Right atrial pressure is not adequate to calculate portal pressure gradient in cirrhosis: A clinical-hemodynamic correlation study

Authors

  • Vincenzo La Mura,

    1. Hepatic Hemodynamic Laboratory, Liver Unit, Hospital Clínic, Institut d'Investigacions Biomèdiques August Pi i Sunyer, University of Barcelona and Centro de Investigación Biomédica en Red de Enfermedades Hepáticas y Digestivas (CIBERehd), Barcelona, Spain
    Search for more papers by this author
  • Juan G. Abraldes,

    1. Hepatic Hemodynamic Laboratory, Liver Unit, Hospital Clínic, Institut d'Investigacions Biomèdiques August Pi i Sunyer, University of Barcelona and Centro de Investigación Biomédica en Red de Enfermedades Hepáticas y Digestivas (CIBERehd), Barcelona, Spain
    Search for more papers by this author
  • Annalisa Berzigotti,

    1. Hepatic Hemodynamic Laboratory, Liver Unit, Hospital Clínic, Institut d'Investigacions Biomèdiques August Pi i Sunyer, University of Barcelona and Centro de Investigación Biomédica en Red de Enfermedades Hepáticas y Digestivas (CIBERehd), Barcelona, Spain
    Search for more papers by this author
  • Eva Erice,

    1. Hepatic Hemodynamic Laboratory, Liver Unit, Hospital Clínic, Institut d'Investigacions Biomèdiques August Pi i Sunyer, University of Barcelona and Centro de Investigación Biomédica en Red de Enfermedades Hepáticas y Digestivas (CIBERehd), Barcelona, Spain
    Search for more papers by this author
  • Alexandra Flores-Arroyo,

    1. Hepatic Hemodynamic Laboratory, Liver Unit, Hospital Clínic, Institut d'Investigacions Biomèdiques August Pi i Sunyer, University of Barcelona and Centro de Investigación Biomédica en Red de Enfermedades Hepáticas y Digestivas (CIBERehd), Barcelona, Spain
    Search for more papers by this author
  • Juan Carlos García-Pagán,

    1. Hepatic Hemodynamic Laboratory, Liver Unit, Hospital Clínic, Institut d'Investigacions Biomèdiques August Pi i Sunyer, University of Barcelona and Centro de Investigación Biomédica en Red de Enfermedades Hepáticas y Digestivas (CIBERehd), Barcelona, Spain
    Search for more papers by this author
  • Jaime Bosch

    Corresponding author
    1. Hepatic Hemodynamic Laboratory, Liver Unit, Hospital Clínic, Institut d'Investigacions Biomèdiques August Pi i Sunyer, University of Barcelona and Centro de Investigación Biomédica en Red de Enfermedades Hepáticas y Digestivas (CIBERehd), Barcelona, Spain
    • Calle Villaroel 170, Barcelona 08036, Spain
    Search for more papers by this author
    • fax: (34)-932279856


  • Potential conflict of interest: Nothing to report.

  • See Editorial on Page 1894

Abstract

Hepatic venous pressure gradient (HVPG), the difference between wedge and free hepatic venous pressure, is the preferred method for estimating portal pressure. However, it has been suggested that hepatic atrial pressure gradient (HAPG)—the gradient between wedge hepatic venous pressure and right atrial pressure (RAP)—might better reflect variceal hemodynamics. The aim of this study was to (1) investigate whether HAPG with nonselective beta-blockers correlates with prognosis in cirrhotic patients with portal hypertension at baseline and during treatment; (2) compare the prognostic value of HAPG with that of HVPG; and (3) investigate the agreement between portoatrial gradient (PAG) and portocaval gradient (PCG) in patients with transjugular intrahepatic portosystemic shunt (TIPS). We included 154 cirrhotic patients with varices with a complete hemodynamic study at baseline and on chronic treatment for primary (n = 71) or secondary (n = 83) prophylaxis for bleeding and 99 patients with TIPS. All patients were followed for up to 2 years; portal hypertensive-related bleeding and bleeding-free survival were analyzed. HVPG was equal or lower than HAPG in all patients (−3.2 mm Hg; P < 0.001). Agreement between HAPG and HVPG was modest, especially in patients with increased intra-abdominal pressure. One hundred two patients were HVPG nonresponders and 52 patients were HVPG responders to nonselective beta-blockers, whereas 101 patients were HAPG nonresponders and 53 patients were HAPG responders (k = 0.610). HVPG response revealed an excellent predictive value for bleeding risk and bleeding-free survival; HAPG did not. In our TIPS patients, 20% had a PCG ≤12 mm Hg and a PAG >12 mm Hg, which may have induced unnecessary overdilation of the TIPS. Conclusion: The excellent prognostic information provided by HVPG response to drug therapy is lost if HAPG response is considered. RAP should not be used for the calculation of portal pressure gradient in patients with cirrhosis. Hepatology 2010;51:2108–2116

