Journal of Hepatology 43 (2005) 167176 www.elsevier.

com/locate/jhep

Special report

Evolving Consensus in Portal Hypertension Report of the Baveno IV Consensus Workshop on methodology of diagnosis and therapy in portal hypertension
Roberto de Franchis*,
Gastroenterology and Gastrointestinal Endoscopy Service, Department of Internal Medicine, University of Milan, IRCCS Ospedale Maggiore Policlinico, Via Pace 9, 20122 Milan, Italy

Portal hypertension is the haemodynamic abnormality associated with the most severe complications of cirrhosis, including ascites, hepatic encephalopathy and bleeding from gastroesophageal varices. Variceal bleeding is a medical emergency associated with a mortality that, in spite of recent progress, is still in the order of 20% at 6 weeks. The evaluation of diagnostic tools and the design and conduct of good clinical trials for the treatment of portal hypertension have always been difcult. Awareness of these difculties has led to the organisation of a series of meetings aimed at reaching consensus on the denitions of some key events related to portal hypertension and variceal bleeding, and at producing guidelines for the management of patients and for the conduct of trials in this eld. Such meetings took place in Groningen, the Netherlands in 1986 [1], in Baveno, Italy in 1990 (Baveno I) [2] and in 1995 (Baveno II) [3,4], in Milan, Italy in 1992 [5], in Reston, USA [6], in 1996 and in Stresa, Italy, in 2000 (Baveno III) [7,8]. All these meetings were successful and produced consensus statements on some important points, although several issues remained unsettled. To continue the work of the previous meetings, a Baveno IV workshop was held on April 2829, 2005. The workshop was attended by many of the experts responsible for most of the major achievements of the last years in this eld. The majority of them had attended the Groningen, Baveno I, Baveno II, Reston and Baveno III meetings as well. The main elds of discussion of the Baveno IV workshop were the same as in Baveno IIII, i.e. the denitions of key
* Tel.: C39 02 5503 5331/2; fax: C39 02 5032 0747. E-mail address: [email protected] (R. de Franchis). On behalf of the Baveno IV Chairpersons and panellists.

events concerning the bleeding episode, the therapeutic options in patients with portal hypertension, and the methodological requirements for future studies in this eld. For each of these topics, a series of consensus statements were discussed and agreed upon. Whenever applicable, the level of existing evidence was evaluated and the recommendations were ranked according to the Oxford System [9] (i.e. level of evidence from 1Zhighest to 5Zlowest; grade of recommendation from AZstrongest to DZweakest). The presentations given during the workshop are reported in extenso in the Baveno IV proceedings [10]. A summary of the most important conclusions is reported here.

1. Denition of key events regarding the bleeding episode Denitions and criteria to evaluate failure to control bleeding and failure to prevent rebleeding were introduced at Baveno II [3,4] and reviewed at Baveno III [7,8]. Since then, these denitions and criteria have been extensively applied in trials; it has been found that some of them are rather difcult to apply and do not reect adequately the situation in clinical practice; therefore, new denitions and criteria were proposed at Baveno IV. Given the lack of validated parameters to dene failure, these new criteria are necessarily arbitrary (level of evidence 5; grade of recommendation D) [9], and must be validated in future studies, in particular as surrogate markers of outcome. It is proposed that current and future studies should incorporate both Baveno IIIII and Baveno IV criteria, and evaluate failure to control bleeding using both sets of criteria. A judgment of the validity of the new criteria will be possible only after their extensive application in such studies.

0168-8278/$30.00 q 2005 European Association for the Study of the Liver. Published by Elsevier B.V. All rights reserved. doi:10.1016/j.jhep.2005.05.009

168

Evolving Consensus in Portal Hypertension / Journal of Hepatology 43 (2005) 167176

The Baveno IIIII and the Baveno IV denitions and criteria are reported below. 1.1. Baveno IIIII denitions and criteria for failure to control bleeding The denition of failure to control bleeding is divided into two time frames: (1) Within 6 h: any of the following factors: (a) transfusion of 4 units of blood or more, and (b) inability to achieve an increase in systolic blood pressure of 20 mmHg or to 70 mmHg or more, and/or (c) a pulse reduction to less than 100/min or a reduction of 20/min from baseline pulse rate. (2) After 6 h: any of the following factors: (a) the occurrence of hematemesis, (b) reduction in blood pressure of more than 20 mmHg from the 6-h point, and/or (c) increase of pulse rate of more than 20/min from the 6-h point on two consecutive readings 1 h apart, (d) transfusion of 2 units of blood or more (over and above the previous transfusions) required to increase the Hct to above 27% or Hb to above 9 g/dL. 1.2. Baveno IV denitions and criteria for failure to control bleeding (1) The time frame for the acute bleeding episode should be 120 h (5 days) (2) Failure signies need to change therapy: one criterion denes failure, whichever occurs rst: (a) Fresh hematemesis R2 h after start of specic drug treatment or therapeutic endoscopy. In the minority of patients who have a naso-gastric tube in place, aspiration of greater than 100 mL of fresh blood represents failure (b) 3 g drop in Hb (z9% drop in Ht) if no transfusion is administered (c) Death (d) Adjusted blood transfusion requirement index (ABRI, see below) R0.75 at any time point. (The threshold of ABRI dening failure requires validation).

