Impact of preoperative liver dysfunction on outcomes in patients with left ventricular assist devices

Abstract

Open in new tabDownload slide

OBJECTIVES

We evaluated the impact of preoperative liver function on early and 1-year postoperative outcomes in patients supported with a left ventricular assist device (LVAD) and subsequent evolution of liver function markers.

METHODS

A retrospective multicentre cohort study was conducted, including all patients undergoing continuous-flow LVAD implantation. The Model for End-stage Liver Disease (MELD) score was used to define liver dysfunction.

RESULTS

Overall, 290 patients with an LVAD [78% HeartMate II, 15% HVAD and 7% HeartMate 3, mean age 55 (18), 76% men] were included. Over 40 000 measurements of liver function markers were collected over a 1-year period. A receiver operating characteristic curve analysis for the 1-year mortality rate identified the optimal cut-off value of 12.6 for the MELD score. Therefore, the cohort was dichotomized into patients with an MELD score of less than or greater than 12.6. The early (90-day) survival rates in patients with and without liver dysfunction were 76% and 91% (P = 0.002) and 65% and 90% at 1 year, respectively (P < 0.001). Furthermore, patients with preoperative liver dysfunction had more embolic events and more re-explorations. At the 1-year follow-up, liver function markers showed an overall improvement in the majority of patients, with or without pre-LVAD liver dysfunction.

CONCLUSIONS

Preoperative liver dysfunction is associated with higher early 90-day and 1-year mortality rates after LVAD implantation. Furthermore, liver function improved in both patient groups. It has become imperative to optimize the selection criteria for possible LVAD candidates, since those who survive the first year show excellent recovery of their liver markers.

INTRODUCTION

In end-stage patients with heart failure (HF), continuous-flow left ventricular assist devices (LVADs) are increasingly used as bridge-to-transplantation or as destination therapy [1]. Although the overall clinical outcomes of the second- and third-generation LVADs are favourable, patient selection for LVAD therapy is still suboptimal given the considerable morbidity and mortality [2]. Many of these end-stage patients with HF have signs of hepatopathy due to the systemic hypoperfusion and passive congestion of the liver, most likely due to increased systemic venous pressure [3, 4]. Prior studies showed that preoperative liver dysfunction in patients with LVAD is associated with worse survival and adverse events including the onset of right ventricle failure, acute kidney injury and bleedings [5, 6].

The Model for End-stage Liver Disease (MELD), a score to assess liver function, was developed to predict mortality in patients undergoing trans-jugular intrahepatic portosystemic shunt procedure. The score includes 3 parameters: serum creatinine, total bilirubin and international normalized ratio (INR) [7]. Following its introduction, the MELD score has been utilized in the cardiac population, predicting mortality in patients undergoing cardiac surgery and heart transplantation [8, 9]. This has led to the utilization of the MELD score in patients with LVAD, in order to asses liver function and to predict mortality [5, 10]. However, the question of whether liver function changes following the initial preoperative MELD-score assessment is unknown.

Therefore, the aim of this study was to determine the association of preoperative liver dysfunction with early (90 days) and 1-year mortality in patients with LVAD. Subsequently, we aimed to depict the evolution of several liver function markers following the initial MELD-score assessment.

MATERIALS AND METHODS

Study design

We conducted a retrospective multicentre cohort study including all patients with available baseline laboratory data implanted with a Heartmate II, Heartmate 3 (Abbott, Chicago, IL, USA), and HVAD (Medtronic HeartWare, Framingham, MA, USA) continuous-flow-LVAD between October 2004 and April 2017 in 2 participating tertiary referral centres. Clinical and laboratory data were obtained from a computerized database and electronic patient files. Preoperative laboratory values were defined as the last available set of results prior to LVAD implantation. This study was approved by the institutional review board of the Erasmus MC, University Medical Centre, Rotterdam, the Netherlands and the Johns Hopkins Hospital, Baltimore, MD, USA.

