Frequency of Poor Outcome (Death or Poor Quality of Life) After Left Ventricular Assist Device for Destination Therapy: Results from the INTERMACS Registry

Abstract

Background

A left ventricular assist device (LVAD) improves survival and quality of life for many, but not all, patients with end-stage heart failure who are ineligible for transplantation. We sought to evaluate the frequency of poor outcomes using a novel composite measure that integrates quality of life with mortality.

Methods and Results

Within the INTERMACS national registry, poor outcome was defined as death or an average Kansas City Cardiomyopathy Questionnaire<45 over the year following LVAD (persistently limiting heart failure symptoms and poor quality of life). Among 1638 LVAD patients, 29.7% had a poor outcome, with death in 22.4% and persistently poor quality of life in 7.3%. Patients who had a poor outcome were more likely to have higher body mass indices (29.3 vs. 28.2 kg/m², p=0.007), lower hemoglobin levels (11.1 vs. 11.4 g/dL, p=0.005), prior cardiac surgery (47.8% vs. 39.8%, p=0.004), history of cancer (13.8% vs. 9.7%, p=0.025), severe diabetes (15.6% vs. 11.5%, p=0.038), and poorer quality of life pre-implant (KCCQ scores: 29.8 vs. 35.3, p<0.001).

Conclusions

Nearly one-third of patients die or have a persistently poor quality of life over the year after LVAD. We identified a number of factors associated with a poor outcome, which may inform discussions prior to LVAD implantation to enable more realistic expectations of recovery.

Patients with end-stage heart failure are highly symptomatic and are expected to have a short life expectancy without intervention.¹ For patients who are not candidates for heart transplantation, a left ventricular assist device (LVAD) may be an ideal treatment option as destination therapy, meaning that the patient is expected to live with the device until death (i.e., not a bridge to heart transplant). While an LVAD markedly improves survival and quality of life in the majority of such patients, the risk of mortality remains high even after implantation, despite continued improvements in device technology and surgical experience.² Furthermore, a number of patients who undergo LVAD survive but continue to have very poor quality of life due to a myriad of potential reasons, including refractory or recurrent heart failure symptoms, failure to recover from major cardiac surgery, or device-related complications such as bleeding, infections, stroke, and pump malfunction.³ Finally, the reasons that these patients are ineligible for heart transplantation—advanced age, non-cardiac comorbidity, and frailty—may persist after LVAD implantation and can limit the ability of the patients to successfully regain a meaningful quality of life.

Prolonging life, while important, does not encompass all of the major treatment goals that matter to patients with end-stage heart failure. Quality of life is substantially diminished among the vast majority of patients eligible for LVADs.⁴ Particularly among elderly patients ineligible for transplantation, a reasonable quality of life—where the patient is no longer profoundly limited by heart failure symptoms—may be an even more important treatment goal than survival.^{5, 6} Consequently, integrating quality of life outcomes into the definition of a poor outcome is particularly relevant in these challenging and complex patients with end-stage heart failure. The goal of this study was to define the frequency of poor outcome after LVAD and to examine patient factors associated with this poor outcome after LVAD. Our hope was that these data could provide patients and their families with more realistic expectations of recovery and could facilitate shared medical decision-making.

METHODS

Study Population and Protocol

Our study cohort was derived from the Interagency Registry for Mechanically Assisted Circulatory Support (INTERMACS)—a multicenter, observational registry of patients receiving mechanical circulatory support in the US.⁷ INTERMACS is administered through a contract from the National Heart, Lung and Blood Institute to the University of Alabama at Birmingham and is focused on quality improvement and scientific research. All consecutive, non-incarcerated, adult patients who receive Food and Drug Administration-approved devices are screened for enrollment in the registry. At the time of device implant, the surgeon indicates the indication for LVAD, with multiple options including destination therapy (i.e., not eligible for heart transplant) and bridge to transplant, with further qualification that includes transplant unlikely/moderate/likely/listed. For this analysis, we excluded patients who received total artificial hearts, right and bi-ventricular assist devices, and LVADs for indication other than destination therapy or bridge to transplant–transplant unlikely. Data were collected through medical chart abstractions and patient interviews. Comorbidities were abstracted from the chart by trained data personnel. All participating sites obtained Institutional Review Board protocol approval. As INTERMACS is considered a quality improvement registry and all patient identifiers were removed, the Institutional Review Board at the University of Alabama at Birmingham granted a waiver of individual patient informed consent for these analyses.