1.3. Notes for the Baveno IV denitions and criteria For the purposes of analysis the following criteria should be adopted: Time to failurerst occurrence of any of the above criteria for failure (cumulative hazard plots and Cox regression analysis) Failure occurring at 120 h is considered as YES or NO The use of both time to failure and nal evaluation at 120 h is encouraged All specic therapeutic procedures should be documented with time points Intention to use further specic therapy should be documented even if not used Transfusion requirements should be recorded as a function of time for the whole interval of acute bleeding if no failure has occurred, e.g. units transfused/120 h or units transfused up to time of failure. 1.4. Baveno IIIII denitions and criteria for failure of secondary prophylaxis Failure to prevent rebleeding is dened as a single episode of clinically signicant rebleeding from portal hypertensive sources Clinically signicant rebleeding: (a) transfusion requirement of 2 units of blood or more within 24 h of time zero (the time of admission of a patient to the rst hospital he is taken to [2]) (b) together with a systolic blood pressure !100 mmHg or (c) a postural change of O20 mmHg and/or (d) pulse rate O100/min at time zero. 1.5. Baveno IV denitions and criteria for failure of secondary prophylaxis Failure to prevent rebleeding is dened as a single episode of clinically signicant rebleeding from portal hypertensive sources Clinically signicant rebleeding: (a) Hematemesis/melaena. In the minority of patients who have a naso-gastric tube in place, aspiration of greater than 100 mL of fresh blood represents failure plus (b) Adjusted Blood Requirement Index (ABRI)R0.5 (The threshold of ABRI dening failure requires validation) or (c) Decrease 3 g of Hb if no transfusion is given

Adjusted blood requirement index (ABRI) ABRI Z Blood units transfused final Ht K initial Ht C 0:01

Ht (or Hb) is measured at least every: B 6 h for the rst 2 days B 12 h for days 35 The transfusion target should be an haematocrit of 24% or a haemoglobin of 8 g/dL

2. Predictive models for portal hypertension Because of the growing importance of prognostic models in hepatology, and particularly in portal hypertension, a session devoted to this topic was introduced in Baveno IV,

Evolving Consensus in Portal Hypertension / Journal of Hepatology 43 (2005) 167176

169

to replace the session that was dedicated to the diagnosis of portal hypertension in Baveno III. Status classication of cirrhosis Varices, ascites and bleeding in patients with cirrhosis identify four clinical statuses of increasing severity: stage 1: no varices, no ascites; stage 2: varices, no ascites; stage 3: ascitesGvarices; stage 4: bleedingGascites [11] The outcome of a clinical status is transition to another status, death or OLT. Prognostic models specic to each clinical status should be developed Indicators of varices, and predictors of their development There are no satisfactory non-endoscopic indicators of the presence of varices While further studies are awaited, endoscopic screening is still the best practice to detect varices The hepatic vein pressure gradient (HVPG) is presently the most reliable predictor of variceal development Outcome prediction in compensated patients In compensated patients the development of ascites and portal hypertensive bleeding are the most relevant outcomes HVPG is the only known predictor of the development of ascites; other potential predictors should be investigated The NIEC score is presently the most reliable predictor of variceal rupture; the contribution of HVPG and other predictors should be investigated Outcome prediction in decompensated patients Child-Pugh and MELD predict overall mortality The additional role of HVPG and other potential predictors (sodium, spontaneous bacterial peritonitis, hepatorenal syndrome, others) should be assessed

