Outcomes with the current generation of continuous flow devices are superior to the prior generation of pulsatile devices (1), which has not only led to continuous flow devices becoming the dominant technology, but has also resulted in an increasing use of mechanical circulatory support (MCS), both as a bridge to transplant (BTT) and destination therapy (DT) (2). Despite the improved outcomes with the current generation of continuous flow pumps (3- 5), proper patient selection remains critical, as patients who are less sick at the time of implantation tend to have superior outcomes (6). Furthermore, understanding and quantifying risk prior MCS is essential for physicians to be able to convey these risks to patients and their families so that they can make informed decisions.

The desire for risk prediction prior to MCS implantation has led to the development of several risk scores (7- 11). However, these scores were largely derived from retrospective analyses of small, single-center experiences using pulsatile flow technology and lacked validation. One of the most widely used risk scores is the Destination Therapy Risk Score (DTRS), which was developed as part of a U.S. Food and Drug Administration–mandated post-market surveillance of the pulsatile-flow HeartMate XVE (Thoratec Corporation, Pleasanton, California) left ventricular assist device (LVAD) implanted as DT (10). This multi-institutional dataset evaluated pre-operative risk factors that predicted 90-day in-hospital mortality. Using multivariate analysis, a weighted risk score consisting of 9 risk factors was developed that stratified patients into low-, medium-, high-, and very high-risk categories. Whereas the score was initially developed to predict 90-day in-hospital mortality, the same risk categories were able to stratify patients by their long-term outcomes as well.

As this was the most comprehensive risk model available, it was applied to not only long-term outcomes for DT, but also to long-term outcomes for BTT patients with pulsatile devices and eventually was extended to patients implanted as both BTT and DT with continuous flow devices. However, the utility of the DTRS in BTT populations, in particular, and those with continuous flow devices, in general, has not been systematically studied. Given the changes in patient selection, device type, and patient management since the DTRS was introduced, we hypothesized that the DTRS would lack significant discriminatory capacity when it was applied to large populations of patients receiving continuous flow MCS.

The details of patient recruitment into the HeartMate II (Advanced Heart Failure Treated With Continuous-Flow Left Ventricular Assist Device) BTT and DT trials have been previously reported (1,3). Patients who received a HeartMate II (HMII) as part of the HMII BTT trial (n = 486), HMII DT trial (n = 638), or who received a HeartMate XVE (XVE) as part of the HeartMate II DT trial (n = 114) were included in this study. Baseline demographics, laboratory values, echocardiographic parameters, and medical therapies were recorded just prior to device implantation.

The DTRS was calculated for each patient and then patients were divided into risk groups as previously described: low risk (DTRS 0–8), medium risk (DTRS 9–16), high risk (DTRS 17–19), and very high risk (DTRS >19). Complete data for calculating the DTRS was available for 95% (1,063 of 1,124) of the patients with a HMII. Patients with missing data were not included in the analysis. The high-risk and very high-risk categories were combined into a single risk category due to the low number of patients in the very high-risk group, and are henceforth referred to as “high risk.”

Analyses of continuous variables for more than 2 groups were performed using either analysis of variance (if data was normally distributed), or the Kruskal-Wallis nonparametric test (for non-normal data). Post hoc pairwise comparisons between 2 groups were either performed using the Tukey or the Mann-Whitney U tests. Categorical variables were compared using the Pearson chi-squared test for >2 groups and the Fisher exact test for 2 groups. The 90-day in-hospital mortality was evaluated for all groups and was the primary outcome variable for the analysis. Area under the receiver-operator characteristic curves were calculated for the 90-day in-hospital mortality to evaluate the discriminatory value of the DTRS for each group studied, and Hosmer-Lemeshow goodness-of-fit statistics were used to evaluate the calibration of the dataset. Discrimination and calibration was performed on DTRS itself and not on risk groups. Survival analysis was performed using the Kaplan-Meier method and comparisons of survival between groups were made using the Mantel log-rank test. Patients were censored at the time of transplantation, recovery, or ongoing support at the time of this analysis (September 2010). All statistical analyses were performed using either Systat (Cranes Software, Chicago, Illinois) or SAS (SAS Institute, Cary, North Carolina).

