INTRODUCTION
Upper gastrointestinal haemorrhage (UGIH) related to portal hypertension (PH) is one of the major complications of cirrhosis, occurring at a rate of 10-15% per year according to the American Association for the Study of Liver Diseases (AASLD) (1).It represents a turning point in the course of cirrhosis and is a major cause of mortality, with a recurrence and an early mortality rates of 7-20% and 10-20% respectively at six weeks, despite a codified management (2–4). Rupture of oesophageal or gastric varices represents the main cause of upper gastrointestinal haemorrhage due to portal hypertension. The management is based on medical measures associated to endoscopic procedures.
However, the challenge of this management highlights the importance of stratifying high-risk patients to guide therapeutic options. Several prognostic scores have been developed to assess the severity of cirrhosis, such as Child-Pugh, MELD score (Model of End stage Liver Disease) and ALBI (albumin- bilirubin) (5), or to stratify the risk of gastroduodenal ulcer bleeding, such as the Rockall and Glasgow-Blatchford scores (6). However, their use in portal hypertension haemorrhage remains limited, as there is no consensus on an appropriate score for this condition.
Recently, the Cirrhosis Acute Gastro-Intestinal Bleeding (CAGIB) score, developed by a Chinese team, has been proposed as a specific model for PH haemorrhage (7). However, its performance has not yet been validated on a large scale. The aim of our study was to evaluate the performance of the CAGIB score in a Tunisian cohort of cirrhotic patients presenting with upper gastrointestinal bleeding due to PH.
METHODS
Characteristics of the study
We conducted a single-centre, retrospective and descriptive study in the gastroenterology department of Mohamed Taher Maamouri Hospital in Nabeul including the records of patients from January 2012 to December 2022.
Study population
Inclusion criteria: We included cirrhotic patients having had an episode of UGIH due to PH. The diagnosis of cirrhosis was based on compatible clinical, laboratory, and radiologic evidence. Endoscopic diagnosis of PH haemorrhage was established in the presence of active bleeding from a varix, a platelet plug or ulcer on a varix, or in case of finding blood in the upper digestive tract with oesophageal or gastric varices, in the absence of other potential sources.
Cirrhotic patients were also included if they presented with hematemesis, melena, or rectal bleeding (with or without deglobulisation), or isolated deglobulisation and the absence of any other lesion that could explain the bleeding episode.
Non-inclusion criteria: Patients with UGIH related to non-cirrhotic PH or to lesions not associated with PH (as peptic ulcer disease or bleeding oesophagitis) confirmed by upper gastrointestinal endoscopy were not included.
Exclusion criteria: Patients with missing data, patients who were lost to follow-up or whose follow-up was less than two months were excluded from the study.
Data collection: Data were collected retrospectively from patients' medical records including: • General data (age, sex, lifestyle habits such as smoking, alcohol or drug abuse, medical history) and the characteristics of the cirrhosis, such as the length of illness, the aetiology and the complications. • The initial assessment of PH: - presence or not of oesophageal varices (OV) and gastric varices (GV) and their classification according to the established classifications (New Italian Endoscopic Club (NIEC) for OV and Sarin for GV) (8,9). - Presence or absence of hypertensive gastropathy. • Previous prophylactic and therapeutic procedures carried out for the management of UGIH due to PH were recorded. • The clinical, biological and endoscopic characteristics of the current UGIH episodes and their therapeutic management according to Baveno VII consensus (10) were also detailed.
Calculation of the scores: The following prognostic scores were calculated for each patient: Child-Pugh, MELD score (Model of End stage Liver Disease) (and its variants MELD-Na and MELD 3.0), ALBI, ROCKALL, Glasgow Blatchford, AIMS 65, neutrophil to lymphocyte ratio (NLR) and the CAGIB score. The ROCKALL, Glasgow Blatchford and AIMS 65 scores were calculated using the calculator available on MDCalc, including several variables: initial heart rate, systolic blood pressure, presentation with melena or syncope, level of haemoglobin, blood urea nitrogen, coexistent hepatic disease, heart failure or other significant comorbidities, age and the results of endoscopic examination. The MELD score and its two variants, the ALBI score and the CAGIB score were calculated using the following formulae: MELD = 3.78 × ln[serum bilirubin (mg/dL)] + 11.2 × ln[INR] + 9.57 × ln[serum creatinine (mg/dL)] + 6.43 (11).