HVPG is predictive of varices formation (1b;A). Recommendations for management All cirrhotic patients should be screened for varices at diagnosis (5;D). Despite some pharmaco-economical analysis, it is not indicated to treat cirrhotic patients with beta-blockers without prior assessment of the presence of esophageal varices (5;D). There is no indication, at this time, to treat patients to prevent the formation of varices (1b;A). Areas requiring further study Basic mechanisms in the development and progression of portal hypertension Natural history of low-risk varices (epidemiology and predictive factors of progression) Routine use of HVPG in clinical trials involved in investigating the complications of portal hypertension Treatment to decrease or prevent the progression and/or prevent the development of varices Biliary atresia (a very interesting entity of pediatric portal hypertension with rapid rate of progression). Non-invasive tests Non-invasive tests might be useful to identify patients at risk of having or prone to develop varices (HVPG O12 mmHg), but prospective studies are required (4;C). 3.2. Prevention of the rst bleeding episode Patients with small varices Patients with small varices could be treated with nonselective beta-blockers to prevent progression of varices and bleeding, but further studies, especially as relates to prevention of bleeding, are required before a formal recommendation on their use can be made (5;D). Patients with small varices with red wale signs or of Child C class have an increased risk of bleeding and may benet from treatment (5;D). Pharmacological treatments Non-selective beta-blockers reduce the risk of rst variceal bleeding in patients with medium and large oesophageal varices (1a;A). Isosorbide mononitrate administered alone must not be used (1a;A). There is not enough data to recommend the use of the combination of beta-blockers plus ISMN or

3. Therapeutic options in patients with portal hypertension 3.1. Pre-primary prophylaxis (prevention of the formation/growth of varices) Background Prevention of the development of complications of portal hypertension is clearly an important area for future research. Portal-systemic collaterals may develop before the appearance of varices, and can be diagnosed noninvasively. However, their clinical importance is uncertain (5;D).

170

Evolving Consensus in Portal Hypertension / Journal of Hepatology 43 (2005) 167176

spironolactone plus beta-blockers for primary prophylaxis (1b;A) [9]. Other pharmacological agents able to reduce portal pressure must be adequately tested before their clinical use (5;D). Use of HVPG measurements HVPG monitoring identies patients with cirrhosis who will benet from non-selective beta-blocker therapy in primary prophylaxis (1b;A). a la carte treatment using HVPG response in primary prophylaxis needs to be evaluated, especially in high-risk patients. Until then, routine use of HVPG cannot be recommended (5;D). Endoscopic treatment

haemodynamic stability and PRBC to maintain the haemoglobin at approximately 8 g/dL, depending on other factors such as patients co-morbidities, age, haemodynamic status, and presence of ongoing bleeding clinically (1b;A). Recommendations regarding management of coagulopathy and thrombocytopenia cannot be made on the basis of currently available data (5;D). Use of antibiotics for preventing bacterial infections/spontaneous bacterial peritonitis Antibiotic prophylaxis is an integral part of therapy for patients presenting with variceal bleeding and should be instituted from admission (1a;A). Prevention of hepatic encephalopathy

Prophylactic endoscopic band ligation (EBL) is useful in preventing variceal bleeding in patients with medium and large esophageal varices (1a;A). EBL is more effective than non-selective beta blockers in preventing rst variceal bleeding but does not improve survival. However, the long-term benets of EBL are uncertain because of the short duration of follow-up (1a;A). EBL should be offered to patients with medium/large varices and contraindications or intolerance to betablockers (5;D). Gastric varices In the absence of specic data on prophylactic studies, RCTs should be performed in patients with gastric varices. Cost-effectiveness analysis Markov models are not a substitute for well designed clinical trials. However, well-designed Markov models are complementary to clinical studies and should be pursued for exploratory purposes and to establish the costeffectiveness of various strategies. Markov models may ll in a void where clinical trials are simply not feasible. Areas requiring further study (5;D) Comparison of EBL and beta-blockers with respect to cost-effectiveness and quality of life to determine the treatment of choice. Studies to clarify whether the use of EBLCbeta-blockers is better than each treatment alone. 3.3. Treatment of acute bleeding from varices Blood volume restitution Blood volume restitution should be done cautiously and conservatively, using plasma expanders to maintain

In patients who present or develop encephalopathy, this should be treated with lactulose/lactitol or other drugs (5;D). There are no studies evaluating the usefulness of lactulose/lactitol for the prevention of hepatic encephalopathy (5;D). Assessment of prognosis No adequate prognostic model has been developed to predict outcomes (2b;B). No individual characteristic sufciently predicts prognosis (2b;B). Child-Pugh class, active bleeding at endoscopy, HVPG, infection, renal failure, severity of initial bleeding, presence of portal vein thrombosis or of HCC, and ALT have been identied as indicators of poor prognosis (2b;B). Timing of endoscopy Endoscopy should be performed as soon as possible after admission (within 12 h), especially in patients with clinically signicant bleeding or in patients with features suggesting cirrhosis (5;D). Use of balloon tamponade Balloon tamponade should only be used in massive bleeding as a temporary bridge until denitive treatment can be instituted. (For a maximum of 24 h, preferably in an intensive care facility) (5;D). Pharmacological treatment In suspected variceal bleeding, vasoactive drugs should be started as soon as possiblebefore diagnostic endoscopy (1b;A),

