Potential conflict of interest: Nothing to report.
Hereditary hemorrhagic telangiectasia (HHT) is a genetic disease characterized by cutaneous, mucosal, and sometimes visceral arteriovenous malformations. Severe hepatic manifestations have been characterized in a subgroup of patients, but few data are available in previously nonscreened patients. We prospectively evaluated liver involvement and its cardiac consequences in such patients. Between 2000 and 2005, we prospectively evaluated the clinical, biological, and hepatic Doppler sonography (DS) characteristics of 102 consecutive HHT patients (mean age, 52.5 years; range, 19-88; 80.4%) with an identified genetic mutation. Patients were segregated into three different severity groups according to DS values. Factors predictive of an abnormal DS, according to predetermined criteria, and of a high cardiac index were identified by logistic and linear regression analysis, respectively. Abnormal liver biology and clinical signs of hepatic involvement were present in 35.3% and 27.5% of cases, respectively. Abnormal DS (defined as at least enlargement of the main hepatic artery) was observed in 56 (54.9%) cases, and direct or indirect signs of significant fistulas were present in 26 (25.5%) cases. Abnormal liver biology and a mutation involving the ACVRL1 gene were predictive of hepatic ultrasound (US) abnormalities. The diameter of the main hepatic artery and the presence of focal nodular hyperplasia (FNH) were predictive of a higher cardiac index. Conclusion: This large prospective series of previously nonscreened HHT patients identified a subgroup at risk of liver involvement (patients with abnormal liver biology and ACVRL1 mutations) and a subgroup with a higher cardiac index: future studies will show whether such patients would benefit from systematic DS screening and long-term cardiac surveillance. (HEPATOLOGY 2008;48:1570–1576.)
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Hereditary hemorragic telangiectasia (HHT) is an autosomal dominant genetic disease driven by mutations in genes such as endoglin (ENG), activin receptor-like kinase type 1 (ACVRL1), and Smad-4, which control the transforming growth factor beta proliferation pathway.1 HHT patients usually develop a wide range of cutaneous, mucosal, and sometimes visceral arteriovenous malformations.2–4 The symptomatology is dominated by epistaxis and anemia, but potentially life-threatening visceral localizations can be present, such as pulmonary, hepatic, and cerebral vascular malformations.5
Hepatic involvement in HHT has been described in several studies.6, 7 Characteristic HHT hepatic localization consists of arteriovenous shunting responsible, in severe cases, for high-output cardiac failure (fistulas between the hepatic artery and hepatic veins), ascites (fistulas between the hepatic artery and portal veins), and cholangiopathy with sepsis. Liver transplantation is the main treatment option at these advanced stages. Previous, mostly small series have shown Doppler sonography (DS) abnormalities in 30% to 70% of HHT patients.8, 9 Early and advanced DS findings include (1) enlargement, increased flow velocity, and tortuosity of hepatic artery branches; (2) increased diameter and flow velocity in hepatic veins; and (3) modification of portal flow.6, 10 Computed tomography may also show very characteristic diffuse intrahepatic telangiectases and early opacification of the hepatic or portal veins indicating arteriovenous shunting.6, 11 Nodular lesions with the radiological characteristics of focal nodular hyperplasia are also frequently observed in HHT patients and have been histologically confirmed in a few cases.12, 13 Little is known about either early liver manifestations and their consequences on patients' health, or the frequency of liver involvement in asymptomatic HHT mutation carriers. There is also little information as to whether certain early hepatic changes are predictive of an unfavorable outcome with evolution into one of the previously described late-stage complications.
The current study was part of a prospective multidisciplinary cohort study initiated by a French HHT network. Its aims were to describe the nature and prevalence of liver lesions in HHT patients without previously identified liver involvement, to look for the hemodynamic consequences of undiagnosed liver disease, and to describe potential phenotype–genotype correlations regarding liver involvement in a large series of HHT patients.
Between October 2000 and October 2005, 110 consecutive patients with HHT were evaluated for hepatic manifestations of the disease as part of a complete clinical (hepatic, digestive, pulmonary, and neurological), morphological, and genetic protocol. All patients gave informed consent, and the study protocol was approved by the Institutional Ethic Committee of the Hospices Civils de Lyon. The Curaçao criteria were used for the diagnosis of HHT.1 Of these 110 patients, 102 had no previously identified liver involvement or liver-related symptoms and were referred for systematic hepatic screening, corresponding to “asymptomatic screened patients.” Eight patients were referred for evaluation and management of known liver abnormalities (biological, clinical, or morphological) and were thus excluded from this study.
