Ventricular Assist Device Utilization in Heart Transplant Candidates
Nationwide Variability and Impact on Waitlist Outcomes
Background: Continuous-flow left ventricular assist devices (CF-LVADs) have become a standard treatment choice in advanced heart failure patients. We hypothesized that practice patterns with regards to CF-LVAD utilization vary significantly among transplant centers and impact waitlist outcomes.
Methods and Results: The United Network for Organ Sharing registry was queried to identify adult patients who were waitlisted for heart transplantation (HT) between 2008 and 2015. Each patient was assigned a propensity score based on likelihood of receiving a durable CF-LVAD before or while waitlisted. The primary outcomes of interest were death or delisting for worsening status and HT at 1 year. A total of 22 863 patients from 92 centers were identified. Among these, 9013 (39.4%) were mechanically supported. CF-LVAD utilization varied significantly between and within United Network for Organ Sharing regions. Freedom from waitlist death or delisting was significantly lower in propensity-score–matched patients who were mechanically supported versus medically managed (83.5% versus 79.2%; P<0.001). However, cumulative incidence of HT was also lower in mechanically supported patients (53.3% versus 63.6%; P<0.001). Congruous mechanical and medical bridging strategies based on clinical risk profile were associated with lower risk of death or delisting (hazard ratio, 0.88; P=0.027) and higher likelihood of HT (hazard ratio, 1.14; P<0.001).
Conclusions: CF-LVAD utilization may lower waitlist mortality at the expense of lower likelihood of HT. Decision to use CF-LVAD and timing of transition should be individualized based on patient-, center-, and region-level risk factors to achieve optimal outcomes.
WHAT IS NEW?
This study demonstrates that significant variation in the use of left ventricular assist devices as a bridge to heart transplant exists among and within United Network for Organ Sharing regions.
The use of ventricular assist devices in waitlisted patients can decrease the likelihood of death or delisting in the first year but may increase time to transplant.
Although overall waitlist survival may improve with ventricular assist device use, patient, as well as center- and region-specific, characteristics must be considered to improve outcomes.
WHAT ARE THE CLINICAL IMPLICATIONS?
Because of the higher priority given to ventricular assist devices in the new allocation system, time to transplant may be reduced. However, overtreating with ventricular assist devices, particularly in patients with unfavorable characteristics, could lead to worsening waitlist outcomes.
More research is needed to better delineate optimal selection of bridging strategy by analyzing patient-, center-, and regional-level data.
Heart transplantation (HT) remains the gold standard of treatment in patients with end-stage heart failure. Although the number of transplants performed annually is on the rise, overall donor supply falls far short of demand.1 As a result, an increasing number of patients require mechanical circulatory support as a bridge-to-transplantation (BTT). In 2000, the International Society for Heart Transplantation reported that 19.1% of transplant recipients were mechanically bridged, a figure that increased to 45.0% in 2012.2 This was accompanied by major advances in the device design evolving from previous generation pulsatile-flow left ventricular assist devices to new generation continuous-flow left ventricular assist devices (CF-LVADs). Prospective clinical trials have demonstrated excellent durability, complication profile, and survival in patients supported with CF-LVADs for the BTT indication.3,4
Mechanical bridging with CF-LVAD remains a reasonable strategy for majority of waitlisted patients, particularly for those with evidence of clinical decompensation, hemodynamic instability, or worsening end-organ function on medical therapy. However, certain patient populations are considered poor candidates for CF-LVAD therapy because of their underlying physiology (restrictive disease), bleeding diathesis, or hypercoagulable states. CF-LVAD implantation may also positively or negatively impact transplant priority and listing status depending on clinical course of the patient. Moreover, region- and center-level factors, such as waitlist time and physician/surgeon preference, may impact the decision to use CF-LVAD. The purpose of this study was (1) to determine the patient-level risk factors associated with CF-LVAD utilization in HT candidates, (2) to investigate region- and center-level differences in CF-LVAD utilization for BTT indication, and (3) to assess the impact of mechanical versus medical bridging strategies on HT waitlist outcomes.