Evolving Consensus in Portal Hypertension / Journal of Hepatology 43 (2005) 167176

171

Vasoactive drug therapy (terlipressin, somatostatin, vapreotide, octreotide) should be maintained in patients with oesophageal variceal bleeding for 25 days (1a;A). Endoscopic treatment Endoscopic therapy is recommended in any patient who presents with documented upper GI bleeding and in whom esophageal varices are the cause of bleeding (1a;A). Ligation is the recommended form of endoscopic therapy for acute esophageal variceal bleeding although sclerotherapy may be used in the acute setting if ligation is technically difcult (1b;A). Endoscopic therapy with tissue adhesive (e.g. N-butylcyanoacrylate) is recommended for acute gastric variceal bleeding (1b;A). Endoscopic treatments are best used in association with pharmacological therapy, which preferably should be started before endoscopy (1a;A). Management of treatment failures Failures of initial therapy with combined pharmacological and endoscopic therapy are best managed by a second attempt at endoscopic therapy or TIPS (preferably with PTFE covered stents) (2b;B). Areas requiring further study (5;D) Optimal duration of vasoactive drug therapy, Effectiveness of early TIPS placement and of covered stents, Best treatment for gastric varices (especially glue vs. TIPS), The potential of rFVIIa, The best treatment of patients with no active bleeding at time of endoscopy on drug therapy, Prognostic factors/models for acute bleeding (MELD score, variceal size, age, etiology of portal hypertension and other comorbidities). 3.4. Prevention of rebleeding Time to start secondary prophylaxis Secondary prophylaxis should start as soon as possible from day 6 of the index variceal bleeding episode (5;D). The start time of secondary prophylaxis should be documented. Patients with cirrhosis who have not received primary prophylaxis Beta blockers (1a;A), band ligation (1a;A)or both (1b;A) should be used for prevention of recurrent bleeding.

Combination of beta blockers and band ligation is probably the best treatment (1b;A) but more trials are needed. Assessment of haemodynamic response to drug therapy provides prognostic information about rebleeding risk (2b;B). Patients with cirrhosis who are on beta blockers for primary prevention and bleed Band ligation should be added (5;D). Patients who have contraindications or intolerance to beta blockers Band ligation is the preferred treatment for prevention of rebleeding (5;D). Patients who fail endoscopic and pharmacological treatment for prevention of rebleeding TIPS or surgical shunts (distal splenorenal shunt or 8 mm H-graft) are effective for those with Child class A/B cirrhosis and should be used (2b;B). In non-surgical candidates, TIPS is the only option (5;D). Transplantation provides good long-term outcomes in Child class B/C cirrhosis and should be considered (2b;B). TIPS may be used as a bridge to transplantation (4;C). Patients who have bled from isolated gastric varices, type 1 (IGV1) [12] or gastro-oesophageal varices, type 2 [12] (GOV 2) N-butyl-cyanoacrylate (1b;A), TIPS (2b;B) or beta blockers (2b;B) are recommended. Patients who have bled from gastro-esophageal varices, type 1 (GOV 1) May be treated with N-butyl-cyanoacrylate, band ligation of oesophageal varices or beta blockers (2b;B). Patients who have bled from portal hypertensive gastropathy Beta blockers (1b;A) should be used for prevention of recurrent bleeding. Patients in whom beta blockers are contraindicated or fail and who cannot be managed by non-shunt therapy TIPS (4;C) or surgical shunts (4;C) should be considered. Areas requiring further study (5;D) Combination of beta blockers plus nitrates. Use of HVPG monitoring for decision making and its effect on patients outcome.

172

Evolving Consensus in Portal Hypertension / Journal of Hepatology 43 (2005) 167176

