The current recommended prophylaxis of variceal. Long-Term Follow-up of Hemodynamic Responders to Pharmacological Therapy After Variceal Bleeding

Transcription

1 Long-Term Follow-up of Hemodynamic Responders to Pharmacological Therapy After Variceal Bleeding Salvador Augustin, 1 Antonio Gonzalez, 1 Laia Badia, 1 Laura Millan, 1 Aranzazu Gelabert, 2 Alejandro Romero, 2 Antoni Segarra, 2 María Martell, 1 Rafael Esteban, 1,3 Jaime Guardia, 1,3 and Joan Genesca 1,3 Although it is assumed that hemodynamic responders to pharmacological therapy after a variceal hemorrhage are adequately protected from rebleeding, there is no evidence that either this response or its protective effect extend beyond the usual 2-year follow-up featured in available studies. We aimed to assess the maintenance of hemodynamic response and its impact on outcomes in a large cohort of hemodynamic responders during a long follow-up. One hundred three patients with cirrhosis admitted with acute variceal bleeding between 2001 and 2010 were prospectively evaluated. The hepatic venous pressure gradient (HVPG) was determined 5 days after the bleeding and repeated 5-7 days after maximal tolerated doses of nadolol and nitrates. Hemodynamic responders (HVPG 12 mm Hg or 20% decrease from baseline) were maintained on drugs and followed up with annual HVPG measurements. Forty-eight patients (47%) were hemodynamic responders. The median follow-up was 48 months (range, months). Long-term HVPG evaluations could not be performed in eight patients (four deaths, two rebleedings, two followups <1 year). Among the remaining 40 patients, hemodynamic response was maintained in 26 (65%) and lost in 14 (35%). There were no baseline differences between the two subgroups. However, 100% of alcoholic patients who remained abstinent maintained long-term response, compared with 36% of nonabstinent alcoholics and 50% of patients with viral cirrhosis. Patients with loss of hemodynamic response rebled more during follow-up and showed a higher incidence of death or liver transplantation. Conclusions: After variceal bleeding, long-term maintenance of hemodynamic response to drug therapy is mainly restricted to patients with alcoholic cirrhosis who remain abstinent. The loss of this long-term response carries worse clinical outcomes. (HEPATOLOGY 2012;56: ) Abbreviations: HR, hazard ratio; HVPG, hepatic venous pressure gradient; LT, liver transplantation; TIPS, transjugular portosystemic shunt. From the 1 Liver Unit, Department of Internal Medicine, and the 2 Interventional Radiology Unit, Department of Radiology, Hospital Universitari Vall d Hebron, Institut de Recerca, Universitat Autonoma de Barcelona, Barcelona, Spain; and 3 CIBERehd, Instituto de Salud Carlos III, Madrid, Spain. Received September 30, 2011; accepted February 13, Salvador Augustin is a recipient of a Río Hortega fellowship grant from the Instituto de Salud Carlos III and a Juan Rodes grant from the Asociacion Espa~nola para el Estudio del Hígado. The study was partially funded by grant SAF from Ministerio de Ciencia e Innovacion. CIBERehd is supported by Instituto de Salud Carlos III. Address reprint requests to: Salvador Augustin, Liver Unit, Department of Internal Medicine, Hospital Universitari Vall d Hebron, Passeig Vall d Hebron 119, 08035, Barcelona, Spain. saugustin@vhebron.net; fax: Copyright VC 2012 by the American Association for the Study of Liver Diseases. View this article online at wileyonlinelibrary.com. DOI /hep Potential conflict of interest: Nothing to report. The current recommended prophylaxis of variceal rebleeding consists of the combined use of pharmacological therapy (nonselective betablockers alone or with nitrates) and endoscopic variceal ligation. 1-3 In patients treated with drug therapy, the evaluation of the hemodynamic response by the measurement of the hepatic venous pressure gradient (HVPG) has been strongly recommended. 1,2,4 Different observational studies and randomized trials have shown that a reduction of HVPG below 12 mm Hg or 20% from baseline in patients under drug therapy is associated with a marked decrease in rebleeding. 5,6 In view of this, it has been proposed that HVPG responders could be maintained on drug therapy only and spared from endoscopic prophylaxis. 1,2 Nevertheless, some investigators have questioned the clinical benefit of using HVPG monitoring to identify hemodynamic responders and guide prophylaxis accordingly. 7,8 Among other issues, one important question that remains unanswered is to what extent it could be assumed from HVPG measurements taken shortly after the bleeding episode that the responder status is maintained in the long term (i.e., beyond the 706

