Secondary Logo

Journal Logo

Adult Circulatory Support

Melding a High-Risk Patient for Continuous Flow Left Ventricular Assist Device into a Low-Risk Patient

Amione-Guerra, Javier*†; Cruz-Solbes, Ana S.*†; Gonzalez Bonilla, Hilda*; Estep, Jerry D.*†; Guha, Ashrith*†; Bhimaraj, Arvind*†; Suarez, Erik E.*†; Bruckner, Brian A.*†; Torre-Amione, Guillermo*†‡; Park, Myung H.*†; Trachtenberg, Barry H.*†

Author Information
doi: 10.1097/MAT.0000000000000591
  • Free


Implantation of left ventricular assist devices (LVADs) for refractory end-stage heart failure, either as destination therapy (DT) or bridge to transplantation (BTT), has grown exponentially in the past few years, and the benefits of LVAD versus medical therapy have been clearly demonstrated.1–3 Although the number of implants continues to increase (>1500 LVADs implanted in the United States according to the latest Interagency Registry for Mechanically Assisted Circulatory Support [INTERMACS] report), the overall survival of these patients (i.e., ≈80% at 1 year, 70% at 2 years) has remained unchanged during the past 5 years.3 With the growing number of patients being treated with LVADs, patient selection and the optimization of high-risk patients (i.e., INTERMACS class 1–2 patients) are critical factors in improving outcomes.

Currently, several models for assessing mortality and morbidity in patients with LVADs exist.4–7 Most of these risk scores have been derived from a mixed population of patients with pulsatile and continuous flow LVADs, single-center studies, or from patients enrolled in clinical trials who may not reflect “real-world” practice.

Notably, the model for end-stage liver disease (MELD) score, originally developed for prognosis assessment of cirrhotic patients undergoing transjugular intrahepatic portosystemic shunts (TIPS)8 and adopted later as a tool for prioritizing patients awaiting liver transplantation,9 has also been used as a predictor of morbidity and mortality in LVAD patients.6,10,11

Interagency Registry for Mechanically Assisted Circulatory Support profiling and other risk scores are all derived from laboratory and clinical variables measured within 48 hours previous to LVAD implantation. Whereas these risk scores assume variables are static, in real practice variables may be modified by careful management or simply with time. It is our aim to determine if assessing changes in MELD classification from the time of admission to implantation could improve risk stratification of patients receiving CF-LVAD for the treatment of end-stage heart failure.


Data Collection

We retrospectively collected clinical data (laboratories, invasive hemodynamics, echocardiographic data, and medication lists) from adult patients implanted with the continuous flow LVAD (CF-LVAD) as BT, DT or bridge to decision from May 2008 to December 2014 at the Houston Methodist Hospital (HMH). This study was approved by the institution’s internal review board.

Risk Score Calculation

Day of admission and preimplantation (≤48 hours before surgery) laboratory values were collected in all patients. Model for end-stage liver disease scores were calculated using the United Network for Organ Sharing-modified MELD formula: MELD = 9.57(loge creatinine) + 3.78(loge bilirubin) + 11.2(loge INR) + 6.43; variable lower limits were set at 1.0 and creatinine upper level was set at 4.0 mg/dl, if the patient received renal replacement therapy within 2 weeks of implant, then the assigned creatinine was 4 mg/dl. We also calculated the MELD-XI (MELD excluding INR) score, given the chance that many patients would be on oral anticoagulation, as this score could be more accurate than MELD alone.12 MELD-XI = 5.11(loge bilirubin) + 11.76(loge creatinine) + 9.44. Dichotomization of high risk or low risk was defined based on area under the curve (AUC) analysis.

Risk Classification Scheme

We developed a classification scheme based on MELD risk at admission and MELD risk ≤48 hours before implant, forming four risk groups (Figure 1): Patients with a low-risk MELD at admission who remained low risk at the time of implant (group LL) or changed to high risk (group LH); and patients with a high-risk MELD at admission who remained high risk (group HH) or changed to low risk (group HL). Likewise, we used MELD-XI to divide the patients into similar categories. The primary outcome was all-cause mortality at 1 year after implant.

Figure 1.:
Risk classification scheme based on MELD scores at admission and at day of implant. Flow chart demonstrating the scheme used to create the four different groups based on MELD score at two different time points. MELD, model for end-stage liver disease.

