Clinical Model for the Prediction of Severe Liver Fibrosis in Adult Patients with Type II Diabetes Mellitus

Abstract

Background and Objectives: Non-alcoholic fatty liver disease (NAFLD)-related severe liver fibrosis is associated with a higher risk of progressing to decompensated cirrhosis and hepatic failure and developing NAFLD-related hepatocellular carcinoma (HCC), particularly in populations with diabetes. Our pilot study aims to evaluate the performances of various noninvasive methods in predicting liver fibrosis in a population of patients with diabetes and to establish a new scoring system for the prediction of severe fibrosis (>F3). Materials and Methods: A total of 175 patients with diabetes were enrolled for liver fibrosis evaluation. Using the degree of agreement (concordance) between a noninvasive score based on serum biomarkers (NAFLD fibrosis score) and point shear-wave elastography (pSWE) as the reference method, we generated receiver operating characteristic (ROC) curves and performed a multivariate analysis to predict severe liver fibrosis. Results: In our population of patients with diabetes, gamma-glutamyltransferase (GGT), age, body mass index (BMI), the homeostatic model assessment of insulin resistance (HOMA-IR), and glycosylated hemoglobin (HbA1C) were significant predictors for the diagnosis of the F3/F4 group (area under the ROC: 0.767, 0.743, 0.757, 0.772, and 0.7, respectively; p < 0.005 for all). Moreover, the combined composite score (the sum of GGT, age, BMI, HOMA index, and HbA1C) had the highest diagnostic performance at a cut-off value of 3 (AUROC—0.899; p < 0001). The sensitivity, specificity, negative predictive value (NPV), and positive predictive value (PPV) were 91.20%, 79%, 79%, and 89%, respectively, and 89% of patients were correctly classified as having severe liver fibrosis. In contrast with the Fibrosis 4 (FIB-4) score and the AST-to-platelet ratio index (APRI), the composite score had the best accuracy in discriminating advanced fibrosis. Conclusions: The proposed composite score had a reliable and acceptable diagnostic accuracy in identifying patients with diabetes at risk of having severe fibrosis using readily available laboratory and clinical data.

Machine learning models including insulin resistance indexes for predicting liver stiffness in United States population: Data from NHANES

“…The most common direct measure of insulin resistance is the high insulin/ normoglycaemic clamp (HEC) technique, which is invasive, complex and impractical. Calapod et al (15) previously developed a Logsitic regression model using HOMA-IR to predict the development of severe liver fibrosis in diabetic patients. However, HOMA-IR is susceptible to the accuracy of insulin measurements and is poorly reproducible (37).…”

“…A growing number of studies have demonstrated a positive correlation between insulin resistancerelated indexes and the degree of liver fibrosis (10,11), However, most previous studies were still limited to demonstrating that insulin resistance may be a risk factor for altered liver stiffness (12-14). Calapod et al (15) previously used homeostasis model assessment of insulin resistance (HOMA-IR) to develop a Logsitic regression model for predicting the development of severe liver fibrosis in diabetic patients. However, this study, which focused first on participants with NAFLD, not only failed to quantitatively fit liver stiffness but also had unsatisfactory predictive performance as a conventional predictive model.…”