End points

The primary end point was all-cause mortality, early (90 days) and 1-year post-LVAD implantation. Secondary outcomes were neurologic events, re-explorations and the evolution of liver function markers in patients with and without pre-LVAD liver dysfunctions.

Liver function assessment was based on pre-implantation values of total bilirubin, albumin as well as the MELD-score modification used by the United Network for Organ Sharing: 3.78 × ln(bilirubin) + 11.2 × ln(INR) + 9.57 × ln(creatinine) + 6.43 [11]. Evaluation for possible liver cirrhosis was done at the discretion of the treating physician. A receiver operating characteristic (ROC) curve analysis was performed for the 1-year mortality rates for the MELD scoring system. The Youden Index was calculated from the MELD-score ROC curve analysis to establish the optimized cohort cut-off point.

Statistical analysis

Continuous parameters are expressed as median and interquartile range or mean and standard deviation, depending on the distribution, and were compared by Student’s t-test or the Mann–Whitney U-test. The normality of data was assessed by performing the Shapiro–Wilk test. Categorical parameters were expressed as number and percentage and compared by the χ² test or Fisher’s exact test (if any of the expected cell sizes was ≤5) for the association. Kaplan–Meier curves stratified by liver function were constructed for the evaluation of mortality in the first year after continuous-flow-LVAD implantation. Differences were compared by the log-rank test. An ROC curve analysis was conducted to determine the optimal cut-off MELD-score value for predicting mortality. A multivariable Cox proportional hazards analysis was performed for the identification of parameters associated with mortality. Variables were included in the multivariable models if P-value ≤0.20 in the univariate analysis. All multivariable models were constructed by using the enter method. Two-tailed value of P-value <0.05 was considered statistically significant. Analyses were performed using the SPSS statistical software package, version 24.0 for Mac (SPSS Inc., IBM, Chicago, IL, USA) and GraphPad Prism version 5.0a for Mac (GraphPad Software, La Jolla, CA, USA).

Mixed modelling

Continuous repeated measurement data were analysed using mixed-models with the maximum likelihood estimator. Flexibility over time was established by using natural splines. Included random effects were intercepts for patients with random slopes for time. A backward selection procedure was applied using the Akaike information criterion and Bayesian information criterion to select the number of splines for time in the random effect. Likelihood ratio tests were used to compare nested models. The model was visualized by effect plots. Statistical analyses were done in R (R Foundation for Statistical Computing, Vienna, Austria) version 3.3.3 with package ‘lme4’ [12].

RESULTS

In total, 290 patients received an LVAD [77% male, mean age 55 (interquartile range 18)]: 225 (78%) patients received a Heartmate II device, 43 (15%) patients received an HVAD and 22 (7%) patients received a Heartmate 3 device. The baseline characteristics of the patients are presented in Table 1. None of the implanted patients had preoperative signs or symptoms of liver cirrhosis. Postoperatively, 15 (6%) patients required a temporary right ventricular assist device, 110 (38%) patients needed re-exploration due to early bleedings and 38 (13%) experienced a neurologic event. After 1 year of LVAD support, 216 (75%) patients were still alive. In total, 41 (14%) patients were successfully transplanted after LVAD implantation.

In a univariable Cox regression analyses, age, gender, body mass index, a previous coronary artery bypass graft, pre-implantation need for intra-aortic balloon pump, interagency registry for mechanically assisted circulatory support (INTERMACS) profile 1 and 2, destination therapy and the HeartMate II were all predictors of mortality within 1 year after LVAD implantation (Table 2). Laboratory data, significantly associated with mortality at 1 year, included total bilirubin, creatinine, albumin, INR and the MELD score.

A ROC curve analysis was performed for both the 90-day and 1-year mortality rates for the MELD scoring system to determine the optimal cut-off value. The 90-day mortality rate gave a cut-off value of MELD 15.0 with a sensitivity of 70% and specificity of 54% with an area under the curve (AUC) of 0.62. The 1-year mortality rate gave a cut-off value for the MELD score was established at 12.6, with a sensitivity of 87% and a specificity of 40% with an AUC of 0.63. Therefore, the cohort was dichotomized, dividing patients with an MELD score of <12.6 or ≥12.6 (Fig. 1). Subsequently, the MELD score <12.6 and ≥12.6 were added to the univariate analysis. The MELD score of ≥12.6 was a significant predictor in the univariable analysis and therefore added to the multivariable analysis.