Statistical Analysis

Data analysis was performed using STATA13 (StataCorp LP, College Station, TX). A p value of <0.05 was considered statistically significant for all our analyses. Continuous variables were analyzed using Student’s t-test or the Wilcoxon rank sum test and are reported as median, mean, and standard deviation. Categorical variables were analyzed using either the χ2 test or the Fisher’s exact test and are presented as percentages. Multiple group comparison was performed using an analysis of variance (ANOVA), Kruskal–Wallis rank, or Dunn’s testing. Post hoc analysis for multiple comparisons was using Bonferroni’s test. Receiver operating characteristic curves were constructed to identify the risk score with the highest AUC and an optimal cutoff was selected to classify patients as either low or high risk. Based on risk group, survival was compared using Kaplan–Meier curves and log-rank testing. Patients who received a heart transplant were censored at the time of transplant. Univariate Cox proportional hazards models were performed on all variables to identify the potential factors for mortality; then a stepwise forward Cox proportional hazards modeling on the univariate predictors (entry criteria was p value <0.1) was performed to identify 1 year survival correlates.


Patient Cohort

A total of 244 consecutive patients implanted with CF-LVADs were retrospectively analyzed; of those, 21 (8%) patients were excluded because of missing variables, thus precluding the ability to calculate MELD scores at the day of implant. Table 1 summarizes the patients’ baseline characteristics. The mean age for implant was 56 ± 12 years, 174 (78%) were males, 111 (49%) were Caucasian, and 134 (60%) had ischemic cardiomyopathy. Device strategy was classified as DT for 133 (59.6%) of the patients, and 34% of the patients were either INTERMACS profile 1 or 2 at time of implant. Average time from admission to implant was 13 ± 10 days.

Table 1.:
Baseline Demographics of All Patients During the Study Period (n = 223)

Risk Score Reclassification Scheme and Outcomes

From our final cohort of 223 patients, receiver operating characteristic curve analysis revealed that MELD score at the day of implant had the highest AUC (0.68, 95% confidence interval [CI]: 0.60–0.76) and that a cutoff of ≥19 had a sensitivity of 40% and a specificity of 86% to predict 1 year mortality. Using MELD ≥19 as a cutoff, 90 (40%) of the patients were classified as high risk (MELD ≥19) and 133 (60%) as low risk (MELD < 19) at the day of admission; of the patients admitted with a low risk score, 122 (92%) remained low risk (group LL) and 11 (8%) were reclassified as high risk (group LH). Conversely, of the patients admitted with a high-risk score, 36 (40%) remained high risk (group HH), whereas 54 (60%) improved to a low-risk score (group HL; Figure 1).

During the study period, 59 (26%) patients died within 1 year of CF-LVAD implant. When stratified by risk score classification, patients who were admitted as high risk but improved to a low risk (group HL) had no difference in survival when compared with patients who remained low risk from admission to implant (group LL, 79% ± 5% vs. 80% ± 4%; p value = 0.9; Figure 2A). Furthermore, patients who were admitted with a low risk and worsened to a high risk (group LH) had the worst survival among all groups, even when compared with those who remained high risk (group HH; 9% ± 8% vs. 59% ± 8%; p < 0.01; Figure 2B). Group LL patients had improved 1 year survival when compared with those of group LH (80% ± 4% vs. 9% ± 8%; p value < 0.01), whereas patients in group HL had improved survival compared with those of group HH (79% ± 5% vs. 59% ± 8%; p value = 0.03). A MELD score >25 at day of implant had a specificity of 98% to predict 1 year mortality.

Figure 2.:
One year survival estimates for HMII-LVAD implanted patients based on MELD risk category. A: Patients in group LL (remained low risk) vs. those in group HL (improved to low risk). B, Patients in group LH (worsened to high risk) vs. those in group HH (remained high risk). LVAD, left ventricular assist device; MELD, model for end-stage liver disease.