Measurements of portal pressure have wide application in chronic liver disease.1-3 Portal pressure is usually expressed as the portal pressure gradient, because this represents the portal perfusion pressure and it is not influenced by changes in intra-abdominal pressure and it is also not influenced by inadequate positioning of the external zero reference point. Currently, the preferred method for determining portal pressure gradient is measurement of the hepatic venous pressure gradient (HVPG), the difference between wedge and free hepatic venous pressures (wedge hepatic venous pressure [WHVP] and free hepatic venous pressure [FHVP], respectively). HVPG measurement is being used increasingly in clinical hepatology, and numerous studies have demonstrated that it is a robust surrogate marker for hard clinical endpoints.4-9 Its main clinical applications include diagnosis, risk stratification, identification of patient candidates for liver resection for hepatocellular carcinoma, monitoring of the efficacy of medical treatment, and assessment of progression of portal hypertension and chronic liver disease.4-7, 10, 11

The technique to obtain HVPG values is relatively straightforward; however, specific guidelines must be followed to achieve accurate measurements, as emphasized in recent studies.1, 2, 12 The HVPG represents the pressure gradient between the portal vein and the inferior vena cava pressure (IVCP) and, therefore, represents the portal perfusion pressure. The WHVP is measured by occluding the hepatic vein, either by wedging the catheter into a small branch of a hepatic vein or by inflating a balloon at the tip of the catheter in a larger hepatic vein; the latter method is recommended because it decreases the variability of measurements. FHVP measurement is advantageous because it acts as an internal zero, eliminating sources of variability due to incorrect placement of the external zero reference level or changes in abdominal pressure caused by ascites or obesity, as these similarly increase FHVP and WHVP.13 The FHVP is measured by maintaining the tip of the catheter free in the hepatic vein, at 2-4 cm from its opening into the inferior vena cava. The FHVP should be close to the IVCP; if the difference between them is >2 mm Hg, it is likely that the catheter has been placed inadequately or that the hepatic vein has been obstructed. In these cases, the IVCP at the level of the hepatic vein should be used to calculate HVPG.2

In recent years, especially since the introduction of transjugular intrahepatic portosystemic shunt (TIPS), many reports have used right atrial pressure (RAP) as the internal zero reference to calculate the portal pressure gradient (hepatic atrial pressure gradient [HAPG]).14-21 There are several reasons for this trend. Many radiologists find it difficult to measure the IVCP during a TIPS procedure, and the RAP may be a closer estimate of the outflow pressure of the esophageal varices than the FHVP, because the varices drain mainly through the azygos vein into the superior vena cava in the vicinity of the right atrium.22-24 However, no study thus far has assessed the prognostic significance of HAPG either in baseline conditions or after pharmacological therapy, as compared with measurement of the HVPG in patients with cirrhosis and portal hypertension. In this study, our goals were: (1) to determine whether the measurements of HAPG and its changes after drug therapy correlate with prognosis in patients with cirrhosis and portal hypertension; (2) to compare the prognostic value of HAPG with that of HVPG; (3) to determine whether HAPG measurement adds significant information to that of HVPG; and (4) investigate the agreement between portoatrial gradient (PAG) and portocaval gradient (PCG) in patients with TIPS.

Abbreviations

CI, confidence interval; FHVP, free hepatic venous pressure; HAPG, hepatic atrial pressure gradient; HVPG, hepatic venous pressure gradient; ICC, intraclass correlation coefficient; IVCP, inferior vena cava pressure; PAG, portoatrial gradient; PCG, portalcaval gradient; RAP, right atrial pressure; TIPS, transjugular intrahepatic portosystemic shunt; WHVP, wedge hepatic venous pressure.

