A simplified prognostic model to predict mortality in patients with acute variceal bleeding

Comparison of 1-day versus 3-day intravenous terlipressin in cirrhosis patients with variceal bleeding: A pilot randomised controlled trial

BACKGROUND

In cirrhosis patients with acute variceal bleeding (AVB), the optimal duration of vasoconstrictor therapy after endoscopic haemostasis is unclear.

AIMS

We aimed to compare efficacy of 1-day versus 3-day terlipressin therapy in cirrhosis patients with AVB post-endoscopic intervention. The primary objective was to compare rebleeding at 5 days between the two arms. Secondary objectives included rebleeding and mortality rates at 6 weeks.

METHODS

In this open-label, randomised controlled trial, cirrhosis patients with AVB were randomised to either 1-day or 3-day terlipressin therapy.

RESULTS

A total of 150 cirrhosis patients with AVB were recruited to receive either 1 day (n = 75) or 3 days (n = 75) of terlipressin therapy. One patient from 1-day arm was excluded. Modified intention-to-treat analysis included 149 patients. Baseline characteristics were comparable between the two groups. Rebleeding at 5 days: 3 (4.1%; 95% confidence interval [CI]: 0.4-9.0) versus 4 (5.3%; 95% CI: 2.0-10.0), risk difference (RD) p = 0.726 and 5-day mortality rates: 1 (1.4%; 95% CI: 0-7.3) versus 1 (1.3%; 95% CI: 0.2-7.0), RD p = 0.960 were similar. Rebleeding at 42 days: 9 (12.2%; 95% CI: 7.0-20.0) versus 10 (13.3%; 95% CI: 7.0-20.0), RD p = 0.842 and mortality at 42 days: 5 (6.8%; 95% CI: 3.0-10.0) versus 4 (5.3%; 95% CI: 2.0-10.0), RD p = 0.704 were also similar. Patients in the 1-day terlipressin therapy arm experienced significantly fewer adverse effects compared with those receiving 3 days of terlipressin therapy: 28 (37.8%) versus 42 (56%), p = 0.026.

CONCLUSIONS

Our results suggest that 1 day of terlipressin therapy is associated with similar 5-day and 42-day rebleeding rates, 42-day mortality and an overall superior safety profile compared with 3-day of terlipressin therapy. These findings require to be validated in double-blinded, larger, multiethnic and multicentre studies across the various stages of cirrhosis (CTRI/2019/10/021771).

Predicting the risk of variceal rehemorrhage in cirrhotic patients with portal vein thrombosis: A two-center retrospective study

OBJECTIVES

Although portal vein thrombosis (PVT) was thought to deteriorate portal hypertension and contribute to poor prognosis, risk stratification remains unclear. This study aimed to evaluate its effect on the risk of variceal rehemorrhage and to develop a competitive risk model in cirrhotic patients with PVT.

METHODS

Cirrhotic patients with and without PVT admitted for acute variceal hemorrhage were retrospectively included after matching (1:1) for age, gender and etiology of cirrhosis from two tertiary centers with 1-year follow-up. Those with PVT were subsequently divided into the training and validation cohorts. Cox regression analysis was performed to identify risk factors and develop a competitive risk model, of which the predictive performance and optimal decision threshold were evaluated by C-index, competitive risk curves, calibration curves and decision curve analysis.

RESULTS

Among 398 patients, PVT significantly increased the variceal rehemorrhage risk. Multivariate Cox regression analysis identified that the Child-Turcotte-Pugh score (P = 0.013), chronic PVT (P = 0.025), C-reactive protein (P < 0.001), and aspartate aminotransferase (P = 0.039) were independently associated with variceal rehemorrhage, which were incorporated into the competitive risk model, with high C-index (0.804 and 0.742 of the training and validation cohorts, respectively), risk stratification ability, and consistency. The optimal decision range of the threshold probability was 0.2-1.0.

