Risk prediction and outcomes in patients with liver cirrhosis undergoing open-heart surgery

Abstract

Objective: There are few data assessing factors, which identify patients with liver cirrhosis (LC) facing high risk for open-heart surgery. We sought to compare the Model for End-Stage Liver Disease (MELD) score, the Child–Turcotte–Pugh (CTP) classification and the European system for cardiac operative risk evaluation (EuroSCORE) for risk prediction in cirrhotic patients. Methods: Fifty-seven consecutive patients with non-cardiac LC, who underwent open-heart surgery with the use of cardiopulmonary bypass between 1998 and 2008, were studied at our institution. Potential preoperative predictors of outcome, as well as preoperative MELD score, CTP classification and EuroSCORE were calculated. The primary study end points were all-cause in-hospital and long-term mortality. Results: MELD score and CTP classification both differed significantly between survivors and non-survivors for in-hospital (P ≪ 0.0001) and long-term mortality (P ≪ 0.0001). Univariate predictors of in-hospital mortality were emergency surgery (odds ratio (OR), 4.9; 95% confidence interval (CI), 1.2–20.6; P = 0.03), ascites (OR, 7.2; 95% CI, 2.0–25.5; P = 0.002), total serum protein (OR, 0.4; 95% CI, 0.2–0.8; P = 0.01), CTP class (OR, 5.5; 95% CI, 1.4–21.5; P = 0.04) and MELD score (OR, 1.4; 95% CI, 1.1–1.6; P = 0.001). Multivariable exact logistic regression analyses revealed MELD score (OR, 1.3; 95% CI, 1.005–1.6; P = 0.04) as the only independent factor associated with in-hospital mortality. Receiver operating characteristic curve (ROC) analysis showed MELD score to be highly predictive with an optimal cut-off value of 13.5 (sensitivity: 82.0%, specificity: 78.5%) for postoperative in-hospital mortality (area under curve (AUC): 85.1 ± 0.05%) and superior compared to the CTP classification (AUC: 75.7 ± 0.08%) and EuroSCORE (AUC: 65.9 ± 0.08%). Conclusions: The mortality of patients with liver cirrhosis undergoing open-heart surgery progressively increases with the severity of liver dysfunction. Therefore, the MELD score most reliably identifies those cirrhotic patients who are at high risk for open-heart surgery.

Liver cirrhosis, Risk prediction, MELD score, Child–Pugh classification, EuroSCORE

1 Introduction

Liver cirrhosis has been shown to be a major preoperative risk factor in general surgery with a strong correlation between the severity of preoperative liver disease and postoperative patient’s outcome [1,2]. Although liver dysfunction and liver cirrhosis are also considered to significantly increase perioperative morbidity and mortality in cardiac surgical patients, there are no data separating patients who benefit from cardiac surgery from those in whom perioperative risk outweighs benefit. In fact, liver disease has not been addressed as an independent preoperative risk factor by any of the routinely used and established cardiac surgical risk score models (e.g., European system for cardiac operative risk evaluation (EuroSCORE); except for Pons et al.) [3–6]. Recent evidence only comes from several small studies showing highly increased short- and long-term mortality rates for those patients, who undergo cardiac surgery having end-stage liver disease with liver cirrhosis of Child–Turcotte–Pugh (CTP) classification B or C [7–12]. On the other hand, over the past several decades, liver transplantation has emerged as a viable treatment option for patients with end-stage liver disease [13], and the prevalence of concomitant cardiac disease in these patients often carry cardiac risk factors, such as ageing, obesity, diabetes mellitus, hypertension and hyperlipidaemia [14]. The above cardiovascular risk factors have been shown to increase post-liver transplantation complications [15], and thus cardiac surgery, especially in huge transplant centres, is faced with increasingly more patients with end-stage liver disease awaiting liver transplantation, who require open-heart surgery prior to liver transplantation. Hence, a preoperative risk assessment is an important clinical requirement and essential in these high-risk patients before they undergo open-heart surgery. While the Model for End-Stage Liver Disease (MELD) was initially derived in a small number of patients undergoing transjugular intrahepatic portosystemic shunt (TIPS) procedures to predict survival [16], to date MELD score has become a standard tool for allocation of organs for liver transplantation [17], as well as for risk prediction in several cohorts of patients with varying severity of liver disease [18,19]. Whether the MELD score, the CTP classification or possibly the EuroSCORE are helpful tools to predict the risk of morbidity and mortality after cardiac surgery in these cirrhotic patients has not yet been evaluated in a larger cohort. We therefore sought to compare the predictive value of MELD, CTP classification, as well as EuroSCORE in patients with liver cirrhosis undergoing open-heart surgery at a huge transplant centre.