4. Non-cirrhotic portal hypertension A session devoted to non-cirrhotic portal hypertension was introduced at Baveno IV, in view of the increasing recognition and growing interest of this clinical entity. Due to time constraints, the discussion was limited to the BuddChiari syndrome [BCShepatic venous outow tract obstruction (HVOTO)] and to extra-hepatic portal vein obstruction (EHPVO). This session replaced the session that in Baveno III was devoted to portal hypertensive gastropathy and gastric varices. 4.1. Budd-Chiari syndrome [BCShepatic venous outow tract obstruction (HVOTO)] Denition Budd-Chiari syndrome (BCS) is an eponym for hepatic venous outow tract obstruction (HVOTO) which can be located from the level of the small hepatic veins to the level of the termination of inferior vena cava into the right atrium. BCS is an heterogeneous condition with regard to causes and pathogenesis. BCS is considered secondary when the mechanism for HVOTO is compression/invasion by a benign or malignant tumor, abscess or cyst. BCS is considered primary otherwise. Hepatic congestion secondary to heart failure and pericardial disease are excluded from the denition of BCS. Obstruction conned to small hepatic veins or sinusoids in the context of liver irradiation, chemotherapy, stem cell transplantation or exposure to toxic agents is excluded from the denition of BCS. The terms veno-occlusive disease and sinusoidal obstruction syndrome require further denition. Etiology Primary BCS is frequently associated with one or several risk factors for thrombosis. These underlying disorders are often occult at presentation with BCS. Myeloproliferative disorders should be investigated in any patient with BCS, irrespective of the peripheral blood picture. When liver synthetic function is impaired, low plasma levels of antithrombin, protein C, and protein S are not specic for an inherited defect. Diagnosis BCS is diagnosed by the demonstration of an obstruction of the venous lumen, or by the presence of hepatic vein collaterals.

Liver biopsy is not necessary to make a diagnosis of BCS when vascular imaging has demonstrated obstruction of the hepatic venous outow tract. Liver biopsy is the only means to make a diagnosis of BCS of the small intrahepatic veins. Clinical trials for therapy of BCS have not been performed so that current therapy is based on less rigorous information. Treatment On the basis of current expert opinion (5;D) Anticoagulation should be recommended to all patients, in the absence of major contra-indications. However, there is no consensus on the optimal duration of anticoagulation. Previous bleeding related to portal hypertension is not considered a major contra-indication for anticoagulation, provided appropriate prophylaxis for recurrent bleeding is initiated. Complications of portal hypertension may be treated as recommended for the other types of liver diseases. Stenoses that are amenable to percutaneous angioplasty/stenting should be actively looked for, and treated accordingly. TIPS insertion should be attempted when angioplasty/stenting is not feasible, and when the patient does not improve on medical therapy. Liver transplantation should be considered in patients with manifestations refractory to the above procedures. Areas requiring further studies (5;D) Accurate diagnostic tests for myeloproliferative disorder and antiphospholipid syndrome. Benet and risk of prolonged anticoagulation therapy. Benet and risk of pharmacological therapy for portal hypertension. Optimal timing of angioplasty and TIPS with respect to severity of symptoms. Indications for thrombolysis. 4.2. Extra-hepatic portal vein obstruction (EHPVO) [13] Denition EHPVO is dened by obstruction of the extra-hepatic portal vein with or without involvement of the intrahepatic portal veins. EHPVO often manifests as portal cavernoma, which is a network of portoporto collaterals and develops as a sequel of portal vein obstruction. Isolated thrombosis of the splenic vein or superior mesenteric vein with patent portal vein is excluded. The denition should be augmented by a statement of presence or absence of cirrhosis and neoplasia.

Evolving Consensus in Portal Hypertension / Journal of Hepatology 43 (2005) 167176

173

Etiology EHPVO is a heterogeneous entity with regards to causes and pathogenesis, particularly between children and adults. EHPVO in adults is frequently associated with one or several risk factors for thrombosis which may be occult at presentation. Presence of cirrhosis, neoplasia and other intra-abdominal causes such as inammation, trauma, etc. do not exclude the presence of systemic risk factors. Clinical presentation EHPVO can be acute or chronic. EHPVO can be assumed to be recent when patients present with symptoms such as abdominal pain, ascites, fever or symptoms suggestive for intestinal ischaemia, in the absence of portal cavernoma and porto-systemic collaterals. Chronic EHPVO is associated with portal cavernoma and may present with variceal bleed, splenomegaly, abnormal blood cell counts and occasionally jaundice. A proportion of children have growth retardation. Diagnosis EHPVO is diagnosed by imaging techniques like Doppler US, CT or MRI which demonstrate portal vein obstruction, presence of intraluminal material or portal vein cavernoma. Natural history Most patients with EHPVO in the absence of cirrhosis and neoplasia have a relatively benign course. Morbidity is mainly related to variceal bleed, recurrent thrombosis, symptomatic portal biliopathy and hypersplenism. The natural course of EHPVO is mainly determined by the presence or absence of associated diseases such as cirrhosis or neoplasia. Treatment (in the absence of cirrhosis and neoplasia) Chronic EHPVO For primary prophylaxis of variceal bleeding there is insufcient data on whether beta-blockers or endoscopic therapy should be preferred. For the control of acute variceal bleeding, endoscopic therapy is effective (1b;A). In the absence of specic data on patients with EHPVO, it is presumed that the same treatments used in bleeding cirrhotic patients could be applied (5;D). For secondary prophylaxis, endoscopic therapy is effective (1b;A). There is insufcient evidence to recommend beta-blockers.

