Machine Learning for Short-Term Mortality in Acute Decompensation of Liver Cirrhosis: Better than MELD Score

Prediction of short-term mortality in patients with acute decompensation of liver cirrhosis could be improved. We aimed to develop and validate two machine learning (ML) models for predicting 28-day and 90-day mortality in patients hospitalized with acute decompensated liver cirrhosis. We trained two artificial neural network (ANN)-based ML models using a training sample of 165 out of 290 (56.9%) patients, and then tested their predictive performance against Model of End-stage Liver Disease-Sodium (MELD-Na) and MELD 3.0 scores using a different validation sample of 125 out of 290 (43.1%) patients. The area under the ROC curve (AUC) for predicting 28-day mortality for the ML model was 0.811 (95%CI: 0.714- 0.907; p < 0.001), while the AUC for the MELD-Na score was 0.577 (95%CI: 0.435-0.720; p = 0.226) and for MELD 3.0 was 0.600 (95%CI: 0.462-0.739; p = 0.117). The area under the ROC curve (AUC) for predicting 90-day mortality for the ML model was 0.839 (95%CI: 0.776- 0.884; p < 0.001), while the AUC for the MELD-Na score was 0.682 (95%CI: 0.575-0.790; p = 0.002) and for MELD 3.0 was 0.703 (95%CI: 0.590-0.816; p < 0.001). Our study demonstrates that ML-based models for predicting short-term mortality in patients with acute decompensation of liver cirrhosis perform significantly better than MELD-Na and MELD 3.0 scores in a validation cohort.

Effect of MELD-Na score on overall survival of periampullary cancer

Periampullary cancers have a short overall survival (OS), and many prognostic factors have been studied for this purpose. They usually present with biliary obstruction, which negatively affects the liver, kidney, immune system, and cardiovascular system. This study aimed to investigate the effect of MELD-Na scores on OS in patients undergoing pancreaticoduodenectomy due to periampullary cancer. Patients who underwent pancreaticoduodenectomy due to periampullary cancer between January 2010 and January 2021 were included in the study. After applying the exclusion criteria, 80 of the 124 patients were included in the study. The demographic, laboratory, and pathologic data of the patients were analyzed retrospectively. Univariate analysis showed that MELD-Na score at admission, age-adjusted Charlson Comorbidity Index, adjuvant treatment, portal vein resection, lymphovascular invasion (LVI), T-stage, and tumor location were significantly associated with OS (p < 0.1). In multivariate analysis, MELD-Na score at admission (HR: 1.051, 95% CI [1.004-1.101]; p = 0.033), adjuvant treatment (HR: 4.717, 95% CI [2.371-9.383]; p < 0.001), LVI (HR: 2.473, 95% CI [1.355-4.515]; p = 0.003), and tumor location (HR: 2.380, 95% CI [1.274-4.445]; p = 0.007) were independent risk factors for OS. MELD-Na score, adjuvant treatment, LVI, and tumor location were independent risk factors for the OS of periampullary cancer. The MELD-Na score may be used to predict OS for patients undergoing pancreaticoduodenectomy due to periampullary cancer.

Combination of Model for End-Stage Liver Disease (MELD) and Sarcopenia predicts mortality after transjugular intrahepatic portosystemic shunt (TIPS)

TIPS is the most effective treatment for portal hypertension. Patient selection remains important to achieving optimal post-TIPS outcomes. The study evaluates 1-year mortality factors in cirrhotic patients receiving TIPS.

METHODS

87 cirrhotic patients received a TIPS between 2015 - 2021. Predictors of 1-year and overall mortality were assessed by estimating cumulative incidence functions and Grey's test to adjust for liver transplantation as a risk competing with mortality. Variables with p < 0.05 were checked for collinearity and included in the multivariate Cox proportional hazards model. Model discrimination was evaluated by calculating the area under the ROC curve.