In multivariable Cox regression analyses, age, body mass index, INTERMACS profile 1, destination therapy, a decrease in albumin and an MELD score of ≥12.6 were all independent predictors of mortality within 1-year after LVAD implantation (Table 3).

Patients with an MELD score ≥12.6 were more often male, had a higher percentage of implanted cardioverter-defibrillators or pacemakers and were more often in need of preoperative intra-aortic balloon pump support. Accordingly, the patients with an MELD score ≥12.6 had worse INTERMACS profiles, higher total bilirubin, INR, serum creatinine and alanine aminotransferase (ALT) prior to implantation (Supplementary Material, Table S1). Patients with an MELD score ≥12.6 had a significantly worse early 90-day and 1-year survival compared to patients with an MELD score <12.6 following LVAD implantation (Fig. 2). Additionally, patients with an MELD score ≥12.6 had a significantly higher neurological event rate and a higher rate of re-exploration (Table 4).

Additionally, a sub-group analysis of patients with available preoperative right heart catheterization measurements was performed (n = 200; 77 patients with MELD score <12.6 and 113 with MELD score ≥12.6). The patients with an MELD score ≥12.6 had higher preoperative right atrial pressures (11.4 ± 6.7 vs 15 ± 7.1, P < 0.001), higher pulmonary artery pressures (34.6 ± 11 vs 38.6 ± 10.3, P = 0.01), higher pulmonary capillary wedge pressures (PCWP; 24.9 ± 9 vs 28.2 ± 9.3, P = 0.02), right atrial/pulmonary capillary wedge pressure (RA/PCWP) (0.45 ± 0.18 vs 0.56 ± 0.22, P < 0.001) and lower pulmonary artery pulsatility index (0.29 ± 0.33 vs 0.19 ± 0.19 P = 0.01). In univariable Cox regression analysis, all of the measurements except for RA/PCWP and pulmonary artery pulsatility index were predictors of 1-year mortality. However, none of the measurements reached significance in multivariate analysis.

In addition, we evaluated the impact of liver dysfunction in patients who did not receive any extracorporeal life support (ECLS) prior to LVAD implantation. In total, 170 (59%) patients were free of ECLS; 69 patients had an MELD score <12.6 and 101 patients had an MELD score ≥12.6. The 1-year survival rates for patients with and without live dysfunction in this subset of patients was 72% versus 91%, respectively (log rank, P = 0.005).

Evolution of liver function

In total, 23 333 repeated measurements of serum total bilirubin (mg/dl) were collected during follow-up: MELD score ≥12.6: 14 858, MELD score <12.6: 8475. The mean follow-up time was 228 ± 147 days for the MELD score ≥12.6 group and 305 ± 118 days for the MELD score <12.6 group, respectively. Initially, patients with preoperative liver dysfunction have higher mean levels of total bilirubin. The evolution of total bilirubin for all patients dichotomized based on preoperative MELD score is plotted in Fig. 3. At 1 year of follow-up, patients with and without preoperative liver dysfunction have similar mean total bilirubin levels. Moreover, both groups have mean total bilirubin within the acceptable range at 1 year of follow-up.

Overall, 21 069 repeated measurements of serum albumin (g/dl) were collected during follow-up: MELD score ≥12.6: 14 747, and MELD score <12.6: 8322. The evolution of albumin, dichotomized based on preoperative MELD score, is plotted in Fig. 4. Although mean albumin levels start at an acceptable range preoperatively, directly postoperative mean albumin levels are substantially lower in both groups. Following implantation, mean albumin levels incrementally increase in both groups, with recovery in the low MELD-score patients after 140 days versus 1 year in the high MELD-score group. At the 1-year follow-up mark, both patients with and without preoperative liver dysfunction have regained their initially lost serum albumin levels.