MELD-Na = MELD + 1.59 × (135 – serum sodium [mmol/L]) (12). MELD 3.0=1.33 × (female sex) + 4.56 × ln[serum bilirubin (mg/dL)] + 0.82 × (137 – serum sodium [mmol/L]) – 0.24 × (137 – serum sodium [mmol/L]) × ln[serum bilirubin (mg/dL)] + 9.09 × ln[INR] + 11.14 × ln[serum creatinine (mg/dL)] + 1.85 × (serum albumin [g/dL] – 3.5) – 1.83 × (serum albumin [g/dL] – 3.5) × ln[serum creatinine (mg/
Ben Mohamed & al. CAGIB score in Gastrointestinal bleeding due to portal hypertension dL)] + 6 (13). For the MELD, MELD Na and MELD 3.0 scores, we set a threshold of 20 to divide the patients into two groups (14).
ALBI score= (log10 bilirubin [µmol/L] × 0.66) + (albumin [g/L] × −0.0852). As a result, ALBI grades 1, 2, and 3 were developed as follows: ALBI score ≤ −2.60 (ALBI grade 1), > −2.60 to ≤ −1.39 (ALBI grade 2), and > −1.39 (ALBI grade 3). CAGIB score = diabetes (yes = 1, no = 0) × 1.040 + Hepatocellular carcinoma (yes = 1, no = 0) × 0.974 + Total bilirubin (μmol/L) × 0.005 - ALB (g/L) × 0.091 + ALT (U/L) × 0.001 + creatinine (μmol/L) × 0.012 - 3.964.
Progression after the UGIH episode: Patient outcomes were monitored by noting the failure of first-line treatment, the early recurrence (within five days), the short-term recurrence (within six weeks), and the early mortality (within six weeks). These criteria were used to assess the performance of the prognostic scores.
Statistical study
Data were entered and analysed using SPSS version 22 software. Qualitative variables were expressed as simple and relative frequencies, while quantitative variables were presented as means, standard deviations and extreme values. Survival curves were generated using the Kaplan-Meier model and compared using the log- rank test.
Score performance was analysed using receiver operating characteristic (ROC) curves and the Mann- Whitney test, with statistical significance set at p<0.05.
Ethical considerations
Strict anonymity of the individual data was respected throughout the study. Given the retrospective nature of the study, it was not possible to obtain informed consent from the patients. There were no conflicts of interest in this work.
RESULTS
Study population
General patients’ characteristics The study included 105 cirrhotic patients having had an episode of upper gastrointestinal bleeding secondary to PH. The mean age of patients was 63.06 ± 13.32 years [19-89]. The F/M ratio was 1.05.
Alcohol consumption was reported in 8.6% of patients, with a median weekly consumption of 25g.
Characteristics of the cirrhosis The mean length of illness was 4 years [0-18]. The main identified aetiologias were hepatitis C (26.7%), hepatitis B (17.1%) and metabolic dysfunction-associated fatty liver disease (11.4%). Fifty-one patients (48.6%) had one or more comorbidities associated with cirrhosis mainly represented by diabetes in 33.3% of cases (n=35) and hypertension in 31.4% of cases (n=33).
In our cohort, 9.5% of patients (n=10 cases) had hepatocellular carcinoma. Hepatocellular carcinoma was classified as BCLC A in one patient, BCLC B in one patient, BCLC C in 3 patients and BCLC D in 5 patients. No patient had chronic hepatic encephalopathy or hepatorenal syndrome.
Fifty-three patients (50.5%) had a previous history of oedematoascitic decompensation. At the time of enrolment, 45.7% of patients (n=48) were known to have endoscopic signs of portal hypertension. All these patients had oesophageal varices classified as grade II in 54.2% of cases (n=26) and grade III in 41.7% of cases (n=20).
Two patients had OV classified as grade I with no red signs. Among these population, eight patients had gastric varices (16.7%) classified as GOV 1 in 7 cases and GOV 2 in only one case. Primary prophylaxis against variceal bleeding was initiated in 83.3% (n=40) of patients with known uncomplicated varices.
It was based on beta-blockers in 35 patients (propranolol with a mean dose of 52 mg per day [40-160]). Endoscopic band ligation for OV as primary prophylaxis was carried out in 5 cases due to contraindication to beta-blockers without achieving eradication of varices. Of the remaining patients who did not receive primary prophylaxis (n=8), six patients had grade II or III OV with a contraindication to non-selective beta-blockers.