Patients and Methods

Study Cohort.

We reviewed the data of cirrhotic patients with esophageal varices enrolled in three previous longitudinal studies4, 5, 25 to assess the prognostic value of HVPG changes induced by drug therapy for portal hypertension. Inclusion criteria for the present study were: diagnosis of cirrhosis (based on liver biopsy and/or unequivocal clinical data and compatible findings on imaging techniques); presence of esophageal varices; baseline HVPG values >12 mm Hg; subsequent chronic treatment with nonselective beta-blockers (propranolol or nadolol) after baseline hemodynamic evaluation; second HVPG measurement after 1-6 months of continued pharmacological therapy; and measurement of RAP in both hemodynamic studies. Patients with hepatocellular carcinoma at baseline, portal vein thrombosis, contraindications to beta-blockers, or cholestatic liver disease were excluded.

The study protocol was approved by the Ethics Committee of the Hospital Clínic, Barcelona.

Hemodynamic Measurements.

Baseline hemodynamic studies were performed before starting prophylaxis for variceal bleeding or rebleeding. Patients who recovered from an acute variceal bleeding were studied under stable conditions, between days 5 and 7 after the index bleed. In brief, under local anesthesia, a venous introducer was placed in the right femoral vein or internal jugular vein by the Seldinger technique. Under fluoroscopy, a 7-F balloon-tipped catheter (Boston Scientific Medi-Tech, Natik, MA) was guided into the right atrium to measure the RAP. The same catheter was then guided into the main right hepatic vein for measurements of WHVP and FHVP as described.1, 2 HVPG was calculated as the difference between WHVP and FHVP. No patient had a difference between FHVP and IVCP >2 mm Hg. Hepatic to RAP gradient (HAPG) was calculated as the difference between WHVP and RAP. All measurements were taken in triplicate, and permanent tracings were obtained in each case in a multichannel recorder (Marquette Electronics, Milwaukee, WI) and were reviewed specifically for this study by experienced investigators (J. B., J. G. A., and J. C. G.-P.) unaware of the clinical data of the patients. All measurements were repeated after titration of pharmacological therapy (1-6 months after baseline measurements).

Follow-up.

Patients were followed according to the standardized protocols of our unit. After the hemodynamic evaluation all patients were started on oral propranolol (20 mg twice daily) or nadolol (20 mg once a day), that were stepwise increased, if tolerated, until heart rate had fallen to 50-55/minute, or systolic blood pressure was <90 mm Hg. All patients were followed in the outpatient clinic at 1, 3, and 6 months and every 6 months thereafter. Medical history, physical examination, biochemistry, hematologic tests, abdominal ultrasound, and continued alcohol abuse were recorded every 6 months. Follow-up data were collected for up to 2 years, at death, or at liver transplantation.

The clinical events analyzed were bleeding and bleeding-free survival. Bleeding from portal hypertension sources (esophago-gastric varices or portal hypertensive gastropathy) was defined according to Baveno criteria.26 Patients who discontinued propranolol/nadolol for intolerance or self-made decision were censored at the time the discontinuation was recorded in the clinical chart. Therefore, our analysis is a per-treatment received analysis. Patients undergoing liver transplantation were censored the day of transplantation. For the analysis of survival patients with an incomplete follow-up were traced by contacting them or their relatives or, when this was not successful, by means of a requirement to local civil registry to assess the exact date and cause of death.

TIPS Patients.

We included a consecutive series of 99 cirrhotic patients treated with TIPS from 2000 to 2007. We excluded patients with TIPS placed for the treatment of Budd-Chiari syndrome.

The TIPS procedure was performed as described.27, 28 Twenty-eight patients received bare stents (Wallstent; Boston Scientific Iberica, Madrid, Spain) and 71 polytetrafluoroethylene-covered stents (Gore-Viatorr; W.L.Gore, Tilburg, Netherlands). Baseline measurements of RAP, portal venous pressure, and IVCP were obtained in duplicate before angiography. IVCP was measured through the venous introducer, by placing its tip just cranially to the stent. Permanent tracings of all pressure measurements were obtained. These were read specifically for this study by V. L. M. and J. G. A. PCG was calculated as the difference between portal pressure and IVCP; PAG was calculated as the difference between portal pressure and RAP. The stent was initially dilated to 8 mm. If a PCG of <12 mm Hg was not achieved, it was further dilated to 10 mm.27 In 10 patients, this threshold was not achieved despite maximum dilation.