All patients underwent a clinical examination, in search of one of the four previously described clinical signs of hepatic involvement in HHT patients6: audible bruit, thrill, pulsatility in the right upper quadrant, or hepatomegaly. Clinical abnormality was defined as the presence of one of these signs. Biological liver tests were performed on all patients, including tests for transaminases, alkaline phosphatases, gamma-glutamyl transferases, and total bilirubin. Biological liver abnormality was defined as the presence of at least one abnormal test. For patients with a history of blood transfusion, hepatitis C serology was checked.
Hepatic DS was performed by experienced radiologists (F.P., N.R.) in accordance with a predetermined protocol. The examinations were performed using the Sonoline Antares (Siemens-Acuson) ultrasound hardware. The following values were obtained prospectively: diameter of the common hepatic artery (between 5 and 10 mm distal to the celiac trunk) and maximum velocity, with the identification of anatomical variants (secondary artery developed from the mesenteric artery); diameter of the extrahepatic portal vein (at the midportion) and of the three hepatic veins (approximately 2 cm proximal to the entry into the inferior caval vein); maximum velocity in the hepatic artery (v max); mean velocity in the portal vein and hepatic veins; and flow direction in the portal vein and intrahepatic right and left portal veins. For each arterial, hepatic venous, and portal venous Doppler examination, three measurements were obtained and averaged. The presence or absence and type of hepatic nodules were also assessed, including those nodules compatible with a diagnosis of focal nodular hyperplasia: the typical focal nodular hyperplasia (FNH) aspect at DS included a slightly hypoechoic, isoechoic, or slightly hyperechoic mass with an arterial spectrum and central vascularization with a centripetal vein and a “spoke wheel” pattern within the lesion.
The cardiac index (L/min/m2) was calculated from the cardiac output, which was obtained in 85 patients using the following algorithm: Qc = π × d2/4 × VTI × Qf. Qc: cardiac output, d: diameter of the left ventricular outflow tract; VTI: sub-aortic velocity time integral; Qf: cardiac frequency.
Mutation analysis was performed in the molecular biology laboratory at Edouard Herriot Hospital (Lyon). DNA was isolated from peripheral blood lymphocytes using a purification kit (QIAamp blood kit, Qiagen, Courtaboeuf, France). The methodology for the mutation analysis of the ENG and ACVRL1 genes has been described previously.14, 15
Interpretation of the Results
Previous reports have described different early and late signs of liver involvement in HHT patients using ultrasonography, but there is no uniform consensus on the specific signs of liver involvement, or for distinguishing between early and advanced liver involvement. Abnormality of the common hepatic artery (diameter, flow velocity, tortuosity) may be one of the earliest signs of liver involvement, and an increased diameter appears to be one of the most specific signs of the disease.10, 16–18 We thus considered, in the first group (G1), patients with no abnormality evidenced at DS in the diameter or maximum velocity of the hepatic artery portal and hepatic veins (Table 1); in group 2, we included patients with “early changes” involving only the diameter or maximum velocity of the hepatic artery. Group 3 patients presented signs of significant arteriovenous fistulas, involving either the portal or hepatic veins, or a directly visible intrahepatic vessel suggestive of arteriovenous fistula, and all patients in group 3 had hepatic artery diameter enlargement or high maximum velocity.
