The majority of patients who die of cirrhosis die due to a complication of increased portal venous pressure, such as variceal hemorrhage, ascites, hepatic encephalopathy, hepatopulmonary syndrome, or hepatorenal syndrome.[1, 2] The hepatic vein pressure gradient (HVPG), an indirect measure of portal pressure, is a prognostic indicator for long-term survival in cirrhosis[1, 2] Furthermore, HVPG can reflect progression of disease in the precirrhosis stage. There is an association between the severity of hepatic inflammation and fibrosis and the HVPG even before cirrhosis develops. In addition, HVPG predicts the response to hepatitis C treatment among patients with cirrhosis.
One of the most frequent severe complications of portal hypertension is hemorrhage from gastroesophageal varices (GEV), which is a significant cause of death in patients with cirrhosis. Reduction of the HVPG below 12 mmHg (normal is 0-5 mmHg), either through spontaneous reversion after the insult is resolved or with medical, radiological, or surgical interventions, effectively prevents recurrent bleeding.[3, 6-8] Currently, there is no established noninvasive test to predict portal pressure among patients who are treated medically and, thus, there is no way to predict either the response to standard of care or the complications of portal hypertension (including potentially lethal esophageal bleeding) other than performing screening esophagogastroduodenoscopy (EGD) with the added costs and morbidity of the procedure. Although transient elastography has a very good predictive value for clinically significant portal hypertension, there are some limitations of this technique in patients with chronic liver diseases and with obesity.
The ability to predict portal pressure with a simple blood test would revolutionize clinical management of patients with chronic liver diseases, as well as aid in the design and performance of clinical research into the complications of cirrhosis.[1, 2] Given that liver inflammation due to liver injury and/or bacterial translocation occurs in liver cirrhosis with portal hypertension,[10-15] we postulated that some inflammatory biomarkers could serve as a noninvasive test to predict the presence of severe portal hypertension at levels associated with the presence variceal bleeding.
In this study we found that the novel inflammatory biomarkers IL-1β, IL-1Rα, Fas-R, VCAM-1, TNFβ, and HSP-70 are significantly correlated with HVPG in a compensated cirrhosis cohort. Further, and as expected, some demographic and clinical parameters correlated significantly with HVPG, including age, MELD, CPS, platelets, and at-risk alcohol use.
The rationale for screening inflammatory serum biomarkers of HVPG is based on the fact that portal hypertension is pathogenically related to liver injury and fibrosis,[10-15] and that in turn these are associated with the activation of inflammatory pathways.[11, 12, 14, 15] Indeed, portal hypertension occurs in the presence of liver injury and inflammation even in the absence of liver fibrosis in fulminant acute liver failure and acute viral hepatitis,[17, 18] indicating that liver injury and inflammation can be sufficient and critical for the development of portal hypertension (with 50% of the patients having portal pressures >12 mmHg). In addition, patients with chronic alcoholic liver disease in the absence of cirrhosis may have HVPG >12 mmHg and develop esophageal varices, suggesting that in addition to and sometimes in the absence of liver fibrosis, hepatocyte injury and inflammation affect the portal pressure.
Inflammatory pathways can be activated by bacterial translocation (or translocation of LPS and DNA) from the intestine to the portal vein circulation that occurs in patients with cirrhosis and portal hypertension.[10, 13, 14] Bacterial/LPS/DNA translocation leads to activation of Toll-like receptors (TLRs) and their induction of signaling pathways, resulting in the secretion of inflammatory mediators into the circulation.[12, 13] In support of our findings, the activation of these signaling inflammatory pathways may be clinically inconspicuous but could be detected by measuring hemodynamic effects or humoral mediators in blood.[10, 12, 13] The increase in HVPG after a meal significantly correlated with serum bacterial DNA concentration, suggesting a causal effect between HVPG and bacterial translocation.
A critical inflammatory signaling pathway is the inflammasome. We have found that IL-1β, a critical cytokine product of the inflammasome, and its receptor IL-1Rα correlated significantly with HVPG.[20, 21] Active caspase-1 is essential for the cleavage of pro-IL-1β into its mature, biologically active form IL-1β. Based on this rationale, anti-caspase drugs are being analyzed in clinical phase 2 studies to ameliorate hepatocyte injury. Similarly, polymorphisms of the TLR-9, which initiates signals activating the Inflammasome,[23-25] have been implicated in rapidly progressing tissue fibrosis. We have also found that TNFβ, a product of activated T and B lymphocytes and a member of the TNFα superfamily, correlates significantly with HVPG. TNFβ is secreted as a soluble inflammatory polypeptide that forms heterotrimers with lymphotoxin-β and mediates a large variety of inflammatory, immunostimulatory, and antiviral responses, which are relevant to our cohort of cirrhosis patients etiologically linked to chronic HCV infection and alcohol use. In addition, the serum Fas-R, another member of the TNFα cell death receptor superfamily, which may be increased with liver injury and inflammation,[28, 29] also correlated with HVPG.
We also found a highly significant correlation between HVPG and serum VCAM-1, a product of endothelial cells.[30-32] The increase in circulating endothelial cells in cirrhosis patients is congruent with our findings. In addition, bacterial DNA translocation is associated with intrahepatic endothelial dysfunction in patients with cirrhosis. Hyaluronan, homocysteine, and angiotensin-II can induce the expression of VCAM-1 synthesis.[30-32] All of these factors are mechanistically related to cirrhosis. Serum hyaluronan and homocysteine are increased in liver fibrosis, while angiotensin-II stimulates liver fibrosis.[34-37] Therefore, in future studies we will analyze the relationship between hyaluronan, homocysteine, and angiotensin II with HVPG.