2 HEPATOLOGY, Vol. 56, No. 2, 2012 AUGUSTIN ET AL year follow-up of most of available studies on secondary prophylaxis). 5,6 This issue could have important clinical consequences, as it is likely that, in an HVPG-guided prophylactic regimen, responders would be maintained on drugs indefinitely, long after a 2- year follow-up. The scarce published evidence on this matter is contradictory, as the only two available studies to date featuring HVPG reassessments were performed in small cohorts with different baseline risks (primary and secondary prophylaxis) and reached opposite conclusions. 9,10 In order to answer this specific question, we performed a longitudinal, prospective, observational study in a large cohort of patients with cirrhosis in which the hemodynamic response was as evaluated after a variceal hemorrhage, and assessed the long-term maintenance of this response as well as its impact on outcomes. Patients and Methods The Strengthening the Report of Observational Studies recommendations for reporting observational studies 11 were applied for the manuscript design. The study was approved by the ethics committee of our institution; all patients gave written informed consent. Cohort Recruitment and Data Collection. We analyzed data of all consecutive patients with cirrhosis admitted to the Bleeding Unit of our tertiary University hospital with acute variceal esophageal bleeding from January 2001 to June These data had been prospectively recorded in our unit for different studies, two of which analyzed the efficacy of an HVPGguided protocol to prevent the recurrence of variceal bleeding. 12,13 Both the data collection protocol and the study aim were prespecified before recruitment of the cohort. For each patient, baseline data were recorded at admission by the team in charge of the patient, and data on treatments and outcomes were recorded during follow-up by the same physicians. Exclusion criteria for the study were any of the following features: age >80 years, Child-Pugh score 13, failure to control the index bleeding, current active therapy with beta-blockers and nitrates or endoscopic variceal obliteration, contraindications to beta-blockers or nitrates, advanced hepatocellular carcinoma, severe associated conditions, portal thrombosis, and a HVPG <10 mm Hg. Patients with previous history of betablocker therapy were considered eligible if they were not receiving active treatment with beta-blockers plus nitrates at the moment of the index bleeding. Study Protocol. The study protocol of this cohort has been already described. 12,13 In summary, acute bleeding was treated with somatostatin, antibiotics (norfloxacin in the period and ceftriaxone in ) and endoscopic therapy (14 patients at the beginning of recruitment did not receive endoscopic therapy due to mild bleeding). A first hepatic hemodynamic study was performed days after the bleeding. The moment of this first hemodynamic study was considered time zero of inclusion in the study, and this first HVPG value was considered the baseline gradient from which initial and long-term hemodynamic response was evaluated. Pharmacological therapy with nadolol and nitrates was subsequently started. Nadolol was given orally at an initial dose of 40 mg/day. The dose was then increased daily for a period of 5-7 days until intolerance appeared, the heart rate decreased to 55 beats per minute, or a maximal dose of 160 mg/day was reached. Oral isosorbide mononitrate was then started at 20 mg once at bedtime, followed by 20 mg twice a day for 1 day, and finally increased to 40 mg BD if tolerated. Once the maximal tolerated drug doses were achieved, a second hemodynamic study was scheduled 5-7 days later. According to the hemodynamic response observed, patients were classified in two groups: responders, when the HVPG was 12 mm Hg and/or had decreased 20% from baseline values; and nonresponders, when these hemodynamic targets were not achieved. According to available recommendations, responders were maintained with the same pharmacological therapy (drugs only) until their first rebleeding episode or the end of follow-up (regardless of the results of follow-up HVPG measurements). Only if a first rebleeding episode occurred was ligation prophylaxis added to drug therapy in responders. Nonresponders were treated from the start with the combination of drugs and endoscopic ligation, except for a few patients at the beginning of the study period who received a transjugular portosystemic shunt (TIPS). Ligation was performed with commercial multiband devices at 2- to 3-week intervals until variceal obliteration had been achieved. Once eradicated, patients underwent a follow-up endoscopy at 3 months and at 6-month intervals thereafter. Additional sessions of ligation were conducted if varices reappeared. None of the patients with hepatitis C in this cohort of decompensated patients with cirrhosis received antiviral therapy during follow-up, in line with available recommendations. Patients were followed at 3-month intervals at the outpatient liver clinic. History and examination evaluated alcohol abstinence, compliance to drug therapy and adverse effects. Alcohol abstinence was assessed by

