Background and aims: Risk factors for mortality and re-bleeding following acute variceal haemorrhage (AVH) are incompletely understood. The aim of this study was to determine risk factors for 6-week mortality, and re-bleeding within 5 days in patients with cirrhosis and AVH.
Methods: Kaplan–Meier and Cox proportional hazards regression analyses were used to determine risk factors among 256 patients with AVH entered into a randomised, prospective trial.
Results: Thirty-five patients (14%) died within 6 weeks of AVH; 14 deaths (40%) occurred within 5 days. Only the Model for End-stage Liver Disease (MELD) score and units of packed red blood cells (PRBCs) transfused in the first 24 h were associated with 6-week mortality univariately (HR 1.11, p<0.001; HR 1.22, p<0.001) and bivariately (HR MELD = 1.10, p<0.001; HR per unit of PRBCs transfused = 1.15, p = 0.005). Re-bleeding within 5 days occurred in 37 patients (15%); MELD score (p = 0.01) and a clot on a varix (p = 0.05) predicted re-bleeding. Patients with a MELD score ⩾18; both MELD score ⩾18 and ⩾4 units of PRBCs transfused; both MELD score ⩾18 and active bleeding at index endoscopy; and variceal re-bleeding had increased risk of death 6 weeks post-AVH (HR = 7.4, p<0.001; 11.3, p<0.001; 9.9, p<0.001; 10.2, p<0.001 respectively).
Conclusions: Patients with AVH and MELD score ⩾18, requiring ⩾4 units of PRBCs within the first 24 h or with active bleeding at endoscopy are at increased risk of dying within 6 weeks. MELD score ⩾18 is also a strong predictor of variceal re-bleeding within the first 5 days.
Statistics from Altmetric.com
Haemorrhage from oesophageal varices is one of the most feared complications of portal hypertension and is associated with significant morbidity and mortality.1 2 Mortality with each episode of acute variceal haemorrhage (AVH) has decreased to current levels of only approximately 20%.3–8 Several risk factors have been identified that are associated with mortality risk for an episode of AVH, including active bleeding at initial endoscopy, haematocrit level, aminotransferase levels, hepatic venous pressure gradient (HVPG), presence of portal vein thrombosis, alcoholic liver disease, serum bilirubin and albumin levels, hepatic encephalopathy, hepatocellular carcinoma (HCC) and the Child–Turcotte–Pugh (CTP) score.4 9 Many of these data have been derived from retrospective studies when mortality rates following an episode of variceal bleeding were up to 50%, from relatively small numbers of patients, or from patient groups in whom therapy was not according to a standard protocol, including variability in the use of antibiotics.4 10 11. In the best study to date on prognostic factors for mortality and re-bleeding following a variceal bleed, antibiotics were not used routinely, 27% of patients did not receive endoscopic therapy during the acute bleeding episode and 17% of patients did not receive vasoactive drugs or endoscopic therapy.4 The use of prophylactic antibiotics during acute bleeding episodes in patients with cirrhosis has resulted in significant decreases in mortality rates, but did not become standard protocol until recent years.12–17
The CTP class has traditionally been used to stratify patients at risk for death after AVH, with a higher CTP score being associated with increased mortality.18 Prospective randomised trials in acute variceal bleeding have stratified patients variably into CTP class A and B versus CTP class C, or CTP class A versus CTP class B and C. However, the CTP classification system has limitations in its application, including the use of subjective parameters such as ascites and encephalopathy, a “ceiling” and “floor” effect with the objective parameters of bilirubin, albumin and prothrombin time, and the lack of standardisation of albumin and prothrombin time measurements.
The Model for End-stage Liver Disease (MELD) score has been shown to be superior to the CTP score as an index of liver disease severity in patients with chronic liver disease awaiting liver transplantation, and accurately predicts 3-month mortality in these patients.19 The MELD score was initially created from a cohort of patients with cirrhosis who underwent transjugular intrahepatic portosystemic shunt (TIPS) procedures for treatment of complications of portal hypertension, mainly variceal haemorrhage.20 Given its accuracy in survival prediction, the MELD score is now used within the USA and other countries to prioritise organ allocation for patients registered on the liver transplant waiting list.21 Because MELD was originally derived from a cohort of patients with complications of portal hypertension, we tested the hypothesis that MELD is an accurate predictor of short-term survival in patients with cirrhosis presenting with AVH. Additionally, previous studies of prognostic factors in AVH have been derived from retrospective studies or from studies employing non-uniform treatment modalities. This has made it difficult to apply conclusions from these study results to current practice where antibiotics are used routinely, and the mortality rates from AVH are only approximately 20%.
