Short- and long-term survival in advanced heart failure patients supported with left ventricular assist device (LVAD) therapy has dramatically improved over time. Average survival is approximately 80% at 1 year for patients in the Interagency Registry for Mechanically Assisted Circulatory Support (INTERMACS) who are on continuous-flow (CF) support.1 Although there are many factors contributing to survival gains, prior studies have demonstrated that patients who are younger and less ill at the time of LVAD implant tend to have better outcomes on device therapy.1,2 Multiple risk scores have been developed with the goal to provide additional risk prediction above and beyond our usual clinical assessment before LVAD implant.3–10 Importantly, these models can also be used in the patient consent process to discuss risks associated with LVAD therapy based on patient-specific characteristics that may make the patient outside quoted average risks for survival.
The HeartMate II risk score (HMRS) was retrospectively derived and validated in HeartMate II (St. Jude Corporation, Saint Paul, MN) bridge to transplant (BTT) and destination therapy (DT) trial patients with modest accuracy based on the area under the receiver operator characteristic curve (AUC).11 However, this score was derived from a cohort of study patients meeting specific clinical trial criteria who may not be representative of a general LVAD candidate population. Recently, a single-center study questioned the accuracy of the HMRS in predicting mortality in general HeartMate II candidates.12 The components of the HMRS include age, serum creatinine, serum albumin, international normalized ratio (INR) of warfarin therapy, and center volume. The first four variables have repeatedly been shown in large analyses to correlate with outcome after LVAD.1 Although center volume may play a significant role in patient survival during the early periods of new device introduction into the mechanical circulatory support field, center experience had not previously been identified as a marker of risk in candidates for CF-LVAD support. Furthermore, the applicability of the HMRS is unknown for patients being considered for other CF devices (e.g., the centrifugal flow HeartWare Ventricular Assist Device [HVAD]).
Using INTERMACS data, we aimed to evaluate the accuracy of an amended form of the HMRS, termed the adjusted HMRS, removing the center volume criterion. We then applied the adjusted HMRS to the HVAD (HeartWare International Inc., Framingham, MA) BTT and Continued Access Protocol (CAP) trial patients to determine if it predicts risk in an alternative CF device modality.13
INTERMACS HMRS Analysis
The INTERMACS database was queried for CF-LVAD implants performed between 2008 and 2014. Baseline patient demographics, characteristics, laboratories, and preoperative clinical information were analyzed. INTERMACS does not presently categorize patients by CF-LVAD type, so patient stratification by model was not possible. Patients were excluded if the CF device followed a prior pulsatile durable LVAD.
The original HMRS formula (HMRS = (0.0274 × [age in years]) − (0.723 × [albumin g/dl]) + (0.74 × [creatinine mg/dl]) + (1.136 × [INR]) + (0.807 × [center volume])) could not be calculated in this analysis because INTERMACS does not provide data on center volume.11,14 Thus, an adjusted HMRS was calculated, omitting center volume, according to the formula: adjusted HMRS = (0.0274 × [age in years]) − (0.723 × [albumin g/dl]) + (0.74 × [creatinine mg/dl]) + (1.136 × [INR]). For patient age, only age groups (per 10 years) were provided by INTERMACS. Thus, the median age of the patient’s assigned age group was used for patient risk score calculation. Of the 11,260 CF-LVAD patients in INTERMACS, 9,733 (86%) had complete data for calculation of the adjusted HMRS.
Because the HMRS was amended, new score thresholds for the purposes of risk groups needed to be devised. Survival probabilities for 90 day survival, the original end-point in HMRS study, were calculated in the INTERMACS cohort using logistic regression.11 These probabilities were used to generate new risk groups based on adjusted HMRS score: very low risk (<5%, 90-day mortality), low risk (6–10%, 90-day mortality), medium risk (10–20%, 90-day mortality), and high risk (>20%, 90-day mortality). Kaplan-Meier survival estimates during CF-LVAD support were compared between adjusted HMRS strata.
HeartWare Ventricular Assist Device Analysis
There were 382 patients enrolled in the HVAD-BTT trial with CAP.13 Details of the trial have been previously published.13 In brief, all patients had class IV heart failure and were listed for transplant and were randomized to either the HVAD or the HeartMate II LVAD (Thoratec Corporation, Pleasanton, CA). Patients with severe right ventricular failure, renal failure (Cr >3.0 mg/dl or dialysis), or those on prolonged ventilator support were excluded.
The adjusted HMRS was calculated as previously defined in the 360 patients with available data. The risk groups identified in the previously mentioned INTERMACS analysis were used for HVAD patient stratification. Survival at 90 days and overall survival were compared between adjusted HMRS risk strata based on mortality status.
