Virtual Hepatic Venous Pressure Gradient with CT ... · hepatic vein (approximately 1–3 cm from the inferior vena cava). Then, as the balloon was inflated for total occlusion of

ORIGINAL RESEARCH • GASTROINTESTINAL IMAGING

Portal hypertension is a common complication of chronic liver disease and is associated with most clin-

ical consequences of cirrhosis, such as variceal hemor-rhage, ascites, and hepatic encephalopathy (1–3). The most reliable method for assessing portal hypertension is the measurement of the hepatic venous pressure gra-dient (HVPG), by which clinically significant portal

hypertension (CSPH) is defined as an HVPG of at least 10 mm Hg. Variceal hemorrhage may occur when the HVPG is at least 12 mm Hg (4,5). CSPH can become symptomatic, and the development of varices and hy-perdynamic circulation puts patients at high risk of de-compensation (6,7). Thus, there is an increasing need to predict and select patients with CSPH who are at risk

Virtual Hepatic Venous Pressure Gradient with CT Angiography (CHESS 1601): A Prospective Multicenter Study for the Noninvasive Diagnosis of Portal HypertensionXiaolong Qi, MD* • Weimin An, MD* • Fuquan Liu, MD • Ruizhao Qi, MD • Lei Wang, MD • Yanna Liu, MD • Chuan Liu, MD • Yi Xiang, MD • Jialiang Hui, MD • Zhao Liu, MD • Xingshun Qi, MD • Changchun Liu, MD • Baogang Peng, MD • Huiguo Ding, MD • Yongping Yang, MD • Xiaoshun He, MD • Jinlin Hou, MD • Jie Tian, PhD • Zhiwei Li, MD

From the Chinese Portal Hypertension Diagnosis and Monitoring Study Group (CHESS), CHESS Frontier Center, Lanzhou University, Lanzhou, China (Xiaolong Qi); Hepatic Hemodynamic Laboratory, Institute of Hepatology, Nanfang Hospital, Southern Medical University, Guangzhou, China (Xiaolong Qi, Y.L., Chuan Liu, Y.X., J. Hui, Z. Liu, J. Hou); Department of Radiology (W.A., Changchun Liu), Department of General Surgery (R.Q., Z. Li), and Center for Therapeutic Research of Hepatocarcinoma (Y.Y.), 302 Hospital of PLA, Beijing, China; Department of Interventional Therapy, Beijing Shijitan Hospital, Capital Medical University, Beijing, China (F.L., L.W.); Department of Gastroenterology, General Hospital of Shenyang Military Area, Shenyang, China (Xingshun Qi); Department of Hepatobiliary Surgery (B.P.) and Organ Transplant Center (X.H.), The First Affiliated Hospital of Sun Yat-sen University, Guangzhou, China; Department of Gastroenterology and Hepatology, Beijing You’an Hospital, Capital Medical University, Beijing, China (H.D.); Key Laboratory of Molecular Imaging of Chinese Academy of Sciences, Institute of Automa-tion, Chinese Academy of Sciences, Beijing, China (J.T.); and Department of Hepatobiliary Surgery, The Third People’s Hospital of Shenzhen, Shenzhen, China (Z. Li). Received February 27, 2018; revision requested March 20; final revision received September 12; accepted September 19. Address correspondence to Xiaolong Qi (e-mail: [email protected]).

Supported by Special Funds for the Cultivation of Guangdong College Students’ Scientific and Technological Innovation (pdjha0096), the Dean’s Fund of Nanfang Hospital Southern Medical University (2017Z012), Beijing Municipal Administration of Hospitals Ascent Plan (DFL20151602), Application Research on Clinical Char-acteristics of the Capital (z161100000516197), Guangzhou Industry-Academia-Research Collaborative Innovation Major Project (201704020015), Beijing Municipal Administration of Hospitals Clinical Medicine Development of Special Funding (ZYLX201610), Program for Development and Cultivation of Science and Technology of Southern Medical University (KJ20161125), National Natural Science Foundation of China (81600510, 81672725), Guangdong Science Fund for Distinguished Young Scholars (2018B030306019), and Capital’s Funds for Health Improvement and Research (2018-1-2081).

* Xiaolong Qi and W.A. contributed equally to this work.

Conflicts of interest are listed at the end of this article.

See also the editorial by Malayeri in this issue.

Radiology 2019; 00:1–8 • https://doi.org/10.1148/radiol.2018180425 • Content code:

Purpose: To develop and validate a computational model for estimating hepatic venous pressure gradient (HVPG) based on CT angiographic images, termed virtual HVPG, to enable the noninvasive diagnosis of portal hypertension in patients with cirrhosis.