There is no consensus on the indication for anticoagulant therapy. However, in those patients with a persistent documented prothrombotic state, anticoagulant therapy can be considered. There is insufcient evidence in favor of interventional therapy such as TIPS and local thrombolysis. Decompressive surgery should only be considered for patients with failure of endoscopic therapy (5;D). For portal biliopathy with obstructive jaundice, endoscopic therapy is recommended (5,D). In case of failure, shunt surgery may be considered (5;D). Recent EHPVO Recent EHPVO rarely resolves spontaneously. The evidence on which to base recommendations for anti-coagulant therapy is weak. On the basis of current expert opinion (5;D), in patients with recent EHPVO Anticoagulation should be given for at least 3 months in all patients. When an underlying persistent prothrombotic state has been documented, life-long anticoagulant therapy is recommended. In patients with EHPVO and associated cirrhosis, hepatocellular carcinoma should be excluded. There is insufcient data on which to base recommendations for giving anticoagulant therapy to these patients. Areas requiring further studies (5;D) Natural history of EHPVO in children vs. adults: hepatic dysfunction, portal biliopathy, growth retardation Etiologyrole of various prothrombotic states in EHPVO (in the East), identication of susceptible population. Assessment of thrombosis, progression and recurrence. Denitions of variceal bleeding and predictors of rst bleed and rebleed. Role of beta-blockers and comparison with endoscopic therapy. Usefulness of long-term anticoagulants, TIPS, shunt surgery. Development of good experimental models.

5. Providing scientic evidence: RCTs and beyond In previous Baveno workshops, a session was devoted to the methodological requirements for future trials in portal hypertension. At Baveno IV this session was replaced by one addressing some aspects of therapy in clinical practice that have not been or cannot be evaluated by RCTs, due to: (a) inadequate quality of trials, (b) uncommon diseases or (c) distinct features of the more common diseases.

174

Evolving Consensus in Portal Hypertension / Journal of Hepatology 43 (2005) 167176

Addressing these issues should contribute to EBM when adequate information from RCTs is not possible to obtain or not yet available. Possible use of per protocol analysis In superiority trials, ITT strategies are preferred and PP analysis regarded only as supportive. In non-inferiority trials, ITT and PP approaches (if appropriately pre-dened) may both contribute. When PP results differ from ITT results, the population excluded from PP analysis should be scrutinized. The applicability of the intervention may be questioned. Assessing changes in therapeutic effects with progression of the disease To assess how treatment effect may change with disease progression, use interaction tests between outcome predictors and the intervention(s). Both unadjusted results and results adjusted for strong outcome predictors should be provided, regardless of baseline comparisons. Any subgroup analyses should be pre-dened, have sufcient power and usually be limited to primary outcome. Otherwise, they are exploratory methods that can help design further studies but should not modify the conclusions of RCTs. Handling the heterogeneity of RCTs in meta-analysis Heterogeneity can be used cautiously to suggest indications for a particular intervention. This requires that: B differences in trial methodology are not present B Clinical source of heterogeneity is identied Stratied analysis of pooled individual data can be done. B Primary/secondary aims should be dened. B Plan for statistical analysis should be pre-dened (including multiple testing). B Subsequent analysis can use the same pooled data as long as the above protocol is followed. Identication of factors that modify therapeutic effects in a clinically signicant way Physicians must learn how to identify the factors that most often modify the clinical outcome at variance with the results of RCTs. The quality of RCTs (internal and external validity) should be evaluated. The internal validity can be assessed according to the CONSORT statement. The external validity can be assessed according to a list of variables which dene the peculiarity of the trial population: differences in demography, co-morbidities, limitations due to inclusion/exclusion criteria, variability