Study Design, Variables, and Definitions
The United Network for Organ Sharing (UNOS) database was queried to identify adult patients (≥18 years old) who were listed for HT between 2008 and 2015. Patients without a designated listing center, as well as those who were listed for multiorgan transplant, were excluded from the current analysis (Figure 1). Center waitlist volume was calculated by identifying the total number of patients listed for HT at a given center during the study period. Patients from centers who listed <10 patients per year during the study period (<80 patients overall) were excluded from the analysis. Patient-level baseline characteristics analyzed included demographics, pathogenesis of heart failure, comorbid conditions, functional status, and UNOS status at listing. Patient who were less likely to receive a CF-LVAD based on their physiology, including those with hypertrophic or restrictive cardiomyopathy, congenital abnormalities, or cardiac allograft failure, were categorized under a nondilated myopathy group. Bridging strategy was considered mechanical for patients who received a durable CF-LVAD before listing or while listed and medical for those who did not receive a CF-LVAD. Primary outcome of the study was freedom from death or delisting for worsening status. Secondary outcome of the study was cumulative incidence of HT.
We first compared waitlisted patients who received mechanical versus medical bridging strategy to identify patient-level determinants of CF-LVAD utilization. Based on these variables, a propensity score was generated for each patient predicting the likelihood of receiving a CF-LVAD before or while waitlisted—also termed VAD Likelihood Score (Figure IA in the Data Supplement). Variables included in the propensity score were sex, ethnicity, history of diabetes mellitus, smoking history, functional status, automatic internal cardiac defibrillator, blood type O, high body surface area (BSA), nondilated myopathy, UNOS region, UNOS status at listing, and listing center. For example, patient A who has a large BSA, blood type O, dilated myopathy, and listed region 6 would have a high VAD likelihood score, thereby an increased chance of receiving a CF-LVAD, as opposed to patient B who has a normal BSA, blood type AB, hypertrophic cardiomyopathy, and listed in region 5 consistent with a low VAD likelihood score. Propensity score matching was then used to identify patients with similar VAD likelihood scores. Impact of mechanical versus medical bridging strategies on transplant waitlist outcomes was assessed in the propensity-score–matched patient cohort.
Next, we determined whether bridging strategy (mechanical versus medical) was consistent with VAD likelihood score of each individual on the transplant waitlist (Figure IB in the Data Supplement). Patients with a high VAD likelihood score (>1 SD above the mean) were considered to have congruous bridging strategy if mechanical support was chosen and considered to have incongruous bridging strategy if medical management was used. Conversely, patients with a low VAD likelihood score (>1 SD below the mean) were considered to have congruous bridging strategy if medical management was chosen and considered to have incongruous bridging strategy if mechanical support was used. Patients who have a VAD likelihood score within 1 SD around the mean were considered to have a congruous bridging strategy irrespective of the strategy used. Waitlist outcomes were then assessed in 4 groups of patients: congruous mechanical bridging (anticipated CF-LVAD and received), congruous medical bridging (did not anticipate CF-LVAD and did not receive), incongruous mechanical bridging (did not anticipate CF-LVAD but received one), and incongruous medical bridging (anticipated CF-LVAD but did not receive one).
To determine region-level differences in CF-LVAD utilization for transplant eligible patients, we calculated CF-LVAD utilization percentage (VAD%) by dividing the number of patients implanted with CF-LVAD (before or during the waitlist period) by the total number of waitlisted patients for each center using the propensity-matched cohort. Variability in mechanical versus medical bridging strategies was then assessed at the center level among and within UNOS regions. The current study was approved by the Columbia University Institutional Review Board. The data used in the study are available to other researchers for purposes of reproducing the results or replicating the procedure via data request from the UNOS/Organ Procurement and Transplantation Network network.