Statistical Analysis.

Statistical analysis was performed with SPSS version 16.0 (SPSS, Chicago, IL). Data are reported as frequencies or means with standard deviation. The comparisons between groups were performed using analysis of variance followed by preplanned polynomial contrast analysis. Multivariate analysis was conducted with ordinal logistic regression. The mean difference between HVPG and HAPG and its confidence interval was evaluated with a paired t test. The agreement between HVPG and HAPG (or between PCG and PAG) was assessed using the Pearson coefficient, intraclass correlation coefficient (ICC) for absolute agreement (which is influenced by any differences between the observed values), and the ICC for consistency (which is not influenced by the systematic differences).29 The ICC can range between 0.00 (total lack of agreement) and 1.00 (perfect agreement). Values >0.80 indicate good agreement and values >0.90 indicate excellent agreement.30 The discriminative ability of HVPG and HAPG to predict bleeding and survival was analyzed using Kaplan-Meier curves, and comparisons were conducted using a log-rank test.

Results

Baseline Characteristics of the Patients (Non-TIPS Cohort).

The original studies4, 5, 25 included 309 patients. Of these, 154 patients had simultaneous HVPG and RAP measurements (60 from Abraldes et al.,4 56 from Turnes et al.,5 and 38 from La Mura et al.25) before and during continued therapy with nonselective beta-blockers and were included in the final analysis. Clinical, endoscopic, and hemodynamic characteristics are reported in Tables 1 and 2. There were no differences in the baseline characteristics between patients with or without RAP measurements in the original databases (data not shown).

Table 1. Clinical and Biochemical Characteristics of the Study Cohort
VariableStudy Cohorts
Medical Treatment (n = 154)TIPS Treatment (n = 99)
Sex (% male)6367
Age (years)57 ± 952 ± 11
Alcoholic etiology (%)3869
Active alcoholism (%)2137
Ascites (%)3069
Hepatic encephalopathy (%)629
Varices (%)  
 Absent13
 Small3138
 Large6950
Risk signs (% of patients)4039
Previous bleeding (n)8390
Child-Pugh score6.58 ± 1.628.96 ± 1.96
Child-Pugh class A/B/C (n)89/56/910/53/36
Albumin (g/L)34.3 ± 6.729.7 ± 6.8
Bilirubin (mg/dL)1.8 ± 1.93.3 ± 3.1
Prothrombin time (%)67.9 ± 15.454.5 ± 13.6
Hematocrit (%)35.0 ± 0.128.5 ± 3.9
Platelet count (103 U/mm3)90.0 ± 36.672.0 ± 70.0
Glucose (mg/dL)123 ± 43149 ± 50
Aspartate aminotransferase (U/L)92 ± 5873 ± 102
Alanine aminotransferase (U/L)82 ± 6247 ± 75
Creatinin (mg/dL)0.90 ± 0.200.90 ± 0.34
Table 2. Hemodynamic Characteristics at Baseline and During Chronic Nonselective Beta-Blocker Treatment
VariableBaselineAfter Beta-Blocker TreatmentP Value
RAP (mm Hg)4.9 ± 2.85.4 ± 2.90.036
FHVP (mm Hg)8.1 ± 3.98.1 ± 3.20.766
WHVP (mm Hg)26.6 ± 5.124.1 ± 4.3<0.001
HVPG (mm Hg)18.5 ± 3.516.0 ± 3.4<0.001
HAPG (mm Hg)21.7 ± 4.418.7 ± 3.9<0.001

Basal Hemodynamic Evaluation: HVPG Versus HAPG.

RAP values were equal (n = 8) or lower (n = 146) than FHVP values in all cases. As a result, HVPG values were, in all cases, equal or lower than those of HAPG (mean difference 3.2 mm Hg [95% confidence interval (CI) 2.8-3.7 mm Hg]; P < 0.001). The correlation between basal HVPG and basal HAPG was, as expected, statistically significant (Pearson correlation coefficient 0.779; P < 0.001). The ICC for absolute agreement was 0.569 (95% CI −0.034-0.809), whereas the ICC for consistency was 0.760 (95% CI 0.684-0.819); this means that even after excluding systematic differences, agreement was only modest (Fig. 1).