The risk factors associated with re-bleeding from varices, as defined by the Baveno criteria, within the 5-day period of an acute variceal bleed are also unclear.2 Therefore, we used prospectively collected unpublished data from one of the largest international, randomised, controlled, clinical trials ever conducted on AVH in cirrhotics to determine the risk factors for 5-day and 6-week survival and 5-day variceal re-bleeding, in patients presenting with AVH.
The data for this investigation were obtained from an international, randomised, double-blinded, placebo-controlled clinical trial investigating the role of the somatostatin analogue, lanreotide, as an adjunct to endoscopic therapy for the control of AVH among patients with cirrhosis and portal hypertension. The clinical trial results were completely negative to the degree that the sponsor stopped the support of further activities on the study drug lanreotide. The results of the main clinical trial—that is, absence of any benefit of the somatostatin analogue in conjunction with endoscopic therapy for the control of acute variceal bleeding—are yet to be published. However, because of the lack of the effect of lanreotide, the blinded data (both treatment arms combined) can be used to model the natural history of patients with cirrhosis and AVH.22 Permission was obtained from the sponsor to publish these data. There are no conflicts of interest among the authors of this publication with respect to the use of these data.
We were able to obtain data on patients enrolled at 16 centres throughout North America between December 1998 and March 2002. Cirrhotic patients between the ages of 18 and 83 years who presented to hospital with melaena or haematemesis suspicious for oesophageal variceal bleeding were included in the study. Cirrhosis was diagnosed based upon previous liver biopsy findings, clinical evaluation (stigmata of chronic liver disease), laboratory data (low serum albumin, elevated serum bilirubin or prolonged prothrombin time), or imaging findings compatible with a diagnosis of cirrhosis. Only those patients who presented within 24 h of the onset of the bleeding episode were included. Exclusion criteria included patients who had received somatostatin, somatostatin analogues or vasopressin within the previous 12 h, and pregnant or breast-feeding females. Patients with HCC, those who had previously had a TIPS placed and those patients who had undergone endoscopic treatment of varices with sclerotherapy or band ligation within the previous 5 days were also excluded.
Therapeutic interventions for variceal bleeding
Patients who met the study inclusion/exclusion criteria were randomised to receive either lanreotide or placebo at least 30 min prior to endoscopy and the infusion was continued for a total of 5 days. The randomisation sequence was stratified by the participating medical centres in the clinical trial. Upper gastrointestinal endoscopy to identify the source of bleeding was performed on all patients within 12 h of presentation; endoscopic variceal band ligation was performed if the source of gastrointestinal bleeding was believed to be from oesophageal varices—that is, in the setting of active variceal bleeding (defined as visible oozing or spurting of blood from a varix) or in the presence of stigmata of recent bleeding, or if high risk varices (defined below) were present. If band ligation was not feasible because blood obscured the visual field, or if band ligation failed to control bleeding, the bleeding oesophageal varices were treated with injection sclerotherapy. All patients in the study were treated with a 5-day course of antibiotics to cover Gram-negative bacteria.12 14 Norfloxacin by mouth or ciprofloxacin intravenously were the usual antibiotics used. Patients with haemodynamic instability or a significant drop in haemoglobin level (<8 g/dl) were given packed red blood cell (PRBC) transfusions to a haemoglobin of 8 g/dl, and the volume transfused was recorded. In the case of failure to control bleeding within the first 24 h of admission, defined as continued bleeding within 6 h requiring transfusion of ⩾4 units of PRBCs, or the recurrence of haematemesis along with haemodynamic changes, the patient was withdrawn from randomised treatment, the study drug stopped and the patient managed with appropriate clinical care, including the use of TIPS. Re-bleeding was defined according to the Baveno criteria as recurrence of bleeding after the first 24 h and within 5 days of admission for the bleeding episode after initial bleeding control evidenced by new melaena or haematemesis, requirement for ⩾2 units of PRBCs in a 24-h time period, and haemodynamic instability.23 Patients meeting these criteria for re-bleeding were also withdrawn from the study and managed according to standard clinical care. All patients, including those who were withdrawn from the study, were followed up at day 5 and again at 6 weeks after the AVH to assess outcome. After the 5-day period, patients were treated with either β-blockers or endoscopic therapy as prophylaxis against variceal re-bleeding. The primary outcome of interest for the current study protocol was 6-week mortality post-AVH. Secondary outcomes of interest were 5-day mortality post-AVH and risk of variceal re-bleeding within 5 days—that is, the Baveno criteria were used to define the duration of the acute bleeding episode (5 days), and the period of increased mortality related to the acute bleeding episode (6 weeks).2