3 708 AUGUSTIN ET AL. HEPATOLOGY, August 2012 direct anamnesis with the patient, separated interview with close relatives and unexpected determinations of alcoholemia. Compliance to drug therapy was assessed in all patients by blood pressure and heart rate measurements at each follow-up visit, in addition to a specific anamnesis. Moreover, responders underwent annual determinations of HVPG for 3 years. The study was continued for 6 months after the enrollment of the last patient. Definitions and Endpoints. Diagnosis of cirrhosis was based on liver biopsy or the usual clinical and imaging criteria. Bleeding was considered from esophageal variceal origin when the emergency endoscopy performed within 12 hours after admission showed any of the accepted criteria defining variceal bleeding. 14 The primary endpoints of the study were maintenance of long-term response, rebleeding and the composite of death and liver transplantation (LT). These endpoints were prespecified before the beginning of the study. An initial hemodynamic responder was considered to maintain long-term response if all follow-up HVPG measurements were 12 mm Hg and/or >20% of baseline value. Clinically significant rebleeding from portal hypertensive sources was defined as in Baveno IV, 15 which is equivalent to failure of secondary prophylaxis. Under these circumstances, an alternative therapy was instituted (e.g., banding in those on drug therapy, TIPS, or transplantation). The composite endpoint death/lt was preferred over mortality, as it was deemed that, in the specific scenario of advanced liver disease, considering only the latter may lead to important potential bias associated to the impact of LT on the natural history of cirrhosis. Statistical Analysis. Differences between categorical variables were assessed by chi-square test or Fisher s exact test when necessary. Continuous variables were compared using the Student t test or Mann-Whitney test as appropriate. A two-sided P value <0.05 was considered statistically significant. Overall rebleeding and death/lt analysis was conducted on an intention-to-treat basis. Kaplan-Meier curves were constructed to evaluate the dynamics of rebleeding and death/lt, and Cox analysis was conducted to identify independent prognostic indicators for long-term response, rebleeding, and death/lt. Moreover, for the case of rebleeding, in order to describe accurately its cumulative incidence, a competing risk analysis was performed. For Kaplan-Meier analysis, follow-up was censored on the date when the patient was last seen, or the date of death or transplantation. However, in this setting, the occurrence of death or LT (competing risk) may preclude the occurrence of a first rebleeding, modifying its probability and the derived calculations (such as the ones needed for Kaplan-Meier analysis). 16,17 The competing risk model allows differentiating censored patients according to their competing risk status (alive versus dead or transplantation) at the end of follow-up. Then, the model calculates first the probability of occurrence of any event (rebleeding or death/lt), and afterward the conditional probability of the event of interest (rebleeding), from which the cumulative incidence is derived. Significance between Kaplan-Meier curves was assessed by log-rank test, and for the case of competing risks curves, the specific method described by Gray was used. 18 The SPSS (version 15.0; SPSS Inc., Chicago, IL) and STATA (version 10.0; StataCorp, College Station, TX) statistical packages were used for the analysis. Results Patient Features and Hemodynamic Studies. During the study period, 304 consecutive patients admitted with acute variceal bleeding were considered. A total of 201 patients were excluded, leaving 103 for hemodynamic assessment. The flow chart of patients in the study is shown in Fig. 1. Basal clinical and hemodynamic features of the 103 patients evaluated are shown in Table 1. The second HVPG measurement was not performed in 13 patients (12.7%). Among these, two patients died from liver failure shortly after the first hemodynamic study, and the remaining 11 patients presented a variceal rebleeding before the second hemodynamic study could be conducted. The remaining 90 patients underwent a second hemodynamic study days after the first study (i.e., approximately 19 days after the index bleeding). No differences in baseline characteristics or treatment received in the acute phase were observed between the 11 patients who bled before the second HVPG measurement and the 90 patients who underwent the second hemodynamic study. As shown in Fig. 1, of the 90 patients who underwent a second HVPG measurement, 48 (53%) were classified as responders and maintained on nadolol. The remaining 42 (47%) patients were nonresponders and had banding ligation added to their drug therapy, except for eight patients recruited at the beginning of the study period who received a TIPS. Table 2 compares the clinical and hemodynamic characteristics of responders and nonresponders. Differences were not detected in any parameter, except for the second