2 Patients and methods

2.1 Data collection and study design

This study was a retrospective single-centre cohort study including all consecutive patients, who underwent open-heart surgery with the use of cardiopulmonary bypass at the West-German Heart Center Essen between January 1998 and July 2008. The study was approved by the Institutional Review Board. All of the patients had previously granted permission for use of their medical records for research purposes.

Patients with end-stage liver disease and a preoperative known diagnosis of liver cirrhosis were retrieved from the institutional cardiovascular database. Once these patients were identified, a retrospective medical chart review was performed to confirm the diagnosis of liver cirrhosis. The diagnosis of liver cirrhosis was finally confirmed either by liver biopsy when applicable or by characteristic clinical signs of portal hypertension of the liver and spleen verified by radiological findings (ultrasound, computer tomography or magnetic resonance imaging). In addition, a confirmed diagnosis of liver cirrhosis was also verified by several diagnostic criteria, such as the presence of encephalopathy, ascites or oesophageal varices. The severity of liver cirrhosis was classified by using the MELD score, as well as the CTP classification. MELD was calculated by incorporating a total of three preoperative laboratory variables: (1) the international normalised ratio (INR), (2) serum creatinine and (3) serum bilirubin into the mathematical formula: MELD = 9.57 × log_e (creatinine) + 3.78 × log_e (total bilirubin) + 11.2 × log_e (INR) + 6.43 [20]. The CTP classification was calculated by incorporating a total of five preoperative variables: (1) severity of hepatic encephalopathy, (2) the presence of ascites and (3) total bilirubin, serum albumin, and prothrombin time. The CTP class was evaluated to class A, B, or C, respectively [21]. To further compare these risk score models, the logistic EuroSCORE was calculated for each patient.

2.2 Data collection

Data used in this analysis were retrieved from the West-German Heart Center cardiovascular surgical database. This database prospectively collects a comprehensive list of prespecified data points, with more than 1800 data items per patient, including demographic, clinical and outcome data in all of the consecutive patients undergoing open-heart surgery at our institution. Within the database, patients were coded as having liver cirrhosis preoperatively before cardiac surgery.

2.3 Definitions and outcome measures

The outcome measures used in this analysis were specified earlier. Given the subject nature of many clinical outcomes, we only prespecified all-cause in-hospital (30 days) mortality as well as long-term survival after surgery as the primary study end point. Prespecified secondary end points were the postoperative complication rate, including (a) perioperative myocardial infarction (PMI), (b) low cardiac output syndrome (LCOS) with high-dose inotropic support or the use of an intra-aortic balloon pump, (c) the necessity for thoracic re-exploration due to mediastinal bleeding or cardiac tamponade and (d) major bleeding, which was defined as drainage loss of >200 ml h⁻¹ during the first 12 h, or (e) the total amount of thoracic drainage loss within the first 48 h after surgery. In-hospital death was defined as all-cause in-hospital or operative mortality within 30 days after surgery. Long-term survival was defined as survival between cardiac surgery and the time point of mortality until July 2008 and obtained by a retrieval of the population registry of the regional registry office. When no death was recorded, the patient was considered to be alive. A PMI was considered to have occurred, if one of the following diagnostic criteria was present: (1) a cardiac troponin I level above 10.5 ng ml⁻¹ after coronary artery bypass grafting (CABG), as previously described [22], (2) a creatine kinase-MB level 3 times above the upper normal level, (3) new persistent ST-segment or T-wave changes (Minnesota code 4-1, 4-2, 5-1, 5-2 or 9-2), or (4) the development of new Q-waves (Minnesota code 1-1-1 to 1-2-7). LCOS was supposed with a cardiac index below 2.0l min⁻¹m⁻² or a systolic arterial pressure below 90 mm Hg, despite high-dose inotropic support (IV dopamine ≥8 μg kg⁻¹ min⁻¹ or dobutamine ≥6 μg kg⁻¹ min⁻¹ or epinephrine >0.1 μg kg⁻¹ min⁻¹ or norepinephrine >0.1 μg kg⁻¹ min⁻¹).