As compared to the HMII DT group, the BTT group was younger and less likely to be white or have an ischemic cardiomyopathy (Table 1). The BTT group also had a higher baseline heart rate and pulmonary capillary wedge pressure and lower systolic blood pressure and pulmonary vascular resistance. Laboratory values also tended to indicate a more advance state of heart failure in the BTT group as evidenced by lower sodium and higher transaminases. The BTT group also had a higher use of inotropes, intra-aortic balloon pump, and mechanical ventilation and was less likely to be treated with beta-blockers or diuretics. However, the BTT group had a higher creatinine clearance and platelet count than the DT group.

The HMII DT group was similar to the XVE DT group. The XVE DT group was more frequently male, had a higher use of inotropes and angiotensin blockade, but had fewer patients on beta-blockade. The HMII DT patients had a slightly lower body mass index and body surface area than the XVE DT group. Otherwise, their hemodynamic profile and laboratory values were similar, with the exception of worse renal function and a lower white blood cell count in the HMII DT group.

The majority of patients had a low- or medium-risk DTRS regardless of indication. Approximately 80% to 90% of patients were considered either low or medium risk, whereas only about 10% to 20% of the patients were considered high risk (Figure 1). This is in contrast to the original DTRS cohort as published by Lietz et al. (10), in which less than 30% of patients were low risk and over 20% were considered high risk. The percentage of patients with each individual component of the risk score is presented in (Table 2). Compared with the HMII DT group, the HMII BTT group had a lower percentage of patients who had a platelet count ≤148 k/μl and who were not on intravenous inotropes and a higher percentage of patients with an aspartate amiotransferase (AST) >45 U/ml. Interestingly, only the percentage of patients who met the AST cutoff was different between the HMII BTT and XVE DT groups. Lastly, the only difference between the DT groups was a higher proportion of patients not treated with inotropes in the HMII DT group. Sixty-one (5%) of the 1,124 HMII patients had missing data for at least 1 of the variables required for calculation of the Lietz-Miller score. Out of these 61 patients, platelet count was missing in 9 (15%), albumin was missing in 18 (30%), international normalized ratio was missing in 13 (21%), pulmonary artery pressure measurement was missing in 38 (62%), AST was missing in 16 (26%), hematocrit was missing in 8 (13%), and blood urea nitrogen was missing in 8 (13%) patients. The 90-day in-hospital mortality in this group was 16%.

There were no differences in 90-day in-hospital mortality between those with low- and medium-risk DTRS in any of the groups (Figure 2). However, in the HMII BTT and DT groups, those with a high-risk DTRS had significantly higher 90-day in-hospital mortality than did those with a low-risk DTRS. Although there were numerically higher rates of 90-day in-hospital mortality in the medium- and high-risk DTRS groups for those who received a XVE, they were not statistically different from the low-risk group. Receiver-operator characteristic curves for each of the following groups, BTT+DT, BTT alone, DT alone, and XVE DT demonstrated the DTRS had only modest discriminatory capacity for the determination of 90-day in-hospital mortality with areas under the curve of <0.6 (Figure 3). Additionally, the DTRS did not appear to provide any meaningful discrimination for HMII BTT (p = 0.30) and the XVE DT (p = 0.93) patient cohorts. (Figure 4) shows the Hosmer-Lemeshow calibration plots for each of the 4 groups studied. All the p values were >0.05, demonstrating that there is no evidence of a lack of fit across the 4 groups. (Table 3) shows the univariate association (p value) of the individual components of DTRS with 90-day in-hospital mortality of the 4 cohorts. The only statistically significant variable in this group that was predictive of 90-day mortality was blood urea nitrogen >51 for the HMII (BTT+DT), HMII (BTT), and the XVE DT cohorts.