4 HEPATOLOGY, Vol. 56, No. 2, 2012 AUGUSTIN ET AL. 709 Fig. 1. Flow chart of the patients in the study. HVPG* stands for hepatic venous pressure gradient. Table 1. Clinical and Hemodynamic Characteristics of Patients Characteristic N ¼ 103 Age, years Sex, male/female 79/24 Etiology, no. (%) Alcohol 48 (47) Alcohol þ virus 16 (15) Virus 39 (38) Child-Pugh class, A/B/C 38/42/23 Child-Pugh score 7 (6-9) MELD score 13 (10-17) Ascites, no. (%) 38 (37) Prior bleeding, no. (%) 11 (11) Prior beta-blocker prophylaxis, no. (%) 25 (24) Variceal grade, 1/2/3* 1/56/46 Active bleeding, no. (%) 34 (33) Treatment of index episode, no. (%) Drugs (somatostatin þ antibiotics) alone 14 (14) Drugs plus sclerotherapy 31 (30) Drugs plus banding 58 (56) Heart rate, bpm Mean arterial pressure, mm Hg Hepatic venous pressure, mm Hg Free Wedged Gradient Continuous variables are presented as the mean 6 SD for normally distributed variables or median (interquartile range) if normality criteria were not met. Abbreviations: bpm, beats per minute; MELD, model for end-stage liver disease. *Grade 1, varices flattened by insufflation; grade 2, varices not flattened by insufflation and separated by areas of normal mucosa; grade 3, confluent varices not flattened by insufflation. At day of first hemodynamic measurement, before initiation of nadolol therapy. hemodynamic study results. Notably, there were only five drug dose reductions (with one drug discontinuation) among responders (see the Outcomes and Prognostic Analysis Among Responders section for more details) and four dose reductions (including two discontinuations) among nonresponders. Outcomes and Prognostic Analysis in the Whole Cohort. The median follow-up was 35 months for the whole cohort (range, 7 days to 108 months), 48 months for responders (range, months), and 26 months for nonresponders (range, months). Table 3 shows the outcomes of the whole cohort and the different subgroups according to hemodynamic response. Among the 11 patients in whom a second HVPG could not be obtained because of early rebleeding, five patients died during follow-up and five underwent transplantation. If only the 90 patients in whom the hemodynamic response could be assessed were considered, rebleeding and mortality rates were 23% and 28%, respectively (median follow-up, 37 months; range, months). As shown, the rebleeding incidence was higher in responders (33% versus 12%; chi-square P ¼ 0.02). The incidence of rebleeding in Table 2. Comparison of Baseline Clinical Characteristics and Hemodynamic Features Between Initial Hemodynamic Responders and Nonresponders to Drug Therapy Characteristic Responders (n ¼ 48) Nonresponders (n ¼ 42) P Age, years Sex, male/female 37/11 33/9 0.9 Etiology, no. (%) Alcohol 28 (58) 17 (40) Alcohol þ virus 7 (15) 7 (17) 0.2 Virus 13 (27) 18 (43) Child-Pugh class, A/B/C 19/20/9 15/15/ Child-Pugh score 7 (6-9) 8 (6-10) 0.3 MELD score 13 (10-17) 14 (10-17) 0.7 Ascites, no. (%) 16 (33) 17 (41) 0.5 Albumin, g/dl 2.9 ( ) 2.8 ( ) 0.4 Bilirubin, mg/dl 1.4 ( ) 2.2 ( ) 0.09 International normalized ratio 1.6 ( ) 1.8 ( ) 0.4 Prior bleeding, no. (%) 5 (10) 5 (12) 1.0 Prior beta-blocker 8 (17) 13 (31) 0.11 prophylaxis, no. (%) Heart rate, bpm Mean arterial pressure, mm Hg First hepatic venous pressure, mm Hg Free Wedged Gradient Nadolol dose, mg/day* 160 (85-160) 160 ( ) 0.5 Nitrates dose, mg/day* 80 (80-80) 80 (80-80) 1 Second hepatic venous pressure, mm Hg Free Wedged <0.001 Gradient <0.001 Continuous variables are presented as the mean 6 SD for normally distributed variables or median (interquartile range) if normality criteria were not met. Abbreviations: bpm, beats per minute; MELD, model for end-stage liver disease. *At the time of second hemodynamic study.

5 710 AUGUSTIN ET AL. HEPATOLOGY, August 2012 Table 3. Overall Rebleeding, Mortality, and Liver Transplantation Rates During the Whole Follow-Up Period Second HVPG Hemodynamic Response All Patients (n ¼ 103) No (n ¼ 13) Yes (n ¼ 90) Responders (n ¼ 48) Nonresponders (n ¼ 42) P* Rebleeding 32 (31) (23) 16 (33) 5 (12) 0.02 Mortality 31 (30) 7 25 (28) 11 (23) 14 (33) 0.3 Transplantation 17 (17) 5 14 (16) 5 (10) 9 (21) 0.2 Mortality or transplantation 48 (47) (44) 16 (33) 23 (54) 0.04 Rebleeding-related mortality 5 (5) 2 3 (3) 2 (4) 1 (2) 1.0 Data are presented as no. (%). *Chi-square comparisons between responders and nonresponders. nonresponders after exclusion of the eight patients who received a TIPS was similar (four [12%] rebled). The composite endpoint death/lt was significantly higher in nonresponders, however (54% versus 33%; chi-square P ¼ 0.04). Moreover, nonresponders showed higher median Child-Pugh scores at the end of follow-up (8 versus 5.5; Mann-Whitney P ¼ 0.022) and percentage of change from baseline ( 16.7% versus 0.0%; Mann-Whitney P ¼ 0.059) than responders. Regarding other decompensating events, 9 (21%) responders presented at least one nonbleeding decompensating episode (five new-onset ascites, four encephalopathy) versus 15 (36%; nine new-onset ascites, six encephalopathy) nonresponders (chi-square P ¼ 0.07). As for readmission rates, 29 (60%) responders and 29 (69%) nonresponders were readmitted at least once during