All patients underwent complete medical assessment upon presentation to the emergency room (designated as time zero), including collection of demographic information, history and physical examination with vital signs, documentation of aetiology of liver disease and presenting clinical symptoms, ascites, and CTP classification (according to the clinical trial protocol, patients were categorised as either class A or B/C). Blood for laboratory testing (complete blood count, serum creatinine, total bilirubin, serum albumin, prothrombin time, international normalised ratio (INR), aspartate and alanine aminotransferase levels (AST and ALT, respectively) and serum sodium) was also drawn before the initial endoscopy. Findings at endoscopy were also documented, including sites of varices, presence of active bleeding or stigmata of recent bleeding or high risk stigmata; and, if endoscopic treatment was performed, the type of therapy applied was recorded. Stigmata of recent haemorrhage or high risk varices were defined as follows: the presence of an adherent clot or white nipple or red signs on varices (cherry red spot, red wale sign or haematocystic spots); or the presence of varices and blood in the stomach; or the presence of large varices in the setting of a patient presenting with haematemesis. Patients were reassessed throughout the study by the investigators at the individual sites, initially at least every 24 h for the first 5 days and then at 6 weeks. MELD scores were calculated according to the following formula:24
We conducted a retrospective analysis of the prospectively collected data from the lanreotide clinical treatment trial for acute variceal bleeding. Continuous data were summarised using the median value and interquartile range (IQR). Patient survival at 5 days and 6 weeks post-AVH was assessed using the Kaplan–Meier method. The log rank test was used to test for differences between strata determined by MELD score.24 The effects of MELD score, CTP class (A vs B/C), liver disease aetiology (categorised as alcohol, viral and other), volume of blood transfused (quantified as 250 ml units of PRBCs), AST and ALT levels (U/l), systolic blood pressure (mm Hg) at the time of presentation, ascites (categorised as present or absent), serum sodium (mmol/l) and findings at the index endoscopy on the risk of mortality at 5 days and 6 weeks post-AVH, and also on the risk of re-bleeding within the first 5 days, were assessed using Cox proportional hazards regression analyses. Patients with all the laboratory parameters required for calculation of the MELD score drawn within the first 24 h of admission were included in the models. The risks estimated from the Cox regression models are expressed as hazard ratios (HRs) with their respective 95% CI. Patients were followed-up until death, liver transplantation, last known follow-up or until 6 weeks post-AVH. The study protocol was approved by the Mayo Clinic Institutional Review Board.
Between December 1998 and March 2002, 368 cirrhotic patients at 16 medical centres in North America hospitalised with features of acute upper gastrointestinal bleeding suspicious for variceal hemorrhage met inclusion criteria and were enrolled in the multicentre phase III clinical trial. Patients underwent randomisation to either lanreotide or placebo prior to endoscopy and within 12 h of admission to the hospital. All patients, except one, submitted to upper endoscopy within the first 12 h of presentation to hospital.
For the purposes of our investigation of the prognostic factors for mortality post-AVH, 108 patients were excluded from the analysis because of bleeding from sources other than oesophageal or gastric varices. Two patients died within the first 24 h of presentation with AVH and were not included in our analyses. Two additional patients were initially enrolled in the clinical trial, but subsequently removed from the study, one due to retraction of consent to participate in the trial and the other for undisclosed reasons. Therefore, the number of patients with cirrhosis presenting with AVH who were analysed in our study was 256. Baseline characteristics of the patient population at presentation to the hospital are given in table 1. The median age of the study group was 51 years (IQR 45–56), with 73% (n = 186) being male. The most common aetiology for cirrhosis was alcoholic liver disease (45%) followed by viral hepatitis (either hepatitis B or C) (41%). Median MELD score at the time of hospital admission was 12 (IQR 8–17), and 75% of patients were CTP class B/C. Of our 256 patients, 208 had the necessary laboratory data required for calculation of MELD available within the first 24 h of admission. The majority of patients (81%) were found to have oesophageal varices alone, with 6% of patients having only gastric varices, and 13% of patients having both oesophageal and gastric varices. Active variceal bleeding at the initial endoscopy was noted in 71 patients (28%). The median haemoglobin level at admission was 9.2 g/dl (IQR 7.7–11.2) and the majority of patients (81%) required blood transfusions during the first 24 h of hospitalisation, with the median number of units of PRBCs transfused being 3 (IQR 2–5).