Continuous data were evaluated for normality and compared between groups using Student’s t-test (for normal data) and nonparametric Mann-Whitney test with data presented as means ± standard error of mean (SEM) or medians (25th, 75th), as appropriate. Categorical data are presented as frequencies and were compared with Fisher’s exact testing or Pearson’s testing for >2 × 2 comparisons.
In both cohorts, odd ratios (95% confidence intervals) for 90 day mortality after LVAD implant based on adjusted HMRS were calculated using logistic regression. The AUC for predicting 90 day mortality was calculated for the adjusted HMRS. Kaplan-Meier methods were used to estimate survival, and these estimates were compared based on adjusted HMRS grouping using log-rank and Breslow testing. The Breslow comparison favors the early part of curve’s hazard. Overall and pairwise comparisons (between each adjusted HMRS strata) for survival were made. Patients were censored at the time of transplant and explant for recovery. Cox modeling was used to calculate hazard ratios for survival, controlling for the HMRS and previously published risk correlates (body mass index [BMI], year of implant, sex, prior sternotomy, INTERMACS profile, need for dialysis or preoperative ventilator support, concomitant surgery).1,2 For all analyses, a p < 0.05 was considered significant.
The INTERMACS Data Access, Analysis, and Publication Committee (DAAP) approved the INTERMACS component of the project herein. Patient’s informed consent and institutional review board approval for a given patient was obtained before entry into the INTERMACS database and before enrollment into the HVAD trial.
Interagency Registry for Mechanically Assisted Circulatory Support Cohort
The baseline characteristics of the 9,733 INTERMACS patients are displayed in Table 1 with further grouping by survival status at 90 days. There were 2,207 (23%) deaths over 14 ± 0.14 months (median 10 months) of follow-up. Survival for the cohort was 91 ± 0.3% at 90 days (n = 884 deaths) and 82 ± 0.4% at 1 year. Patients who died tended to be older, implanted for DT, and were more likely to be in sicker INTERMACS profiles (p < 0.05).
Survival by Adjusted HeartMate II Risk Score
In the entire INTERMACS cohort, the mean ± standard error of the mean adjusted HMRS was 1.7 ± 0.01. Patients who died at 90 days had significantly higher (2.1 ± 0.04) adjusted HMRS than survivors (1.6 ± 0.01, p < 0.001), and significant differences were noted in all variables comprising the risk score (Table 2). The odds of 90 day mortality increased by 1.3 (1.3–1.5) for each unit increase in the adjusted HMRS (model AUC 0.64 ± 0.01, p < 0.001). Using logistic regression survival probabilities from the INTERMACS cohort, four risk groups were identified based on adjusted HMRS: very low (score <0.20), low (score 0.20–1.97), medium (score 1.98–4.48), and high risk (score >4.48) for postimplant LVAD mortality. Within INTERMACS, 469 (4.8%) patients were very low risk, 6,059 (62%) were low risk, 3,065 (31%) were medium risk, and 140 (1.4%) patients were high risk based on adjusted HMRS scoring.
Figure 1 shows Kaplan-Meier survival curves for the INTERMACS cohort stratified by adjusted HMRS categories. The 90 day survivals for the very low-, low-, medium-, and high-risk groups were 97 ± 0.8%, 93 ± 0.3%, 86 ± 0.6%, and 82 ± 3.3%, respectively. Log-rank and Breslow day revealed significant differences between all risk groups (p < 0.001 for both). On pairwise comparisons, there were significant differences between every group on every comparison (all p < 0.005), suggesting the adjusted HMRS provided risk discrimination overall and according to risk grouping.
On multivariable analysis, adjusting for INTERMACS profile, sex, implant year, need for dialysis or ventilator support, prior sternotomy, concomitant cardiac surgery, and BMI, the adjusted HMRS was predictive of mortality (hazard ratio, 1.19 [1.25–1.23] per unit increase; p < 0.001). Patients in the high-risk adjusted HMRS group had a 1.9 (1.4–2.5) higher adjusted risk of death compared with patients in the combined low- and very low-risk groups and a 1.4 [1.3–1.5] higher adjusted risk of death compared with those in the medium-risk group (p < 0.001).
