1 Introduction

Chronic liver disease may induce subclinical myocardial dysfunction in the absence of other cardiac conditions, potentially influencing patient prognosis and outcomes following interventions [1].

Recent evidence, as reported by Liu et al., suggests that the underlying pathogenic and pathophysiological mechanisms are rooted in two distinct pathways: (1) factors related to portal hypertension, hyperdynamic circulation, gut bacterial/endotoxin translocation, and the resulting inflammatory response; (2) liver insufficiency with alterations in the synthesis or metabolism of proteins, lipids, carbohydrates, bile acids, and hormones [2, 3]. As its own entity, cirrhotic cardiomyopathy (CCM) can be diagnosed only when other primary or secondary cardiomyopathies have been excluded.

Research indicates that approximately 50% of individuals with cirrhosis exhibit signs of cardiac dysfunction, the clinical significance of which remains uncertain [4]; furthermore, cardiovascular complications post-liver transplantation are fatal in 7%–21% of cases [5].

Autopsy studies have consistently shown that cardiac histopathological findings in cirrhosis are not contingent upon the aetiology of the underlying liver condition [6, 7].

While previous studies have demonstrated a correlation between hepatic steatosis and increased cardiovascular mortality in patients with metabolic dysfunction-associated steatotic liver disease (MASLD), comprehensive data on cirrhotic cohorts remain limited. Notably, fatty liver alterations have been associated with cardiovascular mortality in up to 16% of cases, exhibiting a dose-dependent relationship [8]. The most recent multidisciplinary European MASLD guideline recommends a risk-adapted, stepwise cardiovascular workup in individuals with MASLD considered for liver transplantation, due to the significantly increased pre-, peri-, and post-operative cardiovascular morbidity and mortality in this population [9]. It is important to note that patients with alcohol-associated liver cirrhosis also have a high burden of concomitant cardiometabolic risk factors [10], rendering them at risk for cardiovascular morbidity and mortality.

The revisited diagnostic criteria for CCM now recommend the routine assessment of global longitudinal strain (GLS) alongside left ventricular (LV) ejection fraction (EF) to evaluate systolic function in patients with cirrhosis [11]. Analysis of speckle patterns offers high sensitivity to early myocardial alterations that precede overt heart failure symptoms and enhances the diagnostic capability for CCM.

However, the mechanisms contributing to GLS impairment in the cirrhotic heart have yet to be fully elucidated. It is crucial to recognise that GLS is not a specific marker for CCM; instead, it reflects the influence of prevalent cardiovascular risk factors or coexisting cardiac conditions. Furthermore, the availability of strain echocardiography is not universal in routine clinical practice, and the routine interdisciplinary collaboration between cardiologists and hepatologists is not always established.

A simple clinical tool would be advantageous for identifying cirrhotic patients at an elevated risk for myocardial dysfunction, particularly those who would benefit most from further cardiological evaluation.

Our research aimed to develop and validate a predictive model based on a scoring system, utilising routinely obtained patient characteristics, laboratory, and clinical variables, including liver stiffness measurements (LSM) and controlled attenuation parameter (CAP) by vibration-controlled transient elastography. The model is designed to reliably differentiate patients with liver cirrhosis at risk for reduced global longitudinal strain from those with a low probability of myocardial dysfunction.

2 Methods

2.1 Study Design and Participants

The study cohort comprised consecutive patients with liver cirrhosis, irrespective of aetiology, prospectively assessed at a university-affiliated tertiary care hospital in Austria. Recruitment occurred during inpatient and outpatient visits between January 2019 and July 2022 as part of a collaborative effort between the Departments of Gastroenterology, Hepatology, and Cardiology. All participants underwent initial screening, which included a comprehensive medical history and physical examination, followed by a thorough evaluation of hepatic and cardiac function (Figure 1).

In addition, we included an independent validation cohort of prospectively examined consecutive patients, recruited after completing the training group, who underwent the same diagnostic procedures.

Consistent with the definition of cirrhotic cardiomyopathy, which is diagnosed by ruling out other potential etiologies [12] and to ensure a more accurate evaluation of the liver-heart axis, individuals with pre-existing non-cirrhotic cardiomyopathy, such as ischemic, valvular or hypertensive heart disease-conditions that could impair strain measurements-were excluded from the final analysis.