Characteristics of Model for End-Stage Liver Disease Risk Groups

General demographics and admission characteristics among the different MELD risk groups are shown in Table 2. There were no differences in age, gender, etiology of heart failure, or device strategy between the groups. Patients in group LL were more often Caucasian when compared with those in group HH (53% vs. 27%; p = 0.04). As expected, patients implanted in the high-risk groups (HH and LH) compared with low-risk groups (LL and HL) were more likely to have an INTERMACS profile 1 or 2 and had worse baseline hemodynamics with higher central venous pressure (CVP; 18 ± 6 vs. 14 ± 7 mm Hg; p < 0.01), pulmonary capillary wedge pressure (PCWP; 29 ± 6 vs. 26 ± 7 mm Hg; p = 0.02), CVP/PCWP ratio (0.63 ± 0.2 vs. 0.53 ± 0.2; p = 0.01) and mean pulmonary artery pressures (41 ± 9 vs. 37 ± 8; p = 0.01). In addition, echocardiography showed an increased right ventricle (RV)/left ventricle (LV) ratio (0.82 ± 0.1 vs. 0.73 ± 0.1; p < 0.01) and a greater prevalence of severe RV dysfunction (32% vs. 16%; p = 0.01). There was no difference in cardiac output, cardiac index (both by invasive catheterization and echocardiography), LV size, tricuspid annular plane systolic excursion (TAPSE), and tricuspid regurgitation (TR) severity between patients with high versus low risk at the time of implantation. Although baseline characteristics (i.e., INTERMACS profile, echocardiography, and invasive hemodynamics) were significantly different between high-risk (HH and LH) and low-risk (LL and HL) patients at implant, pairwise comparisons between groups were usually not. Most of the statistical differences occurred between group LL versus either group HH or group HL. For example, both the CVP and the CVP:PCWP ratio were lower in group LL versus group HH and HL (13 ± 6 mm Hg vs. 18 ± 5 mm Hg and 16 ± 7 mm Hg, respectively; p < 0.01 for both comparisons and group LL: 0.48 + 0.22 vs. 0.64 + 0.18 in group HH and 0.61 + 0.28 in group HL; p < 0.01 for both comparisons).

Table 2.:
Comparison of Characteristics Among the Different MELD Risk Groups

The use of temporary mechanical circulatory support (MCS) before CF-LVAD was more common in patients in group LH (72%), this was followed by groups HH, HL, and LL (55%, 46%, and 28%, respectively). However, there was only a statistical difference when comparing those in group LL versus group HH and group HL. The temporary circulatory support (TCS) device most commonly used was the intraaortic balloon pump, followed by the TandemHeart® (CardiacAssist, Inc., Pittsburgh, PA. Interestingly in our cohort, only two patients received extracorporeal membrane oxygenation (ECMO) support, and both patients were in the LH groups, suggesting these patients were the sickest of all patients. There was no difference in the time of TCS support between the groups with a median time of support of 7 days (range 1–67 days).

These results suggest that improvement in group HL compared with group HH cannot be attributed to the use of TCS. Four (36%) patients in group LH and 16 (44%) in group HH had undergone dialysis during their admission, compared with no patients undergoing dialysis in either groups LL and HL. Of the 14 patients that died on group HH, nine (64%) had dialysis before their LVAD surgery. The proportion of patients managed with therapy tailored to hemodynamics using a Swan-Ganz catheter was lower in the LL patients (25%) when compared with LH (73%), HL (41%), and HH (42%) patients (p < 0.01); however, there was no significance between the HH and HL group (p = 0.9).

Analyses of the changes in the variables that determine MELD from admission to implant are shown in Table 3. Notably, group LH patients had a significant increase in creatinine and bilirubin; in group HL, all variables improved significantly. In group HH, while INR and creatinine improved, bilirubin worsened. The median time from admission MELD to day of implant MELD was 11 days (range: 2–54) for all the groups. There was no statistically difference within group comparisons (p = 0.06); however, there was a trend toward longer time in the HH group (median time of 12 days, range 3–54) when compared with the LL group (median 10 days, range 2–37; p = 0.055).