Descriptive analyses were conducted for all baseline variables and are presented as means and SDs for continuous variables and numbers and percentages for categorical variables. Differences among medical versus mechanical bridging strategy groups were quantified using independent Student t test and χ2 where appropriate. Propensity scores for receiving durable CF-LVAD before listing or while listed (also termed as VAD Likelihood Score) were generated using multivariable logistic regression analysis. Propensity matching was performed using 1 to 1 nearest neighbor matching with specified caliper distance of 0.001.
To ensure balance in propensity-matched cohort, absolute standard differences were assessed before and after matching, with <10% absolute standard differences considered acceptable. Kaplan–Meier survival estimates were used to assess the impact of bridging strategies on freedom from death or delisting and cumulative incidence of transplant, respectively, in the propensity-matched cohort with log-rank test used for comparisons between mechanical versus medical bridging groups. As secondary analysis, cumulative incidence curves were generated to visualize competing event rates of death or delisting versus HT, comparing both medical and mechanical and congruous and incongruous bridging strategies. Multivariable Cox-regression model was used to determine whether congruous utilization predicted waitlist outcomes. All P values were reported as 2-sided tests with P<0.05 considered statistically significant. STATA version 13.1 (Stata corp, College Station, TX) was used to perform statistical analysis.
Patient Population and Nationwide Trends in Durable Device Utilization
A total of 22 863 adult patients from 92 centers were identified as having been listed for single-organ HT during the study period. The number of waitlisted patients who received CF-LVAD has steadily increased from 2008 through 2014 (Figure IIA in the Data Supplement). In parallel with this increase, median time spent on the waitlist has also increased (Figure IIB in the Data Supplement). The baseline characteristics of study population were summarized in Table 1. Among these, 9049 (39.4%) received a mechanical bridging strategy (Figure 1). Patients who received a mechanical bridging strategy were more likely to be male and more likely to have a history of diabetes mellitus and tobacco use. Patients with larger BSA and blood type O were more likely to receive mechanical support (Table 1). Only 10% of patients overall had congenital heart disease, were listed for retransplantation, or had restrictive or hypertrophic cardiomyopathy. These patients only represented 3% of mechanically supported patients versus 15% of medically managed patients. Most patients were UNOS status 1B at the time of listing although both UNOS status 1A and 1B were more common among mechanically supported patients.
Predictors of Bridge to Transplant With CF-LVAD and Propensity Score
In an effort to identify factors that contributed to CF-LVAD utilization at the patient level, we performed univariable and multivariable logistic regression analyses using baseline patient characteristics. Male sex, ethnicity, diabetes mellitus, smoking history, functional status, blood type O, high BSA, pathogenesis of heart failure, UNOS status at listing, and UNOS regions were strongly associated with CF-LVAD utilization as BTT (Table 2). When the variables were used to generate a propensity score, a propensity-matched cohort of 11 888 patients (5944 mechanically supported, 5944 medically bridged) was identified. Propensity scores ranged from 0.024 to 0.869 and showed a normal distribution within the study population (Figure IA in the Data Supplement). Baseline characteristics between patients who received mechanical versus medical bridging strategies were well balanced after propensity score matching (Table I in the Data Supplement).
Nationwide Variability in Center-Level Utilization of CF-LVAD as BTT
After analyzing patient-level determinants of CF-LVAD utilization, we turned our attention to region- and center-specific utilization patterns. In the overall cohort, mean center CF-LVAD utilization was 38.2% and highly variable among centers with a range of 8.1% to 77.4%. The variability among UNOS regions was displayed using a color-coded map in Figure 2A, where the mean VAD% of each region is represented. Significant variability not only existed among regions but also within regions, where Figure 2B demonstrates the variation in size of interquartile range from the mean center-specific VAD% within each region. We hypothesized that, in part, variability in center-to-center CF-LVAD utilization may be explained by wait list times. To this end, we correlated overall wait list days and days as UNOS 1A, 1B, and status 2 with VAD% in each of the 93 centers. In analyzing overall wait list days, we found a weakly positive correlation (R2=0.16) with VAD% (Figure IIIA in the Data Supplement). Although the correlation was similar when UNOS status 1A days were analyzed (R2=0.13; Figure IIIB in the Data Supplement), the correlation was stronger between UNOS 1B days and VAD% (R2=0.25; Figure IIIC in the Data Supplement). There was no correlation between UNOS status 2 days and VAD%.