Materials and Methods: In this prospective multicenter diagnostic trial (ClinicalTrials.gov identifier: NCT02842697), 102 consecu-tive eligible participants (mean age, 47 years [range, 21–75 years]; 68 men with a mean age of 44 years [range, 21–73 years] and 34 women with a mean age of 52 years [range, 24–75 years]) were recruited from three high-volume liver centers between August 2016 and April 2017. All participants with cirrhosis of various causes underwent transjugular HVPG measurement, Doppler US, and CT angiography. Virtual HVPG was developed with a three-dimensional reconstructed model and computational fluid dynamics.

Results: In the training cohort (n = 29), the area under the receiver operating characteristic curve (AUC) of virtual HVPG in the prediction of clinically significant portal hypertension (CSPH) was 0.83 (95% confidence interval [CI]: 0.58, 1.00). The diagnostic performance was prospectively confirmed in the validation cohort (n = 73), with an AUC of 0.89 (95% CI: 0.81, 0.96). Inter- and intraobserver agreement was 0.88 and 0.96, respectively, suggesting the good reproducibility of virtual HVPG measurements. There was good correlation between virtual HVPG and invasive HVPG (R = 0.61, P , .001), with a satisfactory performance to rule out (7.3 mm Hg) and rule in (13.0 mm Hg) CSPH.

Conclusion: The accuracy of a computational model of virtual hepatic venous pressure gradient (HVPG) shows significant correla-tion with invasive HVPG. The virtual HVPG also showed a good performance in the noninvasive diagnosis of clinically significant portal hypertension in cirrhosis.

© RSNA, 2018

Online supplemental material is available for this article.

This copy is for personal use only. To order printed copies, contact [email protected]

Virtual Hepatic Venous Pressure Gradient for Portal Hypertension

2 radiology.rsna.org n Radiology: Volume 00: Number 0— 2018

for variceal hemorrhage. However, the reference standard for the measurement of HVPG for diagnosing CSPH (transjug-ular HVPG) is invasive and impractical for routine clinical practice (4,8). Therefore, noninvasive tests that correlate well with invasive HVPG are urgently needed.

Liver stiffness measurement by means of transient elastography (FibroScan; Echosens, Paris, France) is considered a promising noninvasive tool for predicting portal hypertension and esopha-geal varices; however, concerns exist regarding the unreliable measurement in patients with obesity, intrahepatic inflammatory activity, or severe ascites (2,4). Recent advances in imaging-based three-dimensional modeling combined with computational fluid dynamics analysis have permitted noninvasive calculation of blood flow pressure (9,10). Relevant techniques have been successfully applied in the coronary artery for the diagnosis of ischemia (9). In this study, we aimed to develop a computational model based on CT angiographic images, which we termed vir-tual HVPG, and validated that virtual HVPG could allow the accurate estimation of HVPG in patients with cirrhosis. Herein, we evaluated the correlation between virtual HVPG and transjugular HVPG and the accuracy of virtual HVPG in the diag-nosis of CSPH.

Materials and MethodsThe study protocol conformed to the ethical guidelines of the Declaration of Helsinki and was approved by the institutional review board at each site. All participants provided written in-formed consent. All authors had access to the study data and reviewed and approved the final manuscript.

Study Design and ParticipantsThe study (CHESS-1601 trial) is a prospective multicenter diagnostic trial in three high-volume liver centers in China (ClinicalTrials.gov identifier: NCT02842697). The training cohort enrolled consecutive participants at Beijing Shijitan Hospital (Beijing, China) between August 2016 and November

2016, and the developed virtual HVPG method was further validated in a cohort recruited from two other centers (The 302 Hospital of PLA, Beijing, China; Nanfang Hospital, Guangzhou, China) between December 2016 and April 2017. Inclusion criteria were as follows: (a) participants in whom cir-rhosis was diagnosed by means of liver biopsy; (b) participants who underwent transjugular HVPG measurement, abdominal CT angiography, and Doppler US; (c) adult participants; and (d) participants who provided written informed consent. Exclusion criteria were as follows: (a) participants who had previously undergone creation of a transjugular intrahepatic portosystemic shunt, splenectomy, partial splenic emboliza-tion, balloon-occluded retrograde transvenous obliteration, endoscopic therapies, and liver transplantation; (b) partici-pants with hepatocellular carcinoma; and (c) pregnant partici-pants. Laboratory assessments were conducted the day before HVPG measurement, whereas CT angiography and Doppler US were performed within 14 days before catheterization.