Figure 1.

Concordance analysis between the HVPG and the HAPG at baseline (A) and during chronic treatment with non-selective beta-blockers (B). Solid lines represent the regression lines while dotted lines represent the identity lines. HAPG was, in mean, 3.2 mm Hg higher than HVPG. Even after correcting for this systematic difference agreement between these two measurements was only fair.

In order to explore the clinical factors associated with the differences between the HVPG and the HAPG, we divided the patients in tertiles according to the difference between FHVP and RAP (Table 3). Sixty-four patients had an FHVP-RAP difference ≤2 mm Hg, 57 between 2-4 mm Hg, and 33 >4 mm Hg. A positive association was shown between the increment of the FHVP-RAP difference and Child-Pugh score, glucose levels, alcoholic etiology, history of variceal bleeding and FHVP, and WHVP and HAPG at baseline and after drug therapy. A negative association was found with albumin levels. Multivariable analysis identified baseline FHVP and glucose levels as independently associated with the FHVP-RAP.

Table 3. Factors Associated with Differences Between HVPG and HAPG
VariableFree RAP GradientP Linear
≤2 mm Hg2-4 mm Hg>4 mm Hg
Age (years)57 ± 857 ± 1055 ± 100.381
Weight (kg)72 ± 2169 ± 1376 ± 140.498
Child-Pugh score6.3 ± 1.46.6 ± 1.97.1 ± 1.50.017
Model for end-stage liver disease score9.3 ± 3.59.4 ± 3.510.2 ± 3.60.393
Bilirubin (mg/dL)1.96 ± 2.571.60 ± 1.131.85 ± 1.400.640
Prothrombin time (% of activity)69 ± 1369 ± 1562 ± 190.065
Albumin (g/dL)35.6 ± 5.634.1 ± 7.732.3 ± 6.20.019
Creatinin (mg/dL)0.9 ± 0.90.9 ± 0.20.9 ± 0.20.932
Glucose (mg/dL)111 ± 26126 ± 41138 ± 460.002
Glutamate oxaloacetate aminotransferase (U/L)100 ± 5789 ± 6183 ± 530.321
Glutamate pyruvate aminotransferase (U/L)88 ± 5081 ± 7472 ± 820.464
Hematocrit (%)35.8 ± 0.134.7 ± 0.134.0 ± 0.10.192
Platelet count (103 U/mm3)80.7 ± 34.690.3 ± 30.898.8 ± 40.00.015
Male (%)62.564.960.60.918
Alcohol etiology (%)25.043.951.50.006
Active alcoholism (%)18.815.833.30.161
Patients in secondary prophylaxis (%)50.043.978.80.025
History of ascites at basal evaluation (%)31.129.825.00.581
Presence of encephalopathy at basal evaluation (%)6.63.57.10.926
Presence of large esophageal varices (%)65.070.275.00.330
Presence of red signs (%)38.333.953.60.290
Basal RAP (mm Hg)5.1 ± 2.94.7 ± 2.64.9 ± 2.80.628
Basal FHVP (mm Hg)6.3 ± 2.97.9 ± 2.512.1 ± 4.9<0.001
Basal WHVP (mm Hg)25.0 ± 3.826.1 ± 4.430.6 ± 6.4<0.001
Basal HVPG (mm Hg)18.7 ± 3.518.2 ± 3.718.5 ± 3.50.615
Basal HAPG (mm Hg)19.9 ± 3.421.4 ± 3.625.7 ± 4.8<0.001
During prophylaxis RAP (mm Hg)5.5 ± 3.15.6 ± 2.94.7 ± 2.70.314
During prophylaxis FHVP (mm Hg)7.0 ± 3.18.2 ± 2.59.9 ± 3.6<0.001
During prophylaxis WHVP (mm Hg)23.3 ± 4.424.1 ± 4.025.6 ± 4.10.013
During prophylaxis HVPG (mm Hg)16.4 ± 3.915.9 ± 3.215.7 ± 2.80.357
During prophylaxis HAPG (mm Hg)17.8 ± 3.918.4 ± 3.320.9 ± 4.0<0.001

Hemodynamic Response to Chronic Beta-Blocker Treatment: Comparison Between HVPG Response and HAPG Response.