follow-up (P ¼ 0.4). Figure 2 depicts the actuarial probability of rebleeding and of the composite endpoint death/lt in both groups. The actuarial probability of rebleeding at 2 years was 23% in responders and 11% in nonresponders, and at 4 years it was 33% and 17%, respectively. Due to the different tendency in rebleeding and mortality in these groups, the cumulative incidence of rebleeding considering death/lt as a competing endpoint was calculated and plotted along with the usual Kaplan-Meier estimation. Rebleeding risk was higher among responders in both types of analysis. Multivariate Cox analysis identified viral etiology of cirrhosis (hazard ratio [HR], 2.6; 95% CI [confidence interval] ; P ¼ 0.02), age (HR, 1.04; 95% CI, ; P ¼ 0.006), baseline Child-Pugh score (HR, 1.4; 95% CI, ; P ¼ 0.001), and lack of initial hemodynamic response (HR, 2.0; 95% CI, ; P ¼ 0.05) as statistically significant predictors of death/lt for the whole cohort. Multivariate Cox analysis of rebleeding did not allow the identification of any significant predictor variable. Outcomes and Prognostic Analysis Among Responders. As described above, 48 patients (37 men; median age, 53 years) were classified as hemodynamic responders after the second HVPG measurement. The median follow-up of this subgroup was 48 months (range, 2-108). Long-term HVPG evaluations could not be performed in eight patients (four deaths, two rebleedings, two follow-ups <1 year). Among the remaining 40 patients, 21 had three HVPG measurements, 13 had two HVPG measurements, and six had one HVPG Fig. 2. Kaplan-Meier and competing risk curves for rebleeding (A) and Kaplan-Meier curves for transplantation-free survival (B) during follow-up in hemodynamic responders and nonresponders (n ¼ 90) after a variceal hemorrhage.

6 HEPATOLOGY, Vol. 56, No. 2, 2012 AUGUSTIN ET AL. 711 Fig. 3. Comparison of the mean HVPG of hemodynamic responders who maintained the response during follow-up (long-term responders) and of those who lost that response (long-term nonresponders). measurement. Long-term hemodynamic response was maintained in 26 (65%) patients and lost in 14 (35%) patients. Comparison of the median HVPG measurements in long-term responders and nonresponders is shown in Fig. 3. Long-term response was already lost at the first annual HVPG in most long-term nonresponders (10 of 14 patients). There were no baseline differences between long-term responders and nonresponders. However, all 15 alcoholic patients who remained abstinent maintained long-term response compared with four (36%) of 11 nonabstinent alcoholics (P < 0.001) and seven (50%) of 14 patients with viral cirrhosis (P ¼ 0.002), six of whom were abstinent. During the study period, 14 (35%) of these 40 patients rebled, seven (17.5%) died of liver-related causes, and four (10%) underwent transplantation. Patients with loss of hemodynamic response rebled more (79% versus 11%; chi-square P < 0.001) and showed a higher incidence of death/lt (50% versus 15%; chi-square P ¼ 0.029). All abstinent alcoholics were alive at the end of follow-up (two had rebled and two underwent transplantation). Figure 4 shows the actuarial probability of rebleeding and death/lt in both groups calculated using the Kaplan-Meier method and the respective cumulative incidences estimated by competing risks analysis. Actuarial probability of rebleeding at 2 years was 8% in long-term responders and 44% in long-term nonresponders, and at 4 years it was 8% and 54%, respectively. Only three (11.5%) long-term responders and two (14%) long-term nonresponders had their drug doses reduced due to intolerance or noncompliance during follow-up. Among the three long-term responders, follow-up HVPGs were maintained in one patient and increased slightly (but remained below 10 mm Hg) in the other two patients despite the dose reduction. All three patients (two viral and one alcoholic cirrhosis, who had nadolol stopped for impotence) remained abstinent during all follow-ups (range, months), and none of them rebled. On the other hand, the two patients with dose reduction and loss of hemodynamic response rebled (at 13 and 17 months, respectively). Both patients in this subgroup had mixed viral and alcohol cirrhosis and had resumed alcohol drinking prior to the rebleeding episode. Drug therapy in the remaining 43 patients was well tolerated, and all of them were kept during all follow-up on the same maximal tolerated doses they had at the moment of the second HVPG. To evaluate whether the differences between alcohol abstinents and nonabstinents in long-term response and outcomes may have been related to baseline differences or beta-blocker doses received compliance, a comparison of the most relevant among these Fig. 4. Kaplan-Meier and competing risk curves for rebleeding (A) and Kaplan-Meier curves for transplantation-free survival (B) in hemodynamic responders (n ¼ 40) who maintained the response during follow-up (long-term responders) and in those in which the hemodynamic response was lost (long-term nonresponders).