Table 1. Definition of the Three Groups of Patients According to Hepatic Doppler Ultrasonography Values
HA, common hepatic artery; PV, main portal vein; HV, hepatic veins. Patients were included in group 3 when at least one venous criterion (PV or HV) was present, and all patients in group 3 had group 2 criteria for HA. Cutoff values were chosen accordingly to previous studies.1, 2
HA diameter (mm)
HA maximal velocity (cm/s)
PV flow direction
HV maximal diameter (mm)
HV maximal velocity (cm/s)
Statistical analysis was performed using SAS Institute software (Cary, NC). Data are presented as mean (± standard deviation). All P values were two-sided. Proportions and categorical variables were compared using chi-squared and Fisher's exact tests. Continuous variables were compared using Student t and Pearson's correlation coefficient. We performed logistic models to predict groups by adding age, type of mutation, clinical and biological abnormalities, and calculated the odds ratios and their 95% confidence intervals. Cardiac index values were tested for normal distribution using the Kolmogorov Smirnov goodness-of-fit test for continuous variables. A multiple linear regression analysis was carried out to determine the effects on cardiac index of separate variable sets: sex, mean flow of the portal vein, mean hepatic artery size, hemoglobin, pseudo FNH. The r2 value indicates the overall proportion of variation in cognitive measures that is explained by the model. A P value < 0.05 was considered significant.
Clinical abnormalities revealed on hepatic examination were present in 28 patients (27.5%), including hepatomegaly, four cases; pulsatility, 21 cases; audible bruit, 15 cases; and thrill, three cases.
Biological liver abnormalities (see Patients and Methods) never exceeding twice the upper limit of normal values were present in 36 patients (35.3%). Abnormal biology concerned transaminases (trans glutamyl in all cases) in 10 patients, alkaline phosphatases in 14, gamma-glutamyltransferase in 26, and bilirubin in one. Nine patients (9%) had a history of blood transfusion, of which only one had a positive hepatitis C serology.
Hepatic Doppler Sonography.
According to our predefined criteria, DS showed no liver abnormality in 46 cases (group 1, 45.1%), isolated enlargement or increased velocity of the main hepatic artery in 30 (group 2, 29.4%), and abnormality of the hepatic veins, portal veins, or direct visualization of large arteriovenous fistulas in 26 cases (group 3, 25.5%). Table 2 shows the mean values, for each group, of the six main DS values obtained prospectively.
Table 2. Clinical Characteristics and Doppler Sonography Values in the Three Groups of Patients
HA, common hepatic artery; PV, main portal vein; HV, hepatic veins; FNH, suspected focal nodular hyperplasia.
Values are expressed as mean (range) or number (%);
P value of intergroup comparison using chi-squared (categorical variables) or t test (continuous variables).
Number of patients
Mean hepatic artery size (mm)
HA mean maximal velocity
PV mean diameter
PV mean velocity
HV maximal diameter
HV mean maximal velocity
Mean cardiac index (L/min/m2)
Presence of clinical abnormality
4 (8.7 %)
15 (57.6 %)
Presence of biological abnormality
6 (13.0 %)
12 (40.0 %)
18 (69.2 %)
1 (2.2 %)
4 (13.3 %)
9 (34.6 %)
Mutations in the ACVRL1 and ENG Genes.
A specific mutation in the ACVRL1 or ENG genes was identified in 82 patients (80.4%), in the ACVRL1 gene in 62 (60.8%), and in the ENG in 20 (19.6%, Table 3). The percentage of patients with ACVRL1 mutations increased from group 1 (47.8%) to group 3 (73.1%) patients, whereas the proportion of patients with ENG mutations clearly decreased from group 1 (32.6%) to group 3 (3.8%). Interestingly, 19 of the 20 group 3 patients (95.0%) with an identified mutation were ACVRL1 cases. The proportion of women was comparable in patients with ACVRL1 and ENG mutations (59.6% and 55.0%, respectively). The mean age of patients with an ACVRL1 mutation (53.6 years) was not significantly different compared with that of patients with ENG mutations (46.5 years, P = 0.09).
Table 3. HHT Mutations in the Three Patient Subgroups
Group (number of patients)/Mutation
Group 1 (46)
Group 2 (30)
Group 3 (26)
ENG, endoglin; ACVRL1, activin receptor-like kinase type 1.
DS identified hepatic nodules compatible with FNH in 14 patients in this series (13.7 %), all of them women. The mean number of lesions per patient was 3.5 (range, 1-12), and the mean maximum diameter per patient was 49 mm (range, 10-90 mm). A nuclear magnetic resonance examination was performed in 11 cases and confirmed the presence of one or more nodule compatible with FNH in 10 cases.
Clinical Significance of Doppler Sonography Findings.