RESULTS

87 patients were included (68% men; 22% ≥70 years). ALD was the primary cirrhosis cause. Most patients were Child-Pugh class B, MELD-Na score was 13.6 ± 6.0 points. The most frequent indication for TIPS was bleeding (51.7%), followed by refractory ascites (42.5%). The variables positively associated with mortality in univariate analysis were ascites, clinically overt sarcopenia and MELD-Na score, while ongoing nutritional supplementation improved survival. In the multivariate analysis, only clinically overt sarcopenia and MELD-Na score remained independently associated with mortality. A MELD-Na/sarcopenia model demonstrated a good discrimination, AUROC: 0.86 (95% CI 0.77 - 0.95).

CONCLUSION

MELD-Na score, and sarcopenia were significantly associated with 1-year survival in cirrhotic patients who received TIPS.

Prognostic significance of hyperammonemia in neuroendocrine neoplasm patients with liver metastases

Neuroendocrine neoplasms (NENs) are rare, usually slow-growing tumors, often presenting with extensive liver metastases. Hyperammonemia due to insufficient hepatic clearance has been described in NEN cases; however, no systematic evaluation of risk factors and outcomes of NEN-associated hyperammonemia exists so far. This case report and retrospective review of NEN patients developing hyperammonemia from the years 2000 to 2020 at the Erasmus Medical Center in Rotterdam, the Netherlands, aimed to describe these patients and determine prognostic factors to improve evaluation and treatment. Forty-four NEN patients with documented hyperammonemia were identified. All patients had liver metastases with 30% (n = 13) showing signs of portal hypertension. Patients who developed encephalopathy had higher median ammonia levels, but there was no association between the severity of hyperammonemia and liver tumor burden or presence of liver insufficiency. Eighty-four percent (n = 37) of patients died during follow-up. The median (IQR) time from diagnosis of hyperammonemia to death was 1.7 months (0.1-22.7). Hyperbilirubinemia, hypoalbuminemia, elevated international normalized ratio, presence of liver insufficiency, encephalopathy and ascites were associated with worse outcomes. Their role as independent risk factors for mortality was confirmed using the Child-Pugh score as a summary factor (P < 0.001). No difference was seen concerning overall survival between our hyperammonemia patients and a propensity score-matched control stage IV NEN cohort. In conclusion, hyperammonemia comprises a relevant and potentially underdiagnosed complication of NEN liver metastases and is associated with worse outcomes. Assessment of signs of encephalopathy, risk factors and the Child-Pugh score could be helpful in selecting patients in whom ammonia levels should be measured.

Early transplantation maximizes survival in severe acute-on-chronic liver failure: Results of a Markov decision process model

Background & Aims

Uncertainties exist surrounding the timing of liver transplantation (LT) among patients with acute-on-chronic liver failure grade 3 (ACLF-3), regarding whether to accept a marginal quality donor organ to allow for earlier LT or wait for either an optimal organ offer or improvement in the number of organ failures, in order to increase post-LT survival.

Methods

We created a Markov decision process model to determine the optimal timing of LT among patients with ACLF-3 within 7 days of listing, to maximize overall 1-year survival probability.

Results

We analyzed 6 groups of candidates with ACLF-3: patients age ≤60 or >60 years, patients with 3 organ failures alone or 4-6 organ failures, and hepatic or extrahepatic ACLF-3. Among all groups, LT yielded significantly greater overall survival probability vs. remaining on the waiting list for even 1 additional day (p <0.001), regardless of organ quality. Creation of 2-way sensitivity analyses, with variation in the probability of receiving an optimal organ and expected post-transplant mortality, indicated that overall survival is maximized by earlier LT, particularly among candidates >60 years old or with 4-6 organ failures. The probability of improvement from ACLF-3 to ACLF-2 does not influence these recommendations, as the likelihood of organ recovery was less than 10%.