Quality of Life Data

Although INTERMACS began collecting data on patients with mechanical circulatory support in 2005, quality of life measures were not routinely collected until May 2012 (version 3.0). Assessment of quality of life is attempted on all patients at baseline and at 3 and 6 months after LVAD implantation and every 6 months thereafter for the life of the device (i.e., INTERMACS stops quality of life data collection at the time of transplant or LVAD explant).⁸ Disease-specific quality of life was assessed with the Kansas City Cardiomyopathy Questionnaire (KCCQ),⁹ a 23-item self-administered questionnaire that assesses specific health domains pertaining to heart failure.¹⁰ A 12-item version of the KCCQ was developed in 2013,¹¹ and INTERMACS began collecting this shortened version in 2014 (version 4.0).⁸ Both versions yield an overall summary scale (KCCQ-OS), which was the primary quality of life outcome for our study and has a concordance of >0.98 between the two versions. Values for the KCCQ-OS range from 0 to 100, with higher scores indicating fewer symptoms, better function, and higher quality of life. Previous studies have suggested that KCCQ-OS scores correlate roughly with New York Heart Association Functional Class as follows: Class I ≈ KCCQ-OS 75–100; Class II ≈ 60–74; Class III ≈ 45–59; and Class IV ≈ 0–44.¹² The KCCQ has undergone extensive reliability and validity testing in various heart failure populations^{9, 13, 14} and has been used in studies examining the changes in quality of life with LVADs.^15–17

Definition of Poor Outcome

We have previously examined several potential definitions for poor outcome that combine mortality and quality of life in a population of patients with severe symptomatic aortic stenosis undergoing transcatheter aortic valve replacement.¹⁸ Similar to patients considering LVAD, patients with severe symptomatic aortic stenosis are expected to have limited life expectancy and poor quality of life without intervention.^{1, 19} Moreover, both LVAD and transcatheter aortic valve replacement have substantial effects on both survival and quality of life in the majority of patients^{15, 16, 19} and yet both procedures have high residual morbidity and mortality despite intervention.^{1, 19} For the purposes of this study, we used outcomes at 1 year after LVAD as the basis for our primary endpoint, as we reasoned that survival for at least 1 year with a reasonable quality of life would be the minimum acceptable outcome for patients considering LVAD.

We used heart failure-specific health status to define poor quality of life after LVAD, as heart failure symptoms are a primary indication for LVAD and are therefore expected to be corrected, to a variable extent, in highly symptomatic patients. Although overall health status would also be expected to improve among highly symptomatic patients undergoing LVADs, generic health status measures are less sensitive to change and do not specifically assess symptoms and functional status along the disease pathway of heart failure. As such, they would not be as sensitive or specific as a disease-specific measure for determining whether a patient has had an acceptable outcome after LVAD. Therefore, poor outcome was defined as death or an average KCCQ-OS <45 throughout the year after implantation.^{18, 20} We integrated quality of life over the follow-up time to avoid inadvertently labeling patients as having poor outcome due to temporary dips in quality of life at a single point in time. These outcomes were considered hierarchically—with death as the primary and, if alive, then quality of life was examined. As such, there was no overlap between the outcomes (i.e., quality of life was not considered as an outcome in patients prior to death, which may have increased the proportion of patients with poor quality of life outcomes after LVAD implantation).

Statistical Analysis

To reduce the effect of selection bias due to missing follow-up KCCQ data, we constructed a multivariable logistic regression model among patients who survived 1 year to determine the probability of having missing follow-up KCCQ data. We then weighted each of the surviving patients in the analytic cohort by the inverse probability of the likelihood of having missing data.²¹ In this way, our analyses better reflected the overall LVAD population by giving more weight to the model for patients who were most like those who survived but did not complete follow-up KCCQ questionnaires. All analyses used these weighted estimates.

For the quality of life component of the outcome, we constructed a piecewise linear mixed-effects model to predict KCCQ scores at each follow-up assessment through survival or 1 year following LVAD. The model included baseline patient characteristics, fixed and random effects for month, whether the patient died, and time to death. For each patient, very poor quality of life was then defined as average area under the curve of <45, suggestive of persistent NYHA Class IV symptoms, on average, during the year after LVAD implantation, based on the patient’s predicted KCCQ scores. This approach accounts for transient declines in quality of life without overweighting these events should patients enjoy better quality of life before or after such episodes.