Patients were also excluded if they had a history of portal vein thrombosis, transjugular intrahepatic portosystemic shunt (TIPS) implantation, hepatocellular carcinoma outside Milan criteria, failed transient elastography measurements, or refused to participate in the study (Figure 1).

Enrolled participants underwent longitudinal follow-up with regular clinical evaluations and liver function tests tailored to individual patient conditions. Mortality rates were systematically recorded throughout this period.

Written consent was obtained from all participating subjects. This study adhered to the principles of the 2013 Declaration of Helsinki and was approved by the local ethics committee (Ethikkommission Kärnten No. A21/19).

3 Results

3.2 Prevalence of Cirrhotic Cardiomyopathy and Cardiological Assessment

The prevalence of cirrhotic cardiomyopathy among the participants in the training group was 18.8%, as determined using the 2019 CCC diagnostic criteria. Reduced global longitudinal strain, assessed by speckle-tracking echocardiography, was identified in 13 (9.8%) of the cirrhotic patients in the training group and 8 (19.5%) in the validation group (p = 0.16). Only two presented with reduced LVEF ≤ 50%, indicating that most patients with end-stage liver disease have a subclinical contractile dysfunction.

The mean absolute value of GLS in the validation group was lower than in the training group (20.2% ± 3.1% vs. 21.6% ± 2.9%, p = 0.01), reflecting the non-balanced distribution of covariates, such as aetiology and haemodynamic-related factors between the groups.

In the training group, 10 patients (7.5%) met more than two diagnostic criteria for diastolic dysfunction. An additional 18 patients (13.5%) presented with exactly two criteria and were classified as having “indeterminate” diastolic function. Following comprehensive echocardiographic assessment using additional parameters, 7 patients from the indeterminate group were reclassified as having diastolic dysfunction, resulting in an overall prevalence of 12.8%. Five patients with reduced GLS were also diagnosed with diastolic dysfunction.

Patients presenting reduced GLS underwent further evaluation for the presence of coronary artery disease (CAD) (Figure 2). Among the cirrhotic patients in the training group, a confirmed or suspected CAD, indicated by elevated Agatston scores, was identified in four individuals. In three cases, CAD was definitively excluded. Additionally, three patients died before a cardiological assessment could be conducted within the study period, two declined the examinations, and one developed atrial fibrillation during follow-up (Table 2).

TABLE 2. The diagnosis and outcomes of coronary artery disease during the study period in patients with liver cirrhosis presenting with reduced GLS.

Patient	Diagnosis	BMI	CAP	LSM	CP	DD	CAD work-up/outcome
61 M	HBV	27,8	266	27	A	1	CACS 414, PTCA of RCA
58 M	ALD	29,7	271	21,5	C	0	CACS 98, PTC: CAD excluded/TIPS in follow-up
58 W	MASLD	36,1	271	72	A	0	CACS 501; no PTCA/TIPS in follow-up
77 W	ALD	28,6	281	26	A	1	ICMP; PTC (no stent)
71 W	MASLD	31,9	288	13	A	0	CACS 0
58 M	ALD	28,7	291	12,5	C	1	CACS 1400/died; obduction (right heart and acute liver failure)
72 M	ALD	24,7	302	19,2	B	0	No CAD work-up/atrial fibrillation in follow-up
63 M	ALD, HCC	33,1	314	9,7	A	0	Died before CAD work-up
63 M	ALD	35,6	330	75	B	0	Died before CAD work-up/Covid-19
61 M	HCV	31,1	375	27	A	0	CACS 425, PTCA of LAD and LCX
46 M	HCV	31,2	385	70,6	B	0	PTC: CAD excluded

• Abbreviations: ALD, Alcoholic Liver Disease; BMI, Body Mass Index; CACS, Coronary Artery Calcium Score; CAD, Coronary Artery Disease; CAP, Controlled Attenuation Parameter in dB/m; CP, Child-Pugh Class; DD, Diastolic Dysfunction; HBV, Hepatitis B Virus; HCC, Hepatocellular Carcinoma; HCV, Hepatitis C Virus; LSM, Liver Stiffness Measurement in kPa; MASLD, Metabolic Dysfunction-Associated Steatotic Liver Disease; PTCA, Percutaneous Transluminal Coronary Angioplasty.