HeartWare Ventricular Assist Device Trial Cohort
Baseline characteristics and demographics of the HVAD-BTT with CAP cohort (n = 360) stratified by 90 day mortality status (n = 16 deaths, 4%) are shown in Table 3. Overall, this was a lower risk patient cohort with few profile 1 patients and, per study design, no DT patients. Serum creatinine, serum albumin, and INR were similar between survivors and those who died (Table 2). The average cohort adjusted HMRS was 1.35 ± 0.44 (median, 1.29). There were 22 patients (6%) in the very low-risk category, 266 (74%) in the low-risk, and 72 (20%) in the medium-risk category. There were no patients in the high-risk category. Patients who died (n = 16) had a significantly higher adjusted HMRS (1.76 ± 0.22) than survivors (1.33 ± 0.05) (p = 0.047). The odds of 90 day mortality per unit increase in the adjusted HMRS was 1.7 (1.02–2.95).
Kaplan-Meier survival curves for the HVAD-BTT cohort stratified by adjusted HMRS categories are shown in Figure 2. The 90 day survival for very low-, low-, and medium-risk patients was 100%, 97 ± 1.1%, and 90 ± 3.6%, respectively (p = 0.007). On pairwise comparison, patients in the very low-risk group had significantly better survival than medium-risk patients (p = 0.019), and trends were noted for comparison between the very low- versus low-risk group (p = 0.12). Patients in the low-risk group had better survival than medium-adjusted HMRS risk patients (p = 0.048).
In this study, an adjusted version of the previously devised HMRS was examined in the INTERMACS cohort. The INTERMACS cohort represents a heterogeneous group of patients supported on different commercially approved CF-LVADs with both centrifugal and axial CF configurations. The calculation of the score itself was the same as in the derived model, except for omission of the center volume criterion. At the new score thresholds, the adjusted HMRS successfully stratified patients into risk groups within INTERMACS. In the centrifugal flow HVAD trial cohort, the adjusted HMRS also appeared to stratify LVAD candidate risk.
Although several risk models have been devised, the rapid change in technology and patient selection may limit ongoing applicability of risk scores to all VAD candidates. For example, the destination therapy risk score (DTRS) was derived using data from patients in the post- Randomized Evaluation of Mechanical Assistance for the Treatment of Congestive Heart Failure (REMATCH) era supported with the pulsatile HeartMate XVE.7,15 This score was found to poorly discriminate risk when applied to patients supported with the HeartMate II.16,17 As mentioned previously, the HMRS was derived and validated in HeartMate II BTT and DT patients, who met careful trial inclusion and exclusion criteria.11 Recently, the HMRS was applied to a single-center cohort and was found to appropriately risk stratify patients, with high-risk patients have a 5.7 times high risk of death than the low-risk group (overall model AUC 0.7).18 However, another small (n = 205) single-center study by Thomas et al.12 found that the HMRS did not appropriately stratify risk (AUC 0.56). The authors hypothesized that the HMRS may not have been predictive of risk in their patient cohort because it was a relatively low-risk group with most patients implanted for the BTT indication.12 The cohort from which the original HMRS was derived and validated contained a relatively equal number of BTT and DT patients, and the overall 1 year survival was slightly lower (72–74%) in the HMRS cohort compared with 83% in the study by Thomas et al.1,11,12
In the present analysis of 9,733 patients with an average 1 year survival of 83%, the adjusted HMRS appears to be applicable to a “real-world” cohort of patients and to patients supported with any currently approved CF-LVAD configuration. Importantly, however, the impact of center volume on outcome was omitted in this analysis. In the original HMRS derivation, 90 day mortality in patients implanted at centers enrolling <15 trial patients during the approximately 2 year studies was twofold higher. Lietz et al.19 performed a retrospective analysis of the Thoratec HeartMate Registry (n = 377) and found inferior survival in patients implanted in low versus high volume centers (48% vs. 67%, 1 year survival). However, once preoperative DTRS was taken into account, center volume was not an independent predictor of 1 year survival.19 Outside the HMRS analysis, there are no data examining patient outcomes based on center VAD volumes. In addition to data not being available in INTERMACS, we felt the criterion for center experience in the original HMRS warranted reconsideration. It is reasonable to expect a learning curve exists for a new device, and the criterion established “then” may not apply to “now.” Beyond center experience, individual surgeon experience, cardiology and extended provider support, and device type may play into center-specific outcomes that impact VAD patient survival. Certainly, more studies are needed to more carefully scrutinize this topic.
Risk scores may have utility in the process of informed consent and shared decision making. When questioned about end-of-life treatments for their advanced heart failure, some patients favor an approach that prolongs survival time, whereas others prefer quality of life even if they do not survive as long.20 Unfortunately, current evidence suggests that even after a formalized informed consent process, patients still poorly comprehended complications from LVAD implant and long-term survival rates.21 The use of a risk score may be an additional tool in educating the patient, their family, and other healthcare providers about LVAD survival expectations.