2.4 Surgical methods

Cardiac surgical procedures were performed in all patients using median sternotomy, standard cardiopulmonary bypass (CPB) technique with ascending aortic and either two-stage venous or both venae cavae canulation, mild hypothermia (>32 °C) and cold crystalloid cardioplegic arrest. Heparin was administered to achieve an activated coagulation time >400 s. Bypass graft flow was assessed of each graft by Doppler transit time flowmetry. Protamine was administered to reverse heparin according to standard practice. For CABG surgery, a medication of 500 mg aspirin was routinely administered within the first 6 h after surgery followed by a daily dose of 100 mg.

2.5 Statistical analysis

Descriptive statistics are summarised for categorical variables as frequencies (%) and compared between groups using Pearson’s chi-square exact test. Continuous variables, expressed as mean ± SD, were compared between groups using the Satterthwaite (unequal-variance) t-test. This test was recommended for general use. Univariate and multivariate logistic regression analyses were performed to identify preoperative independent predictors for in-hospital mortality. Those variables identified by univariate regression analysis with a P value ≤0.05 were added to a multivariate logistic regression model. Because of the relative small number of events, an exact multivariate logistic regression was performed. Graphs for in-hospital and long-term survival were generated with the Kaplan–Meier method; differences between Kaplan–Meier curves were tested with log-rank tests. Receiver operating characteristic (ROC) curve analyses were applied to determine optimal cut-off values of MELD for in-hospital mortality and further to evaluate the predictive power of MELD in comparison to CTP classification and the logistic EuroSCORE. The comparisons of ROC curves were performed by using a contrast matrix to take differences of the areas under the empirical ROC curves. A P value less than 0.05 was considered to indicate statistical significance. Statistical analyses were performed using SAS System^®, version 8 (SAS Institute Inc., Cary, NC, USA) and LogXact, version 6 (Cytel Software, Cambridge, MA, USA).

3 Results

The investigation was conducted between January 1998 and July 2008 at the West-German Heart Center Essen. Initially, a total of 82 patients with supposed liver cirrhosis undergoing cardiac surgery were identified from the institutional database. Out of these, 25 patients were excluded from the study: 13 patients were excluded due to unproven or incomplete diagnosis of liver cirrhosis, and five patients due to incomplete laboratory data. Seven patients, who had non-open-heart surgery without CPB, were excluded. Another two patients who underwent off-pump coronary artery bypass surgery (OPCAB) remained consciously in the present study, although they underwent surgery without CPB. Therefore, a total of 57 patients fulfilled the inclusion criteria with a confirmed preoperative diagnosis of liver cirrhosis before open-heart surgery with the use of CPB (except for the two OPCAB cases) and were included into the present retrospective study.

3.1 Demographics and baseline data

Preoperative demographic and baseline characteristics of the patients are shown in Table 1 . Mean age at surgery was 62 ± 10 years; 77% of the patients were male. A total of 10 patients (18%) underwent emergency surgery with significantly more emergency cases in the non-survivor group (32% vs 10%; P = 0.05). The considerable prevalence of patient’s preoperative risk factors and co-morbidities (e.g., diabetes 37%, previous stroke 18%, chronic obstructive lung disease (COPD) 30%, renal failure 40%, previous myocardial infarction (MI) 32%, ascites 40% and oesophageal varices 30%) represents a high-risk group of cardiac surgical patients. Moreover, five (9%) of the patients had acute endocarditis, and 19 (34%) patients underwent CABG with concomitant valve surgery or multiple valve surgery. However, survivors and non-survivors did not differ according to their demographics, risk factors and co-morbidities, except for a significantly higher number of patients in the non-survivor group with ascites and oesophageal varices. Regarding the preoperative laboratory data of the cirrhotic patients listed in Table 1, mostly all of them were pathological because of patient’s end-stage liver disease, significantly differing between survivors and non-survivors. According to the preoperative risk assessment, 39 patients (68%) presented with CTP class A, 14 (25%) with class B, and at least four patients (7%) with CTP class C. As a result of this high-risk group of patients, the mean logistic EuroSCORE was 14.7 ± 16.9 on average, which was not significantly different between survivors and non-survivors. In contrast, the mean MELD score and the CTP classification were significantly different according to the survival status.

