Establishment and verification of a diagnostic model for liver cirrhosis based on three non-invasive indexes

Objective Liver biopsy is the gold standard for assessment of cirrhosis. However, there is a risk of bleeding for liver biopsy. This study aims to screen non-invasive indicators to establish a highly pathology-correlated diagnostic method for liver cirrhosis.

Methods From July 2015 to November 2016, 460 patients from liver surgery department of Zhongshan Hospital Fudan University were enrolled. Using pathological diagnosis as the gold standard, a non-invasive diagnostic model was established by least absolute shrinkage and selection operator (LASSO). Receiver operator characteristic (ROC) curve was used to evaluate the diagnostic effectiveness of the model, and an independent cohort of patients (n=152) was used to verify the model results.

Results Patients with cirrhosis showed significantly high two-dimensional shear-wave elastography (2D-SWE), (13.60±4.98 kPa vs 9.67±3.68kPa, P < 0.01), positive rate of hepatitis B surface antigen (HBsAg63.5% vs 30.4%, P < 0.01), collagen Ⅳ (74.23±45.57ng/mL vs 61.30±41.22ng/mL, P=0.01), alkaline phosphatase (84.21±30.94U/L vs 98.49±68.30U/L, P=0.02), prothrombin time (11.75±0.89s vs 11.39±0.78s, P < 0.01), hyaluronic acid (152.71±143.36U/L vs 99.64±71.08U/L, P < 0.01) than non-cirrhosis patients, while platelet (150.94±70.28×10⁹/L vs195.39±67.99×10⁹/L, P < 0.01) was lower than non-cirrhosis patients. This paper established a diagnostic model based on three non-invasive parameters (2D-SWE, platelet and HBsAg). The area under curve (AUC) for diagnosis of cirrhosis was 0.823(0.776-0.864)for modeling group and 0.860(0.795-0.911)for validation group, which was significantly better than the previous reported non-invasive detection model (P < 0.001).

Conclusions The non-invasive cirrhosis prediction model established in this study correlates well with pathology, and may allow most patients to avoid invasive liver biopsy.