HSP-70 correlated significantly with HVPG in our logistic regression analysis. Of interest, glutamine, an amino acid induced in hepatic acinar zone 3 by hypoxia (characteristic of cirrhosis with portal hypertension), stimulates transcription of heat shock factor (HSF)-1, an inducer of HSP-70.[38, 39] Thus, glutamine and glutamine syntethase may also be biomarkers of HVPG.
We found a significant correlation using Pearson's test of HVPG with novel inflammatory biomarkers (IL-1β, IL-1Rα, Fas-R, and VCAM-1). By using multivariate logistic regression analysis and selected parameters (TNFβ, HSP-70, at-risk alcohol use, and Child class B) we can exclude HVPG ≥12 mmHg with 86% accuracy (significant 95% CI: between 67.78 and 96.16%) and the sensitivity was 87.09% (significant 95% CI: between 69.68 and 96.34%). Therefore, the composite test could identify 86% of compensated cirrhosis patients with HVPG below 12 mmHg and prevent unnecessary EGDs with their associated morbidity and costs in these patients. As is the case for estimating HVPG by measuring liver stiffness (LS) with transient elastography, our diagnostic test was not efficient in predicting HVPG ≥12 mmHg (PPV: 45.76%; specificity: 43.86%). Therefore, as expected our ROC was only moderately accurate (area 0.767 ± 0.057; asymptotic sigma P < 0.0001; 95% CI: 0.656 to 0.879) and similar to the ROC curve (0.76 ± 0.07; 95% CI: 0.60-0.87) reported for the prediction of HVPG by LS-elastography for all cirrhosis patients in their cohort.
Although LS has been proposed for predicting HVPG, the method as currently used has several technical and logistic limitations, making the measurement not interpretable in a large percentage of patients with cirrhosis. The exclusion criteria for LS include obesity, ascites, congestive heart failure, extrahepatic cholestasis and severe liver inflammation related to HCV infection.[9, 42] Also, in cirrhosis patients LS values increased by 25% after a light meal, as compared with fasting patients, suggesting a spurious postprandial increase in the predicted HVPG in cirrhosis.
Vizzutti et al. reported a good correlation between LS and HVPG in the entire cohort (R2 = 0.61; P <0.0001) in 61 consecutive selected patients with HCV-related chronic liver disease. Although the correlation between LS and HVPG was very good for HVPG values less than 10 or 12 mmHg (R2 = 0.72, P = 0.0001 and R2 = 0.67 P <0.0001, respectively) it was poor for HVPG >10 mmHg and >12 mmHg (R2 = 0.35, P = 0.0001 and R2 = 0.17 P <0.02, respectively). Berzigotti et al. have shown that LS provides excellent results when combined with platelets count and spleen size (LSPS). Analyses of LSPS were effective in identifying patients with clinically significant HVPG; they correctly classified 83% of patients in the training set (N: 117) and 85% in the validation set (N: 56). Berzigotti et al. also reported that obesity was present in 30% of a cohort of compensated cirrhosis patients. Thus, in evaluating HVPG by LS including all subjects (an “intention to diagnose” study), the 85% predictive accuracy of LSPS reported by Berzigotti et al. would be applicable to only about 70% of those subjects, resulting in a correct classification of HVPG in about 60% of the patients (85% × 0.70).
Colecchia et al. suggested using spleen stiffness (SS) measurement as a screening test for the indication of esophagogastroduodenoscopy. Using an intention-to-diagnose approach, only 7 of 113 (7.1%) screened patients would have wrongly avoided esophagogastroduodenoscopy. Similarly, Sharma et al. found that SS ≥40.8 kPa had high sensitivity (94%), specificity (76%), PPV (91%), NPV (84%), and diagnostic accuracy (86%) for predicting EV. However, in the latter study, out of 270 patients SS was performed only in 174 patients since 70 patients were excluded before performing the SS measurement (due to ascites, alcohol abuse, and hepatitis reactivation) or in 26 of whom the SS measurement could not be obtained. Thus, the intention-to-treat would markedly reduce the sensitivity of the technique.
Our study also has significant limitations: 1) the presence of only cirrhosis patients without esophageal varices in our cohort may not fully reflect other cirrhosis populations; 2) the lack of a validation cohort, as was the case with the studies using LS or SS by elastography[40, 46, 47]; and 3) the absence of a control group. Additional studies with a larger cohort of cirrhosis patients, including a significant percentage of cirrhosis patients with esophageal varices and a validation cohort, will be needed to confirm our findings. More important, a larger cohort, while introducing additional factors related to the selected members of the specific inflammatory signaling pathways, may allow a more precise prediction of HVPG in the future.
It remains to be established whether a blood test for HVPG would be effective in all patients, including those unsuitable for LS or SS measurements (e.g., patients with obesity, ascites, congestive heart failure, and extrahepatic cholestasis). Our present test has a similar accuracy in predicting HVPG to LS or SS and it may become more accurate with additional biomarkers. Thus, if a test based on blood biomarkers could be developed it would be more accessible worldwide due to low costs and ease of execution.