Thirty-five patients (19%) died within the follow-up period of 6 weeks. Of these deaths, 14 (40%) occurred within the first 5 days of presentation of the AVH. One patient underwent liver transplantation within 48 h of presentation with AVH and was censored at the time of liver transplantation.
Predicting 5-day mortality after AVH
In the univariable analyses, the MELD score, volume of PRBCs transfused within the first 24 h, AST level, presence of ascites, active variceal bleeding at the index endoscopy, clot seen on a varix at the index endoscopy and alcohol-related liver disease were all significantly associated with 5-day mortality after the acute variceal bleeding episode (table 2). Surprisingly, CTP class (A vs B/C) was not predictive of mortality. The limited number of deaths (n = 14) occurring within the first 5 days precluded performance of a meaningful multivariable analysis. A MELD score of <18 or ⩾18 was used to stratify patients based on previous findings that a MELD score of 18 separated out patients undergoing TIPS into lower and higher risk groups. This stratification (higher MELD ⩾18 points; lower MELD <18 points) revealed no significant difference in 5-day post-AVH survival between patients with higher compared with lower MELD scores, (p = 0.2) (fig 1). The c-statistic, reflecting the accuracy of MELD alone in predicting 5-day mortality post-AVH, was 0.63 (95% CI 0.43 to 0.84).
Predicting 6-week mortality after AVH
With respect to predicting survival 6 weeks after AVH, univariable analyses revealed that the MELD score, volume of PRBCs transfused within the first 24 h, AST level, serum sodium, presence of ascites, CTP class and active bleeding seen at the index endoscopy were all significantly associated with 6-week mortality, with MELD score (p<0.001) and volume of PRBCs transfused (p<0.001) being the most significant (table 3). The presence of a clot on the varix fell just short of significance (p = 0.07) as a predictor of 6-week mortality. Stratification of patients according to MELD score (higher MELD ⩾18 points; lower MELD <18 points) revealed a significant increase in 6-week mortality post-AVH between patients with higher compared with lower MELD scores, (p<0.001) (fig 2). The c-statistic, reflecting the accuracy of MELD in predicting 6-week mortality post-AVH, was 0.76 (95% CI 0.65 to 0.88).
Bivariable analyses demonstrated that the MELD score (HR = 1.10, p<0.001) and the volume of PRBCs transfused within the first 24 h (HR = 1.15, p = 0.005) were significantly predictive of 6-week mortality. The accuracy of MELD and volume of blood transfused in the first 24 h in predicting mortality 6 weeks after AVH, as reflected by the c-statistic, was 0.80 (95% CI 0.70 to 0.90). Compared with patients with lower MELD scores (<18) and lower volume of blood transfusion (<4 units of PRBCs), those patients who required a higher volume blood transfusion (⩾4 units of PRBCs) alone, those with higher MELD score (⩾18) alone and those with both higher MELD score (⩾18) and higher volume of blood transfusion (⩾4 units of PRBCs) all had an increased risk of death at 6 weeks, with the latter two being statistically significant (table 4). The CTP class and presence of active bleeding at the index endoscopy were not predictive in the bivariable analyses that also included either MELD or the volume of PRBCs transfused. However, compared with patients with low MELD (<18) without endoscopic evidence of active bleeding, those patients with either a high MELD (⩾18) alone), or both high MELD (⩾18) and endoscopic evidence of active bleeding (HR = 9.9 (95% CI 3.0 to 32.5), p<0.001) had a significantly increased risk of death at 6 weeks (HR = 8.5 (95% CI 3.0 to 24.6), p<0.001), respectively (table 5). Additionally, variceal re-bleeding occurring within the first 5 days after AVH was highly predictive of an increased mortality at 6 weeks (HR = 10.2 (95% CI 3.6 to 28.8), p<0.001).