7 712 AUGUSTIN ET AL. HEPATOLOGY, August 2012 Table 4. Comparison of Baseline Characteristics and Drug Therapy Received Between Abstinent and Nonabstinent Alcoholics with Long-Term Hemodynamic Evaluations Characteristic parameters was performed (Table 4). Both subgroups were comparable. Cox multivariate analysis identified loss of hemodynamic response (HR, 7.5; 95% CI, ; P ¼ 0.002) and history of previous variceal bleeding (HR, 8.6; 95% CI, ; P ¼ 0.002) as risk factors for rebleeding, and viral etiology (HR, 4.1; 95% CI, ; P ¼ 0.05) and Child-Pugh score at 1 year (HR, 1.5; 95% CI, ; P ¼ 0.015) as main determinants of death/lt. Discussion Abstinent Alcoholics (n ¼ 15) Nonabstinent Alcoholics (n ¼ 11) P Age, years Sex, male/female 13/2 10/1 0.6 Child-Pugh class (A/B/C) 5/6/4 3/4/4 0.9 Child-Pugh score 7 (5-10) 8 (6-11) 0.5 MELD score 13 (10-16) 14 (10-20) 0.2 Ascites, no. (%) 8 (53) 4 (36) 0.5 Albumin, g/dl 3.0 ( ) 2.8 ( ) 0.7 Bilirubin, mg/dl 1.5 ( ) 1.8 ( ) 0.7 International normalized 1.6 ( ) 1.7 ( ) 0.2 ratio Prior bleeding, no. (%) 3 (20) 2 (18) 1.0 Prior beta-blocker 2 (13) 3 (27) 0.6 prophylaxis, no. (%) Active bleeding at initial 5 (33) 3 (27) 0.9 endoscopy, no. (%) First HVPG, mm Hg Second HVPG, mm Hg Nadolol dose at second HVPG, mg/day Nadolol dose at end of follow-up, mg/day Continuous variables are presented as the mean 6 SD for normally distributed variables or median (interquartile range) if normality criteria were not met. Abbreviation: MELD, model for end-stage liver disease. It is currently accepted in clinical guidelines 1,2 that, in an HVPG-guided prophylactic strategy after a variceal hemorrhage, those patients meeting the accepted criteria of hemodynamic response are reasonably protected from rebleeding under drug therapy alone. This view is supported by longitudinal and randomized trials in which the responder status is based on HVPG measurements taken shortly (from 2 weeks to 3 months) after the index hemorrhage, with no subsequent reassessments during the usual 2-year follow up of these studies (range, 8-28 months). 5,6 Nevertheless, although it is assumed that these patients should be kept indefinitely on drug therapy only, there is no evidence that the initial hemodynamic response is maintained after this 2-year period. In this longitudinal observational study, we followed up for a median of 4 years a large cohort of hemodynamic responders treated with beta-blockers and nitrates after a variceal bleeding, and found that this response was lost in the long term in one out of three patients, an observation that was associated with a clear negative impact on their outcomes. This is the only available report in which sequential protocol HVPG measurements were performed in responders after a variceal bleeding to check their hemodynamic status. The previous study by Merkel et al. 10 shared a similar approach and reported analogous results, but it was performed in patients with no previous history of bleeding, a setting with clearly lower baseline risk of progression of portal hypertension and incidence of related complications. Therefore, its conclusions cannot be applied to the prophylaxis of variceal rebleeding. The pioneer study by Villanueva et al., 9 on the other hand, did follow-up patients under pharmacologic secondary prophylaxis of rebleeding, but reached opposite conclusions (i.e., hemodynamic response was maintained in most patients, and the outcomes of responders on drugs alone were excellent). Two important features of our study may account for these discrepancies. First, regarding the higher incidence of long-term nonresponders among our patients, it should be noted that, in our study, long-term response was defined taking into account HVPG reassessments over a 3-year period, while patients in the Villanueva et al. study underwent only a single HVPG reassessment months after the index bleeding. If only the first annual HVPG measurement had been considered in our cohort to define long-term response, the percentage of patients with loss of response would have been similar to that reported by Villanueva et al. (21% versus 25% in our cohort). Second, the median follow-up in our study was 48 months, which is almost twice the average median times reported by Villanueva et al. and other cohorts of responders from previous meta-analysis. 5,6 We think that this is the main reason why the overall rebleeding rate of 33% in our cohort of responders is much higher than the 14%-16% reported so far. 5,6 In this regard, the analysis of the actuarial probabilities of rebleeding and the inspection of curves in Fig. 2 in our study nicely illustrate how the outcomes in responders remained within the expected values for the first 2 years, but worsened steadily afterwards, essentially due to the occurrence of rebleeding and death/lt in the long-term nonresponder