Transjugular HVPG MeasurementTransjugular HVPG measurement (reference standard) was performed by F.L., L.W., and R.Q., with 20, 10, and 8 years of experience in interventional radiology, respectively, according to the standard protocol (11). The recording and interpreta-tion of the pressure was supervised by X.Q., Z. Li, and J. Hou (hepatologists with 10, 25, and 30 years of experience, re-spectively). Pressure measurements were conducted by using a balloon catheter (Edwards Lifesciences, Irvine, Calif) with a pres-sure transducer at the tip. A zero measurement with transducer open to air was needed before the transjugular catheterization. The free hepatic venous pressure was measured in the right hepatic vein (approximately 1–3 cm from the inferior vena cava). Then, as the balloon was inflated for total occlusion of the right hepatic vein, the wedged hepatic venous pressure was measured. Continuous recording was necessary until the pressure reached a plateau. HVPG is calculated by subtract-ing the free venous hepatic pressure from the wedged hepatic pressure. All measurements were performed in triplicate and then averaged.

Doppler USDoppler US was performed by two technicians (including W.A.) with more than 10 years of experience in US. Partici-pants fasted for 8 hours before undergoing US, and all mea-surements were conducted with the participants lying supine and breathing normally by using a 3.5-MHz transducer (iU22 Ultrasound System; Philips Healthcare, Reedsville, Pa). The di-ameters of the portal vein were measured by using B-mode US. Doppler examination was used to measure portal vein veloci-ties. All measurements were performed in triplicate and then averaged.

CT Image AcquisitionCT angiography was performed in all participants by Chang-chun Liu (with 12 years of experience in abdominal imaging) and a radiologist with 10 years of experience in abdominal

AbbreviationsAST = aspartate aminotransferase, AUC = area under the receiver operating characteristic curve, CI = confidence interval, CSPH = clini-cally significant portal hypertension, HVPG = hepatic venous pressure gradient

SummaryCalculated from CT angiographic images, the noninvasive virtual he-patic venous pressure gradient correlates well with an invasive hepatic venous pressure gradient and shows good performance in diagnosing clinically significant portal hypertension in patients with cirrhosis.

Implications for Patient Care n The virtual hepatic venous pressure gradient is promising to serve

as a noninvasive surrogate measurement of hepatic venous pressure gradient.

n The virtual hepatic venous pressure gradient may facilitate decision making for individualized diagnosis and monitoring in patients with cirrhosis.

Qi et al

Radiology: Volume 00: Number 0— 2018 n radiology.rsna.org 3

was about 2.5 hours, with 1.5 hours for the human processing and 1.0 hour for machine computation. The virtual HVPG estimation was performed by Y.X., J. Hui, Chuan Liu, and Z. Liu, all with 3 years of experience in hemodynamic analysis. All technicians were blinded to the participants’ baseline char-acteristics, laboratory examination results, and transjugular HVPG values.

Noninvasive ParametersThree imaging-based indexes—liver stiffness measured with transient elastography (FibroScan) (12), CT-based portal pres-sure score (13), and portal diameter measured with US—were assessed. Liver stiffness and portal diameter measurements were conducted within 14 days before HVPG measurement. CT-based portal pressure score was calculated as follows: 17.37 – 4.91 · ln (liver-to-spleen volume ratio) + 3.8 (if perihepatic ascites is present) (13). In addition, serum biomedical indexes, including aspartate aminotransferase (AST), alanine aminotransferase, and platelet count, were measured with a conventional automated analyzer at the day of HVPG measurement. The AST-to–ala-

imaging with use of one of the following systems: Discovery CT750 HD (GE Healthcare, Milwaukee, Wis) or Brilliance iCT (Philips Healthcare, Best, the Netherlands). The follow-ing parameters were used for contrast material–enhanced abdominal CT: tube voltage, 120 kVp or 100 kVp; tube cur-rent, 150–600 mA; section thickness, 1.25 mm; and pitch, 1.375. All participants received intravenous administration of a nonionic contrast agent (io-dine concentration, 370 mg/mL; volume, 1.5–2.0 mL per kilogram body weight; contrast agent type, iopromide injection [Bayer Pharma, Berlin, Ger-many]) at a rate of 3–5 mL/sec. The arterial and portal venous phases were scanned at 21 and 41 seconds, respectively. The following interval between he-patic venous and portal venous phases was 10–30 seconds, depending on the image qual-ity of the arterial and portal venous phases. For participants with poor liver enhancement in arterial or portal venous phases, the hepatic venous phase scan was started at 71 seconds after the start of the contrast agent injection. A volume of 20 mL saline was injected after the in-jection of the contrast agent.