Predicting variceal re-bleeding within 5 days
Thirty-seven patients (16%) re-bled within the first 5 days of their hospital admission. In the univariable analyses, only the MELD score was significantly associated with the risk of re-bleeding (HR = 1.05 (95% CI 1.01 to 1.08), p = 0.01). The presence of a clot on a varix was just significant as a risk factor for variceal re-bleeding (p = 0.05). Active bleeding at initial endoscopy was not a risk factor for variceal re-bleeding (p = 0.9). Bivariable analysis demonstrated that both MELD and the presence of clot on a varix were predictive of re-bleeding within 5 days (HR for MELD = 1.04 (95% CI 1.002 to 1.07), p = 0.04; HR for clot on a varix = 2.43 (95% CI 1.07 to 5.49), p = 0.03). Stratification of patients according to MELD score (higher MELD ⩾18 points; lower MELD <18 points) revealed a significant increase in the risk of re-bleeding after the initial episode of AVH among patients with higher MELD scores, (p = 0.04) (fig 3).
Currently, there is no well-established model for accurate prediction of survival in patients with cirrhosis following an episode of AVH. One of the difficulties with predicting prognosis in such patients has been that their outcome is influenced not only by the severity of the bleeding episode itself, but also by the severity of the underlying liver disease. With the relatively recent development and validation of the MELD scoring system as an index of severity of liver disease, it is now possible to factor an objective measure of liver disease severity into survival prediction models for AVH.
The major finding of this investigation is the demonstration that MELD is a clinically useful and objective predictor of short-term survival after AVH. We found that MELD is significantly predictive of mortality in patients with cirrhosis who are hospitalised with an acute variceal bleed, with every 1-point increase in the MELD score conferring an 8% and11% increased risk of death at 5 days and 6 weeks, respectively. Our findings are consistent with a previously published letter on the utility of the MELD scoring system in patients with acute variceal bleeding. Chalasani et al, in a retrospective study, found that among 239 patients with cirrhosis presenting with variceal bleeding, the c-statistics for MELD in predicting in-hospital and 1-year mortality were 0.83 (95% CI 0.74 to 0.92) and 0.78 (95% CI 0.69 to 0.87), respectively.25 The study included patients from an era when antibiotic use was not a part of standard therapy for variceal bleeding, and did not quantify the increased risk of mortality in relation to the MELD score. Amitrano et al also demonstrated MELD to be an accurate predictor of 6-week and 3-month survival in the setting of acute variceal bleeding, with c-statistics of 0.80 (95% CI 0.74 to 0.86) and 0.79 (95% CI 0.73 to 0.85), respectively.26 Notably, HCC was present in 31.4% of the 172 patients in that study. Additionally, the study stratified patients by MELD score using a cut-off of 15 points and found that mortality was significantly increased among patients with higher MELD scores. Six-week and 3-month mortality rates were 7.6% and 14.3%, respectively, in patients with MELD ⩽15, while these mortality rates were substantially higher, 43% and 55%, respectively, in patients with MELD >15. In our study, which was based upon prospectively collected data with exclusion of patients with HCC, and which used current standard therapy for AVH including antibiotics, we found that with a MELD score cut-off of 18 points, the 6-week mortality rates were 8% and 46%, respectively, for patients with MELD <18 compared with patients with MELD ⩾18.
Our investigation also demonstrated that aminotransferase levels were not clinically useful predictors of mortality in variceal bleeding, consistent with previous studies that have demonstrated the lack of predictive value of ALT levels in patients with variceal bleeding.27 Additionally, we found no clear association between the aetiology of the liver disease (alcohol vs viral vs other aetiologies) and the risk of mortality from the bleeding episode, although alcohol-related liver disease did appear to be associated with a decreased risk of 5-day mortality in the univariable analyses. Similar to previous studies, we found that the presence of active bleeding at index endoscopy is predictive of both 5-day and 6-week mortality.4 Furthermore, patients presenting with both a high MELD (⩾18) and endoscopic evidence of active bleeding had a 10-fold greater risk of dying within 6 weeks of the acute bleeding episode compared with patients with lower MELD (<18) who had no endoscopic evidence of active bleeding. Unlike the previous studies, however, we did not find a significant association between the presence of active bleeding and the risk of re-bleeding within 5 days. It is possible that the limited number of patients who experienced re-bleeding within 5 days influenced these results. However, re-bleeding occurring within the first 5 days after AVH was a significant predictor of 6-week mortality.
