Prognosis of variceal and non-variceal upper gastrointestinal bleeding in already hospitalised patients: Results from a French prospective cohort

New machine-learning models outperform conventional risk assessment tools in Gastrointestinal bleeding

Rapid and accurate identification of high-risk acute gastrointestinal bleeding (GIB) patients is essential. We developed two machine-learning (ML) models to calculate the risk of in-hospital mortality in patients admitted due to overt GIB. We analyzed the prospective, multicenter Hungarian GIB Registry's data. The predictive performance of XGBoost and CatBoost machine-learning algorithms with the Glasgow-Blatchford (GBS), pre-endoscopic Rockall and ABC scores were compared. We evaluated our models using five-fold cross-validation, and performance was measured by area under receiver operating characteristic curve (AUC) analysis with 95% confidence intervals (CI). Overall, we included 1,021 patients in the analysis. In-hospital death occurred in 108 cases. The XGBoost and the CatBoost model identified patients who died with an AUC of 0.84 (CI:0.76-0.90; 0.77-0.90; respectively) in the internal validation set, whereas the GBS and pre-endoscopic Rockall clinical scoring system's performance was significantly lower, AUC values of 0.68 (CI:0.62-0.74) and 0.62 (CI:0.56-0.67), respectively. ABC score had an AUC of 0.77 (CI:0.71-0.83). The XGBoost model had a specificity of 0.96 (CI:0.92-0.98) at a sensitivity of 0.25 (CI:0.10-0.43) compared with the CatBoost model, which had a specificity of 0.74 (CI:0.66-0.83) at a sensitivity of 0.78 (CI:0.57-0.95). XGBoost and the CatBoost models evaluate the mortality risk of acute GI bleeding better, than the conventional risk assessment tools.

Prophylactic Proton Pump Inhibitors in Upper Gastrointestinal Bleeding: Impact and Underprescription in a French Multicentric Cohort

BACKGROUND

Appropriate prescription of Proton pump inhibitors (PPIs) remains an important concern amid the rising overuse. A gap exists in the literature regarding the benefit of PPI prophylaxis and the consequences of underprescription in patients at risk for upper gastrointestinal bleeding (UGIB).

AIMS

This study aims to describe the characteristics of hemorrhage in relation to PPI use in patients experiencing UGIB, with a focus on high-risk individuals requiring gastroprotection.

METHODS

Data from a French multicentric cohort of patients experiencing UGIB were analyzed. Patients using PPI were compared to those without PPI considering bleeding etiologies and outcomes of peptic ulcer disease (PUD)-related hemorrhage. The rate of PPI use and its effect on bleeding characteristics in high-risk populations, defined based on international guidelines, were also assessed.

RESULTS

Among 2497 included patients, 31.1% were on PPI at bleeding onset. PPI users exhibited a significantly lower rate of PUD-related bleeding in comparison with those without PPI (24.7 vs 40.8%, respectively, p < 0.0001). Similar difference was observed in high-risk populations, of whom, only 40.3% had gastroprotection with PPI before bleeding onset. PPI prophylaxis, however, did not influence the severity of bleeding in the general study population or in high-risk groups. Multivariate analysis identified age, comorbidities, and having more than two anti-thrombotic agents as predictors of severe bleeding.

CONCLUSIONS

PPI users appear to have a lower rate of bleeding ulcers compared to non-users. However, underprescription in high-risk groups raises the need for standardized care to ensure appropriate PPI use.

Early prognosis prediction for non-variceal upper gastrointestinal bleeding in the intensive care unit: based on interpretable machine learning

INTRODUCTION

This study aims to construct a mortality prediction model for patients with non-variceal upper gastrointestinal bleeding (NVUGIB) in the intensive care unit (ICU), employing advanced machine learning algorithms. The goal is to identify high-risk populations early, contributing to a deeper understanding of patients with NVUGIB in the ICU.

METHODS

We extracted NVUGIB data from the Medical Information Mart for Intensive Care IV (MIMIC-IV, v.2.2) database spanning from 2008 to 2019. Feature selection was conducted through LASSO regression, followed by training models using 11 machine learning methods. The best model was chosen based on the area under the curve (AUC). Subsequently, Shapley additive explanations (SHAP) was employed to elucidate how each factor influenced the model. Finally, a case was randomly selected, and the model was utilized to predict its mortality, demonstrating the practical application of the developed model.