During chronic treatment with beta-adrenergic blockers, RAP was equal to FHVP in 18 patients, and lower in 136 patients. The mean difference between HAPG and HVPG during chronic treatment was 2.6 mm Hg (95% CI 2.3-3.1 mm Hg), which is not significantly different from the initial basal measurement. There was a significant correlation between the HVPG and the HAPG (Pearson coefficient 0.795; P < 0.001). The ICC for absolute agreement was 0.622 (95% CI 0.011-0.836), whereas the ICC for consistency was 0.788 (95% CI 0.719-0.841), again showing unsatisfactory agreement after excluding systematic differences (Fig. 1).

The mean decrease of the HVPG after chronic beta-blocker treatment was 12.3%. No significant differences were found by the comparison with the decrease of the HAPG which was 12.8% (P = 0.485, paired t test). In 19 patients beta-blockers decreased the HVPG to ≤12 mm Hg, whereas in 47 patients it decreased by more than 20% from baseline. In contrast, the HAPG decreased to ≤12 mm Hg in only eight patients, whereas in 53 patients it decreased by more than 20%. Fifty-two patients were considered HVPG responders according to the standard definitions, whereas 53 patients were HAPG responders. Table 4 shows the concordance between hemodynamic response defined with the HVPG and the HAPG. In 27 patients, the hemodynamic response defined by HVPG was discordant from that defined by the HAPG (k = 0.62).

Table 4. Agreement Between HVPG and HAPG (k = 0.610)
 HAPG NonrespondersHAPG RespondersTotal
HVPG nonresponders8814102
HVPG responders133952
Total10153154

Prognostic Value of HVPG and HAPG Response: Bleeding.

Twenty-five patients had a bleeding episode after 2 years of follow-up; six of those patients were on primary prophylaxis, and 19 were on secondary prophylaxis (P = 0.017). Hemodynamic HVPG responders (a decrease of the HVPG ≥20% from baseline or to ≤12 mm Hg) had an actuarial probability of bleeding significantly lower than HVPG nonresponders (2-year probability of bleeding 23% in nonresponders versus 6% in responders; P = 0.010) (Fig. 2).

Figure 2.

Cumulative probability of remaining free of bleeding according to HVPG response and HAPG response. Hemodynamic HVPG responders (R) had a lower probability of bleeding than HVPG non-responders (NR). In contrast, HAPG response did not allow to separate patients with different bleeding risk. P-values correpond to the log-rank test. Response was defined as a decrease in the pressure gradient ≥ 20% of baseline measurement or to ≤ 12 mm Hg.

Patients defined as responders according to HAPG also showed a lower probability of bleeding as compared with nonresponders, but the difference did not reach statistical significance. The bleeding rate at 2 years was 12% in HAPG responders versus 20% in HAPG in nonresponders (P = 0.207).

In an exploratory analysis, we examined whether HVPG nonresponders who were HAPG responders (14 patients) had a better outcome than HVPG nonresponders who were also HAPG nonresponders (88 patients). The outcome of these two groups of nonresponders was similar (2-year bleeding rate 30% versus 21%), and worse than the outcome of HVPG responders (excellent regardless of whether they were HAPG responders or not). These data suggest that measuring the HAPG does not add relevant prognostic information to the HVPG.

Finally, considering that the mean HAPG was 2.6 mm Hg higher than the HVPG, we also addressed whether HAPG response defined as a decrease of ≥20% from baseline or to ≤14.5 mm Hg had a better prognostic value. No patient whose HAPG decreased to ≤14.5 mm Hg experienced bleeding during follow-up. However, no significant differences were found in the bleeding rate between responders and nonresponders, even when using these modified criteria.

We further assessed whether the presence of ascites could influence the predictive value of the HVPG and HAPG. Whereas the HVPG holds prognostic value in patients with and without ascites, the HAPG response is not predictive of bleeding in patients with or without ascites (Supporting Information Fig. 1). In addition, stratification by different segments of FHVP-RAP difference shows that the higher the FHVP-RAP difference, the lower the prognostic value of the HAPG. In contrast, the prognostic value of the HVPG is maintained at any interval of FHVP-RAP difference (Supporting Information Fig. 2). HAPG response only had clear-cut prognostic value when the difference between FHVP and RAP was <2 mm Hg (that is, when HAPG and HVPG are almost identical).