The other major finding of this investigation is a significant association between 6-week mortality after variceal bleeding and the volume of blood transfusion required within the first 24 h of the bleeding episode. Patients with higher MELD scores (⩾18) who required at least 4 units of PRBCs within the first 24 h of hospital admission had an 11-fold increased risk of death, compared with patients with lower MELD scores (<18) who required <4 units of PRBC transfusion. Intuitively, these findings make clinical sense since a patient with a larger, more haemodynamically significant bleed may be expected to require a greater volume blood transfusion, and patients having more severe liver disease, as measured by the MELD score, would be at an even higher risk of mortality in such a setting.
Finally, our study also investigated the prognostic value of MELD in predicting re-bleeding during the first 5 days after presentation of the initial bleeding episode. We found that for every 1-point increase in the MELD score, there was a 5% increased risk of re-bleeding within 5 days of hospital admission and that patients with a higher MELD (⩾18) had a significantly increased risk of re-bleeding compared with patients with a lower MELD (<18).
One of the strengths of this analysis is that the data were extracted from a randomised, double-blind, placebo-controlled trial in which endoscopic intervention was applied in a systematic fashion to all patients, thus reducing the possibility of non-uniform endoscopic treatment patterns confounding the interpretation of the study results and conclusions. The appropriateness of the management of these patients, including the universal use of antibiotics, is reflected in the low mortality rate (19% at 6 weeks). Conversely, since this was a double-blind, placebo-controlled trial, we were unable rigorously to factor the medical treatment applied (ie, whether the patient received placebo vs infusion of the somatostatin analogue, lanreotide) into our modelling scenarios. Since the randomisation sequence for the study drug was stratified by medical centre, it is likely that there were an equal number of patients who received the somatostatin analogue versus placebo across all centres. Although prior studies have shown that early pharmacological treatment of AVH is associated with a decreased risk of active bleeding at index endoscopy, we cannot comment upon any effect that the study drug may have had on the risk of active bleeding or of re-bleeding. We found no statistically significant association between active bleeding at the index endoscopy and the risk of re-bleeding in univariable analysis.
An additional limitation of our study was that relatively few deaths occurred within the first 5 days, which precluded us from performing a more detailed statistical analysis of the prognostic factors affecting 5-day mortality after AVH. Our findings regarding MELD and 5-day mortality should be corroborated by a prospective investigation using a larger patient population. Finally, the clinical trial from which we obtained the data for this investigation did not include physiological measurements of portal hypertension, such as HVPG, in the participants and, as such, we were not able to assess the predictive ability of HVPG in AVH mortality and re-bleeding. However, given that HVPG has been demonstrated previously to be predictive of poor outcomes among patients with variceal bleeding, it would be of interest to investigate the prognostic utility of MELD along with HVPG in the setting of AVH. Additionally, the clinical trial was not designed to collect the parameters necessary for calculation of the Child–Pugh score. Thus, although we found that the CTP class was not predictive of mortality after AVH, the predictive value of the Child–Pugh score in this setting could not be assessed and may warrant further investigation.
In conclusion, this study demonstrates that MELD, which is an accurate and objective measure of liver disease severity, is also a significant and strong predictor of short-term mortality at 5 days and 6 weeks after an AVH. We demonstrate that patients with a high MELD score (⩾18) are at increased risk of death within 6 weeks after an acute variceal bleeding episode and are also at increased risk of re-bleeding within the first 5 days. Additionally, the severity of the variceal bleeding episode, as indicated by the volume of blood transfusion required within the first 24 h, contributes additional prognostic value to the MELD score at 6 weeks. Together these factors allow for early identification of patients with AVH who are at substantially increased risk of death over the short term. These patients may require care in more specialised units during the bleeding episode, and aggressive follow-up in the immediate postvariceal bleed setting. Such patients would also probably benefit from early referral to a liver transplant centre for consideration of candidacy for transplantation and expedited evaluation. A MELD score of <18 and ⩾18 may be used in future studies instead of CTP class to stratify patients into low risk and high risk for 6-week mortality and re-bleeding.
Competing interests: None.
Ethics approval: The study protocol was approved by the Mayo Clinic Institutional Review Board.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.