CONCLUSION

We confirmed the adverse effect of PVT on variceal rehemorrhage and developed a competitive risk model for variceal rehemorrhage in cirrhotic patients with PVT, which might be conveniently used for clinical decision-making.

Pediatric liver cirrhosis interventional procedures: from biopsy to transjugular intrahepatic portosystemic shunt

Cirrhosis is a complex diffuse process whereby the architecture of the liver is replaced by abnormal nodules because of the presence of fibrosis. Several pediatric diseases such as extrahepatic portal vein obstruction, biliary atresia, alpha-1-antitrypsin deficit and autoimmune hepatitis can lead to cirrhosis and portal hypertension in children. In this article the authors describe interventional radiology procedures that can facilitate the diagnosis and treatment of diseases associated with liver cirrhosis and portal hypertension in the pediatric population. These procedures include image-guided liver biopsy, mesenteric-intrahepatic left portal vein shunts, balloon-occluded retrograde transvenous obliteration, transjugular intrahepatic portosystemic shunts and splenic embolization.

Endotoxin Translocation and Gut Barrier Dysfunction Are Related to Variceal Bleeding in Patients With Liver Cirrhosis

BACKGROUND

Bacterial infections are associated with the risk of variceal bleeding through complex pathophysiologic pathways.

OBJECTIVES

The primary objective of the present case-control study was to investigate the role of bacterial translocation and intestinal barrier dysfunction in the pathogenesis of variceal bleeding. A secondary objective was to determine independent predictors of key outcomes in variceal bleeding, including bleeding-related mortality.

METHODS

Eighty-four (n = 84) consecutive patients participated in the study, 41 patients with acute variceal bleeding and 43 patients with stable cirrhosis, and were followed up for 6 weeks. Peripheral blood samples were collected at patient admission and before any therapeutic intervention.

RESULTS

Child-Pugh (CP) score (OR: 1.868; p = 0.044), IgM anti-endotoxin antibody levels (OR: 0.954; p = 0.016) and TGF-β levels (OR: 0.377; p = 0.026) were found to be significant predictors of variceal bleeding. Regression analysis revealed that albumin (OR: 0.0311; p = 0.023), CRP (OR: 3.234; p = 0.034) and FABP2 levels (OR:1.000, p = 0.040), CP score (OR: 2.504; p = 0.016), CP creatinine score (OR: 2.366; p = 0.008), end-stage liver disease model (MELD), Na (OR: 1.283; p = 0.033), portal vein thrombosis (OR: 0.075; p = 0.008), hepatocellular carcinoma (OR: 0.060; p = 0.003) and encephalopathy (OR: 0.179; p = 0.045) were significantly associated with 6-week mortality.

CONCLUSIONS

Bacterial translocation and gut barrier impairment are directly related to the risk of variceal bleeding. Microbiota-modulating interventions and anti-endotoxin agents may be promising strategies to prevent variceal bleeding.

Development and validation of prognostic model to predict mortality among cirrhotic patients with acute variceal bleeding: A retrospective study

Background and Aim

Acute variceal bleeding (AVB) is a serious complication associated with high mortality. The aim of our study was to investigate mortality predictors and to develop a new simplified prognostic model among cirrhotic patients with AVB.

Methods

A simplified prognostic model was developed using multiple logistic regression after identifying significant predictors of 6‐week mortality.

Results

A total of 713 consecutive patients with AVB were enrolled. The 6‐week overall mortality rate was 18%. Multivariate analysis showed that shock, model for end‐stage liver disease (MELD) score, high‐risk stigmata of esophageal varices on endoscopic finding, and Glasgow Blatchford score were independent predictors of mortality. A new logistic model using these variables was developed. This model (cutoff value ≥ 4) area under the receiver operating characteristics (AUROC) was 0.93 and significantly higher than that of MELD score alone (0.74). Two validation analyses showed that the AUROC of our model was consistently high. The 6‐week rebleeding rate was 25.3%. Multivariate analysis showed that MELD score, Glasgow Blatchford score, history of upper GI bleeding, shock, and alcohol use were independent predictors of rebleeding.