The evolution of ALT in both patient groups shows a similar decrease in mean levels of ALT until 2–3 months after LVAD implantation. Subsequently, both patient groups experience a plateau phase with no significant changes in mean ALT levels. Mean serum levels of ALT, for both groups, remain within the acceptable range during the first year following LVAD implantation. Mean serum levels of aspartate aminotransferase levels, however, are at less than optimal levels early after LVAD implantation (52 U/l in MELD <12.6 and 66 U/l in MELD ≥12.6). Following the initial elevation, mean levels of serum aspartate aminotransferase decrease as the early postoperative period progresses. At the 3 months of follow-up, a substantial decrease of mean aspartate aminotransferase is noticed: 38% in low MELD versus 45% in patients with high MELD score. Following this decrease, a plateau phase follows until the end of the first year of follow-up (Supplementary Material, Figures S2–S7).

DISCUSSION

This study evaluated the impact of preoperative liver function on 90-day and 1-year postoperative outcomes in patients with an LVAD implantation and subsequent evolution of liver function evolution over time. The principal findings of this study are: patients with significant liver dysfunction (MELD score ≥12.6) have worse 90-day and 1-year postoperative survival. Moreover, patients with liver dysfunction have higher rates of adverse events, including higher rates of neurologic events and higher need for re-exploration due to early bleeding/tamponade following LVAD implantation. Regardless of baseline liver dysfunction, all liver function markers improve post-LVAD implantation, and at 1-year follow-up, no significant differences were observed between the separate MELD groups.

To determine the effect of liver dysfunction on LVAD recipients, similar studies have investigated the predictive value of the MELD score for early and late mortality and adverse events [10, 13]. Although the exact value of MELD score used to differentiate liver dysfunction varies between studies, it evidently demonstrates that liver dysfunction has a detrimental impact on outcomes. The optimal cut-off value for MELD score was calculated for both 90-day mortality and 1-year mortality. Though the AUC was similar, the cut-off value for 1-year mortality had a substantially higher sensitivity. Therefore, the cut-off value for MELD score at 1-year was used to define liver dysfunction. Other studies have used the MELD-XI (by excluding INR) instead of the MELD score, given the frequent use of oral anticoagulation. Their results demonstrate that the MELD-XI can similarly be used to predict worse outcomes in patients with LVAD [5, 14]. To account for this, we conducted an ROC curve analysis to determine the predictive value of the MELD, MELD-XI and the MELD-NA (which adds sodium to the MELD equation) in our cohort: 0.63 [confidence interval (CI) 0.56–0.70], 0.63 (CI 0.56–0.70) and 0.62 (0.55–0.69), respectively. Given minimal to no difference in the predictive value, we used the classic MELD score without any modifications. Of note, the use of MELD XI is warranted if patients have received anticoagulants shortly before LVAD surgery. In this cohort, only 13 (4.5%) patients received Warfarin, and 3 patients Dabigatran 7 days prior to LVAD surgery. This constituted a small portion of the cohort and, therefore, the classic MELD score was used.

Thereafter, a mixed-model analysis was performed to illustrate the evolution of several liver function markers. This analysis adjusts for the correlation between multiple measurements of 1 patient and for the correlation between patients. The analysis notes the improvement of liver function markers in both patient groups. Although both groups show different time frames of improvement, this finding underlines the benefits of LVAD therapy for the failing secondary organ systems in end-stage patients with HF.

The differences between outcomes in patients with and without liver dysfunction are probably multifactorial. However, the main factor causing liver dysfunction most likely is chronic right-sided congestion due to chronic primary or secondary right ventricular failure [15]. Preoperative liver dysfunction is a predictor of postoperative right-sided HF in patients with LVAD [16]. Subsequently, the onset of right-sided HF in LVAD recipients has been associated with worse outcomes including worse survival [17]. To evaluate the severity of preoperative right ventricle dysfunction we analysed the available 200 right heart catheterization measurements. We sought for the possible impact of higher preoperative mean right atrial pressures, pulmonary artery pressures, PCWP, RA/PCWP and pulmonary artery pulsatility index on mortality. However, higher preoperative pressures were not predictors of mortality following LVAD transplantation in our cohort. The analysis in patients who received no ECLS prior to LVAD surgery showed that a preoperative MELD score of ≥12.6 is still an accurate predictor of mortality in this population. Previous work from Maxhera et al. [18] showed that in the patients receiving ECLS, a higher preoperative MELD score was associated with increased mortality following LVAD implantation. This remains true for the patient population without an ECLS prior to LVAD implantation.