Table 3.:
Changes in MELD Variables

Univariate and Multivariate Analysis of Predictors of 1 Year Mortality

Univariate predictors of mortality are shown in Table 4. Variables that significantly increased the hazards of death were patients on INTERMACS profile 1 (hazard ratio [HR]: 3.57, 95% confidence interval [CI]: 1.48–8.63; p < 0.01), profile 2 (HR: 2.55, 95% CI: 1.06–6.10; p = 0.03), hemodialysis before implant (HR: 5.12, 95% CI: 2.74–9.5; p value < 0.01), temporary MCS before implant (HR: 1.58, 95% CI: 0.7–2.57; p = 0.06), decreased day of implant albumin (HR: 1.81, 95% CI: 1.24–2.63; p < 0.01), patients in MELD group LH (HR: 8.20, 95% CI: 3.83–17.4; p < 0.01), and risk group HH (HR: 2.46, 95% CI: 1.27–4.76; p < 0.01). Variables that increased the hazards of mortality were included in a stepwise multivariate analysis (entry criteria: p value < 0.1; Table 5) which demonstrated that after adjusting for INTERMACS profile, day of implant albumin levels and temporary MCS, the only variables that remained with a statistically significant hazard for mortality were the MELD group LH (HR: 5.48, 95% CI: 2.40–12.51; p < 0.01) and group HH (HR: 2.08, 95% CI: 1.02–4.24, p = 0.04; Somer’s D = 0.4). Hemodialysis was not included in the model because it is already accounted for in the MELD score calculation.

Table 4.:
Univariate Correlates of 1 Year Mortality (Number of Events = 66)
Table 5.:
Multivariate Analysis of Predictors of 1 Year Mortality*

Model for End-Stage Liver Disease Reclassification in the Sickest of the Patients: INTERMACS 1 and 2 Profiles

From our total study population, 32 (15.2%) of the patients were classified as INTERMACS profile 1 and 41 (19.5%) profile 2 at the day of LVAD implant. Applying the MELD reclassification scheme to this subgroup of patients yielded the following results (Figure 3): Patients who improved to low risk (group HL) had the same survival as those who remained low risk (group LL; 76% ± 12% vs. 73% ± 8%; log rank p = 0.7; Figure 3A). Both patients in group HH and group LH had the worst outcomes; however, those patients in group LH had significantly worse survival versus those in group HH (0% vs. 43% ± 11%; p = 0.04; Figure 3B). Table 6 provides specific data of each of the patients in group LH. Six of the patients had worsening heart failure despite multiple vasopressors and TCS, two patients had an episode of ventricular tachycardia resulting in hemodynamic collapse, two patients had hospital acquired pneumonia, and one patient had severe right ventricular failure and pulmonary hypertension. Notably 10 of the 11 patients in this group had an INTERMACS profile 1 or 2 and expired within 1 year. The most common cause of death was severe right ventricular failure with subsequent multiorgan failure.

Table 6.:
Specific Characteristics of Patients in the LH Group
Figure 3.:
Survival in patients with an INTERMACS profile 1 or 2 by MELD risk group. A, Patients in group LL (remained low risk) vs. those in group HL (improved to low risk). B, Patients in group LH (worsened to high risk) vs. those in group HH (remained high risk). INTERMACS, Interagency Registry for Mechanically Assisted Circulatory Support; MELD, model for end-stage liver disease.

Anticoagulation and Its Effects on Model for End-Stage Liver Disease Scoring: Using the Model for End-Stage Liver Disease-XI Score

Using the MELD-XI, patients were dichotomized into high- and low-risk groups using a cutoff of ≥19 as we did with the MELD score: patients who improved to a low-risk score (group HL) had the same survival as those patients who remained low risk (group LL; 88% ± 5% vs. 77% ± 3%; p = 0.1). Patients who remained high risk (group HH) had worse survival versus those in group LL (61% ± 8% vs. 77% ± 3%; p = 0.02), as well as those in group HL (61% ± 8% vs. 88% ± 5%; p = 0.01). However, although there was a twofold increase in survival in group HH compared with group LH, this did not reach statistical significance (61% ± 8% vs. 30% ± 11%; p = 0.07).