in the schedules and dosages of drugs, usage of interfering drugs, low compliance, duration of treatment. Approach to the diagnosis and treatment of uncommon cases where evidence from RCT is not forthcoming Consensus-driven, clinical protocols are required to dene the optimal methods for clinical management of uncommon conditions where RCTs cannot be performed Treatment of uncommon manifestations of portal hypertension with evidence-based medicine awaits the identication of biologically plausible surrogate markers Alternative study designs (clinical databases, N of 1 trials) should be adapted to identify effective treatments for uncommon manifestations of portal hypertension Observational studies of treatment effect require statistical techniques to minimize confounding by indication Continuous monitoring of the clinical outcome of treatments in so-called clinical databases Development of a database to monitor outcomes is desirable. Goals should include monitoring outcome in: B Three major clinical areas in portal hypertension B Specic sub-groups Funding mechanisms should be identied. B Focus on complications of cirrhosis rather than portal hypertension. B Selected mix of institutions B Potential interest from both government and industry for funding such a database. Survival analysis for competing end-points other than death The KaplanMeier plot is often used to estimate the probability of survival free of other end-points, e.g. variceal bleeding. This produces non-interpretable results that may also be biased. The cause is that analysis using censoring of patients assumes that those who die or reach other competing end-points are still at risk for the primary end-point, which is not true [14]. For this type of analysis, cumulative hazard plots and Cox regression analysis are better. Need for international collaboration on clinical trials Almost all, if not all, RCTs in portal hypertension are underpowered. This applies to uncommon but also common types of conditions associated with portal hypertension. In cardiology and oncology, very large international multinationaltrials are conducted, so it is feasible! We should do the same for solving our problems in management of portal hypertension.

Evolving Consensus in Portal Hypertension / Journal of Hepatology 43 (2005) 167176

175

6. Conclusions The purpose of the consensus denitions about the variceal bleeding episode is to use them in trials and other studies on portal hypertension, as well as in clinical practice. This does not mean that authors cannot use their own denitions, but they are encouraged to use and evaluate in parallel these Baveno IV consensus denitions. This should result in some measure of standardisation and increased ease of interpretation among different studies. Equally important, if there are uniformly dened endpoints, meta-analyses will be based on more homogeneous studies, which is an essential pre-requisite of this methodology. It is desirable that future studies be reported using these denitions as part of the evaluation. Change or renement can then take place, as they have at Baveno IV with respect to Baveno II, Reston and Baveno III, to ensure that the consensus denitions do have clinical relevance and are easily applied in practice. Several denitions agreed upon in Baveno I [2], II [3,4] and III [7,8] were taken for granted and not discussed in Baveno IV. Interested readers can refer to the Baveno I, II [24] and III [7,8] reports. The suggestions about the topics of future studies reect the opinions of the experts about the areas were new information is most needed. As long as new diagnostic tools and new treatments appear, they will have to be assessed in comparison with present-day standards.

M.D., Poitiers; D. Thabut, M.D., Paris; J.P. Vinel, M.D., Toulouse; Germany: M. Schepke, M.D., Bonn; India: Y.C. Chawla, M.D., Chandigarh; Italy: M. Angelico, M.D., Rome; G. Barosi, M.D., Pavia, M. Merli, M.D., Rome, A. Morabito, M.D., Milano; M. Primignani, M.D., Milan; F. Salerno, M.D., Milan; F. Schepis, M.D., Modena; M. Zoli, M.D., Bologna; Spain: J.G. Abraldes, M.D., Barcelona; A. ` Albillos, M.D., Madrid, R. Banares, M.D., Madrid; P Gines, M.D., Barcelona; R. Planas, M.D., Badalona; C. Villanueva, M.D., Barcelona; Switzerland: A. Hadengue, M.D., Geneva; Taiwan: H.C. Lin, M.D., Taipei; G.H. Lo, M.D., Kaohsiung; The Netherlands: H. Janssen, M.D., Rotterdam; H. van Buuren, M.D., Rotterdam; United Kingdom: E. Elias, M.D., Birmingham; D. Patch, M.D., London; USA: A. Blei, M.D., Chicago, IL; T. Boyer, M.D., Atlanta, GA; N. Chalasani, M.D., Indianapolis, IN; J.M. Henderson, M.D., Cleveland, OH; Y. Iwakiri, M.D., New Haven, CT; W.R. Kim, Rochester, MN; D. Kravetz, San Diego, CA; A. Sanyal, M.D., Richmond, VA; V. Shah, M.D., Rochester, MN; B. Shneider, M.D., New York, NY; J. Talwalkar, Rochester, MN. The following gave review lectures during the Workshop Michael Fallon, M.D., Birmingham, AL, USA; Pere ` Gines, M.D., Barcelona, Spain; Christian Gluud, M.D., Copenhagen, Denmark; Pier Mannuccio Mannucci, M.D., Milan, Italy; Miguel Navasa, M.D., Barcelona, Spain; Luigi Pagliaro, M.D., Palermo, Italy.