Survival over 2 years for all patients with a HMII, combining both BTT and DT populations, and stratified by DTRS is shown in (Figure 5)A, with significantly worse survival in those with higher DTRS (p = 0.014). For those in the lowest risk DTRS group, survival at 6, 12, and 24 months was 90 ± 1%, 79 ± 2%, and 66 ± 3%, respectively, whereas for those in the highest risk group, the survival was 76 ± 4%, 66 ± 4%, and 56 ± 5% at the same time intervals. For the HMII BTT population, there was no survival difference over 2 years between the 3 DTRS groups (p = 0.461). In contrast, there was a significant difference in survival over 2 years for the HMII DT patients stratified across DTRS categories (p = 0.041) (Figure 5C). Lastly, survival was not significantly different between risk categories for patients who received a XVE as DT (Figure 5D), but only a small number of patients (n = 12) were in the high-risk group.

In this study, the utility of the Destination Therapy Risk Score (DTRS), originally derived for a pulsatile flow LVAD, was evaluated in a contemporary cohort of patients with a continuous flow device. The application of the DTRS to a large group of patients who received a continuous flow LVAD demonstrated a shift toward lower risk patients than seen in the original DTRS cohort. The score discriminated between those with high and low 90-day in-hospital mortality, but failed to distinguish between the low- and medium-risk groups. The receiver-operator characteristic curves further demonstrated that DTRS had only a modest ability to discriminate risk of 90-day in-hospital mortality in patients with continuous flow LVAD. DTRS did not provide significant discrimination in those implanted as DT with the XVE, the indication and device for which DTRS was initially developed. Two-year survival was not significantly different among the DTRS risk categories for those who had a HMII as BTT, but did differ for the HMII DT population. When the baseline characteristics were compared, there were substantial differences between the HMII BTT and DT groups, but there were fewer differences between the HMII DT and XVE DT groups. The prevalence of 3 of the 9 individual components of the DTRS differed between the HMII BTT and HMII DT groups, but only differed in 1 of the categories between the HMII DT and XVE DT groups.

Risk assessment prior to mechanical support not only allows for improved selection of candidates, but also allows physicians to communicate these risks to patients and their families in order to make more informed decisions, particularly in the setting of DT. Much of the mortality with MCS is perioperative and related to baseline physiologic markers of advanced heart failure, including renal and hepatic dysfunction, malnutrition, and right heart failure. While many individual characteristics are associated with risk, clinicians have sought multivariable risk prediction models that provide an overall assessment of risk. In addition, identification of risk factors for adverse outcomes provides targets for interventions that may further improve outcomes with mechanical support. However, risk prediction models in the literature have fairly substantial limitations. Many of them were derived from single-center experiences, included devices using previous generation pulsatile technology, investigated only a limited number of pre-operative variables, or had not undergone formal validation (8,11- 12). Even though the DTRS used XVE patients from a multicenter experience and derived the risk factors from a rather large set of pre-implantation variables collected as part of a U.S. Food and Drug Administration–mandated registry, it is still based on a DT population using an earlier generation pulsatile device and was not prospectively validated.

Many of the pre-implantation risk factors, which comprise the DTRS are indicative of the acuity of patient illness at the time of implantation such as renal and hepatic dysfunction, nutritional state, and surrogates of right ventricular function. Interestingly, there were few substantial differences in the percentage of patients with each of the individual components of the risk score for those implanted with a HMII or XVE as DT. In contrast to the original DTRS cohort, there were fewer patients categorized as high risk and more patients categorized as low risk. This may reflect the evolution of practice over time with patients being implanted earlier in their disease course since the DTRS was published. Data from the INTERMACS (Interagency Registry for Mechanically Assisted Circulatory Support) has demonstrated that patients who are less ill, and thus have a higher INTERMACS profiles, have superior outcomes (13). This has also been demonstrated in subsequent studies. In a report of 86 patients with continuous flow devices from a single center, the 1-year survival for those who were INTERMACS profiles 1 and 2 was 56% and was significantly worse than the 84% 1-year survival for those who were INTERMACS profiles 3 to 7, p = 0.004 (14). In a separate study of 101 patients who received continuous flow devices, the 1-year survival for the 24 patients who had a INTERMACS profile of 4 to 7 at implantation was over 95% (6). The growing recognition of superior outcomes in those who are less sick has led to a reduction in the number of patients receiving MCS who are INTERMACS profile 1 and an increase in the number who have less advanced profiles at the time of MCS (15). In the era of the DTRS, virtually all patients were on inotropes prior to MCS and thus the lack of inotropes was found to impart increased risk, perhaps due to inotrope-induced arrhythmias or hypotension, whereas in the current era, the lack of inotropy, and hence a higher INTERMACS profile, may actually be a marker for improved outcomes.