Clinical implications

Our study emphasizes the impact of liver dysfunction on outcomes following LVAD implantation. Despite worse outcomes in patients with liver dysfunction, our study shows that LVAD therapy can facilitate the recovery of liver dysfunction in end-stage patients with HF. Furthermore, our results suggest that preoperative MELD score can aid in the selection process of high-risk potential LVAD candidates, who are older (>65), suffer from advanced renal failure and have additional comorbidities. In addition, the MELD score can probably identify patients at risk for bleeding complications following LVAD implantation. It has become imperative to optimize the selection criteria for possible LVAD candidates since those who survive the first year show excellent recovery of their liver markers. The liver status is essential for an optimal decision-making; however, this study only gave a glimpse of what the possible impact of liver dysfunction could be. More prospective research is needed.

Limitations

The study has several limitations that should be taken into consideration while interpreting the results. Firstly, due to the retrospective nature of our study, causality could not be established. Furthermore, an AUC of 0.63 is less than optimal. Moreover, despite standardized treatment plans for all patients, individual changes in medication and therapy could potentially have influenced the findings. The strengths of the study are the relatively large sample size and the multicentre design incorporating both European and American patients. The use of advanced mixed modelling enables a more accurate depiction of liver function markers.

CONCLUSION

Preoperative liver dysfunction is associated with higher early and 1-year mortality rates after LVAD implantation. However, improvement of liver function markers is noticed at 1-year follow-up, regardless of preoperative MELD score. The increasing age and number of comorbidities of potential LVAD candidates, especially in DT candidates, warrants continuous validation and improvement of selection criteria. By considering patients’ pre-implantation MELD score in addition to other comorbidities, could improve the shared decision-making process, preoperative optimization, and probably the postoperative management in this high-risk surgery.

SUPPLEMENTARY MATERIAL

Supplementary material is available at EJCTS online.

Conflict of interest: none declared.

Author contributions

Yunus C. Yalcin: Data curation; Formal analysis; Investigation; Methodology; Writing—original draft. Rahatullah Muslem: Conceptualization; Formal analysis; Methodology; Supervision; Writing—review & editing. Kevin M. Veen: Formal analysis; Software; Visualization; Writing—review & editing. Osama I. Soliman: Formal analysis; Writing—review & editing. Olivier C. Manintveld: Writing—review & editing. Sarwa Darwish Murad: Writing—review & editing. Ahmet Kilic: Writing—review & editing. Alina A. Constantinescu: Writing—review & editing. Jasper J. Brugts: Writing—review & editing. Fatimah Alkhunaizi: Data curation. Ozcan Birim: Writing—review & editing. Ryan J. Tedford: Writing—review & editing. Ad J.J.C. Bogers: Writing—review & editing. Steven Hsu: Conceptualization; Writing—review & editing. Kadir Caliskan: Conceptualization; Methodology; Project administration; Supervision; Writing—original draft.

REFERENCES

ABBREVIATIONS

ABBREVIATIONS
 
• ALT

  Alanine aminotransferase

 
• AUC

  Area under the curve

 
• CI

  Confidence interval

 
• ECLS

  Extracorporeal life support

 
• HF

  Heart failure

 
• INR

  International normalized ratio

 
• INTERMACS

  Interagency registry for mechanically assisted circulatory support

 
• LVAD

  Left ventricular assist device

 
• MELD

  Model for End-stage Liver Disease

 
• PCWP

  Pulmonary capillary wedge pressures

 
• ROC

  Receiver operating characteristic