Conclusion

Our new simplified model accurately and consistently predicted 6‐week mortality among patients with AVB using objective variables measured at admission. Patients with higher MELD scores should be closely monitored due to the higher probability of 6‐week rebleeding.

Risk stratification in acute variceal bleeding: Far from an ideal score

OBJECTIVES

Acute variceal bleeding (AVB) results from rupture of esophageal or gastric varices. It is a life-threatening complication of portal hypertension. Nevertheless, it remains unclear how to predict adverse outcomes and identify high-risk patients. In variceal hemorrhage, high Child-Turcotte-Pugh (Child) and Model for End-stage Liver Disease (MELD) scores are associated with a worse prognosis. The Rockall system (Rockall), Glasgow-Blatchford (Blatchford), and AIMS65 scores have been validated for risk stratification for nonvariceal upper gastrointestinal bleeding; however, their use is controversial in AVB. The aim of this study was to compare the performance of Child, MELD, Rockall, Blatchford, and AIMS65 scores in risk stratification for rebleeding and/or mortality associated with AVB.

METHODS

This retrospective study was conducted at a tertiary care hospital over 42 months. The outcomes were 6-week rebleeding and mortality. The AUROC was calculated for each score (1-0.9, 0.9-0.8, and 0.8-0.7, indicating excellent, good, and acceptable predictive power, respectively).

RESULTS

In total, 222 patients were included. Six-week rebleeding and mortality rates were 14% and 18.5%, respectively. No score was useful for discriminating patients at a higher risk of rebleeding. The AUROCs were 0.59, 0.57, 0.61, 0.63, and 0.56 for Rockall, Blatchford, AIMS65, Child, and MELD scores, respectively. Prediction of 6-week mortality based on Rockall (AUROC 0.65), Blatchford (AUROC=0.60), and AIMS65 (AUROC=0.67) scores were also not considered acceptable. The AUROCs for predicting mortality were acceptable for Child and MELD scores (0.72 and 0.74, respectively).

CONCLUSION

Rockall, Blatchford, and AIMS65 scores are not useful for predicting 6-week rebleeding or mortality in patients with AVB. Child and MELD scores can identify patients at higher risk for 6-week mortality but not for 6-week rebleeding.

A partial imputation EM-algorithm to adjust the overestimated shape parameter of the Weibull distribution fitted to the clinical time-to-event data

BACKGROUND AND OBJECTIVES

Typical clinical data can suffer routine information loss when event times are rounded to the nearest day and right-censored at the end of follow-up. Because of the daily basis recording system, for the first 24 h, there are no events, which can damage the estimation of the Weibull survival model. Its estimation bias is inevitable since, for this short period, massive events might have occurred, the data is missing, and the fitted Weibull model is to show a steep slope. This phenomenon of estimation bias caused by the information loss caused by the problem of measurement resolution has not been properly discussed so far.

METHODS

We propose a partial imputation Expectation Maximization (PIEM)-algorithm to estimate missing lifetimes only for day 1 at the mode among the whole clinical follow-up days. Based on various Weibull distributions, we simulated clinical sets after rounding and censoring raw event times and prepared chimera sets by partially substituting the imputed lifetimes only for the 24 h at the mode among the entire clinical sets.

RESULTS

For shape parameter ≤ 1, almost all the 95% prediction intervals (PIs) of both parameters in the chimera sets include their true values, while those in the clinical sets miss most of the true shape parameters and some of the true scale parameters. Estimating a small proportion of missing data only for the 24 h period, while keeping the rest as they are, greatly reduces biases of both scale and shape parameters. For shape parameter >1, the chimera sets consistently outperform the clinical sets.

CONCLUSIONS

The PIEM-algorithm may be applied as an intuitive tool for time-to-event modeling of survival data with this kind of information loss.