CT Image Analysis and Virtual HVPG EstimationFor CT image analysis and virtual HVPG estimation, original images from CT angiography (Digital Imaging and Commu-nications in Medicine format) were split into thin layers (1.25 mm) and imported into MIMICS 10.0 and 17.0 (Materialise, Leuven, Belgium). Then, a three-dimensional model of the hepatic-portal venous system could be reconstructed from two-dimensional composite images and surrounding tissue on the basis of CT angiographic images, which were then meshed into internal tetrahedra in the computational fluid dynamics solver ANSYS13.0 (Ansys, Canonsburg, Pa). Portal vein velocity was measured with Doppler US and set as a boundary condition. Afterward, finite element analysis and computational fluid dy-namics analysis were applied to compute the pressure distribu-tion in a three-dimensional model. The right hepatic vein was virtually blocked to simulate wedged hepatic venous pressure measurement with balloon occlusion (see Appendix E1 [online] for details). The average time for a virtual HVPG calculation

Figure 1: Flowchart of study enrollment. vHVPG = virtual hepatic venous pressure gradient, 3D = three-dimensional.



maximal sum of sensitivity and specificity when sensitivity is greater than 90% regarding different aims of the test, that is, to diagnose, rule in, and rule out CSPH. The Jackknife cross-validation test was used to assess the accuracy of virtual HVPG. During the process of the test, each case was singled out in turn as a test sample and the remaining cases were used as training samples. The intraclass correlation coefficient was used to ana-lyze the inter- and intraobserver agreement. Spearman correla-tion coefficient analysis (R value) and the Bland-Altman plot were used to assess the correlation and the agreement, respec-tively, between virtual and transjugular HVPG. Two-sided P , .05 was considered indicative of a statistically significant differ-ence. The analyses were performed by using R language 3.0.2 (R Core Team, 2013) and SPSS 20.0 (IBM, Armonk, NY).

Results

Participant CharacteristicsIn the training cohort, a total of 36 consecutive participants were screened and 34 eligible patients recruited at Beijing Shijitan Hospital (Beijing, China). One patient was excluded because of nonevaluable CT images, and four patients were excluded during three-dimensional remodeling. Of the 34 eligible patients, 29 (85%; 18 men, 11 women; mean age, 51 years; age range, 21–75 years) were included in the fi-nal analysis (Fig 1). In addition, 92 consecutive participants

nine aminotransferase ratio was calculated as follows: AST/alanine aminotransferase. The AST-to–platelet count ratio index was calculated as follows: AST (upper limit of normal)/platelet count (3109/L) 3 100. The fibrosis index based on four fac-tors was calculated as follows: [age 3 AST (U/L)]/[platelet count (3109/L)3 alanine aminotransferase (U/L)1/2] (14,15).

Assessment of Inter- and Intraobserver AgreementThe interobserver agreement was analyzed in 16 of 29 (55%) randomly chosen participants by using a computer in a blinded fashion. Then, four technicians (Y.X., J. Hui, Chuan Liu, Z. Liu) conducted the virtual HVPG estimation. To study intraobserver agreement, one of the technicians (J. Hui) re-peated the interpretation process twice in the same participants included in the intraobserver agreement analysis with a 1-month interval between the two readings to reduce the recall bias.

Statistical AnalysisCategorical data are expressed as numbers and percentages, and continuous variables are expressed as means 6 standard deviations. The diagnostic performance of virtual HVPG was assessed by using receiver operating characteristic curves and the areas under the receiver operating characteristic curves (AUCs). The cutoff values were defined as the maximal sum of sensitivity and specificity, the maximal sum of sensitivity and specificity when specificity is greater than 90%, and the

Table 1: Baseline Characteristics of Enrolled Participants

Variable All Participants (n = 102) Training Cohort (n = 29) Validation Cohort (n = 73)Mean age (y)* 47 (21–75) 51 (21–75) 45 (22–73) M 44 (21–73) 47 (21–72) 43 (22–73) F 52 (24–75) 58 (41–75) 49 (24–69)No. of men 68 (66.7) 18 (62) 50 (68)BMI (kg/m2)† 22.7 6 3.1 22.5 6 3.7 23.1 6 3.5HVPG (mm Hg)† 17.1 6 8.8 21.6 6 12.4 15.3 6 6.2Cause of cirrhosis Hepatitis B virus 66 (64.7) 15 (52) 51 (70) Alcohol 8 (7.8) 4 (14) 4 (5.5) Hepatitis C virus 5 (4.9) 5 (17) 0 Hepatotoxic drugs 5 (4.9) 2 (6.9) 3 (4.1) Other 18 (17.6) 3 (10) 15 (20)Child-Pugh score Class A 61 (59.8) 1 (3.4) 60 (82) Class B 39 (38.2) 26 (90) 13 (18) Class C 2 (2.0) 2 (6.9) 0Laboratory values†