Table 1

Demographics and baseline data.

3.2 Intra-operative data

A total of 57 patients with liver cirrhosis underwent open-heart surgery. Thirty-five patients had coronary artery disease and thus, underwent CABG surgery, 24 (42%) had isolated CABG and 11 (20%) combined CABG with valve surgery, whereas 30 (53%) patients had heart-valve disease and thus underwent isolated valve surgery (n = 11) and multiple valve surgery (n = 8). Finally, another three (5%) patients underwent aortic surgery or cardiac tumour extirpation. Intra-operative data, such as aortic cross-clamp time, CPB time and reperfusion time are listed in Table 2 , showing a trend towards slightly longer aortic cross clamp time (ACC) time in the non-surviving group.

Table 2

Intra-operative and postoperative data.

3.3 Postoperative in-hospital outcome

In-hospital mortality at 30 days was 29.8% (n = 17; Fig. 1(a) ), caused by acute hepatic decompensation with multisystem organ failure (n = 8), sepsis (n = 5), postoperative myocardial infarction (n = 2), uncontrollable haemorrhage (n = 1) and necrotic bowl ischaemia (n = 1). Postoperative morbidities included respiratory failure (n = 12), renal failure requiring dialysis (n = 25), major bleeding (n = 10) and thoracic re-exploration (n = 14), as well as other postoperative complications shown in Table 2. As a result, according to the patients’ hepatic dysfunction with impaired haemostasis and pathologic coagulation status, postoperative thoracic drainage loss differed significantly between survivors and non-survivors, leading to a significant number of patients with major bleeding and thoracic re-exploration (Table 2).

Fig. 1

(a and b) In-hospital and long-term survival.

3.4 Long-term survival

All surviving patients had a mean follow-up of 1528 ± 256 days after surgery. Long-term survival for all cirrhotic patients, including in-hospital mortality, was 55.0% at 6 months, 53.0% at 1 year, 38.3% at 6 years, 26.3% at 10 years and at least 13.1% at 12 years’ follow-up (Fig. 1(b)).

3.5 MELD score and CTP classification

To evaluate preoperative predictors of death for all cirrhotic patients, a logistic regression analysis model was constructed. Several univariate factors such as emergency status, total serum protein, preoperative ascites, the CTP classification as well as the MELD score were associated with in-hospital death. The multivariate exact logistic regression analysis, testing the contribution of the univariate risk predictors, is shown in Table 3 . After risk adjustment, only the MELD score was significantly predictive for in-hospital death.

Table 3

Univariate and multivariate logistic regression analysis of variables associated with in-hospital death.

Comparing the discriminative power of the MELD score, the CTP classification, and the EuroSCORE for in-hospital mortality by using the ROC curve analyses yielded MELD score as the most predictive score with an area under curve (AUC) of 0.85 ± 0.05 (Fig. 2 ). The optimal cut-off level for the MELD score was set at 13.5 with a sensitivity of 82.0% and a specificity of 78.5% (Table 4 ). By contrast, ROC analyses of CTP classification and EuroSCORE in the same patient cohort both revealed a lower predictive power with an AUC of 0.76 ± 0.07 and 0.66 ± 0.08, respectively (Table 4 and Fig. 2). The difference between the AUCs based on MELD score and EuroSCORE was significant (P = 0.04, Table 4).

Fig. 2

ROC curve analyses of MELD, Child–Turcotte–Pugh class, and EuroSCORE for in-hospital mortality.

Table 4

Cut-off levels, test characteristics and comparison of ROC curves for MELD, CTP class, and EuroSCORE.