Waitlist Outcomes Based on Patient-Level Device Utilization
In the propensity-matched cohort, a total of 1944 (16.4%) patients died or were delisted for worsening status during the first year on the wait list. Kaplan–Meier analysis demonstrated that patients who received mechanical support had significantly greater freedom from death or delisting as compared with patients who were managed medically (83.5% versus 79.2%; P<0.00; Figure 3A). On the contrary, mechanically supported patients had significantly lower rates of HT compared with those who were medically managed (53.3% versus 63.6%; P<0.001; Figure 3B). Cumulative incidence curves treating death or delisting and transplantation as competing events for patients who had mechanical or medical bridging are represented in Figure IVA in the Data Supplement.
Impact of Congruous Bridging Strategy on Waitlist Outcome
Next, we assessed the impact of congruous bridging strategy on waitlist outcomes. Each patient was assigned 1 of 4 categories based on their VAD likelihood score and bridging strategy used (eg, congruous mechanical bridging, congruous medical bridging, incongruous mechanical bridging, and incongruous medical bridging). We defined incongruous utilization based on bridging strategy and propensity score greater than or less than 1 SD from the mean. In the propensity-matched cohort, we identified 2065 patients (17.4%) who had an incongruous bridging strategy based on their propensity score. Among these, 1063 had incongruous medical bridging and 1002 had incongruous mechanical bridging. When freedom from death or delisting at 1 year was assessed, patients who had congruous mechanical bridging had the best 1-year survival while those who had incongruous medical bridging had the worst 1-year survival (Figure 4A). When cumulative incidence of transplant at 1 year was assessed, patients who had incongruous mechanical bridging had the lowest likelihood of transplantation. Those who had incongruous medical bridging also had a lower rate of transplant than those who had congruous medical bridging (Figure 4B). Congruous bridging was an independent predictor of death or delisting (hazard ratio, 0.84 [0.75–0.95]; P=0.004) and HT (hazard ratio, 1.08 [1.01–1.15]; P=0.026; Table II in the Data Supplement). Taking into account both outcomes, congruous utilization improved rates of death or delisting and increased likelihood of transplant in both mechanical and medical bridging groups (Figure IVB in the Data Supplement).
The current study examines the impact of CF-LVAD utilization on patient-specific waitlist outcomes, specifically death or delisting and HT. The important findings of the present study are as follows: (1) The number of heart transplant candidates who were implanted with a CF-LVAD has steadily increased nationwide after the approval of this technology for the BTT indication, (2) Significant variability exists in CF-LVAD utilization among and within UNOS regions, (3) Both patient-level and center-level factors impact the decision to use CF-LVAD in transplant candidates, (4) Bridging with a CF-LVAD seems to decrease the likelihood of waitlist mortality or delisting for worsening status at the expense of decreased likelihood of transplantation within the first year of listing, and (5) Improved waitlist outcomes can be achieved by congruous utilization of medical and mechanical bridging strategies based on region-, center-, and patient-level factors.
After recent improvements in durability and energy efficiency, CF-LVADs have become a mainstay of advanced heart failure therapy.5,6 Although recent The International Society For Heart and Lung Transplantation registry data suggest that the percent of HT recipients who are bridged to transplant with CF-LVAD are increasing, their data does not address the waitlist population.1 The data in the current study suggest that, since 2008, the use of CF-LVADs has increased steadily in this population, likely reflective of increased wait list time and static organ supply in the contemporary era.