AST level (IU/L) 38 6 21 39 6 21 37 6 21 ALT level (IU/L) 28 6 19 27 6 16 28 6 20 Albumin level (g/L) 36 6 5 37 6 6 35 6 7 Total bilirubin level (mmol/L) 23.1 6 12.9 27 6 0.8 21.5 6 12.1 INR 1.22 6 0.21 1.27 6 0.23 1.21 6 0.21 Platelet count (109/L) 78 6 92 75 6 60 79 6 101

Note.—Except where indicated, data are numbers of participants, with percentages in parentheses. ALT = alanine aminotransferase, AST = aspartate aminotransferase, BMI = body mass index, HVPG = hepatic venous pressure gradient, INR = international normalized ratio.* Numbers in parentheses are the range.† Data are means 6 standard deviations.

Qi et al


were screened and 89 eligible participants were enrolled from two additional centers (The 302 Hospital of PLA, Beijing, China; Nanfang Hospital, Guangzhou, China) with use of the same inclusion and exclu-sion criteria. Six participants were excluded because of nonevaluable CT images and 10 were excluded during three-dimensional re-modeling. Of the 89 eligible participants, 73 (82%) were included in the validation cohort (Fig 1). There were 50 men and 23 women with a mean age of 45 years (range, 22–73 years). The median dura-tion between HVPG measurement and CT angiography plus Doppler US was 8 days (range, 3–14 days). No adverse events were identified during these procedures. Baseline demographic characteristics of the study population are summarized in Table 1. The most common cause of cirrhosis was hepa-titis B virus infection, which was found in 15 of the 29 participants (52%) in the train-ing cohort and in 51 of the 73 participants (70%) in the validation cohort.

Virtual HVPG Interpretation and Diagnostic PerformanceVirtual HVPG was successfully calculated on the basis of the three-dimensional re-constructed model and computational fluid dynamics analysis. A representative example of virtual HVPG interpretation is shown in Figure 2. In the training co-hort, the AUC of virtual HVPG for the noninvasive prediction of CSPH (HVPG 10 mm Hg) was 0.83 (95% confidence interval [CI]: 0.58, 1.00). The perfor-mance of virtual HVPG was prospectively confirmed in the validation cohort, with an AUC of 0.89 (95% CI: 0.81, 0.96). In the diagnosis of CSPH, a virtual HVPG of more than 11.03 mm Hg resulted in a sensitivity of 74% (95% CI: 62%, 89%) and specificity of 93% (95% CI: 80%, 100%) (Fig 3, A).

Correlation between Virtual and Transjugular HVPGWe found a statistically significant cor-relation between virtual and transjugular HVPG in the overall participants (n = 102, R = 0.605, P , .001) (Fig 4, A). A simi-lar result was achieved with the Bland-Altman plot (Fig 4, B). The accuracy of virtual HVPG with the Jackknife cross-validation test was 75.5%. By using two cutoff values, one with a sensitivity of

Figure 2: Interpretation of virtual hepatic venous pressure gradient (HVPG). Arrows show measurement location of free hepatic venous pressure (FHVP) or wedged hepatic venous pressure (WHVP). (a) Examples of virtual and invasive HVPG interpretation in study partici-pant with clinically significant portal hypertension (CSPH). Virtual HVPG was 22.7 mm Hg, and invasive HVPG was 23.2 mm Hg. (b) Examples of virtual and invasive HVPG interpre-tation in study participant without CSPH. Virtual HVPG was 7.4 mm Hg and invasive HVPG was 6.2 mm Hg.



Figure 3: Diagnostic performance of virtual hepatic venous pressure gradient (HVPG) for clinically significant portal hyper-tension (CSPH). Numbers in parentheses are 95% confidence intervals. A, Receiver operating characteristic curves of virtual HVPG for predicting CSPH in training and validation cohorts (n = 29 and n = 73, respectively). AUC = area under receiver operating characteristic curve. B, Receiver operating characteristic curves of virtual HVPG (vHVPG) and other noninvasive mod-els for predicting CSPH in overall participant group (n = 102). AAR = aspartate aminotransferase–to–alanine aminotransferase ratio, APRI = aspartate aminotransferase–to–platelet count ratio index, AUC = area under receiver operating characteristic curve, FIB-4 = fibrosis index based on four factors, HVPGCT score = CT-based portal pressure score. Transient elastography was performed with FibroScan.