Of the total patient population, 66 patients (30%) were on oral anticoagulation at time of admission. Group HL had the highest proportion of subjects receiving warfarin (52%) and this was statistically significant when compared with group LL (16%; p < 0.01). Importantly, MELD parameters other than INR differed between these groups, as well as other characteristics associated with high risk for LVAD implantation. Overall, group HL had worse renal (creatinine: 2.0 ± 0.1 vs. 1.2 ± 0.4 mg/dl, p < 0.001 and blood urea nitrogen (BUN): 49 ± 25 vs. 36 ± 41 mg/dl, p = 0.01) and hepatic function (bilirubin: 2.1 ± 1.4 vs. 1.1 ± 0.7 mg/dl; p < 0.001) compared with group LL. When compared with group LL, group HL had a higher right ventricular preload (CVP: 16 ± 17 mm Hg vs. 13 ± 0.6; p < 0.01 and CVP/PCWP ratio: 0.61 ± 0.28 vs. 0.48 ± 0.22; p < 0.01). With the exception of group LH (no change), the rest of the groups improved their INR by the time of surgery. High-risk patients (HH and LH) at the time of LVAD implant had higher INR compared with those with low risk (LL and HL; 1.5 ± 0.4 vs. 1.2 ± 0.2; p < 0.01), likely reflecting hepatic congestion. There was no difference in the amount of perioperative transfusion requirements between the high-risk (HH and LH) and low (LL and HL)-risk groups: the median number of red blood cell (RBC) units for the high- and low-risk patients was 6 (IQR: 3–9) and 5 (IQR: 2–9; p = 0.2), respectively, the median number of fresh frozen plasma units for high- and low-risk patients was 2 (IQR: 1–4) and 2 (IQR: 1–5; p = 0.1), respectively, and there was no statistically significant difference for number of platelet transfusions and overall all blood product exposure during CF-LVAD implant. Pairwise group comparison did not show a statistically significant difference in the use of perioperative blood products.


Our study showed that using the change in MELD score to reclassify patients from admission to the time of implantation is an independent and powerful predictor of survival in patients undergoing LVAD implantation. Importantly, patients who were identified as high risk by MELD on admission, when improved to low risk at time of implantation, had similar survival as patients who were low risk at both time points. Also, those whose profile changed from low- to high-risk scores had worse survival than those who remained high risk, suggesting that a change in MELD may be more specific than a single MELD score at day of implant. Finally, our MELD reclassification score was an independent predictor of survival even in patients at highest risk for mortality, as determined by INTERMACS profiles 1 and 2.

To our knowledge, our study is the first to demonstrate the utility of assessing the dynamic nature of the MELD or any other risk score in prognostication. A decrease in the AUC-derived cutoff score of ≥19 from day of admission to <19 on day of implant predicted an improved 1 year survival, whereas the lack of improvement in MELD score or a score that changed from <19 to ≥19 was associated with a higher risk of death. Furthermore, when adjusting for other variables commonly known to increase mortality in this group of patients (i.e., albumin, INTERMACS profile, and the use of temporary MCS), our reclassification scheme remained the sole independent predictor of mortality.

Because INR is a key component of the MELD score, we studied the impact of anticoagulation on MELD score in our study. Although INR improved in all of our groups with the exception of those that worsened to a high risk (group LH), the greatest improvement was in the high risk at admission (HH and HL) groups. This suggests a possible confounding role of anticoagulation therapy on MELD scores because a higher proportion of these patients (45%) were on Coumadin therapy versus the low-risk groups (17%). Use of warfarin in our patient population could have biased the MELD, especially for those patients with a HL profile that were on oral anticoagulation as they might have been falsely adjudicated a “high risk” at admission because of a high INR. Both the number of patients on warfarin and the INR at admission was significantly higher in the HL group compared with the LL group. Nonetheless there were other significant differences in the HL group besides INR. Compared with group LL, group HL patients had worse renal and hepatic function at admission and this correlated with worse right ventricular function (higher CVP and CVP/PCWP ratio). These results suggest that despite the potential effect of warfarin on MELD calculation, group HL patients had a higher risk profile at admission compared with those in group LL. Additionally, INR was not the only variable to determine the change in MELD score in any of the groups. For example, in group HL, all variables improved markedly, whereas in group HH, bilirubin was the major determinant of the persistently high MELD score.

An alternative option to MELD in patients receiving oral anticoagulation is the MELD-XI score, which has a good correlation with the traditional MELD and is predictive of outcomes in patients requiring LVAD.12 In our patient population, the use of the MELD-XI had similar discrimination capacity as that of the MELD using a cutoff ≥19: high-risk patients who improved to low risk (group HL) had the same survival as those who remained low risk (group LL). Group HH patients had worse survival when compared with groups LL and HL. However, in contrast to MELD scoring, MELD-XI group LH had only a trend (p = 0.07) toward decreased survival when compared with group HH. Given the high numerical difference between these groups (i.e., twofold difference in survival), it is reasonable to postulate that this lack of statistical difference might be explained by a type 2 error from a small sample size. Nonetheless, whether MELD or MELD-XI performs better in a larger cohort of patients is yet to be determined. However, it is important to consider that compared with MELD-XI, MELD score is more accessible and convenient to the general clinician, with more online and smartphone calculators available compared with MELD-XI.