The next interesting finding of the current study was the dramatic variability in CF-LVAD utilization not only among but also within UNOS regions. Center-level analysis has demonstrated a correlation between longer transplant waitlist times and increased CF-LVAD utilization, suggesting that the average time to transplant at a given center may impact the decision to use CF-LVADs. However, this finding should be interpreted with caution because CF-LVAD utilization may stabilize sick patients and prolong waitlist times by downgrading their status to UNOS 1B in absence of device complications. In addition, the strength of the correlation between waitlist times and CF-LVAD utilization was moderate at best, suggesting that other region- and center-level factors may contribute to the nationwide variability in utilization. For example, insurance coverage for CF-LVAD as BTT varies at the state level, which most certainly influences the way in which patients are cared for before transplant. In addition, physician and surgeon preference likely has a significant impact on the decision to use CF-LVAD, type of support device used, and the timing of device implantation in reference to patient’s clinical status. Other center-level factors, such as hospital resources, infrastructure, and referral patterns, may certainly play a role in the observed variability. It is important to note that many of the aforementioned factors are inherently difficult to quantify and poorly represented in nationwide registries, such as UNOS or the Interagency Registry for Mechanically Assisted Circulatory Support. Nevertheless, future observational studies incorporating center-specific data at the granular level are required to identify precise reasons underlying nationwide variability in CF-LVAD utilization.
The decision to use a CF-LVAD as BTT is also related to patient-specific factors. We identified male sex, ethnicity, diabetes mellitus, history of smoking, current automatic internal cardiac defibrillator, blood type O, large BSA, cardiomyopathy diagnosis, UNOS status at listing, and UNOS region as patient-level factors that were significantly associated with the decision to bridge a patient with a CF-LVAD. Although prior studies have elucidated many of these patient characteristics as risk factors, the current study also takes into account UNOS region and listing center, which dictate organ availability and expected wait list time and, thus, contributes to decision making on CF-LVAD utilization in waitlisted patients.
When the primary outcomes were compared between medically and mechanically bridged patients among the propensity-matched cohort, mechanically supported patients were less likely to die or be delisted but also less likely to receive a transplant than those who were medically managed. This survival benefit is consistent with data from the REMATCH trial (Randomized Evaluation of Mechanical Assistance for the Treatment of Congestive Heart Failure), which demonstrate superior outcomes with CF-LVAD as compared with optimal medical management in the destination therapy population.7 These results are also similar to those published by Trivedi et al,8 who studied the impact of BTT with Heartmate II as compared with medical management in a smaller, less contemporary cohort. They found that BTT patients had lower waitlist mortality and a longer waiting time to transplant than those who were medically bridged. This study, however, only included patients who were listed at UNOS status 1A and 1B and excluded patients with other types of CF-LVADs from the analysis. Similarly, Wever-Pinzon et al9 demonstrated that CF-LVAD patients in the contemporary era had improved freedom from death or delisting when compared with UNOS status 1A and 1B patients who were medically managed.
Last, we analyzed congruity of CF-LVAD utilization based on patient’s location and clinical profile and assessed its impact on waitlist outcomes. Those patients who, based on their propensity scores and bridging method (mechanical versus medical management), had a congruous bridging strategy appeared to have superior outcomes as compared with those who had incongruous bridging. This suggests that improved patient selection—both for CF-LVAD utilization as BTT and medical therapy—could improve waitlist outcomes and increase the number of patients successfully and safely bridged to transplantation.
Importantly, the findings presented in the present study must be viewed in the context of the upcoming changes in the UNOS organ allocation system for HT, which will give higher priority to CF-LVADs than in the previous schema. By prioritizing transplants to CF-LVAD patients, the new system might attenuate some of the advantage to medically bridged patients with regards to likelihood of transplantation seen in the current study. However, it could also incentivize centers to use CF-LVAD more often as BTT, further highlighting the importance of appropriate utilization given the suboptimal outcomes seen in those patients who inappropriately received a CF-LVAD as BTT.
