Accurate computed tomography-based portal pressure assessment in patients with hepatocellular carcinoma Pouya Iranmanesh 1,⇑ , Oscar Vazquez 1 , Sylvain Terraz 2 , Pietro Majno 1 , Laurent Spahr 3 , Antoine Poncet 4 , Philippe Morel 1 , Gilles Mentha 1 , Christian Toso 1 1 Department of Digestive Surgery, Geneva University Hospitals and Faculty of Medicine, Switzerland; 2 Department of Radiology, Geneva University Hospitals and Faculty of Medicine, Switzerland; 3 Department of Gastroenterology and Hepatology, Geneva University Hospitals and Faculty of Medicine, Switzerland; 4 Department of Clinical Epidemiology, Geneva University Hospitals and Faculty of Medicine, Switzerland Background & Aims: Liver resection is generally restricted to patients without clinically significant portal hypertension (Hepatic Venous Pressure Gradient – HVPG – 610 mmHg) and several teams perform transjugular HVPG measurements as part of the pre-operative work-up. The present study investigates whether a non-invasive Computed Tomography (CT)-based assessment could be as accurate as the invasive transjugular measurement. Methods: A cohort of patients with hepatocellular carcinoma (HCC) treated by resection (n = 36) or transplantation (n = 39) was selected (mean age: 61 ± 9.2 years, male/female ratio: 4/1). Pre-operative CTs were read by two independent investigators, and potential CT-based HVPG predictors were compared to the transjugular HVPG measurements. A validation was conducted on another cohort of 70 non-surgical patients. Results: The invasive HVPG values were significantly correlated to liver/spleen volume ratio, spleen volume, platelet count, and peri-hepatic ascites (p <0.001), which all showed high inter-observer agreements (intra-class correlation coefficients P0.927, Kappa P0.945). The presence of a HVPG >10 mmHg was best predicted by the liver/spleen volume ratio (AUC: 0.883 [0.805–0.960]) and the peri-hepatic ascites (p <0.001). These two variables were combined into an accurate model for predicting HVPG >10 mmHg (AUC: 0.911 [0.847–0.975]), with sensitivity, specificity, and positive and negative predictive values of 92%, 79%, 91%, and 81%. The model was also accurate in the validation cohort with an AUC of 0.820 [0.719–0.921]. The computed formula was: HVPG score ¼ 17:37  4:91  lnðLiver=Spleen volume ratioÞ þ 3:8 ½if presence of peri-hepatic ascites Conclusions: The proposed CT-based model showed a high accu- racy in the prediction of HVPG and, if further confirmed by pro- spective validation, could replace the invasive transjugular assessment in patients not requiring a biopsy of the non-tumoral liver. Ó 2013 European Association for the Study of the Liver. Published by Elsevier B.V. All rights reserved. Introduction Liver resection and transplantation are effective treatments for selected patients with hepatocellular carcinoma (HCC) [1–4]. The choice between the two options is based on cancer stage (size, number and alpha-fetoprotein level), hepatocellular func- tion and the presence of portal hypertension [3,5–8]. Among other available markers, the hepatic venous pressure gradient (HVPG >10 mmHg) accurately predicts the risk of peri-operative morbidity and death and many Western centers perform transjugular pressure measurements as part of their standard pre-surgical work-up [3,9,10]. However, the transjugular exploration is invasive and not available at all institutions. In addition, a biopsy of the non-tumoral liver (which can be per- formed during the transjugular assessment) is not always neces- sary. As a result, a number of less invasive portal pressure assessment techniques have been tested over the recent years [11,12]. They were based on ultrasonography with the explora- tion of splanchnic vessels size [13], blood velocity [14–16] or blood-flow resistance [17] and on the assessment of the liver stiffness [18]. Overall, none of these techniques gained clinical acceptance, because of the small sample size, the lack of external validation and/or simply because of their low accuracy in the Journal of Hepatology 2014 vol. 60 j 969–974 Keywords: Hepatic venous pressure gradient; Portal hypertension; Liver transplantation; Liver resection; Hepatocellular carcinoma; Computed tomography. Received 27 July 2013; received in revised form 18 November 2013; accepted 10 December 2013; available online 19 December 2013 ⇑ Corresponding author. Address: Department of Digestive Surgery, Geneva University Hospitals and Faculty of Medicine, Rue Gabrielle-Perret-Gentil 4, 1205 Geneva, Switzerland. Tel.: +41 79 55 33 189. E-mail addresses: Pouya.Iranmanesh@hcuge.ch (P. Iranmanesh), Christian. Toso@hcuge.ch (C. Toso). Abbreviations: HCC, hepatocellular carcinoma; HVPG, hepatic venous pressure gradient; CT, computed tomography; ROI, region of interest; WHVP, wedged hepatic venous pressure; FHVP, free hepatic venous pressure; ICC, intra-class correlation coefficient; ROC, receiver operator characteristic; AUC, area under the curve; HCV, hepatitis C virus; PPV, positive predictive value; NPV, negative predictive value. Research Article