3.4 Prediction of Reduced GLS and Model Building

Independent variables significantly correlated with the reduced GLS were entered into the multiple logistic regression and weighted based on their respective beta coefficients (Table 3). Subsequently, these weighted variables were combined into a composite scoring algorithm, termed the Liver-heart score, designed to differentiate between cirrhotic patients with and without reduced global longitudinal strain.

Various combinations of CAP > 260 dB/m, BMI ≥ 28 kg/m², age > 57 years, and smoking were selected for multiple logistic regression and assessed using AUC. The optimal model included CAP > 260 dB/m, BMI ≥ 28 kg/m² and age > 57 years, yielding an AUC of 0.84. On multivariable analysis, smoking was not independently associated with myocardial dysfunction (Beta-Coefficient: −0.47, 95% CI: 0.87–0.59, p = 0.59). Including smoking in the model reduced its predictive performance (AUC of 0.83) and was therefore not included in the final model.

Based on the beta coefficients for the final model, CAP > 260 dB/m and BMI ≥ 28 kg/m² were each assigned 2 points, and age > 57 years 1 point, to derive the Liver-heart score, which can reach a maximum value of 5 points.

By applying a cut-off value of > 2 points, we observed a sensitivity of 92.3% (95% CI: 64%–99.8%) and a specificity of 66.7% (95% CI: 57.5%–75%). Among the 81 patients scoring ≤ 2 points, only one (1.2%) was incorrectly classified, suggesting a high negative predictive value (98.8%, 95% CI: 93.3%–100%) for ruling out myocardial dysfunction in cirrhotic patients using the proposed scoring system (Table 4, Figure 3).

TABLE 4. Predictive value of risk factors in cirrhotic patients and performance of the predictive score for reduced GLS in the training and validation groups.

Variable	Cutoff	Group	AUC (95% CI)	Se (%)	Spe (%)	PPV (%)	NPV (%)
CAP	> 262	Training	0.68 (0.58–0.77)	90.0	57.9	18.4	98.2
		Validation	0.67 (0.46–0.84)	80.0	50.0	28.6	90.9
BMI	> 27.6	Training	0.78 (0.70–0.85)	76.9	79.2	28.6	96.9
		Validation	0.68 (0.51–0.82)	50.0	78.6	40.0	84.6
Age	> 57	Training	0.65 (0.56–0.73)	84.6	51.7	15.9	96.9
		Validation	0.64 (0.47–0.79)	62.5	58.1	27.8	85.7
Scorea	> 2	Training	0.84 (0.77–0.90)	92.3	66.7	23.1	98.8
		Validation	0.83 (0.65–0.94)	80.0	75.0	57.1	90.0

• Abbreviations: AUC, Area Under the Curve; BMI, Body Mass Index; CAP, Controlled Attenuation Parameter; NPV, Negative Predictive Value; PPV, Positive Predictive Value; Se, Sensitivity; Spe, Specificity.
• ^a Liver-heart score: CAP > 260 dB/m (2 pts) + BMI ≥ 28 kg/m² (2 pts) + Age > 57 years (1 pt).

For a high cut-off of > 3 points, a sensitivity of 53.9% (95% CI: 25.1%–80.8%), a specificity of 90% (95% CI: 83.2%–94.7%), and a negative predictive value of 94.7% (95% CI: 88.9%–98%) were observed.

3.7 Survival Analysis

During a median follow-up of 40 months (range: 0.5–65), all-cause mortality was 34.5% (46/133). Four patients who developed malignancy (two hepatocellular carcinomas, one lung cancer, and one colon cancer) and subsequently died from tumour progression were excluded from the survival analysis. Patients diagnosed with CCM were associated with higher mortality (55% vs. 28.4%) compared to those with cirrhosis but negative echocardiographic findings (HR 2.5, 95% CI: 1.2–4.9, p = 0.01). Another strong predictor of higher mortality in the Cox regression analysis was the Child-Pugh classification (HR: 2.8, 95% CI: 1.5–5.3, p < 0.01). In the multivariable model, both variables remained significant, suggesting an independent association of CCM with mortality.