Moreover, as shown by Kaplan–Meier survival curves in Fig. 3(a) and (b) , in-hospital mortality and long-term mortality was significantly lower for patients with a MELD score ≪13.5, as compared to those patients with a MELD score >13.5 (P ≪ 0.0001). According to the MELD score, survival at 1 year and 5 years was 23.8% with MELD >13.5 as compared to 74.6% at 1 year and 52.1% at 5 years with MELD ≪13.5 (P ≪ 0.0001). Similarly, as shown in Fig. 4(a) and (b) , by stratifying the patients with the use of CTP classification, Kaplan–Meier survival curves again significantly differed according to in-hospital (P ≪ 0.0001) and long-term mortality (P ≪ 0.0001) between CTP classes A, B and C. Long-term survival with CTP class A was 67.9% at 1 year, 45.6% at 5 years, 28.5% at 10 years and 14.3% at 12 years, whereas survival in patients with CTP class B was 26.8% at 1 year and 5 years and remained unchanged thereafter. All patients with CTP class C died during the 30-days in-hospital period (Fig. 4(b)).

Fig. 3

(a and b) In-hospital and long-term survival according to MELD score ≪ (----) or >13.5 (—).

Fig. 4

(a and b) In-hospital and long-term survival according to Child–Turcotte–Pugh classification A (—), B (– –), and C (----).

4 Discussion

In the present study, we analysed the in-hospital outcome and long-term survival of patients with liver cirrhosis undergoing open-heart surgery with the use of CPB at a busy transplant centre – this is to our knowledge the largest series of patients published to date. We sought to determine whether specific preoperative predictors of survival exist in patients with liver cirrhosis by incorporating all preoperative risk factors and co-morbidities and assessed several established risk score models, including the MELD score, the CTP classification, as well as the EuroSCORE. Regarding the predictive value of MELD score, univariate and multivariate logistic regression analyses and c-statistics showed MELD as the most predictive risk model, which was clearly superior to CTP class and the EuroSCORE in predicting in-hospital mortality and long-term survival; the best cut-off value for MELD score was found to be at 13.5 with an optimal sensitivity of 82% and a specificity of 79%. In addition, the CTP classification was also found to be predictive for short- and long-term mortality in our patients’ series; the hospital and long-term outcome was significantly better in patients with CTP class A as compared to classes B and C, and conversely, CTP class C was significantly worse compared to classes A and B. However, c-statistics demonstrated that the predictive power of CTP class was inferior as compared to MELD score but at least superior to the EuroSCORE. Finally, the present study surprisingly did also show that one of the most commonly used and well-established risk score models in cardiac surgery, the EuroSCORE, was helpful neither for risk prediction of in-hospital outcome nor for prediction of long-term survival in these cirrhotic patients.

Patients with liver disease who undergo surgical procedures are in general at higher risk for postoperative complications and mortality [1,2]. It is important to note that the majority of studies evaluating patient-specific risk have focussed on patients with liver cirrhosis or end-stage liver disease. Thus, not much is known regarding the risks for patients with more mild liver disease. Two risk stratification schemes predominate in the literature with regard to assessing perioperative risk of patients with cirrhosis – the CTP classification and the MELD score. The CTP classification was developed empirically in 1973 among patients with bleeding oesophageal varices [21]. CTP classification relies on albumin, prothrombin time, serum bilirubin, ascites and encephalopathy to assess the severity of liver disease. Subsequently, CTP class has been shown in multiple studies to correlate with the perioperative mortality of patients undergoing extrahepatic abdominal surgery. Perioperative mortality rates by class were similar across these studies: CTP class A: 7–10% mortality, CTP class B: 23–30% mortality and CTP class C: 75–100% mortality [1,2]. MELD was developed in 2000 primarily to aid in stratifying patients post-TIPS procedure and calculates 3-month survival rates [20–22]. Variables found to be related to mortality included: (1) cause of liver disease, (2) bilirubin, (3) INR and (4) creatinine. An elaborate equation was developed relating these variables to mortality. A subsequent study evaluated the MELD score as predictive of 3-month survival in end-stage liver disease and found that the cause of disease did not alter survival. In both studies, the patient population had advanced-stage liver disease. Subsequent studies have been performed evaluating MELD as a risk predictor for surgical morbidity and mortality [23] and a MELD score ≥8 was identified to be predictive of poor outcome in patients with liver disease when undergoing surgery.

