Evaluating the Sensitivity and Specificity of Promising Circulating Biomarkers to Diagnose Liver Injury in Humans

• Published in: Toxicological sciences 2021-04, Vol.181 (1), p.23-34
• Main Authors: Llewellyn, Heather P, Vaidya, Vishal S, Wang, Zhenyu, Peng, Qinghai, Hyde, Craig, Potter, David, Wang, Jianying, Zong, Qing, Arat, Seda, Martin, Matt, Masek-Hammerman, Katherine, Warner, Roscoe, Johnson, Kent, Kullak-Ublick, Gerd A, Aithal, Guruprasad P, Dear, James W, Ramaiah, Shashi K
• Format: Article
• Language: English
• ISSN: 1096-6080; 1096-0929
• DOI: 10.1093/toxsci/kfab003

Abstract Early diagnosis of drug-induced liver injury (DILI) continues to be a major hurdle during drug development and postmarketing. The objective of this study was to evaluate the diagnostic performance of promising biomarkers of liver injury—glutamate dehydrogenase (GLDH), cytokeratin-18 (K18), caspase-cleaved K18 (ccK18), osteopontin (OPN), macrophage colony-stimulating factor (MCSF), MCSF receptor (MCSFR), and microRNA-122 (miR-122) in comparison to the traditional biomarker alanine aminotransferase (ALT). Biomarkers were evaluated individually and as a multivariate model in a cohort of acetaminophen overdose (n = 175) subjects and were further tested in cohorts of healthy adults (n = 135), patients with liver damage from various causes (n = 104), and patients with damage to the muscle (n = 74), kidney (n = 40), gastrointestinal tract (n = 37), and pancreas (n = 34). In the acetaminophen cohort, a multivariate model with GLDH, K18, and miR-122 was able to detect DILI more accurately than individual biomarkers alone. Furthermore, the three-biomarker model could accurately predict patients with liver injury compared with healthy volunteers or patients with damage to muscle, pancreas, gastrointestinal tract, and kidney. Expression of K18, GLDH, and miR-122 was evaluated using a database of transcriptomic profiles across multiple tissues/organs in humans and rats. K18 mRNA (Krt18) and MiR-122 were highly expressed in liver whereas GLDH mRNA (Glud1) was widely expressed. We performed a comprehensive, comparative performance assessment of 7 promising biomarkers and demonstrated that a 3-biomarker multivariate model can accurately detect liver injury.