Prognostic Value of HVPG and HAPG Response: Bleeding Free Survival.

Fourteen patients died during follow-up. Two additional patients underwent liver transplantation and were censored the day of the intervention. Only HVPG response (not HAPG response) was predictive of bleeding-free survival (Fig. 3). The 2-year actuarial probability of bleeding-free survival was 90% in HVPG responders versus 72% in HVPG nonresponders (P = 0.012); corresponding values for the HAPG were 85% and 75% (P = 0.185).

Figure 3.

Bleeding-free survival according to HVPG response and HAPG response. Hemodynamic HVPG responders (R) had a higher bleeding-free survival than HVPG non-responders (NR). Bleeding-free survival was not different between HAPG responders (R) and non-responders (NR). P-values correpond to the log-rank test. Response was defined as a decrease in the pressure gradient ≥ 20% of baseline measurement or to ≤12 mm Hg.

TIPS Cohort.

TIPS was placed in 70 patients for the prevention of variceal rebleeding, in 20 patients as rescue therapy due to early rebleeding, and in nine patients for refractory ascites. Clinical, endoscopic, and hemodynamic characteristics are reported in Table 1.

RAP values were equal (n = 12) or lower (n = 87) than IVCP in all cases. As a result, PCG was equal or lower than PAG (mean difference 2.5 mm Hg [95% CI 2.0-2.9 mm Hg]; P < 0.001). The correlation between the PCG and the PAG was statistically significant (Pearson correlation coefficient 0.801; P < 0.001). The ICC for absolute agreement was 0.616 (95% CI 0.034-0.829), whereas the ICC for consistency was 0.771 (95% CI 0.678-0.840). Once again, agreement was not excellent, even after the exclusion of systematic differences (Fig. 4).

Figure 4.

Concordance analysis between the porto-caval gradient (PCG) and the porto-atrial gradient (PAG) after TIPS placement. Solid lines represent the regression lines while dotted lines represent the identity lines. PAG was, in mean, 2.5 mm Hg higher than PCG. Even after correcting for this systematic difference agreement between these two measurements was not excellent.

We further explored the agreement of PCG and PAG in classifying patients according to achieving the target of hemodynamic response of a PCG ≤12 mm Hg. Eighty-nine of 99 patients reached this target; however, 20 of them had a PAG >12 mm Hg (k = 0.411). Therefore, using the PAG, 20% of the patients would have been unnecessarily subjected to further dilation of the stent or to other procedures.

Discussion

In patients with cirrhosis treated with nonselective beta-blockers to prevent variceal bleeding or rebleeding, the evaluation of HVPG response has been shown to provide invaluable prognostic information. Several longitudinal studies and two meta-analyses have consistently shown that patients achieving an HVPG reduction of ≥20 % from baseline or to levels ≤12 mm Hg (HVPG responders) have a very low risk of first bleeding or rebleeding during follow-up and have improved survival.8, 9 Indeed, HVPG response is considered one of the best surrogate outcome measurements in hepatology.31 As such, HVPG measurements should yield consistent and comparable results from center to center, which calls for standardization of the technique and specialist training. This has prompted the publication of several reports detailing the technical aspects of the procedure.1, 2, 12