Conclusion

During the first week after listing for patients with ACLF-3, earlier LT in general is favored over waiting for an optimal quality donor organ or for recovery of organ failures, with the understanding that the analysis is limited to consideration of only these 3 variables.

Lay summary

In the setting of grade 3 acute-on-chronic liver failure (ACLF-3), questions remain regarding the timing of transplantation in terms of whether to proceed with liver transplantation with a marginal donor organ or to wait for an optimal liver, and whether to transplant a patient with ACLF-3 or wait until improvement to ACLF-2. In this study, we used a Markov decision process model to demonstrate that earlier transplantation of patients listed with ACLF-3 maximizes overall survival, as opposed to waiting for an optimal donor organ or for improvement in the number of organ failures.

The Model for End-Stage Liver Disease-Sodium Score at Admission Is Prognostic of Covid-19 Disease Severity

Covid-19 is a systemic viral respiratory illness that can cause gastrointestinal manifestations. There is evidence that Covid-19 can infect liver tissue and may cause transaminemia. A prognostic model is needed to aid clinicians in determining disease severity. The Model for End-Stage Liver Disease-Sodium (MELD-Na) score is a mortality assessment tool in liver transplant patients that has been found to be prognostic in other clinical situations. This study aimed to determine if the MELD-Na score was associated with disease severity in patients with Covid-19, as assessed by multiple clinical outcomes including death within 30 days of discharge and development of an acute kidney injury (AKI). This is a retrospective cohort study that analyzed patients admitted to a community academic hospital with the diagnosis of Covid-19. The 30-day MELD-Na score was found to be significantly higher in those who died (14.38 ± 6.92) relative to those who survived (9.68 ± 5.69; p = 0.03). Additionally, patients with a MELD-Na score greater than 10 were found to have higher risk of developing an AKI (odds ratio (OR) 3.31 (1.08, 10.17); p = 0.03), need for hemodialysis (OR 9.69 (1.74, 53.96); p = 0.007), require vasopressors (OR 4.55 (1.22, 16.99); p = 0.02), and have a longer hospital stay (OR 4.17 (1.05, 16.47); p = 0.03). The MELD-Na score may serve as a useful clinical scoring system for prognosis in patients admitted to the hospital with Covid-19.

Assessing the prognostic scores for the prediction of the mortality of patients with acute-on-chronic liver failure: a retrospective study

Background

Acute-on-chronic liver failure (ACLF), which is characterized by rapid deterioration of liver function and multiorgan failure, has high mortality. This study was designed to identify prognostic scores to predict short-term and long-term outcome in patients with ACLF to facilitate early treatment and thereby improve patient survival.

Materials and Methods

We retrospectively analyzed 102 ACLF patients who were hospitalized in the gastroenterology department. The EASL-CLIF criteria were used to define the ACLF. The demographic characteristics and biochemical examination results of the patients were acquired, and seven scores (CTP score, MELD score, MELD-Na, CLIF ACLF score, CLIF-C OF score, and CLIF SOFA score) were calculated 24 h after admission. All patients were observed until loss to follow-up, death, or specific follow-up times (28 days, 3 months, and 6 months), which were calculated after the initial hospital admission. The receiver operating characteristic (ROC) curve was employed to estimate the power of six scores to forecast ACLF patients’ outcome.

Results

All scores were distinctly higher in nonsurviving patients than in surviving patients and had predictive value for outcome in patients with ACLF at all time points (P < 0.050). The areas under the ROC curve (AUROCs) of the CLIF-SOFA score were higher than those of other scores at all time points. The comparison of the AUROC of the CLIF-SOFA score with other scores was statistically significant at 28 days (P < 0.050), which was the only time point at which it was greater than 0.800.

Conclusion

Patients with ACLF have high mortality. These six scores are effective tools for assessing the prognosis of ACLF patients. The CLIF-SOFA score is especially effective for evaluating 28-day mortality.