We considered the three groups of patients according to the proposed classification based on DS results. As shown in Table 2, the proportion of patients with clinical or biological abnormalities significantly increased from group 1 to group 3. Of the 28 patients with an abnormal clinical liver examination, 24 (85.7%) were in group 2 or 3, versus 32 of 74 (43%) of patients without clinical abnormality (P < 0.01). Of the 36 patients with abnormal liver biology, 30 (83.3%) corresponded to group 2 or group 3 patients, versus 26 of 66 (39.4%) patients without abnormal liver biology (P < 0.01). Moreover, because one of the main long-term consequences of hepatic fistulas in HHT is the risk of high-output cardiac insufficiency, we tested whether the cardiac index increased from group 1 to group 3 patients in our series. As shown in Table 2 and Fig. 1, the mean calculated cardiac index increased significantly from group 1 to group 3 patients (P < 0.01). These results thus suggest that the proposed classification, based on DS criteria, reflects different grades of liver involvement with potentially important clinical consequences.
Abnormal Liver Biology and the Presence of an ACVRL1 Mutation Are Independent Predictive Factors of a Pathological Hepatic Doppler Sonography.
Although DS is a noninvasive test, screening all asymptomatic HHT patients would amount to considerable cost, justifying the search for criteria predictive of liver involvement in this disease. We used logistic regression analysis to determine which patients, in the current cohort, were at higher risk of hepatic involvement, thus justifying screening using DS. We included the type of gene responsible (ACVRL1 or ENG), the presence or absence of clinical or biological abnormality, and age in the logistic model. Multivariate analysis showed that abnormal liver biology (OR 6.69, 95% confidence interval 2.21-20.28) and a mutation involving the ACVRL1 gene (OR 5.45, 95% confidence interval: 1.76-17.02) were predictive of DS abnormalities. The sensitivity of both criteria in detecting patients with liver abnormalities at DS was 89%. Our results thus show that patients with abnormal liver biology and ACVRL1 mutations have a high frequency of hepatic involvement at DS.
FNH and Increased Diameter of the Hepatic Artery Are Independent Determinants of a High Cardiac Index.
One major long-term risk of hepatic arteriovenous shunting in HHT patients is the development of high-output cardiac insufficiency. We hypothesized that a high cardiac index may be a predictive criterion of this unfavorable evolution. We thus looked for criteria obtained from DS that may predict a high cardiac index. In this setting, a linear regression analysis was performed with the cardiac index as the dependent variable and age, sex, and different criteria obtained at DS as independent variables. In univariate analysis, female sex (P = 0.003), the presence of FNH (P = 0.001), mean portal flow velocity (P = 0.006), diameter and maximum velocity of the hepatic artery (P = 0.0003 and 0.002, respectively), maximum velocity of the hepatic veins (P = 0.004), and hemoglobin level (P = 0.002) were correlated with cardiac index. In multivariate analysis, the presence of FNH and the diameter of the hepatic artery were predictive of a higher cardiac index (r = 0.42).
Patients With Focal Nodular Hyperplasia at DS.
Of the 14 women with FNH suspected at DS, 13 of 14 (92.9%) were classified in group 2 (four cases, 28.6%) or three (nine cases, 64.3%, P = 0.02 group 2 or group 3 versus group 1). Interestingly, the mean cardiac index in these 14 patients was 4.2 L/min/m2 as compared with a mean 2.9 value for patients without FNH (P < 0.01). These data thus suggest a close correlation between the presence of FNH detectable at DS and liver vascular involvement in HHT women.
Several studies have described a subgroup of HHT patients developing well-characterized severe liver involvement responsible for cardiac failure, portal hypertension, and biliary complications.6, 19 One important issue in HHT patients is now to identify which patients, in a cohort of asymptomatic patients, are at higher risk of the late and severe complications of hepatic involvement. One important advantage of the current study is to describe a prospective, exhaustive study of “naïve” patients with HHT disease and to propose a classification, based on simple measures at DS, which we showed to be correlated with clinical, biological, and cardiac abnormalities. The value of cardiac output in most of the patients in the current series was only an estimation based on data obtained from cardiac ultrasonography characteristics, which is a limitation. However, this method is the only acceptable way of screening asymptomatic patients, in contrast with measurement of cardiac output through direct vascular access. Moreover, a close correlation between cardiac output measured at echocardiography and after catheterization of the pulmonary artery has been demonstrated in a series including HHT patients.10 The question remains as to whether higher cardiac output, as estimated from echocardiography, will be associated on a long-term basis with a higher risk of cardiac failure: clarifying this point is a major objective for the long-term follow-up of the current cohort of patients; however, previous series have clearly shown very high values for the cardiac index in HHT patients with severe liver involvement and complete resolution after liver transplantation.19, 20 We are now following these patients at 1-year to 3-year intervals, which appears to be the only way to determine whether a high cardiac index is responsible for negative cardiac evolution, and we also now use the B-type natriuretic factor to monitor patients with a high cardiac index (not shown). Interestingly, in this series, the three patients considered for liver transplantation because of major bleeding and a modified general state of health, associated with advanced hepatic fistulas, had a relatively high cardiac index (3.4, 3.6, and 6.6 L/min/m2, respectively).