After defining the outcomes, the demographic and clinical characteristics were compared between patients who had a poor outcome and patients who had an acceptable outcome after LVAD implantation using t-tests for continuous variables and chi-square tests for categorical variables. All statistical analyses were performed with SAS, version 9.4 (SAS Institute, Inc., Cary, NC) and R version 3.2.0.²²

RESULTS

Patient Population

From May 2012 through September 2013, 3922 adult patients received durable LVADs and were enrolled in the INTERMACS registry. We excluded patients who received bi-ventricular devices (n=31) and patients with device strategies of bridge to transplant (n=2208) or bridge to recovery, rescue therapy or other indications (n=45; Figure). As such, our analytic cohort included 1638 patients who received LVADs for destination therapy. The most common age group was 60–69 years of age, 19% were females, and 42% were INTERMACS profile 1–2 at implant. Follow-up KCCQ data were missing on 16% of surviving patients. There were few clinical differences between those with and without follow-up KCCQ data (Supplemental Table 1). Notably, patients with missing follow-up data were more likely to have missing baseline KCCQ data (missing vs. not missing follow-up data: 53% vs. 38% missing baseline data, p<0.001), mostly for administrative reasons (e.g., patient was not deemed too sick to complete the KCCQ; Supplemental Table 1).

Figure. Patient flow.

“BTT/BTT possible” indicates patients with an implant strategy of “bridge to transplant - listed”, “bridge to transplant - likely”, and “bridge to transplant - moderate”. Patients with the implant strategy of “bridge to transplant - unlikely” were included in the analytic cohort. VAD=ventricular assist device; BTT=bridge-to-transplant; BTR=bridge-to-recovery; BiVAD=biventricular ventricular assist device; KCCQ=Kansas City Cardiomyopathy Questionnaire

Frequency of and Factors Associated with Poor Outcome

Among the 1638 patients with LVADs, 29.7% had poor outcome at 1 year following device implantation. This was due to death in 22.4% of patients and persistently poor quality of life in 7.3% of surviving patients. Among patients who died in the first year after LVAD, 70.6% died within the first 6 months, 23.4% died between 6–12 months and had a reasonable quality of life prior to death, and 6.0% died between 6–12 months with poor quality of life.

Compared with patients who survived 1 year with an acceptable quality of life after LVAD, those with poor outcomes were more likely to have higher body mass indices (29.3 vs. 28.2 kg/m², p=0.007), lower hemoglobin levels (11.1 vs. 11.4 g/dL, p=0.005), prior cardiac surgery (47.8% vs. 39.8%, p=0.004), history of cancer (13.8% vs. 9.7%, p=0.025), and severe diabetes (15.6% vs. 11.5%, p=0.038; Table). There was also a non-significant trend toward more frequent heart failure hospitalizations in the past year among patients with poor outcomes. Among patients who completed the KCCQ, pre-treatment scores were lower among those with a poor outcome (poor vs. acceptable outcome: 29.8 vs. 35.3, p<0.001). A comparison of the 5 groups (died <6 months after LVAD, died 6–12 months with poor quality of life, died 6–12 months with reasonable quality of life, survived 12 months with poor quality of life, survived 12 months with reasonable quality of life) is shown in Supplemental Table 2.

Table.