They were candidates for oesophageal band ligation, which was not performed.
History of gastrointestinal variceal haemorrhage: Eight patients (7.61%) had one or more episodes of gastrointestinal haemorrhage due to rupture of oesophageal varices. Endoscopic treatment was performed in all of these patients and the varices’ eradication was not achieved in any case. Six patients were on secondary beta-blocker prophylaxis with propranolol with a mean dose of 68 mg per day [40-160].
Characteristics of the episodes of upper gastrointestinal haemorrhage
Clinical and biological presentation: In 92.4% of patients, the current episode represented a first variceal haemorrhage, inaugural of the cirrhosis in 46.6% of cases. Haematemesis associated with melena was the most common presentation (51.4%), followed by isolated haematemesis (37.2%). The median consultation time after haemorrhage was 25.2 hours ±41,32 [1- 72].
On the other hand, it was a bleeding recurrence in 7.61% of cases (n=8) with a mean delay of 8.37±7.78 months [2-21] from the last episode of UGIH. Only seven patients (6.7%) were included with haemorrhagic shock on admission. Eight cases (7.6%) of renal failure were noted in our patients in the acute phase.
We observed ascites in 44 patients (41.9%) associated with lower limb oedema in 30 patients (28.6%). Forty patients (38%) had a haemoglobin level ≤8g/dl. Volemic expansion with crystalloids or macromolecules was indicated in all patients.
Catecholamines were required in 8 patients (7.6%). Forty-five patients (42.9%) were transfused with packed red blood cells with an average number at 1.9
[1-6]. All patients received octreotide by electric syringe pump for a period of 5 days and were given preventive antibiotics such as ofloxacin 200 mg x 2/day or ceftriaxone 1 g/day. A combination of lactulose and rifaximin was used in 30 patients (28.6%), for curative purposes in 19 cases (18.1%) and as a preventive measure in 11 cases (10.5%).
Management and endoscopic data: Endoscopic exploration was performed in all patients, with a mean delay of 29.8 ± 42.4 hours [1-144]. Stigmata of recent bleeding were present in 58% of patients. Ruptured oesophageal varices (VO) were the dominant aetiology (94.3%), followed by gastric varices (17.1%).
Endoscopic management was performed in 96 patients (91.4%) with a delay of 31.6 ± 40.3 hours [12-192]. According to gastrointestinal exploration, 85.7% (n=90) of patients underwent endoscopic band ligation and one patient had sclerotherapy injection of the OV. Gastric varices obturation with glue injection was performed in 5.8% of patients (six patients).
No complications were reported. Nine patients (8.6%) did not receive endoscopic treatment due to a lack of technical limitations: eight with acute esophageal variceal bleeding, in whom band ligation was not possible due to unavailability of band ligation devices, and one with type II gastroesophageal varices, in whom cyanoacrylate injection (Glubran®) could not be administered.
The main outcomes: The control of UGIH was not possible in 3.8% of patients (n=4) despite medical and endoscopic management complicated by instant death. An early recurrence of bleeding was noted in 4 patients with an average delay of 72 hours [10-120]. Two cases were fatal within 24 hours.
Short-term recurrences were observed in 21 patients (20%) with a mean time to onset equal to 18 days and extremes ranging from 10 hours to 6 weeks. A second endoscopic haemostasis procedure was performed in 48.8% of them (n=9). Following the acute episode, 77 of the 91 surviving patients were started on β-blockers for secondary UGIH prophylaxis.
It was propranolol in all cases with a mean dose of 80mg [40-160mg]. Eighty-three patients underwent secondary prophylaxis with endoscopic band ligation. The mean time between sessions was 25 days [21-128 days] with a mean number of sessions of 2 [0-4].
Secondary prophylactic glue injection was performed in two patients. Fourteen cases of death were recorded in our study during the first 6 weeks following the haemorrhagic episode, representing a mortality rate of 13,3%. The median time to death was 11 days [2-30].
Survival rates at 1, 2 and 3 years were 91.3%, 85% and 80.2% respectively.
Performance of prognostic scores
Prediction of early recurrence: The CAGIB score showed an excellent performance in predicting early recurrence, with an AUROC of 0.995 (Table 1). When looking at the parameters constituting the
CAGIB score, the presence of hepatocellular carcinoma (HCC) was strongly associated with early recurrence (20% in case of early recurrence (group 1) versus 2.1% in the group without early recurrence (group 2) (p = 0.005) and as well as high total bilirubin levels (182.35 UI/L in group 1 versus 42.55 UI/L in group 2 p = 0.007, respectively).