The mortality rate was also higher among patients with cirrhosis and reduced absolute GLS values (53.8% vs. 30.2%, HR: 2.2, 95% CI: 1–4.7, p = 0.05), although only seven events were reported in this subgroup (Figure 4).

Additionally, a Liver-heart score of 5 points was associated with increased mortality, observed at both 2-year follow-up (44.4% vs. 17.3%) and the end of the follow-up period (66.7% vs. 37.7%, HR: 1.3, 95% CI: 1.1–1.5, p < 0.01).

4 Discussion

Delayed recognition of myocardial dysfunction in cirrhotic patients carries unfavourable prognostic implications; therefore, improving diagnostic approaches to identify those at risk for cardiovascular (CV) events early in the disease course is desirable and should be encouraged [18].

Cirrhotic cardiomyopathy presents with a diverse range of structural and functional cardiac abnormalities, including electrophysiological changes in asymptomatic individuals, variations in cardiac output, and clinical manifest heart failure. The diagnostic workup of this condition requires a methodical exclusion of other prevalent cardiac pathologies, employing criteria similar to those applied in our study. CCM has previously been associated with hepatorenal syndrome, major CV events, and increased mortality, particularly during the post-liver transplant period [19]. A recent study utilising revised consensus criteria for CCM reported a hazard ratio of 2.6 (95% CI: 1.2 to 5.5) for CV events following liver transplantation [20].

The reported prevalence of CCM according to these revisited criteria is estimated at approximately 20%, and may increase up to 47% when metabolic factors, fatty liver disease, and alcohol abuse coexist [20]. However, the existing data have been inconsistent across studies, partly due to a critical error in the initial publication of the updated diagnostic criteria. The original publication suggested both an absolute GLS < 18% and > 22% as markers of systolic dysfunction, with the upper threshold later removed in an erratum [21]. This error contributed to several studies reporting an inaccurately high prevalence of CCM [4 22]. For instance, Cesari et al. reported an overall CCM prevalence of 29%, with systolic dysfunction defined by either reduced or elevated GLS-accounting for 25%, regardless of cirrhosis aetiology or the presence of ascites. Consistent with our findings, advanced diastolic dysfunction was identified in 10% of cases, although no information was provided on “indeterminate” diastolic function [22]. In contrast, Izzy and colleagues, applying the same criteria, identified reduced GLS in only two patients (1.7%) but reported a much higher prevalence of diastolic dysfunction (33.3%) compared to other authors, suggesting significant methodological differences between these studies [20]. In another study, only 4.5% of patients met more than two diagnostic criteria for diastolic dysfunction, while 11.1% were classified as having indeterminate diastolic function. After further evaluation, 6 of the 10 patients in this indeterminate group were reclassified as having diastolic dysfunction, resulting in a prevalence of 11.1%, closely reflecting our results [23]. The lack of clear recommendations for managing these cases, which often rely on less reliable additional parameters, represents a limitation in the accurate diagnosis of CCM.

GLS has been proposed as the main parameter for routine clinical assessment of systolic function in cirrhotic patients and should be utilised whenever possible. However, the additional strain measurements can be time-consuming and the required software is not universally available, particularly in resource-limited settings. Furthermore, collaboration between haepatology and cardiology departments is not always well-established.

Values below 20.5% were linked to a six-fold increase in the incidence of heart failure following liver transplantation [20]. In another cohort study, GLS at rest was identified as an independent predictor of mortality among liver transplant candidates over a three-year follow-up period [24].

This study demonstrates a significant correlation between hepatic steatosis and subclinical myocardial dysfunction in patients with liver cirrhosis, irrespective of the disease's aetiology. For the first time, we identified a linear relationship between the degree of hepatic steatosis, quantified by the controlled attenuation parameter, and myocardial contractility as measured by global longitudinal strain. Our analysis further confirms that this correlation persists as statistically significant even after adjustment for shared cardiovascular risk factors, such as elevated BMI. These findings underscore the profound and independent impact of hepatic steatosis on myocardial contractility in cirrhotic patients. Targeting coexisting hepatic steatosis in liver cirrhosis with the objective of enhancing myocardial contractility represents a potentially paradigm-shifting strategy in the clinical management of these patients. Such a strategy would integrate therapeutic interventions that address the multifaceted nature of cirrhosis, potentially redefining treatment protocols and improving patient outcomes.