Acknowledgements 7. Participants The following chaired sessions during the Workshop: Jaime Bosch, M.D., Barcelona, Spain; Andrew K Burroughs, M.D., London, U.K.; Gennaro DAmico, ` M.D., Palermo, Italy; Juan Carlos Garcia-Pagan, M.D., Barcelona, Spain; Guadalupe Garcia-Tsao, M.D., New Haven, CT, USA; Norman D Grace, M.D., Boston, MA, USA; Roberto Groszmann, M.D., New Haven, CT, USA; Patrick Kamath, M.D., Rochester, MN, USA; Loren Laine, M.D., Los Angeles, CA, USA; Didier Lebrec, M.D., Clichy, ` France; Carlo Merkel, M.D., Padua, Italy; Juan Rodes, M.D., Barcelona, Spain; Shiv K Sarin, New Delhi, India; Tilman Sauerbruch, M.D., Bonn, Germany; Thorkild I.A. Srensen, M.D., Copenhagen, Denmark; Dominique Valla, M.D. Clichy, France. The following participated in the presentations and the discussion as panellists: Argentina: J. Vorobioff, M.D., Rosario; Belgium: F. Nevens, M.D., Leuven; Canada: J. Heathcote, M.D., Toronto; N. Marcon, M.D., Toronto; I. Wanless, M.D., Toronto; Denmark: F. Bendtsen, M.D., Copenhagen; E. Christensen, M.D., Copenhagen; Egypt: G. Shiha, M.D., al Mansoura; France: B. Bernard-Chabert, M.D., Reims; P. ` Cales, M.D., Angers, R. Moreau, M.D., Clichy, C. Silvain, The Baveno IV workshop was endorsed and supported in part by a unrestricted educational grant of the European Association for the Study of the Liver (EASL). The Baveno IV workshop was also endorsed by the Italian Association for the study of the liver (AISF), the Italian Association of hospital gastroenterologists and endoscopists (AIGO), the Italian association for digestive endoscopy (SIED) and the Italian Society of Gastroenterology (SIGE).

References
[1] Burroughs AK, editor. Methodology and review of clinical trials in portal hypertension. Excerpta medical congress service no. 763. Amsterdam, New York, Oxford. [2] de Franchis R, Pascal JP, Ancona E, Burroughs AK, Henderson JM, Fleig W, et al. Denitions, methodology and therapeutic strategies in portal hypertension. A consensus development workshop. J Hepatol 1992;15:256261. [3] de Franchis R. Developing consensus in portal hypertension. J Hepatol 1996;25:390394. [4] de Franchis R, editor. Portal hypertension II. Proceedings of the second Baveno international consensus workshop on denitions, methodology and therapeutic strategies. Oxford: Blackwell Science; 1996.

176

Evolving Consensus in Portal Hypertension / Journal of Hepatology 43 (2005) 167176 diagnosis and treatment in portal hypertension. Oxford, UK: Blackwell Science; in press. DAmico G. Esophageal varices: from appearance to rupture; natural history and prognostic indicators. In: Groszmann RJ, Bosch J, editors. Portal hypertension in the 21st century. Dordrecht: Kluwer; 2004. p. 147154. Sarin SK, Lahoti D, Saxena SP, Murthi NS, Makwane UK. Prevalence, classication and natural history of gastric varices: long-term follow-up study in 568 patients with portal hypertension. Hepatology 1992;16:13431349. Sarin SK, Agarwal SR. Extraheptic portal vein obstruction. Semin Liver Dis 2002;22:4358. Thomsen BL, Mller S, Srensen TIA. The Copenhagen Esophageal Varices Sclerotherapy Project. Optimised analysis of recurrent bleeding and death in patients with cirrhosis and oesophageal varices: application of a multi-stage competing risk model. J Hepatol 1994;21: 367375.

[5] Spina GP, Arcidiacono R, Bosch J, Pagliaro L, Burroughs AK, Santambrogio R, Rossi A. Gastric endoscopic features in portal hypertension: nal report of a consensus conference. J Hepatol 1994; 21:461467. [6] Grace ND, Groszmann RJ, Garcia-Tsao G, Burroughs AK, Pagliaro L, Makuch RW, et al. Portal hypertension and variceal bleeding: an AASLD single topic symposium. Hepatology 1998;28:868880. [7] de Franchis R. Updating consensus in portal hypertension: report of the Baveno III consensus workshop on denitions, methodology and therapeutic strategies in portal hypertension. J Hepatol 2000;33: 846852. [8] de Franchis R, editor. Portal hypertension III. Proceedings of the IIIrd Baveno international consensus workshop on denitions, methodology and therapeutic strategies. Oxford: Blackwell Science; 2001. [9] http://www.cebm.net/downloads/Oxford_EBM_Levels_5.rtf. [10] de Franchis R, editor. Portal hypertension IV. Proceedings of the IVth Baveno international consensus workshop on methodology of

[11]

[12]

[13] [14]