To simplify the complex and time-consuming DS protocol used in this study, we looked for criteria predictive of high cardiac output and identified the diameter of the main hepatic artery and the presence of focal nodular hyperplasia as the two independent variables correlated in multivariate analysis. Interestingly, the diameter of the hepatic artery has previously been shown to be the most specific criteria for liver involvement in HHT patients with advanced liver disease, without overlapping values compared with different control groups.10 Confirming these data by direct intracardiac measurement of cardiac output, as is our general rule before liver transplantation,20 would be of considerable interest.
Systematic DS, although a relatively satisfying screening test (available in all centers, simple, risk-free) for liver fistulas, is not recommended on a national or international basis, as stated by Dr Buscarini in a recent consensus work in Lyon.8 One important issue is thus to specify which patients should undergo hepatic screening. Our work suggests that the best predictors of liver involvement are the presence of ACVRL1 mutations and abnormal liver biology, with 89% sensitivity of the coupled criteria. Mutations in the ACVRL1 or ENG genes appear to be relatively significant determinants of visceral involvement in HHT patients, because mutations in the ENG gene have been clearly associated with a higher frequency of pulmonary involvement and lower severity of epistaxis. The link between ACVRL1 mutations and a higher prevalence of liver arteriovenous shunts has been suggested from previous, although not all, studies.21–23 We confirm here a close link between ACVRL1 mutations and hepatic involvement but also demonstrate for the first time that the proportion of patients with ACVRL1 mutations increases as the severity of the liver disease increases. Moreover, in our experience, almost all patients investigated for severe liver involvement or referred for liver transplantation carry ACVRL1 mutations (not shown). It must be outlined that the proportion of ACVRL1 mutations is high in the French cohort of HHT patients, because this gene is involved at least twice as often as the ENG.15 For this reason, our results should be confirmed in other populations with a higher proportion of ENG mutations. Identification of an ACVRL1 mutation in families with HHT disease justifies hepatic screening because these patients represent 71% of cases in groups 2 and 3 in our study, and 63% of patients with ACVRL1 mutations, independent of age, had a liver localization of the disease. This clear genotype–phenotype correlation makes DS an interesting method for genotype characterization in future research studies.
Hepatic nodules with imaging criteria corresponding to FNH have often been identified in patients with HHT and systematic DS.11, 12 In the study by Buscarini, 2.9% of affected patients presented such hepatic nodules, according to a relatively stringent definition: presence of suggestive findings in at least two separate imaging modalities, including DS. The observed frequency (13.5%) in this series was much higher, which may reflect the use of a more simple definition (presence of compatible hepatic nodules at DS only), and the fact that a high proportion of the patients in our series (54% versus 33% in the study by Buscarini) had obvious liver involvement in HHT at DS. However, 11 patients had a magnetic resonance imaging scan performed to confirm these nodules: this examination confirmed the presence of nodules in all cases, and showed criteria compatible with FNH in all but one case. Interestingly, almost all (13/14) patients with suspected FNH at DS were in groups 2 (four cases) or 3 (nine cases; 64%), and there was a close correlation between FNH and a higher cardiac index. A possible explanation for this is based on the accepted physiology of sporadic FHN, said to develop from the preexisting liver vascular malformations responsible for hypervascularization of a liver segment.24
The authors thank Pr D. Valla, Pr C. Vilgrain, and Dr. E. Buscarini for their careful reading of, and comments on, this paper.