Our findings are supported by previous studies, which have demonstrated that the degree of hepatic steatosis in patients with both alcoholic and non-alcoholic fatty liver disease is independently associated with impaired cardiac function, even after adjusting for confounding factors such as type 2 diabetes, age, gender, and BMI [25]. Although we found no correlation between liver stiffness values and either reduced GLS or overall mortality in our cirrhotic cohort, a prospective study on critically ill patients demonstrated that liver stiffness exceeding 18 kPa at ICU admission was associated with increased ICU and long-term mortality, even in non-cirrhotic patients [26].

Obesity has been identified as a significant risk factor for reduced GLS in cirrhotic patients, corroborating previous findings in liver transplant candidates [24] and supported by population-based studies [27]. Our analysis revealed that patients with a BMI of 28 kg/m² or higher exhibited significantly lower absolute GLS values compared to those with a BMI below this cutoff (−20.4% vs. −21.9%, p < 0.01).

Utilising the positive correlations observed among parameters routinely assessed during hepatological examinations, a novel scoring system, termed the Liver-heart score, was developed to differentiate cirrhotic patients at risk for reduced global longitudinal strain from those with a low probability of myocardial dysfunction. This predictive model demonstrated excellent diagnostic accuracy, with a sensitivity of 92.3% and a negative predictive value of 98.5%. Cirrhotic patients with a score of 2 or less seem to be at low risk for myocardial dysfunction, with only 1.2% of such cases misclassified. Those exhibiting a CAP below 260 dB/m are categorised as low risk for cardiac dysfunction, regardless of BMI or age.

Furthermore, cirrhotic patients with a Liver-heart score > 3 points were found to be nine times more likely to exhibit reduced GLS compared to those with lower scores. 60% of patients meeting all three criteria demonstrated reduced GLS, indicating that this subgroup might particularly benefit from comprehensive cardiological assessment. However, practitioners must keep in mind that the Liver-heart score cannot replace a conventional echocardiographic examination, as it is still essential to evaluate diastolic function, particularly in cirrhotic patients with pre-existing cardiac conditions or hypertensive disease.

The score is intended to complement existing diagnostic tools in clinical practice. It can be easily calculated using readily available parameters during routine assessments of cirrhotic patients, and is useful particularly in those with no prior cardiac history and negative findings on conventional echocardiography. Furthermore, it could serve as a decision support tool to optimise the utilisation of often limited resources.

Patients exhibiting reduced GLS underwent further evaluation for coronary artery disease using either a coronary calcium scan to calculate the Agatston score or coronary angiography, based on their clinical presentation. CAD was confirmed or suspected in four individuals, while three were definitively cleared of the condition. Additionally, three patients died before a cardiological assessment could be conducted within the study period, two declined the examinations, and one developed atrial fibrillation during follow-up. Overall, a clinically significant outcome was observed in 72.7% of evaluated cirrhotic patients with reduced GLS, underscoring the critical importance of early detection of impaired GLS and highlighting the potential impact of the score.

A prospective study by Wang et al. assessed long-term clinical outcomes in patients screened for fatty liver disease who subsequently underwent coronary angiography. They reported that patients with MASLD exhibited a significantly higher risk of developing coronary stenosis (84.6% vs. 64.1%; p < 0.001) and an increased need for percutaneous coronary intervention. In this cohort, older age and diabetes emerged as independent predictors of cardiovascular events [28]. Additionally, a recent animal study highlighted that advanced liver fibrosis in mice induces oxidative stress, vascular inflammation, and dysfunction, potentially contributing to the development of coronary artery disease [29].

In addition to pro-inflammatory and pro-coagulant mediators, metabolic factors critically influence the pathogenetic mechanism underlying the interaction between liver and heart function. A comprehensive study demonstrated that the degree of hepatic steatosis, along with impaired fatty acid and glucose metabolism and insulin resistance in the myocardium, was associated with cardiac dysfunction [30]. Moreover, impairment of GLS in cirrhotic patients could also be mediated by factors such as epicardial adipose tissue, which has been previously linked to liver fibrosis and steatosis, cardiac dysfunction, and the development of CAD [31].