LIRER score - a valuable tool to predict medium-long-term outcomes in hepatic cirrhosis decompensation

BACKGROUND

The liver-renal-risk (LIRER) score was developed to predict adverse outcomes in cirrhotic patients with Model for End-stage Liver Disease (MELD)<18, helping the allocation to liver transplantation in this population. We aimed to assess its prognostic performance compared to other prognostic scores in first admission for hepatic cirrhosis decompensation.

METHODS

Retrospective study that included patients admitted for initial decompensation of cirrhosis between January 2010 and February 2017. The LIRER, Child-Pugh (CP), MELD and MELD-sodium (MELD-Na) scores were calculated at admission.

RESULTS

One-hundred and forty-six patients were included, 65.1% with MELD < 18. LIRER was a predictor of in-stay (AUC 0.70; p = .04), first-year (0.70; p < .001), two-years (0.72; p < .001) and overall mortality (0.70; p < .001), being the only score with an acceptable discriminating ability (AUC ≥ 0.70). Stratifying patients in MELD < 18 and ≥18, LIRER was found to be an independent predictor of first-year, two-years and overall-mortality only in MELD < 18 patients (AUC 0.67; 0.70; 0.72), being superior to all other scores predicting first-year mortality and the only with an AUC with a reasonable discriminating ability for predicting two-years and overall-mortality. The LIRER was also a predictor of 30-days hospital readmission (AUC 0.75; p < .001), independently of MELD, with patients with LIRER > 15.9 having a significantly higher probability to be readmitted at 30 days.

CONCLUSIONS

The LIRER score is a predictor of first-year, two-years and overall-mortality in decompensated cirrhosis, particularly in patients with MELD < 18. LIRER is therefore an important tool to predict medium-long-term outcomes in this population. Besides, it allows predicting the 30-days readmission probability in overall patients, independently of MELD.

Effect of the clinical course of acute-on-chronic liver failure prior to liver transplantation on post-transplant survival

BACKGROUND & AIMS

Patients with acute-on-chronic liver failure (ACLF) can be listed for liver transplantation (LT) because LT is the only curative treatment option. We evaluated whether the clinical course of ACLF, particularly ACLF-3, between the time of listing and LT affects 1-year post-transplant survival.

METHODS

We identified patients from the United Network for Organ Sharing database who were transplanted within 28 days of listing and categorized them by ACLF grade at waitlist registration and LT, according to the EASL-CLIF definition.

RESULTS

A total of 3,636 patients listed with ACLF-3 underwent LT within 28 days. Among those transplanted, 892 (24.5%) recovered to no ACLF or ACLF grade 1 or 2 (ACLF 0-2) and 2,744 (75.5%) had ACLF-3 at transplantation. One-year survival was 82.0% among those transplanted with ACLF-3 vs. 88.2% among those improving to ACLF 0-2 (p <0.001). Conversely, the survival of patients listed with ACLF 0-2 who progressed to ACLF-3 at LT (n = 2,265) was significantly lower than that of recipients who remained at ACLF 0-2 (n = 17,631) at the time of LT (83.8% vs. 90.2%, p <0.001). Cox modeling demonstrated that recovery from ACLF-3 to ACLF 0-2 at LT was associated with reduced 1-year mortality after transplantation (hazard ratio0.65; 95% CI 0.53-0.78). Improvement in circulatory failure, brain failure, and removal from mechanical ventilation were also associated with reduced post-LT mortality. Among patients >60 years of age, 1-year survival was significantly higher among those who improved from ACLF-3 to ACLF 0-2 than among those who did not.

CONCLUSIONS

Improvement from ACLF-3 at listing to ACLF 0-2 at transplantation enhances post-LT survival, particularly in those who recovered from circulatory or brain failure, or were removed from the mechanical ventilator. The beneficial effect of improved ACLF on post-LT survival was also observed among patients >60 years of age.

