Evaluation of drug-induced serious hepatotoxicity (eDISH): Application of this data organization approach to phase III clinical trials of rivaroxaban after total hip or knee replacement surgery

Background: The most specific indicator of a drug-induced liver injury signal in a clinical trial database is believed to be the occurrence of subjects experiencing drug-associated elevations in both serum ALT and serum total bilirubin (TB) without a significant elevation in serum alkaline phosphatase (ALP). eDISH (evaluation of Drug-Induced Serious Hepatotoxicity) is a recently described tool that organizes liver laboratory data by graphically displaying peak serum ALT and TB levels for each subject, and can also provide direct links to the pertinent clinical and laboratory data for each subject. Objective: To illustrate the usefulness of the eDISH approach in the presentation of liver safety data by using phase III clinical trial data for rivaroxaban. Methods: Four randomized, active-controlled studies were conducted worldwide in subjects undergoing elective hip or knee replacement surgery to compare the efficacy and safety of the anticoagulant rivaroxaban, an oral, direct Factor Xa inhibitor, with the low-molecular-weight heparin, enoxaparin. Liver laboratory assessments, including ALT, AST, TB and ALP, were performed frequently during the studies. Data were incorporated into eDISH and linked data for selected subjects were analysed. Results: In the pooled analysis of the four studies, a total of 12 262 subjects (6131 rivaroxaban, 6131 enoxaparin) received at least one dose of study drug and had at least one central and/or local laboratory assessment during the study. A total of 143 (2.33%) rivaroxaban subjects and 223 (3.64%) enoxaparin subjects experienced a peak ALT >3 × upper limit of normal (ULN) but did not experience an elevation of TB >2 × ULN; these subjects are displayed in the right lower quadrant of the eDISH plot, termed the 'Temple's Corollary quadrant'. There were ten rivaroxaban and ten enoxaparin subjects with a peak ALT >3 ×ULN and a peak TB >2 × ULN; these subjects were displayed in the right upper quadrant of the eDISH plot, termed the 'Hy's Law quadrant'. eDISH allowed efficient examination of the relevant data for each of these subjects. Conclusions: The eDISH approach is an efficient and effective way to organize and examine large liver safety databases for randomized controlled clinical trials. It greatly facilitates a systematic and transparent examination of the relevant liver safety laboratory data. We believe eDISH should become a standard approach for assessing and studying liver safety issues in clinical trials.