It is also important to take the findings of the current study in the context of its many limitations. First, the data were extracted from a large registry, which subjects the data to both missingness and error in data entry. Second, because data points are routinely collected at the time of listing and the time of transplant, a substantial amount of patient information on time while listed (awaiting HT, eg) is not available. This is specifically challenging when it comes to capturing patients who received mechanical circulatory support during the waiting period but who died or were delisted before transplant. Last, data on acuity of patients at the time of CF-LVAD implantation (such as Interagency Registry for Mechanically Assisted Circulatory Support profile) were not available, which may vary among centers and could impact outcomes analyzed in this study.
In conclusion, CF-LVAD utilization in transplant eligible patients has steadily increased in the United States with significant variability among and within UNOS regions. Patient-, center-, and region-level factors impact the decision to use CF-LVAD. Based on the discordance in outcomes seen in those congruously and incongruously bridged, the risks and benefits of CF-LVAD therapy, as well as timing of its initiation, must be carefully weighed in each individual. As CF-LVAD use as BTT becomes more widespread, and particularly in light of the new UNOS organ allocation system, ongoing research is necessary to identify which patients would derive the greatest benefit, and the least harm, from CF-LVAD utilization as BTT.
Sources of Funding
This study was supported by the National Center for Advancing Translational Sciences, National Institutes of Health, through Grant Number UL1TR001873 (Dr Topkara) and KL2TR001874 (Dr Garan) as well as by Lisa and Mark Schwartz and the Program to Reverse Heart Failure at New York Presbyterian Hospital/Columbia University.
Dr Naka received consulting fees from Abbott and Medtronic. The other authors report no conflicts.
The Data Supplement is available at http://circheartfailure.ahajournals.org/lookup/suppl/doi:10.1161/CIRCHEARTFAILURE.117.004586/-/DC1.
- Received April 27, 2017.
- Accepted March 1, 2018.
- © 2018 American Heart Association, Inc.
- Lund LH,
- Edwards LB,
- Dipchand AI,
- Goldfarb S,
- Kucheryavaya AY,
- Levvey BJ,
- Meiser B,
- Rossano JW,
- Yusen RD,
- Stehlik J
- 2.↵Transplantation ISHLT. Adult Heart Transplantation Statistics. 2014: Slide 41. https://www.ishlt.org/registries/slides.asp?slides=heartLungRegistry&year=2014. Accessed November 23, 2017.
- Slaughter MS,
- Pagani FD,
- McGee EC,
- Birks EJ,
- Cotts WG,
- Gregoric I,
- Howard Frazier O,
- Icenogle T,
- Najjar SS,
- Boyce SW,
- Acker MA,
- John R,
- Hathaway DR,
- Najarian KB,
- Aaronson KD
- Miller LW,
- Pagani FD,
- Russell SD,
- John R,
- Boyle AJ,
- Aaronson KD,
- Conte JV,
- Naka Y,
- Mancini D,
- Delgado RM,
- MacGillivray TE,
- Farrar DJ,
- Frazier OH
- Rose EA,
- Gelijns AC,
- Moskowitz AJ,
- Heitjan DF,
- Stevenson LW,
- Dembitsky W,
- Long JW,
- Ascheim DD,
- Tierney AR,
- Levitan RG,
- Watson JT,
- Meier P,
- Ronan NS,
- Shapiro PA,
- Lazar RM,
- Miller LW,
- Gupta L,
- Frazier OH,
- Desvigne-Nickens P,
- Oz MC,
- Poirier VL
- Wever-Pinzon O,
- Drakos SG,
- Kfoury AG,
- Nativi JN,
- Gilbert EM,
- Everitt M,
- Alharethi R,
- Brunisholz K,
- Bader FM,
- Li DY,
- Selzman CH,
- Stehlik J