Figure 4: Correlation between virtual hepatic venous pressure gradient (HVPG) and invasive HVPG in overall participant group. vHVPG = virtual HVPG. = par-ticipants with clinically significant portal hypertension (HVPG 10 mm Hg), = participants without clinically significant portal hypertension (HVPG ,10 mm Hg). A, Scatterplot shows agreement between virtual HVPG and invasive HVPG (n = 102). B, Bland-Altman plot shows difference between virtual HVPG and invasive HVPG (n = 102). C, Distribution of virtual HVPG values (n = 102).

more than 90% (sensitivity: 90%, specificity: 58%) to rule out CSPH (virtual HVPG = 7.3 mm Hg) and one with a specificity of more than 90% (sensitivity: 64%, specificity: 95%) to rule in CSPH (virtual HVPG = 13.0 mm Hg), 75 of 102 participants (74%) showed definite results cat-egories, with 56 of the 102 participants (55%) classified as

having CSPH and 19 as having no CSPH (19%). Among the 75 participants with definite results, results for 65 of the 102 participants (64%) were well classified and those for 10 (10%) were misclassified. Conversely, 27 of the 102 partici-pants (26%) had an indeterminate result that necessitated further evaluation (Fig 4, C).

Qi et al


Comparison between Virtual HVPG and Other Reported Noninvasive ParametersWe further compared the diagnostic performance of vir-tual HVPG with that of other noninvasive parameters in all cohorts (n = 102). Virtual HVPG exhibited the high-est diagnostic performance for predicting CSPH, with an AUC of 0.88 (95% CI: 0.80, 0.95). In addition, the AUCs for other noninvasive tests, including transient elastography (n = 30), CT-based portal pressure score (n = 85), AST-to–alanine aminotransferase ratio (n = 100), portal diameter (n = 100), fibrosis index based on four factors (n = 99), and AST-to–platelet count index (n = 100) were 0.71 (95% CI: 0.48, 0.95), 0.60 (95% CI: 0.46, 0.74), 0.58 (95% CI: 0.43, 0.72), 0.58 (95% CI: 0.45, 0.70), 0.57 (95% CI: 0.43, 0.72), and 0.51 (95% CI: 0.36, 0.66), respectively (Fig 3, B). The performance of all noninvasive tests in the diagnosis of CSPH is summarized in Table 2.

Reproducibility of Virtual HPVGIntraclass correlation coefficients for inter- and intraobserver agreement assessment were 0.877 and 0.959, respectively, sug-gesting the robust reproducibility of virtual HVPG.

DiscussionIn this prospective multicenter study, we developed a CT an-giography–based virtual HVPG from a three-dimensional hepatic portal vein model and computational fluid dynamics analysis and further validated its performance in the noninva-sive diagnosis of portal hypertension in patients with cirrho-sis. As expected, virtual HVPG showed good performance for CSPH detection in both training and validation cohorts, with AUCs of 0.83 and 0.89, respectively, and had significant agree-ment with invasive HVPG.

Because transient elastography (FibroScan) was recom-mended by the Baveno VI Consensus Workshop to identify CSPH (4,16), we further evaluated the performance of tran-sient elastography in this study. According to our results, the performance of liver stiffness measured with transient elastography was just fair in our study (AUC for CSPH,

0.71 [95% CI: 0.48, 0.95]), likely because of the limited sample size and wider-ranging causes of cirrhosis compared with previous studies (16). Due to the increasing relevance of extrahepatic factors in portal hypertension progression, HVPG cannot be reliably estimated by means of liver stiff-ness in severe portal hypertension (6,7,17). Therefore, we suggest that virtual HVPG could serve as an auxiliary pa-rameter for liver stiffness, especially in patients with obesity, liver necrotic inflammation, or severe ascites.

Another CT-based model, namely the CT-based portal pressure score, was proposed by Iranmanesh and colleagues (13,18) and showed promising results for diagnosing CSPH in patients with hepatocellular carcinoma, with an AUC of 0.911 (95% CI: 0.847, 0.975). However, this model showed insufficient performance in hepatitis B virus–related cirrho-sis, with an AUC of 0.57 (95% CI: 42%, 79%) (17). Thus, we must be aware that the cause of cirrhosis may have an influence on the performance of a diagnosis method.

Virtual HVPG is a noninvasive approach that combines individual anatomic characteristics and hemodynamic changes of hepatic-portal veins, which allows direct simula-tion of invasive HVPG that imaging-based models and se-rum markers fail to do (12–15). In addition, virtual HVPG shows a robust reproducibility and is available for repeated applications in routine clinical practice. The reported rule-in and rule-out cutoff values for identifying CSPH could help make decisions on whether to pursue further treatment easier and quicker. More important, it could help identify patients who could safely avoid invasive HVPG procedures. Virtual HVPG may also be helpful in evaluating drug ther-apeutic efficacy and intervention outcome, which requires the repeated monitoring of HVPG changes.

