Evaluation of liver fibrosis in patients with metabolic dysfunction-associated steatotic liver disease using ultrasound controlled attenuation parameter combined with clinical features

Abstract

Objective To explore the value of constructing a predictive model using ultrasound controlled attenuation parameter (CAP) combined with clinical features in diagnosing fibrosis in patients with metabolic dysfunction-associated steatotic liver disease (MASLD).Methods This retrospective study analyzed adult samples from the National Health and Nutrition Examination Survey (NHANES) database between 2017 and 2020. MASLD was defined as CAP ≥ 248 dB/m, and fibrosis was defined as liver stiffness measured by transient elastography ≥ 8.2 kPa. Patients were divided into fibrosis and non-fibrosis groups. Features were selected using the Boruta algorithm, and a predictive model combining CAP and clinical features was constructed. The receiver operating characteristic curve and area under curve (AUC), sensitivity, specificity and accuracy were used to evaluate the model.Results A total of 1,472 MASLD patients were identified, with 213 patients in the fibrosis group and 1,259 in the non-fibrosis group. The features screened by the Boruta algorithm included waist circumference, body mass index, CAP, blood glucose, combined diabetes, ALT, AST, GGT, hs-CRP, age, ALB, ALP, STB and gender. AUC for CAP alone in predicting liver fibrosis was 0.727[95%CI(0.690, 0.765)] with a sensitivity of 62.4%, specificity of 70.2%, and accuracy of 69.1%. The AUC increased to 0.842[95%(0.813, 0.871)] when combining CAP with clinical features, with a sensitivity of 75.5%, specificity of 76.7%, and accuracy of 75.6%. Delong's test comparing the AUC values of CAP alone and CAP combined with clinical indicators indicated a statistically significant difference (Z=-6.877, P

Keywords

• metabolic dysfunction-associated steatotic liver disease
• liver fibrosis
• ultrasound
• controlled attenuation parameter
• machine learning
• diagnose