RESULTS

In total, 2716 patients with NVUGIB were deemed eligible for participation. Following selection, 30 out of a total of 64 clinical parameters collected on day 1 after ICU admission remained associated with prognosis and were utilized for developing machine learning models. Among the 11 constructed models, the Gradient Boosting Decision Tree (GBDT) model demonstrated the best performance, achieving an AUC of 0.853 and an accuracy of 0.839 in the validation cohort. Feature importance analysis highlighted that shock, Glasgow Coma Scale (GCS), renal disease, age, albumin, and alanine aminotransferase (ALP) were the top six features of the GBDT model with the most significant impact. Furthermore, SHAP force analysis illustrated how the constructed model visualized the individualized prediction of death.

CONCLUSIONS

Patient data from the MIMIC database were leveraged to develop a robust prognostic model for patients with NVUGIB in the ICU. The analysis using SHAP also assisted clinicians in gaining a deeper understanding of the disease.

Impact of peptic ulcer bleeding on the in-hospital outcomes of cirrhotic patients with acute gastrointestinal bleeding: an international multicenter study

OBJECTIVES

Peptic ulcer is the most common source of non-variceal bleeding. However, it remains controversial whether the outcomes of cirrhotic patients with peptic ulcer bleeding differ from those with variceal bleeding.

METHODS

Cirrhotic patients with acute gastrointestinal bleeding (AGIB) who underwent endoscopy and had an identifiable source of bleeding were retrospectively screened from an international multicenter cohort. Logistic regression analyses were performed to explore the impact of peptic ulcer bleeding on in-hospital death and 5-day failure to control bleeding. Propensity score matching (PSM) analysis was performed by matching age, gender, Child-Pugh score, and model for end-stage liver disease score between the peptic ulcer bleeding and variceal bleeding groups.

RESULTS

Overall, 1535 patients were included, of whom 73 (4.7%) had peptic ulcer bleeding. Multivariate logistic regression analyses showed that peptic ulcer bleeding was not independently associated with in-hospital death (OR = 2.169, p = 0.126) or 5-day failure to control bleeding (OR = 1.230, p = 0.680). PSM analyses demonstrated that both in-hospital mortality (9.7% vs. 6.3%, p = 0.376) and rate of 5-day failure to control bleeding (6.9% vs. 5.4%, p = 0.787) were not significantly different between the two groups.

CONCLUSIONS

The impact of peptic ulcer bleeding on the in-hospital outcomes of cirrhotic patients is similar to that of variceal bleeding.

Sepsis as a Potential Risk Factor for Upper Gastrointestinal Bleeding in Critically Ill Patients: A Systematic Review and Meta-analysis

Purpose: Sepsis is a common and critical condition in intensive care units (ICUs) known to complicate patient outcomes. Previous studies have indicated an association between sepsis and various ICU morbidities, including upper gastrointestinal bleeding (UGIB). However, the extent of this relationship and its implications in ICU settings remain inadequately quantified. This study aims to elucidate the association between sepsis and the risk of UGIB in ICU patients. Methods: A comprehensive meta-analysis was conducted, encompassing nine studies with a total of nearly 9000 participants. These studies reported events for both sepsis and nonsepsis patients separately. Pooled odds ratios (ORs) were calculated to assess the risk of UGIB in septic versus nonseptic ICU patients. Subgroup analyses were conducted based on age and study design, and both unadjusted and adjusted ORs were examined. Results: The pooled OR indicated a significant association between sepsis and UGIB (OR = 3.276, 95% CI: 1.931 to 5.557). Moderate heterogeneity was observed (I² = 43.9%). The association was significant in adults (pooled OR = 4.083) but not in children. No difference in association was found based on the study design. Unadjusted and adjusted ORs differed slightly, indicating the influence of confounding factors. Conclusion: This meta-analysis reveals a significant association between sepsis and an increased risk of UGIB in ICU patients, particularly in adults. These findings highlight the need for vigilant monitoring and proactive management of septic ICU patients to mitigate the risk of UGIB. Future research should focus on understanding the underlying mechanisms and developing tailored preventive strategies.