In the present study, we provide a thorough analysis of one technical aspect of HVPG measurement that has received little attention so far: whether RAP, rather than FHVP, could be used as the internal reference to calculate the portal pressure gradient. As expected, FHVP was consistently higher than RAP and, therefore, HVPG was consistently lower than HAPG. However, there was a significant lack of agreement between the two values, even after excluding systematic differences. In addition, HVPG and HAPG responses to beta-blockers were different. When considering the HAPG response, fewer patients achieved a final value of ≤12 mm Hg and more patients achieved a ≥20% reduction compared with the HVPG response. As a result, 27 (18%) patients were differently classified as responders or nonresponders by HVPG or HAPG. The main finding of our study is that, contrary to the excellent prognostic performance of HVPG response, the hemodynamic response calculated with HAPG did not allow the separation of patients with different risk of bleeding or of bleeding-free survival. Therefore, our data clearly and firmly support the use of FHVP (not RAP) in the calculation of the portal pressure gradient (HVPG). Because the HAPG was a mean of 2.6 mm Hg higher than the HVPG, we further evaluated whether a modified definition of HAPG response (a decrease of ≥ 20% or to ≤14.5 mm Hg) could improve its prognostic value. Even with the use of this correction, the prognostic value of HAPG response was still not significant. These data are of relevance in daily practice, because the use of RAP to calculate the portal pressure gradient has increased, particularly in the setting of TIPS placement. In many hospitals, the PAG rather than the HVPG (or its equivalent, the portal pressure to FHVP pressure gradient) is used to determine whether TIPS has decreased portal pressure to safe values (≤12 mm Hg). Indeed, previous data from our unit clearly show that if such reduction in the portal pressure gradient is achieved, the patient is protected from variceal rebleeding and ascites.27 However, this pressure gradient should be calculated by subtracting the FHVP (or the IVCP) from the portal pressure. This can be easily done during a TIPS procedure by measuring the FHVP through the venous introducer, by placing its tip just cranially to the stent. Because RAP is lower than IVCP, the use of RAP to calculate the portal pressure gradient may result in excessive dilatation of the TIPS, which increases the risk of complications such as encephalopathy or liver failure.27 This is shown here in our series of patients with TIPS placement. Indeed, a considerable proportion of patients (one fifth) would have been classified as having high risk of rebleeding or ascites if the PAG was considered, although they were clearly protected because the PCG was in the safe range (≤12 mm Hg). The use of PAG is likely to have induced an unnecessary overdilation of the stent, with an increased risk of encephalopathy or to the use of associated procedures/treatments to prevent a nonexistent risk of recurrence of the complication of portal hypertension for which the TIPS was indicated.

In the present study, we also addressed an additional question: whether HAPG response could help to classify HVPG nonresponders in terms of risk of bleeding. It is well established that although the criteria for defining a good response to nonselective beta-blockers are very specific to detect clinical responders (those patients that will not experience bleeding), its sensitivity is rather low (many nonresponders have an uneventful evolution) and belong to a so-called gray zone of response.32 In that regard, previous studies from our unit have shown that some nonresponders by HVPG criteria exhibit an effective decrease in variceal pressure (and a similar protection from bleeding), presumably through a decrease in portal-collateral blood flow.33-35 The fact that the esophageal varices drain into the azygos system suggests that measurement of HAPG response could help to further classify HVPG nonresponders and decrease the gray zone in a way similar to that of variceal pressure response measurement. Unfortunately, our results did not support this contention, because the bleeding rates of the HVPG nonresponders with a good HAPG response were similar to that of patients who were both HVPG and HAPG nonresponders.

In this study, we also explored the factors associated with HVPG-HAPG disagreement by dividing the patients into tertiles according to the differences between FHVP and RAP. A consistent finding was that the more advanced the liver disease, the higher the difference between HAPG and HVPG. As should be expected, one of the factors having a strong association with this discrepancy was the baseline FHVP: the higher the FHVP, the higher the difference between the HAPG and the HVPG, suggesting that conditions increasing intra-abdominal pressure are the main source of disagreement between the HAPG and the HVPG. We did not detect differences in the body weight of our patients or in the prevalence of ascites across groups, which may be due—at least in part—to the fact that in this study, the body mass index was not recorded; therefore, we cannot directly assess the influence of obesity. One factor that suggests a higher prevalence of obesity in patients with higher HAPG-HVPG differences is their higher blood sugar. On the other hand, most patients with severe ascites had therapeutic paracentesis before hepatic vein catheterization, which explains why the presence of ascites was not different across tertiles of FHVP-RAP difference and suggests that the discrepancy between HVPG and HAPG would have been even greater if paracentesis had not been performed.

In conclusion, there is no good agreement between HAPG and HVPG in patients with cirrhosis, especially in those patients with increased intra-abdominal pressure. The excellent prognostic information provided by HVPG response to drug therapy is lost if HAPG response is considered. Furthermore, in TIPS patients, use of the PAG could result in unnecessary overdilation of the TIPS. RAP should not be used for the calculation of portal pressure gradient in patients with cirrhosis.

Acknowledgements

We thank Rosa Saez, Angels Baringo, and Laura Rocabert for expert logistic support and Clara Esteva for secretarial assistance.

Ancillary

Advertisement