Prediction of Adverse Events in Stable Non-Variceal Gastrointestinal Bleeding Using Machine Learning

Clinical risk-scoring systems are important for identifying patients with upper gastrointestinal bleeding (UGIB) who are at a high risk of hemodynamic instability. We developed an algorithm that predicts adverse events in patients with initially stable non-variceal UGIB using machine learning (ML). Using prospective observational registry, 1439 out of 3363 consecutive patients were enrolled. Primary outcomes included adverse events such as mortality, hypotension, and rebleeding within 7 days. Four machine learning algorithms, namely, logistic regression with regularization (LR), random forest classifier (RF), gradient boosting classifier (GB), and voting classifier (VC), were compared with the Glasgow-Blatchford score (GBS) and Rockall scores. The RF model showed the highest accuracies and significant improvement over conventional methods for predicting mortality (area under the curve: RF 0.917 vs. GBS 0.710), but the performance of the VC model was best in hypotension (VC 0.757 vs. GBS 0.668) and rebleeding within 7 days (VC 0.733 vs. GBS 0.694). Clinically significant variables including blood urea nitrogen, albumin, hemoglobin, platelet, prothrombin time, age, and lactate were identified by the global feature importance analysis. These results suggest that ML models will be useful early predictive tools for identifying high-risk patients with initially stable non-variceal UGIB admitted at an emergency department.

Validation of a Machine Learning Model That Outperforms Clinical Risk Scoring Systems for Upper Gastrointestinal Bleeding

BACKGROUND & AIMS

Scoring systems are suboptimal for determining risk in patients with upper gastrointestinal bleeding (UGIB); these might be improved by a machine learning model. We used machine learning to develop a model to calculate the risk of hospital-based intervention or death in patients with UGIB and compared its performance with other scoring systems.

METHODS

We analyzed data collected from consecutive unselected patients with UGIB from medical centers in 4 countries (the United States, Scotland, England, and Denmark; n = 1958) from March 2014 through March 2015. We used the data to derive and internally validate a gradient-boosting machine learning model to identify patients who met a composite endpoint of hospital-based intervention (transfusion or hemostatic intervention) or death within 30 days. We compared the performance of the machine learning prediction model with validated pre-endoscopic clinical risk scoring systems (the Glasgow-Blatchford score [GBS], admission Rockall score, and AIMS65). We externally validated the machine learning model using data from 2 Asia-Pacific sites (Singapore and New Zealand; n = 399). Performance was measured by area under receiver operating characteristic curve (AUC) analysis.

RESULTS

The machine learning model identified patients who met the composite endpoint with an AUC of 0.91 in the internal validation set; the clinical scoring systems identified patients who met the composite endpoint with AUC values of 0.88 for the GBS (P = .001), 0.73 for Rockall score (P < .001), and 0.78 for AIMS65 score (P < .001). In the external validation cohort, the machine learning model identified patients who met the composite endpoint with an AUC of 0.90, the GBS with an AUC of 0.87 (P = .004), the Rockall score with an AUC of 0.66 (P < .001), and the AIMS65 with an AUC of 0.64 (P < .001). At cutoff scores at which the machine learning model and GBS identified patients who met the composite endpoint with 100% sensitivity, the specificity values were 26% with the machine learning model versus 12% with GBS (P < .001).

CONCLUSIONS

We developed a machine learning model that identifies patients with UGIB who met a composite endpoint of hospital-based intervention or death within 30 days with a greater AUC and higher levels of specificity, at 100% sensitivity, than validated clinical risk scoring systems. This model could increase identification of low-risk patients who can be safely discharged from the emergency department for outpatient management.

Validation of a Machine Learning Model That Outperforms Clinical Risk Scoring Systems for Upper Gastrointestinal Bleeding

BACKGROUND & AIMS

Scoring systems are suboptimal for determining risk in patients with upper gastrointestinal bleeding (UGIB); these might be improved by a machine learning model. We used machine learning to develop a model to calculate the risk of hospital-based intervention or death in patients with UGIB and compared its performance with other scoring systems.

