Automated identification of drug-induced liver injury for patients taking antihyperuricemic drugs using estimated medication history.

Abstract

Enormous costs are often required to identify drug-induced liver injury (DILI) through clinical trials or post-marketing product surveillance. By using Electronic Health Record (EHR) data, it is expected to easily reduce the costs with its readiness for computer analysis.We developed a computer program to estimate the patient medication history, which provided start-date and end-date of the medication for analyses, by considering residual medicines that patient did not take to combine prescription order data extracted from EHR. This study aims to obtain the DILI frequency by a comparison between the estimated medication history and ALT of laboratory test data for 1,777 patients taking allopurinol, febuxostat and benzbromarone.DILI was identified by three criteria as follows: (1)ALT elevated more than before medication (40 IU/l), (2) ALT during medication period did not decrease to less than a quarter of peak ALT and (3) ALT after the periods decreased to less than a half of peak ALT. The identified DILI were 18/1,224 (1.5%) cases for allopurinol, 5/266 (1.9%) cases for febuxostat and 3/287 (1.00%) cases for benzbromarone. The program approximately required 50 minutes to identify DILI. This method enabled us to more readily identify the occurrence of adverse events and is considered to contribute to studies in pharmacoepidemiology.