As patient survival continues to improve and patient selection is shifting to less-sick patients, there will always be a sizeable population of patients who remain at the highest risk. In our study, 32% of our patients were classified as either profile 1 or 2. The use of durable CF-LVADs in patients who are classified as INTERMACS 1 is controversial because these patients have the highest mortality. Nevertheless, 15% of implants are done on INTERMACS 1 patients in the current era,13 whether because of patients who present to the hospital in cardiogenic shock or because of delayed referrals. Useful discriminators to predict survival in this sicker subpopulation would be enormously helpful. These patients have near certain mortality without LVAD but tools are needed to select those with a reasonable chance of survival. Therefore, we tested our MELD reclassification scoring system in patients who were classified as INTERMACS 1 and 2, and it remained a powerful predictor of outcomes. For example, in high-risk patients on admission, those who improved to <19 had a 1 year survival which was similar to those who remained low risk. More importantly, all of the profile 1 and 2 patients who were originally classified as low risk but were reclassified as high risk before implantation had died within 1 year. Thus, our MELD reclassification scheme can be very useful to aid in decision making even in patients already known to be at the highest risk of mortality.

Although previous studies have demonstrated the importance of MELD score within 48 hours of implantation,12 ultimately we have shown that these scores are not static and that our data may help to guide decisions of when and whether to implant an LVAD in patients at higher risk. For example, for a patient who presents with a high-risk MELD score, our data demonstrate that these patients may subsequently undergo LVAD implantation with a similar outcome as patients who are admitted with a low-risk profile if their MELD parameters improve sufficiently during the course of their admission. And conversely, if a patient with a low risk has an acute change in condition that worsens the MELD score, deferring LVAD implantation should be considered. However, delaying LVAD surgery might not be feasible, or might not lead to improvement in MELD scores in certain patients. Patients who failed to improve their MELD (group HH) were sicker at admission compared with those patients who did (group HL). For instance, 44% of the patients in group HH were already on dialysis before admission (vs. 0% in group HL; p < 0.01). For MELD, any patient who had dialysis in the past 2 weeks is allocated a creatinine value of 4 (regardless of current values). Whether supporting these patients for longer periods of time (e.g., >2 weeks) with inotropic or MCS to assess renal recovery before the decision to place a durable LVAD would alter the risk profile is unknown. Certainly, the high proportion of patients in the HH group receiving dialysis in our study was a large factor in their remaining in the high-risk MELD group.

Study Limitations

Of note our study has the following limitations: 1) this is a single-center retrospective study with a limited number of patients, especially patients who transitioned from the low- to the high-risk group (2). Also, our study involved predominantly patients admitted to the hospital who needed an LVAD on the index admission, and thus may not be applicable to a broader population of outpatients. This is an important distinction as implantation is shifting toward less-sick patients.

Importantly, it may seem incongruous that some patients that were categorized as low risk by MELD at day of implant also were high risk by according to an INTERMACS profile of 1 (i.e., Group LL had 14 patients classified as profile 1 and group HL had 5). However, this is not uncommon given the independent nature of the risk scores. For example, in a recent study by Adamo et al.,14 33% of their INTERMACS profile 1 patients were classified as “low risk” using the HeartMate Risk Score, a score that shares some of the same variables as MELD. In addition, our MELD score can offer further classification across different INTERMACS profiles.

Given the retrospective nature of the study, we do not have specific details to determine the interventions that led to an improvement in MELD or MELD-XI scores. The use of therapy tailored toward invasive hemodynamics is frequent at our institution; however, our analysis revealed that there was no difference if patients received treatment tailored to hemodynamics or not; nevertheless, our analysis does not include the specifics of types and dosages of drugs, such as inotropes or vasodilators. Therefore, we are unable to identify a specific strategy that may have been associated with a change in risk profile. We do know that the use of temporary MCS (most often IABP), for example, was fairly common (40% of our cohort), but on analysis there was no difference in those patients who remained high risk and those who improved. Despite this limitation, the risk score reclassification is able to predict outcomes in a robust fashion independent of the management strategy, and this may aid clinicians in determining LVAD candidacy in hospitalized patients with advanced heart failure.