8 HEPATOLOGY, Vol. 56, No. 2, 2012 AUGUSTIN ET AL. 713 subpopulation (Fig. 4A,B). The main explanation for this dynamic is probably that the longer the follow-up, the higher the proportion of patients who lost their hemodynamic response and were no longer protected with drug therapy alone. In fact, the overall rebleeding rate of 33% in our responders is strikingly similar to the 33%-37% overall (responder and nonresponder) rebleeding rates found in most studies of patients treated with drug therapy after a variceal bleeding. 7 In our study, the main determinants of loss of longterm response were viral etiology of cirrhosis and active alcohol consumption. On one hand, we observed that one out of two patients with viral cirrhosis lost their long-term response. In a cohort of patients with decompensated viral cirrhosis not receiving antiviral therapy, such as variceal bleeders, and after a very long follow-up, such as the one in our study, a significant percentage of subjects would be reasonably expected to experience a progression of their disease, even with mild or no evident deterioration of liver function. 10 This progression may ultimately result in increasing of HVPG and worse outcomes. 19 On the other hand, the negative effect of alcohol intake in patients with alcohol cirrhosis is also evident in our study. It is well known that alcohol abstinence is related to maintenance or even reductions of HVPG values in patients receiving or not receiving drug therapy, 9,20 whereas alcohol consumption clearly worsens portal hypertension both in the short 21 and long term. 9 Lastly, it is worth remarking that, in contrast to the study by Villanueva et al., 9 the loss of long-term response in our responders could not be attributed to a reduction of drug doses during follow-up due to intolerance or noncompliance. Only five patients (12.5%) in our cohort had their doses reduced, three of whom maintained the initial response. The present study was designed to evaluate the hemodynamic evolution and outcomes of responders. Consequently, the comparison between the outcomes of initial responders and nonresponders is not suitable, because there are relevant differences between those groups in terms of baseline risks, treatments received, and follow-up times. Nonetheless, two secondary observations derived from the analysis of the whole study cohort deserve consideration. First, the prognosis of those patients who rebled before the second HVPG was dismal (seven deaths and five transplants of 13 patients), which confirms data from previous studies. Second, the protection from rebleeding of nonresponders, which were kept from the beginning with endoscopic ligation combined with drug therapy, was excellent. The low rebleeding rate of initial nonresponders (12%) could be related at least in part to a shorter follow-up (26 months) or a higher incidence of competing events in this subcohort, although the competing risk analysis suggested otherwise. However, from our results and from that of recent observations, 22 we clearly feel that adding ligation to drug therapy in nonresponders (instead of switching them to ligation alone, as in the majority of previous studies) could be an effective approach, which should be nevertheless evaluated in a randomized controlled trial. The potential practical implications of the present study are straightforward. Our results suggest that, in an HVPG-guided prophylactic regimen, responders could be safely treated with drug therapy alone during the first 1-2 years, but whether this strategy remains effective in the long term is unknown. Consequently, it would be reasonable to reassess HVPG regularly in patients with viral cirrhosis and/or active alcohol consumption and to protect patients who have lost their response. Early rebleeders (those who rebleed before the second HVPG) may be regarded as candidates for more aggressive therapies (such as early TIPS) 23 or liver transplantation, and initial or long-term nonresponders may be considered to have ligation added to drugs. All these potential implications should be ideally tested in randomized controlled trials. Before these results could be transferred to patient care, several limitations related to the design of our study should be taken into account. First, the main limitation of our study is obviously its sample size, which forces cautious interpretations regarding specific subgroup outcomes. An external validation of our results would indeed help to narrow the confidence intervals for our observations. Unfortunately, the difficult logistics associated with a very long follow-up and repeated HVPG measurements in a large cohort makes this possibility unlikely, at least in the near future. Second, whether our cohort could be regarded as representative of a given whole population of variceal bleeders would depend mainly on the local resources and the availability of HVPG measurements. The greater this availability is, the larger the percentage of eligible patients. However, it is worth remarking that, like almost all available studies on HVPG-guided strategies, our cohort included only patients with viral and/ or alcohol cirrhosis, so it should be emphasized that our observations apply specifically to these populations. Finally, although great care was taken to adequately collect data on treatment compliance and alcohol abstinence, there is always a significant risk of underestimating their occurrence, because there are no consensuated objective means to measure them. For the specific case of alcoholics, it could even be