Our study had some limitations. First, virtual HVPG in-terpretation is relatively time consuming (about 2.5 hours per case). Second, the CT angiography procedure for virtual HVPG should be further personalized. Finally, the number of participants without CSPH was low in our study, and virtual HVPG should be further validated in rigorously de-signed studies with more patients with mild portal hyper-tension and well-defined causes.

Table 2: Performance of Virtual HVPG and Other Noninvasive Models in the Diagnosis of Clinically Significant Portal Hypertension

ParameterVirtual HVPG (mm Hg)

Transient Elastography (kPa)

CT-based Portal Pressure Score* AAR

Portal Diameter (mm) FIB-4 APRI

AUC 88 (80, 95) 71 (48, 95) 60 (46, 74) 58 (43, 72) 58 (45, 70) 57 (43, 72) 51 (36, 66)Cutoff 11.03 13.45 19.11 1.13 14.05 4.50 0.85Sensitivity 76 (64, 93) 62 (45, 79) 43 (31, 54) 29 (21, 40) 63 (52, 73) 42 (31, 53) 28 (19, 38)Specificity 90 (74, 100) 80 (40, 100) 82 (65, 100) 94 (83, 100) 56 (31, 81) 83 (67, 100) 83 (67, 100)PPV 97 (93, 100) 95 (85, 100) 91 (82, 100) 96 (87, 100) 88 (83, 95) 92 (83, 100) 89 (76, 100)NPV 46 (37, 58) 27 (14, 42) 26 (21, 33) 23 (20, 26) 23 (14, 32) 24 (19, 30) 21 (16, 25)

Note.—Except for cutoff, data are given as percentages. Numbers in parentheses are 95% confidence intervals. Clinically significant portal hypertension is defined as hepatic venous pressure gradient (HVPG) of at least 10 mm Hg. AAR = aspartate aminotransferase–to–alanine aminotransferase ratio, APRI = aspartate aminotransferase–to–platelet count ratio index, AUC = area under the receiver operating charac-teristic curve, FIB-4 = fibrosis index based on four factors, NPV = negative predictive value, PPV = positive predictive value. * CT-based portal pressure was calculated as follows: 17.37 – 4.91 · ln (liver-to-spleen volume ratio) + 3.8 (if perihepatic ascites is present).



In conclusion, we developed and prospectively validated the accuracy of a computational model of virtual HVPG that showed significant correlation with invasive HVPG. Virtual HVPG also showed good performance in the noninvasive diagnosis of CSPH in cirrhosis. It has the potential of dramatically increasing the availability of this important information if routinely calculated in patients with compensated chronic liver disease.

Acknowledgments: We thank Guadalupe Garcia-Tsao, MD, for constructive comments and critical revision of the manuscript. We thank Chunqing Zhang, MD, Zaiyi Liu, MD, Yongxiang Yi, MD, Guoxin Li, MD, Jinjun Chen, MD, and Yajin Chen, MD, for general administrative and technical support and Jiahong Dong, MD, Di Dong, PhD, Jingwei Wei, PhD, Li Liu, MD, and Jinghui Dong, MD, for the general supervision of a research group.

Author contributions: Guarantors of integrity of entire study, Xiaolong Qi, W.A., F.L., R.Q., L.W., Z. Liu, C.L., H.D., Y.Y., J.H., J.T., Z. Li; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manu-script drafting or manuscript revision for important intellectual content, all authors; approval of final version of submitted manuscript, all authors; agrees to ensure any questions related to the work are appropriately resolved, all authors; literature re-search, Xiaolong Qi, W.A., F.L., R.Q., Y.X., J.H., Z. Liu, Xingshun Qi, C.L., H.D., Y.Y., X.H., Z. Li; clinical studies, Xiaolong Qi, W.A., F.L., R.Q., L.W., Y.X., J.H., Z. Liu, Xingshun Qi, C.L., B.P., H.D., Y.Y., X.H., J.H., Z. Li; experimental stud-ies, Xiaolong Qi, W.A., F.L., R.Q., Y.X., J.H., Z. Liu, C.L., H.D., Y.Y., X.H., Z. Li; statistical analysis, Xiaolong Qi, W.A., F.L., R.Q., Y.L., C.L., Y.X., Z. Liu, C.L., B.P., H.D., Y.Y., X.H., J.T.; and manuscript editing, Xiaolong Qi, W.A., F.L., R.Q., Y.L., Y.X., Z. Liu, Xingshun Qi, C.L., H.D., Y.Y., X.H., J.T.