Strain imaging has been shown to identify LV dysfunction earlier than conventional echocardiography, thereby offering a new perspective on the diagnostics of CCM and the potential for early therapeutic interventions before irreversible myocardial dysfunction occurs [32]. A peak GLS in the range of −20% is considered normal in a healthy person, with lower absolute values indicating a higher probability of myocardial dysfunction. Generally, a lower limit of normal in the range of −18% is accepted and has recently been adopted in the hepatological criteria [11 14]. However, the application of these standard values to cirrhotic patients is limited by systemic vasodilation and decreased afterload, leading to a hyperdynamic circulation that complicates the interpretation of GLS values. Consequently, higher absolute cutoff values have been considered when assessing LV contractility in cirrhotic patients compared to the general population [20]. Our results show significantly lower absolute GLS values in the validation group compared to the training group (20.2% vs. 21.6%, p = 0.01), reflecting the unbalanced distribution of covariates such as aetiology, Child-Pugh classification, and haemodynamic factors including portal hypertension, cardiac index, and systemic vascular resistance index between the groups. The fact that our model performed well on the validation cohort, despite these differences, underscores its robustness and its ability to reflect real-world clinical scenarios with heterogeneous patient characteristics. Future studies should address the impact of hypercirculatory changes on GLS values and the prevalence of CCM, particularly in the context of covariates such as age, BMI and CAP.

Our well-characterised prospective study has some limitations. The strict exclusion criteria are consistent with the current definition of CCM, which is diagnosed by ruling out other possible etiologies [12]. However, this approach may introduce selection bias. Hypertension-mediated cardiac dysfunction, a common cause of impaired GLS in the general population, was excluded along with hypertension-related heart disease, which is prevalent in unselected cohorts. This was intended to facilitate a more accurate examination of the liver-heart axis. The meta-analysis conducted by Tadic et al. demonstrated that GLS is more sensitive than conventional ejection fraction measurements in detecting hypertension-induced cardiac damage [33].

A prospective validation cohort better reflects real-world scenarios by accounting for possible temporal changes in patient populations, diagnostic practices, and referral pathways. The aim of our study was to develop a predictive model that can be prospectively applied in clinical practice, making the use of a prospective validation cohort both more generalisable and clinically relevant compared to randomised validation. Furthermore, retrospective 5-fold cross-validation of the regression model with the selected variables in the combined cohort demonstrated consistently high predictive performance.

Nevertheless, external validation in more heterogeneous populations, including diverse ethnicities and utilising various healthcare systems and software versions, is essential to ensure the score's applicability and reliability.

In summary, we identified simple-to-measure clinical predictors for cirrhotic cardiomyopathy as determined by reduced GLS and developed a predictive model termed the Liver-heart score. This model was subsequently validated in an independent cohort, demonstrating excellent performance in differentiating patients with reduced GLS from those with a low probability of myocardial dysfunction. The relevance of the incorporated parameters for cardiac dysfunction is supported by previous well-conducted studies. The Liver-heart score may enhance the utility of conventional echocardiography by enabling more precise cardiovascular risk stratification in cirrhotic patients. It may also facilitate timely referral for advanced diagnostics or intervention in contexts where specialised testing is inaccessible.

Author Contributions

Conceptualization: M.R., S.B., M.P-R.; Methodology: M.R., S.B., M.P-R., H.A., F.T., C.E.; Investigation: M.R., S.B., P.W., P.H., M.H., M.F.; Data Curation: M.R., S.B., P.W.; Formal Analysis: S.B., M.R., N.W.; Writing original draft: M.R., S.B., M.P.-R.; Critical revision of the manuscript: F.T., M.D., C.E., A.W., R.M., N.W.; Project administration: M.R.; Supervision: M.P-R., F.T.

Ethics Statement

This study adhered to the principles of the 2013 Declaration of Helsinki and was approved by the local ethics committee (Ethikkommission Kärnten No. A21/19).

Conflicts of Interest

The authors declare no conflicts of interest.