Derivation and Validation of a Model to Predict Clinically Significant Portal Hypertension Using Transient Elastography and FIB-4
Bubu A. Banini, Samarth Patel, Jonathan W. Yu, Le Kang, Christopher Bailey, Brian J. Strife, Mohammad S. Siddiqui, Vaishali Patel, Scott C. Matherly, Hannah Lee, Shawn Lewis, Reena Cherian, Richard T. Stravitz, Velimir Luketic, Arun J. Sanyal, Richard K. Sterling
Abstract
BACKGROUND: Liver biopsy and hepatic venous pressure gradient (HVPG), the gold standard for assessing advanced fibrosis (AF) and clinically significant portal hypertension (CSPH), are invasive, costly, and time-consuming. GOAL: We investigated if the combination of fibrosis index based on 4 factors (FIB-4) and liver stiffness measure (LSM) can identify AF and more importantly, CSPH. PATIENTS AND METHODS: Patients with chronic liver disease referred for transjugular liver biopsy were analyzed retrospectively. FIB-4 and LSM were compared with liver histology for diagnosing AF. FIB-4, LSM, and platelet count were compared with HVPG for diagnosing CSPH. Optimal cutoffs for predicting CSPH were determined by grid search. A composite log-odds to predict CSPH was derived from logistic regression using LSM, FIB-4, and gender. Internal bootstrap validation and external validation were performed. RESULTS: A total of 142 patients were included in the derivation; 42.3% had AF, and 11.3% had CSPH using the current gold standards. The area under the receiver operating characteristic curve (AUROC) for LSM, FIB-4, and their combination to predict AF were 0.7550, 0.7049, and 0.7768, respectively. LSM, FIB-4, and platelet count predicted CSPH with AUROC 0.6818, 0.7532, and 0.7240, respectively. LSM plus FIB-4 showed the best performance in predicting CSPH with AUROC 0.8155. Based on LSM, FIB-4, and gender, a novel model-the Portal Hypertension Assessment Tool (PHAT)-was developed to predict CSPH. PHAT score ≥-2.76 predicted CSPH with sensitivity 94%, specificity 67%, positive predictive value 27%, negative predictive value 99%, and accuracy 70%. In internal and external validation, AUROCs for the model were 0.8293 and 0.7899, respectively. CONCLUSION: A model consisting of FIB-4, LSM, and gender can identify CSPH among patients with chronic liver disease.