ² ≥ 4".

In patients with kidney injury, the AR method had a sensitivity of 99.58%, specificity of 94.99%, and Youden's index of 0.946, while in patients with liver injury, the sensitivity, specificity, and Youden's index were 99.57%, 94.87%, and 0.944, respectively. Additionally, the lift value and the strength of the signal were positively correlated.

It was suggested that computation using AR might be simple with the detection power equivalent to that of the conventional signal detection method as PRR. In addition, AR can theoretically be applicable to SRS other than JADER. Therefore, complicated conditions (patient's background etc.) that must take factors other than the ADE into consideration can be easily explored by selecting the AR as the first screening for ADE exploration in pharmacovigilance using SRS.