Predicting Clinically Significant Portal Hypertension in MASLD Using Routine Ultrasound and Clinical Data: A Retrospective Cohort Study

Background Clinically significant portal hypertension (CSPH) is a major determinant of adverse outcomes in metabolic dysfunction–associated steatotic liver disease (MASLD), but reliable noninvasive identification remains challenging. We evaluated whether routinely available ultrasound and clinical variables could help identify CSPH in adults with MASLD. Methods We performed a retrospective cohort study of adults with MASLD who underwent HVPG measurement at a tertiary referral centre. Routinely available demographic, laboratory, routine ultrasound and liver stiffness variables were evaluated using logistic regression and internally validated machine learning models to identify CSPH, defined as HVPG ≥ 10 mmHg. Results Among 218 adults with MASLD, 44 (20.2%) had CSPH. Compared with patients without CSPH, those with CSPH showed less favourable noninvasive profiles, particularly in ultrasound‐ and stiffness‐based measures (liver stiffness: 9.2 vs. 7.8 kPa; spleen length: 119.3 vs. 115.0 mm), whereas most routine biochemical markers were broadly similar between groups. Overall predictive performance was modest across all evaluated models; random forest achieved the highest discrimination (AUC = 0.67), followed by SVM (AUC = 0.65) and logistic regression (AUC = 0.604). In multivariable analysis, portal vein calibre (OR = 1.60) and liver stiffness (OR = 1.134) showed the strongest independent associations with CSPH, although overall predictive accuracy remained limited. Conclusions In adults with MASLD, routinely available ultrasound and clinical variables showed only modest ability to identify CSPH. Although some noninvasive markers were independently associated with CSPH, model performance was insufficient for stand‐alone clinical triage. These findings support further refinement of MASLD‐specific noninvasive prediction strategies rather than immediate clinical implementation.