MH-STRALP: A scoring system for prognostication in patients with upper gastrointestinal bleeding

BACKGROUND

Upper gastrointestinal bleeding (UGIB) is a common medical emergency and early assessment of its outcomes is vital for treatment decisions.

AIM

To develop a new scoring system to predict its prognosis.

METHODS

In this retrospective study, 692 patients with UGIB were enrolled from two centers and divided into a training (n = 591) and a validation cohort (n = 101). The clinical data were collected to develop new prognostic prediction models. The endpoint was compound outcome defined as (1) demand for emergency surgery or vascular intervention, (2) being transferred to the intensive care unit, or (3) death during hospitalization. The models' predictive ability was compared with previously established scores by receiver operating characteristic (ROC) curves.

RESULTS

Totally 22.2% (131/591) patients in the training cohort and 22.8% (23/101) in the validation cohort presented poor outcomes. Based on the stepwise-forward Logistic regression analysis, eight predictors were integrated to determine a new post-endoscopic prognostic scoring system (MH-STRALP); a nomogram was determined to present the model. Compared with the previous scores (GBS, Rockall, ABC, AIMS65, and PNED score), MH-STRALP showed the best prognostic prediction ability with area under the ROC curves (AUROCs) of 0.899 and 0.826 in the training and validation cohorts, respectively. According to the calibration curve, decision curve analysis, and internal cross-validation, the nomogram showed good calibration ability and net clinical benefit in both cohorts. After removing the endoscopic indicators, the pre-endoscopic model (pre-MH-STRALP score) was conducted. Similarly, the pre-MH-STRALP score showed better predictive value (AUROCs of 0.868 and 0.767 in the training and validation cohorts, respectively) than the other pre-endoscopic scores.

CONCLUSION

The MH-STRALP score and pre-MH-STRALP score are simple, convenient, and accurate tools for prognosis prediction of UGIB, and may be applied for early decision on its management strategies.

Prognosis of variceal and non-variceal upper gastrointestinal bleeding in already hospitalised patients: Results from a French prospective cohort

OBJECTIVES

Patients who develop upper gastrointestinal bleeding (UGIB) while in hospital appear to have a poor prognosis. Our study aims at analysing the difference in outcome between in-patients (IPs) and out-patients presenting with variceal and non-variceal UGIB.

METHODS

We conducted a multicentre prospective study by collecting data about variceal and non-variceal UGIB cases through 46 hospitals in France between November 2017 and October 2018. We then compared baseline demographic features, endoscopic findings and outcome between patients who developed variceal and non-variceal UGIB on admission (OPs) and those at least 24 h after hospitalisation (IPs). Our primary end-point was mortality and re-bleeding rates at 6 weeks of bleeding onset.

RESULTS

A total of 2498 UGIB cases were identified, of whom 634 (25.4%) occurred in IPs. IPs were older than OPs (72.5 vs. 67.2 years old, p < 0.001) and had a higher rate of comorbidities (38.9% vs. 26.6%, p < 0.0001). Their bleeding was more severe with a Rockall score of >5 present in 40.9% (vs. 30.3% in OPs, p < 0.0001). The 6-week mortality rate was significantly higher in IPs when compared to OPs (21.7% vs. 8%, p < 0.0001). Prothrombin time <50% and rebleeding were the only independent predictors of mortality (p = 0.001 and 0.003, respectively). Six-week rebleeding occurred more frequently among IPs (18.6% vs. 14.4%, p = 0.015) and predictors included female sex, active bleeding upon endoscopy and a Blatchford score >11 (p = 0.017, 0.011 and 0.008, respectively).

CONCLUSION

IPs who develop variceal and non-variceal UGIB are more likely to be elderly with more comorbidities. They have a higher rate of mortality and rebleeding. Independent predictors of mortality were underlying coagulopathy and bleeding recurrence. An optimal bleeding management and efficient rebleeding prevention may improve outcome in these patients.