In summary, our study has demonstrated the reproducibility of the MELD score to predict outcomes in patients who undergo LVAD implantation. Furthermore, we have shown for the first time the critical importance of assessing the score in a dynamic scheme to fully capture the clinical changes that impact outcome. Specifically, we have shown that reclassifying the patient according the MELD score from the time of admission to the time of implantation is highly predictive of patient survival. Our results suggest that patients undergoing LVAD implant with a declining MELD score (low to high risk) and have an INTERMACS profile 1 or 2 should be deferred from LVAD surgery as they have dismal outcomes with 0% survival. Whether any particular intervention could alter the risk profile and hence improve these patients’ outcomes needs to be determined in a larger cohort. Our data are additive to the current risk score literature and can help aid in the timing of LVAD and the decision on whether or not a patient can safely undergo implantation.


1. Miller LW, Pagani FD, Russell SD, et al.; HeartMate II Clinical Investigators: Use of a continuous-flow device in patients awaiting heart transplantation. N Engl J Med 2007.357: 885896.
2. Rose EA, Gelijns AC, Moskowitz AJ, et al.; Randomized Evaluation of Mechanical Assistance for the Treatment of Congestive Heart Failure (REMATCH) Study Group: Long-term use of a left ventricular assist device for end-stage heart failure. N Engl J Med 2001.345: 14351443.
3. Kirklin JK, Naftel DC, Pagani FD, et al. Seventh INTERMACS annual report: 15,000 patients and counting. J Heart Lung Transplant 2015.34: 14951504.
4. Lietz K, Long JW, Kfoury AG, et al. Outcomes of left ventricular assist device implantation as destination therapy in the post-REMATCH era: Implications for patient selection. Circulation 2007.116: 497505.
5. Kormos RL, Teuteberg JJ, Pagani FD, et al.; HeartMate II Clinical Investigators: Right ventricular failure in patients with the HeartMate II continuous-flow left ventricular assist device: Incidence, risk factors, and effect on outcomes. J Thorac Cardiovasc Surg 2010.139: 13161324.
6. Matthews JC, Pagani FD, Haft JW, Koelling TM, Naftel DC, Aaronson KD. Model for end-stage liver disease score predicts left ventricular assist device operative transfusion requirements, morbidity, and mortality. Circulation 2010.121: 214220.
7. Cowger J, Sundareswaran K, Rogers JG, et al. Predicting survival in patients receiving continuous flow left ventricular assist devices: The HeartMate II risk score. J Am Coll Cardiol 2013.61: 313321.
8. Malinchoc M, Kamath PS, Gordon FD, Peine CJ, Rank J, ter Borg PC. A model to predict poor survival in patients undergoing transjugular intrahepatic portosystemic shunts. Hepatology 2000.31: 864871.
9. Ravaioli M, Grazi GL, Ballardini G, et al. Liver transplantation with the Meld system: A prospective study from a single European center. Am J Transplant 2006.6: 15721577.
10. Bonde P, Ku NC, Genovese EA, et al. Model for end-stage liver disease score predicts adverse events related to ventricular assist device therapy. Ann Thorac Surg 2012.93: 15411547; discussion 1547.
11. Deo SV, Daly RC, Altarabsheh SE, et al. Predictive value of the model for end-stage liver disease score in patients undergoing left ventricular assist device implantation. ASAIO J 2013.59: 5762.
12. Yang JA, Kato TS, Shulman BP, et al. Liver dysfunction as a predictor of outcomes in patients with advanced heart failure requiring ventricular assist device support: Use of the Model of End-stage Liver Disease (MELD) and MELD eXcluding INR (MELD-XI) scoring system. J Heart Lung Transplant 2012.31: 601610.
13. Kirklin JK, Naftel DC, Pagani FD, et al. Sixth INTERMACS annual report: a 10,000-patient database. J Heart Lung Transplant 2014.33: 555564.
14. Adamo L, Nassif M, Tibrewala A, et al. The Heartmate Risk Score predicts morbidity and mortality in unselected left ventricular assist device recipients and risk stratifies INTERMACS class 1 patients. JACC Heart Fail 2015.3: 283290.

CF-LVAD; risk scores; MELD; outcomes

Copyright © 2017 by the ASAIO