Cardiac surgery in patients with end-stage liver disease having liver cirrhosis is, in general, infrequently performed. Therefore, only limited data are available to assess specific indicators that identify and stratify those patients who are at high risk for open-heart surgery. Certainly, several reports have shown that open-heart surgery with the use of CPB in patients with liver cirrhosis is associated with adverse clinical outcome and an increased risk for morbidity and mortality, particularly in patients with CTP classes B and C. While the mortality rates of patients in CTP class A were reported to be 10% and less, mortality in CTP class B patients ranges between 0% and 80% and goes up to 100% for CTP class C patients [7–12]. These results demonstrate that CTP classes B and C are a heterogeneous group of patients; a more precise definition of the risk in liver cirrhosis patients is thus warranted. Recently, the MELD score was introduced and analysed in cardiac surgery in a series of 44 patients [24] and compared with the CTP class and/or score by Suman et al. Although the majority of their patients were in CTP class A, 12 patients in CTP class B and only one patient in CTP class C, the authors found a significant association of CTP and MELD score with hepatic decompensation and 3-months’ survival. While a highly predictive cut-off level (>7) could be assessed for the CTP score, the authors failed to find a similar predictive cut-off level for MELD because MELD lacked sensitivity. Nevertheless, the best possible cut-off level was identified at 13 with an AUC of 87 ± 0.09, which is quite comparable with the present study. Although Suman et al. could identify significant results to propose relationships between some preoperative variables and outcome, it was not possible for them to perform multivariate risk analyses due to the small number of adverse events. More recently, Filsoufi et al. analysed a series of 27 cirrhotic patients, who underwent cardiac surgery, and compared the predictive value of CTP class and MELD score. As a result, again the predictive value and reliability of CTP class could be confirmed, but the authors could not find a significant association between the MELD score and in-hospital mortality. However, the authors admitted that a validation of MELD score as a predictor of outcome following cardiac surgery would be difficult due to the small sample size of patients in their study [25].

The present study is the largest series of patients with liver cirrhosis undergoing cardiac surgery to date. By comparing our patients’ data with those of prior studies, the number of our patients’ preoperative co-morbidities was considerably higher and thus our patient cohort clearly represents a ‘high-risk group’ as compared to the cohort presented in the literature. In addition, it has to be taken into account that 18% of our patients underwent cardiac surgery on an emergency basis. As a result, major postoperative complications occurred in 86% of patients with advanced liver cirrhosis and in-hospital mortality at 30 days was 15% in CTP class A patients, 50% in CTP class B and 100% in CTP class C patients. These results are in accordance with other published reporting similar morbidity and mortality rates. By stratifying our patients with a MELD cut-off level of 13.5, mortality was 56% for patients with MELD>13.5 and 9% for patients with MELD ≪13.5%.

In conclusion, our results suggest that mortality of patients with liver cirrhosis undergoing open-heart surgery progressively increases with the severity of liver dysfunction. Open-heart surgery with the use of CPB can be performed with acceptable in-hospital mortality and long-term survival in cirrhotic patients with mild or moderate liver disease – having a preoperative MELD score ≪13.5 or are in CTP class A. In-hospital mortality and long-term survival remains high in those patients with CTP class B and extraordinarily high in CTP class C, as well as in patients with a preoperative MELD of >13.5. Preoperative risk stratification with MELD score best identifies those patients who are at highest risk for open-heart surgery. Therefore, MELD score seems to be essential for preoperative risk evaluation besides the CTP classification and may further help the surgeon to decide whether or not cardiac surgery should be offered or should be avoided, especially in patients with CTP class C and/or with MELD >13.5. Furthermore, preoperative MELD scoring may also help the clinician to optimise medical therapy of patients with liver cirrhosis (by correcting prothrombin time, platelets and electrolytes, minimising ascites, optimising renal function, addressing increased nutrition, etc.) prior to an elective open-heart surgery procedure. Whether ‘off-pump’ procedures in coronary surgery or trans-catheter heart-valve techniques are capable to decrease perioperative morbidity and mortality has not yet been proven and has to be elucidated in future studies.

4.1 Limitations of the study

The present study was a retrospective, observational single-centre study, with a limited number of patients. However, to our knowledge, the present study is the largest series to date. Another limitation of our study due to the retrospective design is the fact that we were neither aware of the former decision for surgery by the respective surgeon, nor if and how many patients denied or were not accepted for a cardiac surgical procedure.

☆

Presented at the 23rd Annual Meeting of the European Association for Cardio-thoracic Surgery, Vienna, Austria, October 18–21, 2009.