The 90-day in-hospital mortality, the original outcome measure for which DTRS was created, was significantly higher in the high-risk cohort for both the HMII BTT and DT groups. However, there were no differences between the low- and medium-risk cohorts. The receiver-operator characteristic curves demonstrated that DTRS provided no meaningful discrimination for the BTT cohort and only modest discriminatory capacity for the BTT+DT and DT cohorts. Survival over 2 years was significantly different between risk groups for the entire HMII cohort, particularly between the low- and high-risk groups. However, the differences in survival were much more modest than for the original DTRS cohort and the survival differences with the HMII population were driven entirely by the DT group, as there were no differences in 2-year survival in the BTT group. For the BTT group, the medium-risk group had a 74% 1-year survival, which is comparable to previously published 1-year survival rates of 73% using HMII as BTT (4), but both are lower than the currently published survival of 85% for this device in the post-trial era (5,16). For the DT population, those considered at medium risk by DTRS had 70% 1-year survival, which is similar to the 68% overall survival with HMII in the DT trial (1) and the 74% 1-year survival for those with continuous flow devices as DT from INTERMACS (2). The highest risk HMII DT population had a 1-year survival of 62%, worse than the other risk groups, but substantially better than the 1-year survival of less than 30% seen in the high-risk cohort in the original DTRS. Thus, while the DTRS allows for some discrimination in the HMII DT cohort, it neither identifies a group that does particularly well, for whom mechanical support might be more strongly recommended, nor does it describe a group whose outcomes are futile, for whom mechanical support might not be appropriate. Lastly, there was no difference in outcome based on DTRS in the XVE DT population, despite the fact that the overall outcomes for the XVE population in the HeartMate II DT trial were not different than the XVE population from the original REMATCH (Randomized Evaluation of Mechanical Assistance for the Treatment of Congestive Heart Failure) trial (17). As the field of MCS continues to evolve, novel risk prediction models are needed that are derived using patients with continuous flow devices from the current era of implantation and patient selection. Furthermore, with the growing population of patients who have these devices, the risk prediction models must also have a validation cohort to ensure their validity. As with all risk models, there are some factors that are not accounted for by the model that may still affect outcomes and thus the model's predictive ability. Therefore, the application of such risk models must still be used within the context of the patients and their clinical presentations, but hopefully will allow for improved decision making and improved communication of potential risks to patients and their families.

This is a retrospective study of prospectively collected data, but it represents implantation with the current generation of devices, surgical techniques, and patient management across multiple centers. There are clear differences between the BTT and DT populations, particularly for those who received a HMII; however, the purpose of the study was not to compare outcomes between indications, but rather to assess the applicability of DTRS to predict outcomes in the current era of MCS for different indications. Furthermore, patients are no longer receiving the XVE device as either BTT or DT, but the results from this population were presented to demonstrate the lack of predictive capacity of the DTRS even in this population. DTRS was originally derived to determine 90-day in-hospital mortality, but its use was further extended to risk stratify overall survival by the MCS community. Despite this, the discriminatory capacity of DTRS in the current era of continuous flow for 90-day in-hospital mortality MCS is still modest. Improved risk prediction models developed from the current era of continuous flow LVAD are in development.

Application of DTRS to a large cohort of patients with a continuous flow device demonstrates only a modest ability to predict 90-day in-hospital mortality. For patients with continuous flow devices implanted as BTT, DTRS provides poor discrimination of 2-year mortality. However, for patients implanted as DT, DTRS stratifies patient outcomes over 2 years, but it does not characterize a group of patients with a futile outcome.