Baseline characteristics of patients according to 1-year outcome after LVAD

	Poor Outcome	Acceptable Outcome	P-Value
Age group (years)			0.041
<40	3.2%	6.2%
40–49	7.7%	8.9%
50–59	17.8%	16.2%
60–69	32.8%	36.0%
70+	38.6%	32.8%
Female	18.6%	18.8%	0.939
High school education	29.3 ± 8.0	28.2 ± 6.7	0.007
Body mass index (kg/m2)	14.0%	13.3%	0.738
Pulmonary disease	25.1%	23.2%	0.445
Atrial arrhythmia	15.6%	11.5%	0.038
Severe diabetes1	5.4%	3.6%	0.137
History of major stroke	10.5%	8.1%	0.152
Peripheral vascular disease	13.8%	9.7%	0.025
History of solid organ cancer	13.5%	12.4%	0.580
History of illicit drug use or alcohol abuse	49.9%	48.1%	0.608
Number of cardiac hospitalizations in past year			0.087
0–1	18.6%	24.5%
2–3	40.9%	36.2%
4 or more	18.8%	17.3%
Unknown	21.7%	21.9%
Previous cardiac surgery	47.8%	39.8%	0.004
Mean arterial pressure (mmHg)	77.4 ± 10.8	78.1 ± 11.6	0.290
Pulmonary systolic pressure (mmHg)	50.4 ± 13.8	50.4 ± 14.7	0.963
Heart rate (bpm)	86.0 ± 16.4	85.6 ± 16.6	0.694
Moderate/severe mitral regurgitation	55.9%	59.8%	0.191
Moderate/severe tricuspid regurgitation	44.2%	45.2%	0.739
Creatinine (mg/dL)	1.5 ± 0.6	1.5 ± 0.7	0.532
Albumin (mg/dL)	3.3 ± 0.7	3.4 ± 0.6	0.467
Hemoglobin (mg/dL)	11.1 ± 1.9	11.4 ± 2.1	0.005
INTERMACS patient profile			0.137
1 Critical Cardiogenic Shock	12.1%	8.5%
2 Progressive Decline	34.5%	32.2%
3 Stable but Inotrope Dependent	28.2%	34.3%
4 Resting Symptoms	19.5%	18.3%
5 Exertion Intolerant	4.4%	4.5%
6 Exertion Limited	0.6%	1.1%
7 Advanced NYHA Class 3	0.7%	1.0%
Pre-Implant KCCQ Status			0.557
Completed	61.1%	62.4%
Missing: too sick	9.9%	10.8%
Missing: admin or patient reason	29.1%	26.8%
KCCQ-OS Pre-Implant	29.8 ± 19.3	35.3 ± 21.2	<0.001

NYHA, New York Heart Association; KCCQ-OS, Kansas City Cardiomyopathy Questionnaire-overall summary score

¹As determined by site personnel

DISCUSSION

In a large, national registry of patients who received LVADs, we found that nearly one-third of patients have a poor outcome over the year after device implantation. This was primarily driven by death, although 1 in 13 surviving patients had persistently poor quality of life throughout the year after treatment. We identified a few factors that were more common in patients who had a poor outcome, including prior cardiac surgery, anemia, obesity, and poorer heart failure symptoms and functional status at baseline. These data may be useful to calibrate patients’ expectations of recovery after LVAD for destination therapy.

Importantly, this was a study examining patients who are not candidates for heart transplantation, and the factors that make these patients ineligible for transplantation are most commonly advanced age or comorbidities. Cardiac and non-cardiac comorbidities, other than obesity and a history of cancer, were not as strongly associated with poor outcome, highlighting that the reasons that a patient may be ineligible for heart transplantation do not necessarily place the patient at high risk for poor outcome after LVAD. In addition, indices of right heart failure (e.g., tricuspid regurgitation, pulmonary systolic pressure), which are often considered high-risk features for LVAD implantation, were not associated with a poor outcome after LVAD. However, it is important to note that all patients in our study received an LVAD, and thus the highest-risk patients, based on these factors, may have been appropriately treated with alternative strategies (e.g., biventricular VAD, medical therapy) and were not included.

Prior Studies

Importantly, prior studies have focused on factors associated with mortality in isolation. Lower hematocrit was a significant predictor in the destination therapy risk score.²³ However, this was the only factor that overlapped with our data. Worse quality of life prior to LVAD implantation, as assessed with the KCCQ, has been shown not to be associated with mortality in two studies, and is likely associated with poor outcome primarily due to its strong association with subsequent quality of life.^{4, 24} By broadening our definition of poor outcome to include both poor quality of life and death, we believe we have better encompassed the treatment goals of patients considering LVAD for destination therapy and, in the process, identified some novel factors associated with poor outcomes after LVAD.

Challenges in Defining Poor Outcome

We believe that our definition of poor outcome reflects a failure of the device to help some patients achieve its anticipated benefits. One could argue for different KCCQ score thresholds for poor quality of life, different follow-up timeframes, or inclusion of additional endpoint components (e.g., cognitive decline, stroke). Furthermore, there is almost certainly variability in how individual patients weigh the importance of death and quality of life in its regard to being a poor outcome after LVAD. Nevertheless, we believe that this is an important initial step in defining how best to communicate the benefits of LVAD therapy to individual patients considering this treatment option. Importantly, by embracing a composite definition of poor outcome—however defined—providers can more consistently and accurately engage patients in shared medical decision-making around LVAD therapy. This is important as it is well documented that the majority of educational materials explain potential LVAD treatment in ways that may not be correct for individual patients.²⁵ As a result, a recent study found that most patients expect greater improvements in their quality of life than they often experience.²⁶ There is a pressing need to improve communication with patients prior to LVAD and to provide more realistic estimates of recovery. The use of an outcome that incorporates both survival and quality of life is likely to be an efficient means for communicating to patients the information they most value in making a decision, particularly because the KCCQ integrates numerous health status characteristics that patients care about, including heart failure symptoms, functional limitations, social limitations, and quality of life. This integration of quality of life outcomes may become even more relevant as LVADs begin to be tested and used in lower risk/less sick patients, who may have similar absolute survival with medical therapy.