Prediction of short-term recurrence Concerning the recurrence at six weeks, the CAGIB score did not show significant performance (AUC = 0.625, p = 0.078) (Table 2).
Prediction of mortality: The CAGIB score had an excellent performance with an AUC of 0.891, a sensitivity and a specificity of 85.7% and 83.5% respectively, in predicting mortality from values greater than or equal to -4.648 with an OR equal to 18. By studying the components of the CAGIB score separately, we found that mortality was significantly higher in cases of HCC (80% versus 6.3%; p<0.001) and significantly associated with higher levels of ALT and total Bilirubin (77.5 UI/L versus 29 UI/L (p=0.001) and 45.3 UI/L versus 19.6 UI/L (p=0.001) respectively, and with lower levels of albumin (30.8 g/L versus 33.4 g/L ; p=0.031). Comparing to the other scores, the CAGIB score was the best predictor of mortality (AUC=0.891).
The AUROC for the different scores are summarised in table 3.
Survival, calculated by the Long Rank test, for the different grades of the CAGIB score (< - 4.6646 and ≥ - 4.6646) was 115.402 and 22.785 months respectively, with a p<0,001 (Figure 1).
Table 4 summarises the results of mean survival according to the different prognostic scores.
DISCUSSION
The purpose of this study was to evaluate the performance of the CAGIB score in the prediction of early recurrence and mortality during a variceal upper GI bleeding episode and according to our results, we could conclude its superiority comparing to the previous scores. In fact,Bai et al (7) conducted a multicentre study in China including cirrhotic patients with upper GI haemorrhage over nine years. A total of 1682 patients were enrolled.
Multivariate logistic regression analysis demonstrated that HCC, diabetes, total bilirubin, albumin, ALT and creatinine were independent predictors of early mortality. A prognostic model called CAGIB was developed using a formula combining each of the parameters found with coefficients. In our study, we studied the six clinical and biological parameters included in the CAGIB score.
The presence of HCC was significantly associated with mortality. In fact, HCC is one of the most frequent causes of cancer- related death (15) and 80% of HCCs are diagnosed in cirrhosis (16).HCC can worsen portal pressure either by compression or by tumour thrombosis and is therefore considered to be an independent factor of recurrence and early mortality in cirrhotic patients with UGIH, according to several studies (17–21).Among the biological parameters studied (total bilirubin, albumin, creatinine, ALT), a high level of ALT and bilirubin were significantly associated with early mortality. Bilirubin levels are included in various prognostic scores for cirrhosis, as the Child Pugh and MELD scores.
It represents also an indicator of liver failure and was associated in several studies to early recurrence and mortality (17,22– 27).Furthermore, during UGIH related to PH, the acute blood loss leads to a liver ischaemia (28).Hypoxic hepatitis, characterised by a quick rise in serum transaminases due to hepatic cell necrosis (29) can be observed in acute GI haemorrhage in cirrhotic patients (21) and worsen the patient's prognosis (30).In the studies by Bai et al (7), Zhao et al (31)and D'amico et al (17), it was noted that the ALT level was an independent factor of mortality. Moreover, according to the literature (7,26,31), elevated creatinine is an independent risk factor for mortality. We conducted literature research on the performance of the CAGIB score in predicting recurrence and mortality.
Few studies were found. The results are summarised in table 5.
Ben Mohamed & al. CAGIB score in Gastrointestinal bleeding due to portal hypertension Our study has a number of advantages. It is the first work carried out in Tunisia to study the prognostic performance of the CAGIB score in UGIH related to PH.
The management of this complication is standardised in our hospital, which avoids the risk of bias. Nevertheless, there are a few limitations to our study that need to be acknowledged. Firstly, this is a retrospective and a single-center study; the lack of data from multicenter may cause potential bias.
Furthermore, the population included was heterogeneous regarding aetiology as well as cirrhosis characteristics, and patients were followed up for inconsistent lengths of time.
CONCLUSION
Upper gastrointestinal haemorrhage related to portal hypertension is a serious complication in cirrhotic patients that requires well-controlled management and risk stratification to reduce recurrence and mortality. According to our results, we can propose the CAGIB score as an easy to calculate tool, including objective clinical and biological parameters, for predicting early recurrence and mortality with a threshold greater than -4.648. However, further large-scale studies are needed to validate this score.