RESUMO
AIMS: To assess the utility of existing metabolomics scores to classify liver disease in patients with type 2 diabetes mellitus (T2DM). MATERIALS AND METHODS: A total of 220 patients with T2DM were recruited. Patients underwent routine laboratory tests, liver proton magnetic resonance spectroscopy (1 H-MRS), a 75-g oral glucose tolerance test, and liver biopsy if 1 H-MRS findings indicated non-alcoholic fatty liver disease. A serum sample was blindly provided to OWL Metabolomics on which to run the OWLiver Care and OWLiver tests. RESULTS: When compared with liver biopsy, the OWLiver Care and OWLiver tests had a suboptimal performance in patients with T2DM (areas under the receiver-operating characteristic [AUROC] curve both <0.70). Given the discordance of these results in this heterogeneous, multiethnic cohort compared with those of a previous report in predominantly white patients without diabetes, we examined the influence of age, ethnicity and other variables on test performance. A specific subset of patients was selected to mirror the characteristics of the population used for the development of this model (ie, white patients without T2DM). Among white patients with good glycaemic control (glycated haemoglobin <53mmol/mol [or <7%]) and without cirrhosis, the AUROC curve was significantly improved (0.79 [CI 95%: 0.68-0.90]). Among white patients with lower insulin resistance (homeostatic model assessment of insulin resistance <3) and without cirrhosis, the AUROC was even higher: 0.87 (CI 95%: 0.76-0.97). CONCLUSIONS: There is a great need to develop non-invasive approaches to diagnose non-alcoholic steatohepatitis in patients with T2DM; models originally developed for patients without diabetes cannot be directly applied to patients with T2DM.