Most patients with chronic liver diseases develop different degrees of liver fibrosis and portal hypertension, the latter being accompanied by the formation of portosystemic collateral vessels like esophageal varices [18]. Varices are a potentially life-threatening source of bleedings and are regularly monitored and treated via endoscopy [19]. Endoscopy itself, however, poses several risks, not only in patients with cirrhosis [20, 21]. A noninvasive risk-stratification for potential bleeding events and the degree of esophageal varices would thus be desirable.

Several laboratory risk factors for esophageal bleeding in cirrhotic patients and predictors for short-term mortality have been identified. Amitrano et al. promoted the MELD score as an indicator for short-term mortality among cirrhotic patients at their first episode of bleeding from esophageal varices [2]. Sanyal et al. connected the risk of developing varices to decreased platelet counts, increased bilirubin, and increased INR [22].

Other studies have identified and evaluated CT findings (size of esophageal and gastric varices, protrusion of gastric varices, hepatomegaly, splenomegaly, and ascites), most of them relying on manual measurements [6, 23, 24].

However, a more objective noninvasive stratification with radiographic methods such as perfusion computed tomography based on accurate quantification of the dual blood supply to the liver has not been investigated yet. As several patients at risk for HCC regularly undergo PCT at few institutions [8, 25], we hypothesized that the hepatic perfusion index could serve as a surrogate parameter for the risk of bleeding in patients with portal hypertension and esophageal varices and might correlate with the extent of esophageal varices.

However, our study data show that this hypothesis must be rejected. Accordingly, the eta correlation coefficient showed neither a positive nor a negative correlation of HPI and bleeding events or the need for variceal band ligation, respectively. Also, the presence of esophageal varices itself did not correlate with the HPI. Grading of the esophageal varices according to Paquets and the degree of liver fibrosis according to Child–Pugh did not show a correlation with the HPI. Analysis of variance revealed that the HPI could not estimate the extent of the esophageal varices according to Paquets.

As patients with a prior history of bleeding esophageal varices and/or variceal band ligation might have biased the correlation analysis [17], we performed a subgroup analysis (n = 33), including only patients with no prior history of bleeding esophageal varices and variceal band ligation. However, even in this subgroup analysis, there was no significant correlation between the investigated parameters and the HPI.

As mentioned in the introduction, we hypothesized that the HPI would correlate with the degree of esophageal varices and bleeding events, as the extent of varices is directly connected to portal hypertension [11]. However, this was not the case in our patient population. There are several potential explanations. The hemodynamic principles of portal hypertension, including increased intrahepatic resistance and hyperdynamic circulation, are complex with subsequent collateral circulation and varices. Feldman, Friedman, and Brandt described the underlying pathophysiological processes in detail [26].

Portal hypertension mainly results from changes in portal resistance in combination with changes in portal inflow. Increased portal resistance is, in principle, a result of mechanical factors that reduce vessel diameter. Hepatic vasoconstriction and resistance to vasodilatory stimuli (such as NO) also increase the portal resistance. Hyperdynamic circulation is an additional factor for portal hypertension. The total blood volume draining into the portal circulation, not necessarily the portal vein, is caused particularly through vasodilatation in the splanchnic bed.

Under these circumstances, the collateral circulation subsequently develops and expands in response to the increased portal pressure. Progression of portal hypertension results from the prominent obstructive resistance in the liver, resistance within the collaterals themselves, and continued increase in portal vein inflow.. Moreover, in the context of the law of Laplace, other local factors that increase variceal wall tension (transmural pressure gradient between the variceal lumen and esophageal lumen, the variceal radius, and variceal wall thickness) are also required for varices to form and bleed. However, the changes in portal pressure and local variceal factors are dynamic and influenced by several physiologic (i.e., increase in intra-abdominal pressure, meal-induced increases in portal pressure), diurnal (circadian changes in portal pressure), and pathophysiologic (acute alcohol use) factors. Therefore, portal pressure and esophageal variceal pressure may vary over time [26,27,28,29].

In the context of all these different complex mechanisms behind portal hypertension, it becomes clear that our approach quantifying the hepatic perfusion index only covers a part of the underlying causes for the development of esophageal varices and potential bleeding.

A rationale for the exclusion of patients with portal venous thrombosis was a publication that emphasized the presence of portal venous thrombosis as an independent risk factor for aggravation of esophageal varices in patients with hepatocellular carcinoma [3]. It would have, therefore, been difficult to rule out thrombosis as a possible confounder. However, this means that patients with an extremely high risk for esophageal bleeding were excluded from the cohort, and the study might therefore be underpowered.

Our study has other limitations. First, the reason for the conduction of PCT in the present cohort was to rule out HCC or exclude additional HCC manifestations before transplantation or resection. Therefore, the small present dataset was analyzed in a single-center retrospective study design. This resulted in only 66 eligible patients for inclusion. The sample size was sufficient for detecting strong correlations (> 0.5). However, the detection of significant moderate-sized correlations was not possible, which would have required a minimum of 84 patients. Furthermore, there was a disproportioned distribution of Child–Pugh class within the sample. Thus, the cohort consisted of patients with liver cirrhosis mainly in Child–Pugh class A (35/66, 53%) and B (25/66, 38%), and the main proportion of patients had only low grade or no present esophageal varices in endoscopy, which might be the reason why only 12% had bleeding events and 18% underwent variceal ligation. Unfortunately, due to the retrospective study design, it was impossible to include more patients to account for the mentioned disproportions. In summary, the sample size, especially of patients with liver cirrhosis in Child–Pugh class C (6/66, 9%), might have been too small and underpowered to detect a significant correlation of HPI and Child–Pugh class or variceal grades. Therefore, the results must be limited to patients in Child–Pugh class A and B. Furthermore, the study used only indirect data for the extent of portal hypertension (Child–Pugh score for the extent of liver cirrhosis and HPI). Additional clinical data such as Doppler sonography of the portal vein and invasive measurements of the portal venous and hepatic venous pressures using percutaneous transhepatic catheterization and venous catheterization, respectively, are missing.

In summary, our study data did not show a strong correlation between HPI and the degree of esophageal varices and variceal bleeding, potentially due to a lack of statistical power. Visual identification of the degree of esophageal varices via endoscopy and risk stratification for bleeding with MELD score, Child–Pugh score, or visual CT-parameters such as intraluminal varix protrusion, varix size as well as liver and spleen volume seems to be more robust than noninvasive parameters using PCT.