METHODS

We analyzed data collected from consecutive unselected patients with UGIB from medical centers in 4 countries (the United States, Scotland, England, and Denmark; n = 1958) from March 2014 through March 2015. We used the data to derive and internally validate a gradient-boosting machine learning model to identify patients who met a composite endpoint of hospital-based intervention (transfusion or hemostatic intervention) or death within 30 days. We compared the performance of the machine learning prediction model with validated pre-endoscopic clinical risk scoring systems (the Glasgow-Blatchford score [GBS], admission Rockall score, and AIMS65). We externally validated the machine learning model using data from 2 Asia-Pacific sites (Singapore and New Zealand; n = 399). Performance was measured by area under receiver operating characteristic curve (AUC) analysis.

RESULTS

The machine learning model identified patients who met the composite endpoint with an AUC of 0.91 in the internal validation set; the clinical scoring systems identified patients who met the composite endpoint with AUC values of 0.88 for the GBS (P = .001), 0.73 for Rockall score (P < .001), and 0.78 for AIMS65 score (P < .001). In the external validation cohort, the machine learning model identified patients who met the composite endpoint with an AUC of 0.90, the GBS with an AUC of 0.87 (P = .004), the Rockall score with an AUC of 0.66 (P < .001), and the AIMS65 with an AUC of 0.64 (P < .001). At cutoff scores at which the machine learning model and GBS identified patients who met the composite endpoint with 100% sensitivity, the specificity values were 26% with the machine learning model versus 12% with GBS (P < .001).

CONCLUSIONS

We developed a machine learning model that identifies patients with UGIB who met a composite endpoint of hospital-based intervention or death within 30 days with a greater AUC and higher levels of specificity, at 100% sensitivity, than validated clinical risk scoring systems. This model could increase identification of low-risk patients who can be safely discharged from the emergency department for outpatient management.

Machine Learning to Predict Outcomes in Patients with Acute Gastrointestinal Bleeding: A Systematic Review

Risk stratification of patients with gastrointestinal bleeding (GIB) is recommended, but current risk assessment tools have variable performance. Machine learning (ML) has promise to improve risk assessment. We performed a systematic review to evaluate studies utilizing ML techniques for GIB. Bibliographic databases and conference abstracts were searched for studies with a population of overt GIB that used an ML algorithm with outcomes of mortality, rebleeding, hemostatic intervention, and/or hospital stay. Two independent reviewers screened titles and abstracts, reviewed full-text studies, and extracted data from included studies. Risk of bias was assessed with an adapted Quality in Prognosis Studies tool. Area under receiver operating characteristic curves (AUCs) were the primary assessment of performance with AUC ≥ 0.80 predefined as an acceptable threshold of good performance. Fourteen studies with 30 assessments of ML models met inclusion criteria. No study had low risk of bias. Median AUC reported in validation datasets for predefined outcomes of mortality, intervention, or rebleeding was 0.84 (range 0.40-0.98). AUCs were higher with artificial neural networks (median 0.93, range 0.78-0.98) than other ML models (0.81, range 0.40-0.92). ML performed better than clinical risk scores (Glasgow-Blatchford, Rockall, Child-Pugh, MELD) for mortality in upper GIB. Limitations include heterogeneity of ML models, inconsistent comparisons of ML models with clinical risk scores, and high risk of bias. ML generally provided good-excellent prognostic performance in patients with GIB, and artificial neural networks tended to outperform other ML models. ML was better than clinical risk scores for mortality in upper GIB.

Development and Validation of a Novel Model for Outcomes in Patients with Cirrhosis and Acute Variceal Bleeding

BACKGROUND

Acute variceal bleeding (AVB) in patients with cirrhosis is associated with high mortality, ranging from 12 to 20% at 6 weeks. The existing prognostic models for AVB lack precision and require further validation.