9 714 AUGUSTIN ET AL. HEPATOLOGY, August 2012 speculated that medication adherence was more likely in abstinents than nonabstinents. To this regard, it should be noted that the specific comparison of these two subpopulations showed that they were similar in terms of baseline status and drug doses received, clearly suggesting that their different outcomes were mainly related to alcohol consumption. In spite of this, an unnoticed and significant difference in adherence could not be completely ruled out. Yet, in the end, while these concerns could bias our analysis on the reasons to account for the loss of long-term response, we consider that they do not alter the main pragmatic implications derived from our results. Recommending high beta-blocker doses, alcohol abstinence, and regular monitoring of hemodynamic response is, ultimately, consistent with current knowledge on the physiopathology of the disease, and certainly not against clinical common sense. In this longitudinal study, we found that the longterm maintenance of hemodynamic response to drug therapy after a variceal bleed is mainly restricted to patients with alcoholic cirrhosis who remain abstinent, but is lost in a significant proportion of patients with viral cirrhosis and nonabstinent alcoholics. In addition, the loss of this long-term response carries worse clinical outcomes. These findings may have important clinical and research implications regarding the use of HVPG measurements in the prophylaxis of variceal rebleeding. References 1. De Franchis R, Baveno V Faculty. Revising consensus in portal hypertension: report of the Baveno V consensus workshop on methodology of diagnosis and therapy in portal hypertension. J Hepatol 2010;53: García-Tsao G, Bosch J, Groszmann R. Portal hypertension and variceal bleeding unresolved issues. Summary of an American Association for the Study of Liver Diseases and European Association for the Study of the Liver single-topic conference. HEPATOLOGY 2008;47: García-Tsao G, Bosch J. Management of varices and variceal hemorrhage in cirrhosis. N Engl J Med 2010;362: Bosch J, Abraldes JG, Berzigotti A, García-Pagan JC. The clinical use of HVPG measurements in chronic liver disease. Nat Rev Gastroenterol Hepatol 2009;6: D Amico G, Garcia-Pagan JC, Luca A, Bosch J. Hepatic vein pressure gradient reduction and prevention of variceal bleeding in cirrhosis: a systematic review. Gastroenterology 2006;131: Albillos A, Ba~nares R, Gonzalez M, Ripoll C, Gonzalez R, Catalina MV, et al. Value of the hepatic venous pressure gradient to monitor drug therapy for portal hypertension: a meta-analysis. Am J Gastroenterol 2007;102: Thalheimer U, Mela M, Patch D, Burroughs AK. Targeting portal pressure measurements: a critical reappraisal. HEPATOLOGY 2004;39: Thalheimer U, Mela M, Patch D, Burroughs AK. Monitoring target reduction in hepatic venous pressure gradient during pharmacological therapy of portal hypertension: a close look at the evidence. Gut 2004; 53: Villanueva C, Lopez-Balaguer JM, Aracil C, Kolle L, Gonzalez B, Mi~nana J, et al. Maintenance of hemodynamic response to treatment for portal hypertension and influence on complications of cirrhosis. J Hepatol 2004;40: Merkel C, Bolognesi M, Berzigotti A, Amodio P, Cavasin L, Casarotto IM, et al. Clinical significance of worsening portal hypertension during long-term medical treatment in patients with cirrhosis who had been classified as early good-responders on haemodynamic criteria. J Hepatol 2010;52: von Elm E, Altman DG, Egger M, Pocock SJ, Gotzsche PC, Vandenbroucke JP. STROBE Initiative. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. Lancet 2007;370: Gonzalez A, Augustin S, Perez M, Dot J, Saperas E, Tomasello A, et al. Hemodynamic response-guided therapy for prevention of variceal rebleeding: an uncontrolled pilot study. HEPATOLOGY 2006;44: Gonzalez A, Augustin S, Dot J, Perez M, Abu-Suboh M, Romero A, et al. Adding banding ligation is effective as rescue therapy to prevent variceal rebleeding in hemodynamic non-responders to pharmacological therapy. Dig Liv Dis 2012;44: De Franchis R, Pascal JP, Ancona E, Burroughs AK, Henderson M, Fleig W, et al. Definitions, methodology and therapeutic strategies in portal hypertension. J Hepatol 1992;15: De Franchis. Evolving consensus in portal hypertension. Report of the Baveno IV consensus workshop on methodology of diagnosis and therapy in portal hypertension. J Hepatol 2005;43: Satagopan JM, Ben-Porat L, Berwick M, Robson M, Kutler D, Auerbach AD. A note on competing risks in survival data analysis. Br J Cancer 2004;91: Kim WR, Therneau TM, Benson JT, Kremers WK, Rosen CB, Gores GJ. Deaths on the liver transplant waiting list: an analysis of competing risks. HEPATOLOGY 2006;43: Gray RJ. A class of K-sample tests for comparing the cumulative incidence of a competing risk. Ann Stat 1988;16: van Leeuwen DJ, Howe SC, Scheuer PJ, Sherlock S. Portal hypertension in chronic hepatitis: relationship to morphological changes. Gut 1990;31: Vorobioff J, Groszmann RJ, Picabea E, Gamen M, Villavicencio R, Bordato J, et al. Prognostic value of hepatic venous pressure gradient measurements in alcoholic cirrhosis: a 10-year prospective study. Gastroenterology 1996;111: Luca A, García-Pagan JC, Bosch J, Feu F, Caballeria J, Groszmann RJ, et al. Effects of ethanol consumption on hepatic hemodynamics in patients with alcoholic cirrhosis. Gastroenterology 1997;112: Dell Era A, Cubero Sotela J, Fabris FM, Petazzi G, Reati R, Iannuzzi F, et al. Primary prophylaxis of variceal bleeding in cirrhotic patients: a cohort study. Dig Liver Dis 2008;40: Garcia-Pagan JC, Caca K, Bureau C, Laleman W, Appenrodt B, Luca A, et al. Early use of TIPS in patients with cirrhosis and variceal bleeding. N Engl J Med 2010;362: