Society: AASLD

Background: Cirrhosis is a major contributor of morbidity and mortality in patients with chronic liver disease globally. The disease burden of viral-related cirrhosis has decreased, related to advancements in prevention and treatments while the prevalence of non-alcoholic fatty liver disease and alcohol-associated cirrhosis has increased mirroring the metabolic disease epidemic. Nevertheless, the relative implication of the transition in the predominant etiology of cirrhosis has yet to be examined between males and females. This study aims to present sex differences in the disease burden of cirrhosis from 2010-2019.

Methods: Using the Global Burden of Disease Study 2019, we evaluated the temporal trends of cirrhosis incidence between the sexes. Yearly frequencies and age-standardized rates (ASR) of incidence, death, and disability-adjusted life-years (DALYs) associated with cirrhosis from 2010-2019 were estimated and stratified by sex, region, country, socio-demographic index (SDI), and cirrhosis etiology.

Results: In 2019, the overall burden of cirrhosis was higher in males compared to females. There were 1,206,125 incident cases, 969,068 deaths, and 31,781,079 DALYs due to cirrhosis in males, and 845,429 incident cases, 502,944 deaths, and 14,408,336 DALYs due to cirrhosis in females, respectively (Figure 1A-D). From 2010-2019, there was an 11% increase in incident cirrhosis cases in males and a 16% increase in females. The frequency of cirrhosis-related deaths also increased by 9% in males and 12% in females. Incidence ASR of cirrhosis increased in females while remaining stable in males. However, death ASRs declined in both males and females in all regions except for the Americas where it remained stable. The largest increase in incident cirrhosis cases was in low SDI countries for both males (+37%) and females (+32%). While the leading cause of incident cirrhosis cases was HCV in both sexes, NASH was the fastest-growing etiology of incident cirrhosis cases in both males (+23%) and females (+30%). The leading cause of cirrhosis-related deaths was alcohol and HCV in males and females respectively (Figure 1E). Furthermore, death ASRs decreased for all etiologies in both sexes, except for NASH-related cirrhosis which remained stable.

Conclusion: In summary, the global burden of cirrhosis is substantially higher in males compared to females. Age-adjusted mortality declined for all etiologies of liver disease for males and females, except in NASH. The age-adjusted incidence rates of NASH cirrhosis in females exceeded that of males, while age-adjusted mortality rates in females approached that of males. Measures are required to tackle obesity and diabetes to reduce the global burden of NASH cirrhosis in both males and females.

<b>Figure 1 </b><br /> <b>(A</b>): Frequency of incident cirrhosis cases in males versus females from 2010 to 2019, by etiology of liver disease<br /> <b>(B</b>): Frequency of cirrhosis-related deaths in males versus females from 2010 to 2019, by etiology of liver disease<br /> <b>(C): </b>Frequency of cirrhosis-related deaths in males versus females in 2019, by etiology of liver disease<br /> <b>(D): </b>Frequency of cirrhosis-related deaths in males versus females in 2019, by World Health Organization region<br /> <b>(E): </b>Contribution of global cirrhosis-related deaths in male versus females in 2019, by etiology of liver disease

Figure 1
(A): Frequency of incident cirrhosis cases in males versus females from 2010 to 2019, by etiology of liver disease
(B): Frequency of cirrhosis-related deaths in males versus females from 2010 to 2019, by etiology of liver disease
(C): Frequency of cirrhosis-related deaths in males versus females in 2019, by etiology of liver disease
(D): Frequency of cirrhosis-related deaths in males versus females in 2019, by World Health Organization region
(E): Contribution of global cirrhosis-related deaths in male versus females in 2019, by etiology of liver disease

Background: Despite broad acceptance of Baveno VI criteria for non-invasive prediction of varices needing treatment (VNT), its adoption in community practices is limited due to lack of easy access to transient elastography (TE) in the United States. Our group has recently established a deep learning model using 12-lead electrocardiograms (ECG) to accurately detect the presence and severity of cirrhosis. Given its wide availability and low cost, we aimed to determine whether an ECG-enabled model in addition to platelet count could be used to spare screening endoscopies (EGD) while maintaining an acceptable missed rate of VNT to less than 5%.

Methods: Adult patients with cirrhosis (based on ICD10 codes) who underwent screening EGD at Mayo Clinic, Rochester from 1/2020-1/2022 were identified. Data were then extracted from the ProVation reporting database to identify two distinct groups: those without or with small (< 5 mm) varices, and those with large (>5 mm) varices. For each patient, the Detection of Undiagnosed Liver Cirrhosis via ECG (DULCE) score was generated using a 12-lead ECG completed nearest to the date of their EGD. Logistic regression models were fit to predict the presence of large varices based on platelets and DULCE score. An ROC curve was produced and a threshold to maximize specificity while retaining a sensitivity >=0.95 was identified.

Results: A total of 1,086 patients with cirrhosis undergoing screening EGD were included. Median age 61, 58.6% males and mean MELD-Na score 16. A significant association between platelet count (OR 0.68, CI 0.56, 0.82) and DULCE score (OR 2.24, CI 1.40, 3.60) with presence of large esophageal varices was found (table 1). A combined model with platelet count and DULCE score was predictive of large esophageal varices with an area under the receiver operating characteristic curve (AUC) of 0.636. For this study, large varices were considered VNT. To maintain the acceptable missed VNT rate, we identified a threshold to achieve 95% sensitivity for VNT. When applying the determined threshold, the combined DULCE and platelet model would result in 129 spared endoscopies (12.5%) and 11 missed VNT (1.1%) within this cohort (Table 2).

Conclusions: An electrocardiogram-based deep learning model may be a useful adjunctive tool for predicting the presence of high-risk esophageal varices in patients with cirrhosis. Compared to the Baveno VI criteria, our model offers the potential to use a widely-available test (ECG) rather than specialized liver stiffness measurement for non-invasive prediction of esophageal varices and thus reduce the need for screening EGD in low-risk patients.

<b>Table 1:</b> Baseline characteristics and DULCE scores for patients in the two defined groups.

Table 1: Baseline characteristics and DULCE scores for patients in the two defined groups.

<b>Table 2: </b>Performance of the combined model (DULCE + platelet) for avoiding upper endoscopy.

Table 2: Performance of the combined model (DULCE + platelet) for avoiding upper endoscopy.

Peter Noseworthy

Appearances