AIM

In this prospective study, we aimed to develop and validate a new prognostic model for AVB, and compared it with the existing models.

METHODS

We included 285 patients from March 2017 to November 2017 in the derivation cohort and 238 patients from December 2017 to June 2018 in the validation cohort. Two prognostic models were developed from derivation cohort by logistic regression analysis. Discrimination was assessed using area under the receiver operator characteristic curve (AUROC).

RESULTS

The 6-week mortality was 22.1% in derivation cohort and 22.3% in validation cohort, P = 0.866. Model for end-stage liver disease (MELD) [odds ratio (OR) 1.106] and encephalopathy (E) (OR 4.658) in one analysis and Child-Pugh score (OR 1.379) and serum creatinine (OR 1.474) in another analysis were significantly associated with 6-week mortality. MELD-E model (AUROC 0.792) was superior to Child-creatinine model (AUROC) in terms of discrimination. The MELD-E model had highest AUROC; as compared to other models-MELD score (AUROC 0.751, P = 0.036), Child-Pugh score (AUROC 0.737, P = 0.037), D'Amico model (AUROC 0.716, P = 0.014) and Augustin model (AUROC 0.739, P = 0.018) in derivation cohort. In validation cohort, the discriminatory performance of MELD-E model (AUROC 0.805) was higher as compared to other models including MELD score (AUROC 0.771, P = 0.048), Child-Pugh score (AUROC 0.746, P = 0.011), Augustin model (AUROC 0.753, P = 0.039) and D'Amico model (AUROC 0.736, P = 0.021).

CONCLUSION

In cirrhotic patients with AVB, the novel MELD-Encephalopathy model predicts 6 weeks mortality with higher accuracy than the existing prognostic models.

Comparison of various prognostic scores in variceal and non-variceal upper gastrointestinal bleeding: A prospective cohort study

BACKGROUND AND AIMS

Various prognostic scores like Glasgow-Blatchford bleeding score (GBS), modified Glasgow-Blatchford bleeding score (mGBS), full Rockall score (FRS) including endoscopic findings, clinical Rockall score (CRS), and albumin, international normalized ratio (INR), mental status, systolic blood pressure, age >65 (AIMS65) are used for risk stratification in patients with upper gastrointestinal bleeding (UGIB). The utility of these scores in variceal UGIB (VUGIB) is not well defined. In this prospective study, we aimed to assess the performance of these scores in patients with non-variceal (NVUGIB) and VUGIB.

METHODS

We included 1011 patients (during March 2017 and August 2018) including 439 with NVUGIB and 572 VUGIB. Performance of GBS, mGBS, FRS, CRS, and AIMS65 for various outcome measures was analyzed using the area under receiver operator characteristic curve (AUROC).

RESULTS

The accuracy of prognostic scores in predicting the composite outcome including the need of hospital-based intervention and 42-day mortality was higher in NVUGIB as compared with VUGIB, AUROC: CRS: 0.641 vs. 0.537; FRS: 0.669 vs. 0.625; GBS: 0.719 vs. 0.587; mGBS: 0.711 vs. 0.594; AIMS65: 0.567 vs. 0.548. GBS and mGBS at a cut-off score of 1 had the highest negative predictive value, 91.7% and 91.3%, respectively, for predicting composite outcome in NVUGIB. Similarly, these scores had better accuracy for predicting 42-day rebleeding in NVUGIB as compared to VUGIB, AUROC: CRS: 0.680 vs. 0.537; FRS: 0.698 vs. 0.565; GBS: 0.661 vs. 0.543; mGBS: 0.627 vs. 0.540; AIMS65: 0.695 vs. 0.606.

CONCLUSION

The prognostic scores such as CRS, FRS, GBS, mGBS, and AIMS65 predict the need for hospital-based management, rebleeding, and mortality better among patients with NVUGIB than VUGIB.