The Utility and Diagnostic Accuracy of Transient Elastography in Adults with Morbid Obesity: A Prospective Study

Abstract

Patients with morbid obesity are at high risk for nonalcoholic fatty liver disease (NAFLD) complicated by liver fibrosis. The clinical utility of transient elastography (TE) by Fibroscan in patients with morbid obesity (body mass index (BMI) ≥ 40 kg/m2) is not well-defined. We examined the diagnostic accuracy of Fibroscan in predicting significant liver fibrosis (fibrosis stage ≥2) in morbidly obese patients (BMI ≥ 40 kg/m2). Patients scheduled for bariatric surgery were prospectively enrolled. Intraoperative liver biopsy, liver-stiffness measurement (LSM) by Fibroscan (XL probe), and biochemical evaluation were all performed on the same day. The endpoint was significant liver fibrosis defined as fibrosis stage ≥2 based on the Nonalcoholic Steatohepatitis Clinical Research Network. The optimal LSM cutoff value for detecting significant fibrosis was determined by using the Youden Index method. Routine clinical, laboratory, and elastography data were analyzed by stepwise logistic regression analysis to identify predictors of significant liver fibrosis and build a predictive model. An optimal cutoff point of the new model’s regression formula for predicting significant fibrosis was determined by using the Youden index method. One hundred sixty-seven patients (mean age, 46.4 years) were included, of whom 83.2% were female. Histological assessment revealed the prevalence of steatohepatitis and significant fibrosis of 40.7% and 11.4%, respectively. The median LSM was found to be significantly higher in the significant fibrosis group compared to those in the no or non-significant fibrosis group (18.2 vs. 7.7 kPa, respectively; p = 0.0004). The optimal LSM cutoff for predicting significant fibrosis was 12.8 kPa, with an accuracy of 71.3%, sensitivity of 73.7%, specificity of 70.9%, positive predictive value of 24.6%, negative predictive value of 95.5%, and ROC area of 0.723 (95% CI: 0.62–0.83). Logistic regression analysis identified three independent predictors of significant fibrosis: LSM, hemoglobin A1c, and alkaline phosphatase. A risk score was developed by using these three variables. At an optimal cutoff value of the regression formula, the risk score had an accuracy of 79.6% for predicting significant fibrosis, sensitivity of 89.5%, specificity of 78.4%, positive predictive value of 34.7%, negative predictive value of 98.3%, and ROC area of 0.855 (95% CI: 0.76–0.95). Fibroscan utility in predicting significant liver fibrosis in morbidly obese subjects is limited with accuracy of 71.3%. A model incorporating hemoglobin A1c and alkaline phosphatase with LSM improves accuracy in detecting significant fibrosis in this patient population.