Disclosures of Conflicts of Interest: Xiaolong Qi disclosed no relevant relationships. W.A. disclosed no relevant relationships. F.L. disclosed no relevant relationships. R.Q. disclosed no relevant relationships. L.W. disclosed no relevant relationships. Y.L. disclosed no relevant relationships. Chuan Liu disclosed no rel-evant relationships. Y.X. disclosed no relevant relationships. Jialiang Hui disclosed no relevant relationships. Zhao Liu disclosed no relevant relationships. Xingshun Qi disclosed no relevant relationships. Changchun Liu disclosed no relevant rela-tionships. B.P. disclosed no relevant relationships. H.D. disclosed no relevant rela-tionships. Y.Y. disclosed no relevant relationships. X.H. disclosed no relevant rela-tionships. Jinlin Hou disclosed no relevant relationships. J.T. disclosed no relevant relationships. Zhiwei Li disclosed no relevant relationships.

References 1. Tsochatzis EA, Bosch J, Burroughs AK. Liver cirrhosis. Lancet 2014;383(9930):

1749–1761. 2. Garcia-Tsao G, Bosch J. Management of varices and variceal hemorrhage in cirrho-

sis. N Engl J Med 2010;362(9):823–832 [Published correction appears in N Engl J Med 2011;364(5):490. Dosage error in article text.].

3. Qi X, Berzigotti A, Cardenas A, Sarin SK. Emerging non-invasive approaches for diagnosis and monitoring of portal hypertension. Lancet Gastroenterol Hepatol 2018;3(10):708–719.

4. de Franchis R; Baveno VI Faculty. Expanding consensus in portal hypertension: re-port of the Baveno VI Consensus Workshop—stratifying risk and individualizing care for portal hypertension. J Hepatol 2015;63(3):743–752.

5. Garcia-Tsao G, Groszmann RJ, Fisher RL, Conn HO, Atterbury CE, Glickman M. Portal pressure, presence of gastroesophageal varices and variceal bleeding. Hepatol-ogy 1985;5(3):419–424.

6. Berzigotti A. Non-invasive evaluation of portal hypertension using ultrasound elas-tography. J Hepatol 2017;67(2):399–411.

7. Shung DL, Garcia-Tsao G. Liver capsule: portal hypertension and varices— pathogenesis, stages, and management. Hepatology 2017;65(3):1038.

8. Dong J, Qi X. Liver imaging in precision medicine. EBioMedicine 2018;32:1–2. 9. Min JK, Leipsic J, Pencina MJ, et al. Diagnostic accuracy of fractional flow reserve

from anatomic CT angiography. JAMA 2012;308(12):1237–1245. 10. Lu MT, Ferencik M, Roberts RS, et al. Noninvasive FFR derived from coronary CT

angiography: management and outcomes in the PROMISE trial. JACC Cardiovasc Imaging 2017;10(11):1350–1358.

11. Groszmann RJ, Wongcharatrawee S. The hepatic venous pressure gradient: anything worth doing should be done right. Hepatology 2004;39(2):280–282.

12. Vizzutti F, Arena U, Romanelli RG, et al. Liver stiffness measurement predicts se-vere portal hypertension in patients with HCV-related cirrhosis. Hepatology 2007; 45(5):1290–1297.

13. Iranmanesh P, Vazquez O, Terraz S, et al. Accurate computed tomography–based portal pressure assessment in patients with hepatocellular carcinoma. J Hepatol 2014;60(5):969–974.

14. Wong VW, Vergniol J, Wong GL, et al. Diagnosis of fibrosis and cirrhosis using liver stiffness measurement in nonalcoholic fatty liver disease. Hepatology 2010; 51(2):454–462.

15. Vallet-Pichard A, Mallet V, Nalpas B, et al. FIB-4: an inexpensive and accurate marker of fibrosis in HCV infection. comparison with liver biopsy and fibrotest. Hepatology 2007;46(1):32–36.

16. Bureau C, Metivier S, Peron JM, et al. Transient elastography accurately predicts presence of significant portal hypertension in patients with chronic liver disease. Ali-ment Pharmacol Ther 2008;27(12):1261–1268.

17. Qi X, Liu F, Li Z, et al. Insufficient accuracy of computed tomography-based portal pressure assessment in hepatitis B virus–related cirrhosis: an analysis of data from CHESS-1601 trial. J Hepatol 2018;28(1):210–211.

18. Iranmanesh P, Toso C. Reply to: “Insufficient accuracy of computed tomography-based portal pressure assessment in hepatitis B virus-related cirrhosis: an analysis of data from CHESS-1601 trial”. J Hepatol 2018;69(1):211–212.

Documents

Virtual Hepatic Venous Pressure Gradient with CT ... · hepatic vein (approximately 1–3 cm from the inferior vena cava). Then, as the balloon was inflated for total occlusion of