LAY SUMMARY

Liver transplantation (LT) for patients with acute-on-chronic liver failure grade 3 (ACLF-3) significantly improves survival, but 1-year survival probability after LT remains lower than the expected outcomes for transplant centers. Our study reveals that among patients transplanted within 28 days of waitlist registration, improvement of ACLF-3 at listing to a lower grade of ACLF at transplantation significantly enhances post-transplant survival, even among patients aged 60 years or older. Subgroup analysis further demonstrates that improvement in circulatory failure, brain failure, or removal from mechanical ventilation have the strongest impact on post-transplant survival.

Metabolomic biomarkers are associated with mortality in patients with cirrhosis caused by primary biliary cholangitis or primary sclerosing cholangitis

Aim:

To assess the ability of signature metabolites alone, or in combination with the model for end-stage liver disease-Na (MELD-Na) score to predict mortality in patients with cirrhosis caused by primary biliary cholangitis or primary sclerosing cholangitis.

Materials & methods:

Plasma metabolites were detected using ultrahigh-performance liquid chromatography/tandem mass spectrometry in 39 patients with cirrhosis caused by primary biliary cholangitis or primary sclerosing cholangitis. Mortality was predicted using Cox proportional hazards regression and time-dependent receiver operating characteristic curve analyses.

Results:

The top five metabolites with significantly greater accuracy than the MELD-Na score (area under the receiver operating characteristic curve [AUROC] = 0.7591) to predict 1-year mortality were myo-inositol (AUROC = 0.9537), N-acetylputrescine (AUROC = 0.9018), trans-aconitate (AUROC = 0.8880), erythronate (AUROC = 0.8345) and N6-carbamoylthreonyladenosine (AUROC = 0.8055). Several combined MELD-Na-metabolite models increased the accuracy of predicted 1-year mortality substantially (AUROC increased from 0.7591 up to 0.9392).

Conclusion:

Plasma metabolites have the potential to enhance the accuracy of mortality predictions, minimize underestimates of mortality in patients with cirrhosis and low MELD-Na scores, and promote equitable allocation of donor livers.

To receive a liver transplant, patients with cirrhosis need to be listed on the US liver transplant waiting list based on a score called the model for end-stage liver disease-Na (MELD-Na) score that is expected to accurately rank the patients based on urgency for a liver transplant. However, MELD-Na score is not sufficiently accurate to identify many patients with cirrhosis with the highest urgency, and this results in longer waiting times on the liver transplant list, and therefore higher death rates. We identified several metabolomic biomarkers that can increase the accuracy of the MELD-Na score, and optimize the allocation of donor livers for transplantation of patients with cirrhosis.

Child-Turcotte-Pugh Score, MELD Score and MELD-Na Score as Predictors of Short-Term Mortality among Patients with End-Stage Liver Disease in Northern India

Background and Objectives

Child-Turcotte-Pugh (CTP), Model for End-Stage Liver Disease (MELD) and MELD-sodium (MELD-Na) scores are used for predicting disease severity and mortality among patients with end-stage liver disease. However, data regarding their usefulness in predicting the short-term outcome of end-stage liver disease are not available in India. This prospective study compared the CTP score, MELD score and MELD-Na score as predictors of short-term outcome among patients with end-stage liver disease.

Methods

CTP, MELD and MELD-Na scores were determined in 171 patients with end-stage liver disease at a tertiary healthcare centre in India at the time of admission, and the concordance (C-) statistics of the three scores for 3-month mortality were assessed and compared. The aetiology of end-stage liver disease and the clinical presentation were determined.

Results

The CTP score, MELD score and MELD-Na score on day 1 were significantly higher among non-survivors than among survivors (p < 0.0001 each). The C-statistic for 3-month mortality for the CTP score was 0.93 (p < 0.0001), that for the MELD score was 0.86 (p < 0.0001) and that for the MELD-Na score was 0.83 (p < 0.0001). The C-statistics of these scores differed significantly for 3-month mortality, and the CTP score was better than the MELD (p < 0.0001) and MELD-Na (p < 0.0001) scores in predicting 3-month mortality.