A few limitations of this study are worth mentioning. The applicability of the adjusted HMRS to all models of CF-LVAD support remains understudied. Although center volume was omitted from this analysis because it was not available in INTERMACS, this variable was a significant correlate of risk in the original HMRS. Further investigation of the contribution of center risk to VAD candidate outcome should be completed in separate study. Finally, no risk model is perfect, and the adjusted HMRS has limitations. The AUC of the adjusted HMRS suggests only modest discrimination (AUC 0.639). This AUC is similar to that of the validation cohort in the original (unadjusted) published model.11 Models with excellent discrimination have AUCs >0.70. This suggests that further work is still needed to improve on patient risk prediction before mechanical circulatory support implant. A constant obstacle to accurate operative risk prediction, however, will be unanticipated device complications (strokes, device infection, device malfunction), which may account for a significant percentage of LVAD operative deaths.
The adjusted HMRS provided risk discrimination in a general LVAD population supported on commercially approved CF-LVADs for both BTT and DT indications. Although previously tested in the HeartMate II population, the amended version of the risk score also appears to modestly discriminate risk in the HVAD-BTT cohort. Although the impact of center experience on VAD patient outcomes was not evaluated in this analysis, the topic warrants future consideration.
1. Kirklin JK, Naftel DC, Pagani FD, et al.: Sixth INTERMACS annual report: A 10,000-patient database. J Heart Lung Transplant 2014.33: 555564.
2. Boyle AJ, Ascheim DD, Russo MJ, et al.: Clinical outcomes for continuous-flow left ventricular assist device
patients stratified by pre-operative INTERMACS classification. J Heart Lung Transplant 2011.30: 402407.
3. Long EF, Swain GW, Mangi AA: Comparative survival and cost-effectiveness of advanced therapies for end-stage heart failure. Circ Heart Fail 2014.7: 470478.
4. Aaronson KD, Schwartz JS, Chen TM, Wong KL, Goin JE, Mancini DM: Development and prospective validation of a clinical index to predict survival in ambulatory patients referred for cardiac transplant evaluation. Circulation 1997.95: 26602667.
5. Fitzpatrick JR 3rd, Frederick JR, Hsu VM, et al.: Risk
score derived from pre-operative data analysis predicts the need for biventricular mechanical circulatory support. J Heart Lung Transplant 2008.27: 12861292.
6. Levy WC, Mozaffarian D, Linker DT, et al.: The Seattle Heart Failure Model: Prediction
of survival in heart failure. Circulation 2006.113: 14241433.
7. 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.
8. 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.
9. 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.
10. Rao V, Oz MC, Flannery MA, Catanese KA, Argenziano M, Naka Y: Revised screening scale to predict survival after insertion of a left ventricular assist device
. J Thorac Cardiovasc Surg 2003.125: 855862.
11. 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.
12. Thomas SS, Nahumi N, Han J, et al.: Pre-operative mortality risk
assessment in patients with continuous-flow left ventricular assist devices: Application of the HeartMate II risk
score. J Heart Lung Transplant 2014.33: 675681.
13. Aaronson KD, Slaughter MS, Miller LW, et al.; HeartWare Ventricular Assist Device (HVAD) Bridge to Transplant ADVANCE Trial Investigators: Use of an intrapericardial, continuous-flow, centrifugal pump in patients awaiting heart transplantation. Circulation 2012.125: 31913200.
14. Interagency Registry for Mechanically Assisted Circulatory Support. www.uab.edu/intermacs
. Retrieved September 7, 2015.
15. 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.
16. Schaffer JM, Allen JG, Weiss ES, et al.: Evaluation of risk
indices in continuous-flow left ventricular assist device
patients. Ann Thorac Surg 2009.88: 18891896.
17. Teuteberg JJ, Ewald GA, Adamson RM, et al.: Risk
assessment for continuous flow left ventricular assist devices: does the destination therapy risk
score work? An analysis of over 1,000 patients. J Am Coll Cardiol 2012.60: 4451.
18. 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.
19. Lietz K, Long JW, Kfoury AG, et al.: Impact of center volume on outcomes of left ventricular assist device
implantation as destination therapy: Analysis of the Thoratec HeartMate Registry, 1998 to 2005. Circ Heart Fail 2009.2: 310.
20. MacIver J, Rao V, Delgado DH, et al.: Choices: A study of preferences for end-of-life treatments in patients with advanced heart failure. J Heart Lung Transplant 2008.27: 10021007.
21. Edlund JE, Edlund AE, Carey MG: Patient understanding of potential risk
and benefit with informed consent in a left ventricular assist device
population: A pilot study. J Cardiovasc Nurs 2014.30: 435439.