Challenges in Predicting a Poor Outcome

Our original goal was to build a model of poor outcome that could be used prior to treatment to help guide patient selection and calibrate patients’ and providers’ expectations of recovery. However, we found this very challenging to accomplish. First, one of the factors that was most strongly associated with poor outcomes was pre-LVAD quality of life. Unfortunately, due to the acute illness of patients prior to LVAD in combination of administrative concerns of collecting pre-LVAD data for the registry, baseline quality of life has been challenging to collect in the sickest patients (e.g., INTERMACS Profile 1). While we could use techniques of imputation and various assumptions and sensitivity analyses, these would be inherently limited in truly understanding the baseline quality of life of patients with missing data. Second, the frequent occurrence of post-LVAD complications, such as stroke and bleeding, impact outcomes of both survival and quality of life and are, unfortunately, relatively unpredictable based on patient characteristics prior to LVAD. As technology and treatments improve, it is expected that these complications will decrease, increasing our ability to predict poor outcome based solely on baseline characteristics. Hopefully in the future, we will be able to construct valid models to estimate a patient’s risk of poor outcome based on his or her individual characteristics and use this estimate to support decision-making prior to this life-changing operation, particularly in patients with higher INTERMACS profiles where there is more discretion in using LVADs.

Limitations

Our findings should be interpreted in the context of several potential limitations. First, there are many aspects of quality of life that may be affected by the LVAD that are not captured by the KCCQ (e.g. challenges with device management, stroke, frequent bleeding). While a KCCQ-OS score <45 reflects a very poor quality of life for patients with heart failure, it is likely that there are patients who report KCCQ-OS scores ≥45 but who have a very poor quality of life for reasons not related to their heart failure. As such, it is likely that we have underestimated the proportion of patients who have poor quality of life after LVAD. Conversely, while most patients strongly value quality of life,⁴ some patients may consider survival with a poor quality of life an acceptable outcome. In contrast, others may believe that having persistent heart failure symptoms that are moderately limiting (i.e., KCCQ-OS 45–60) represents a poor outcome. For these reasons, we may be over- or under-estimating the percentage of patients with poor outcome due to poor quality of life. Second, follow-up KCCQ scores were missing on 15% of surviving patients. While there were few meaningful differences between those with and without follow-up KCCQ data, we had to account for these differences by using inverse propensity weighting methods for all of our analyses.

Conclusion

Using a novel definition that integrates both survival and quality of life, we found that nearly one-third of patients have poor outcome over the year following LVAD. Patients who had a poor outcome were more likely to have had prior cardiac surgery, anemia, obesity, and poorer functional status prior to LVAD implantation. Future efforts to build on these initial insights to develop risk models to support shared decision-making, particularly in patients with higher INTERMACS profiles, should be considered, although these efforts may be challenging due to a number of factors, including unpredictable post-LVAD complications. Further research is needed to understand how best to communicate these risks to patients, so that the decision to proceed with LVAD is concordant with patients’ goals and values.

CLINICAL PERSPECTIVE.

A left ventricular assist device (LVAD) improves survival and quality of life for many, but not all, patients with end-stage heart failure who are ineligible for transplantation. We sought to define the frequency and predictors of a poor outcome after destination LVAD, using a novel definition that integrates mortality and quality of life. By broadening our definition of poor outcome to include both poor quality of life and death, we believe we have better encompassed the treatment goals of patients considering LVAD for destination therapy. Within a national LVAD registry, we found that 29.7% of patients who received a destination LVAD had a poor outcome over the year following implant. This was due to death in 22.4% and persistently poor quality of life in 7.3%. Patients who had a poor outcome were more likely to have had prior cardiac surgery, anemia, obesity, and poorer functional status prior to LVAD implantation. Although there are clearly challenges with identifying pre-operatively which patients will experience these poor outcomes after LVAD implantation, future efforts should focus on building upon these initial insights through the development of risk models to support shared decision-making, particularly in patients with higher INTERMACS profiles.