Conclusions

The CTP, MELD and MELD-Na scores were very good predictors of mortality at 3 months among patients with end-stage liver disease. The CTP score was superior to the MELD and MELD-Na scores in predicting 3-month mortality.

Predictors of Acute Liver Failure in Patients With Acute Hepatitis A: An Analysis of the 2016-2018 San Diego County Hepatitis A Outbreak

Background

Between 2016 and 2018, San Diego County experienced a hepatitis A outbreak with a historically high mortality rate (3.4%) that highlighted the need for early recognition of those at risk of developing acute liver failure (ALF).

Methods

A retrospective case series of adult hospitalized patients with acute hepatitis A.

Results

One hundred six patients with hepatitis A were studied, of whom 11 (10.4%) developed ALF, of whom 7 (6.6%) died. A history of alcohol abuse, hyperbilirubinemia, hypoalbuminemia, hyponatremia, and anemia were associated with increased odds of developing ALF. Initial Maddrey’s and Model of End-Stage Liver Disease Sodium (MELD-Na) scores were also associated with the development of ALF. Multivariable analysis showed that a higher initial MELD-Na score (odds ratio [OR], 1.205; 95% confidence interval [CI], 1.018–1.427) and a lower initial serum albumin concentration (OR, 9.35; 95% CI, 1.15–76.9) were associated with increased odds of developing ALF. Combining serum albumin and MELD-Na (SAM; C-statistic, 0.8878; 95% CI, 0.756–0.988) yielded a model that was not better than either serum albumin (C-statistic, 0.852; 95% CI, 0.675–0.976) or MELD-Na (C-statistic, 0.891; 95% CI, 0.784–0.968; P = .841). Finally, positive blood cultures were more common among patients with ALF compared with those without ALF (63.6% vs 4.3%; P < .00001).

Conclusions

Hypoalbuminemia was associated with an increased risk of ALF in patients with acute hepatitis A. Positive blood cultures and septic shock as a cause of death were common among patients with ALF. Providers caring for patients with acute hepatitis A should monitor for early signs of sepsis and consider empiric antibiotics, especially in patients presenting with hypoalbuminemia.

Single center validation of mortality scores in patients with acute decompensation of cirrhosis with and without acute-on-chronic liver failure

OBJECTIVES

Acute decompensation (AD) of cirrhosis is characterized by high mortality. We aimed to validate the performance in predicting mortality of both the chronic-liver-failure-consortium (CLIF-C) acute-on-chronic liver failure (ACLF) and CLIF-C AD scores in a cohort of patients admitted for AD.

METHODS

In this prospective cohort study, patients were followed-up during their hospital stay and for 365 days thereafter.

RESULTS

About 182 patients with AD were enrolled including 78 (42.8%) who met the criteria for ACLF (ACLF-group) while the remaining had AD without ACLF (AD-group). 56.4% and 56.7% of the ACLF- and AD-groups, respectively, had alcoholic cirrhosis and 85.9% of the ACLF-group hepatic encephalopathy. Only few patients were hospitalized in the intensive care unit (ICU) or transplanted. The probabilities of death estimated for both scores were similar to the overall mortality rates observed at all time points. The model had a good fit in the AD-group at 90 days (p = .974) but a worse, yet adequate, in the ACLF-group at 28 days (p = .08). The CLIF-C ACLF or AD scores had an adequate, predictive discrimination ability for mortality at all time points, with Harrel's concordance index-C ranging between 0.64 and 0.65 or 0.64 and 0.68, respectively. Both scores showed a similar predictive accuracy for mortality compared to those of MELD, MELD-Na and Child-Pugh.

CONCLUSIONS

In this population without access to appropriate ICU treatment, the CLIF-C ACLF and AD performed worse than in studies with patients having ICU access. In addition, the CLIF scores were not superior to classical ones in this setting.