Cardiac Cachexia in Left Ventricular Assist Device Recipients and the Implications of Weight Gain Early After Implantation

Abstract

Background

Severe cardiac cachexia or malnutrition are commonly considered relative contraindications to left ventricular assist device (LVAD) implantation, but post‐LVAD prognosis for patients with cachexia is uncertain.

Methods and Results

Intermacs (Interagency Registry for Mechanically Assisted Circulatory Support) 2006 to 2017 was queried for the preimplantation variable cachexia/malnutrition. Cox proportional hazards modeling examined the relationship between cachexia and LVAD outcomes. Of 20 332 primary LVAD recipients with available data, 516 (2.54%) were reported to have baseline cachexia and had higher risk baseline characteristics. Cachexia was associated with higher mortality during LVAD support (unadjusted hazard ratio [HR], 1.36 [95% CI, 1.18–1.56]; P<0.0001), persisting after adjustment for baseline characteristics (adjusted HR, 1.23 [95% CI, 1.0–1.42]; P=0.005). Mean weight change at 12 months was +3.9±9.4 kg. Across the cohort, weight gain ≥5% during the first 3 months of LVAD support was associated with lower mortality (unadjusted HR, 0.90 [95% CI, 0.84–0.98]; P=0.012; adjusted HR, 0.89 [95% CI, 0.82–0.97]; P=0.006).

Conclusions

The proportion of LVAD recipients recognized to have cachexia preimplantation was low at 2.5%. Recognized cachexia was independently associated with higher mortality during LVAD support. Early weight gain ≥5% was independently associated with lower mortality during subsequent LVAD support.

Nonstandard Abbreviations and Acronyms

HFrEF

heart failure with reduced ejection fraction

Intermacs

Interagency Registry for Mechanically Assisted Circulatory Support

RHF

right heart failure

Clinical Perspective.

What Is New?

• The preimplantation variable of cachexia/malnutrition in the Intermacs (Interagency Registry for Mechanically Assisted Circulatory Support) data set is independently associated with higher mortality during left ventricular assist device (LVAD) support but was identified only in 2.5% of LVAD recipients.

• A sensitivity analysis incorporating the presence of body mass index <18.5 kg/m² pre‐LVAD implantation increased the prevalence of cachexia to 5.0%.

• A ≥5% weight gain in the first 3 months of LVAD support is independently associated with lower subsequent mortality.

What Are the Clinical Implications?

• Cardiac cachexia appears to be underrecognized in the advanced heart failure population, potentially due to the lack of validated and easily applied diagnostic criteria.

• This analysis supports current clinical practice of considering severe cachexia to be at least a relative contraindication to LVAD implantation, although 91.9% of recipients with cachexia left the hospital alive, and the median survival for this group was 32.9 months.

• Early weight gain in the range of 5% to 10% is associated with favorable outcomes during LVAD support, highlighting the importance of metabolic recovery from cardiac cachexia and the need for clinicians to focus on nutritional strategies that facilitate weight regain after LVAD implantation.

Cardiac cachexia is a complex wasting syndrome characterized by unintentional weight loss accompanied by anorexia, inflammation, and abnormal biochemistry. ¹ The prevalence of cachexia in patients with ambulatory heart failure with reduced ejection fraction (HFrEF) is in the range of at least 10% to 20%. ² , ³ Cachexia has been associated with a 2‐fold higher mortality rate for patients with HFrEF. ⁴ Malnutrition and micronutrient insufficiencies may occur with or without cachexia and are also associated with increased risk of mortality. ⁵ , ⁶ Severe cachexia or malnutrition is often considered a contraindication to life‐sustaining advanced HFrEF interventions such as left ventricular assist device (LVAD) implantation or heart transplantation, but data on the mortality risk associated with baseline cardiac cachexia among patients who undergo LVAD implantation are lacking.

Despite the lack of contemporary data on the prognostic implications of cachexia for patients with advanced HFrEF receiving an LVAD, it has been established that the associated metabolic features of low albumin, low cholesterol, increased insulin resistance, and systemic inflammation pre‐LVAD implantation are associated with lower subsequent survival during LVAD support. ⁷ , ⁸ , ⁹ , ¹⁰ Decision making on LVAD eligibility is largely based around the assessment of comorbidities and their anticipated impact on short‐ and medium‐term outcomes after LVAD implantation, as well as the likelihood for improvement. Therefore, quantification of the relationship between baseline cachexia and outcomes during LVAD support would be highly valuable for clinical decision making. It has also been recognized that LVAD recipients tend to regain weight, specifically skeletal muscle mass, early during LVAD support. ¹¹ , ¹² Early improvements in metabolic markers of nutritional status within the first 3 months of support are also associated with greater survival during subsequent LVAD support. ¹³ Whether a similar association exists between early weight gain and subsequent survival is currently unknown.

Thus, the purpose of this national cohort study was to define (1) the proportion of LVAD recipients recognized to have cachexia/malnutrition and their associated clinical profile at LVAD implantation, (2) the relationship between cachexia and major clinical outcomes during LVAD support, and (3) the trajectory of weight changes early after LVAD implantation and to determine the association with subsequent survival.

Methods

The data set was accessed through the publicly available National Heart, Lung, and Blood Institute Biologic Specimen and Data Repository Information Coordinating Center service. The full analytic methods will be made available to other researchers upon reasonable request, and the details of coding for events and censoring variables are presented in Tables S1–S3.

Patient Cohort

This retrospective cohort study used data routinely collected by Intermacs (Interagency Registry for Mechanically Assisted Circulatory Support), a North American registry of data for adults who received a Food and Drug Administration–approved mechanical circulatory support device for the treatment of advanced HFrEF. Eligible subjects were adults aged ≥19 years who received a primary surgical LVAD at an active Intermacs center between March 1, 2006 and December 31, 2017. Device exchange surgeries and biventricular ventricular assist device implantations or isolated right ventricular assist device implantations did not qualify for cohort inclusion. The Tufts Health Sciences Institutional Review Board determined that the project qualified for exempt status, and that there was no requirement for informed consent due to the deidentified nature of the data.

Baseline Clinical Profile of Cachexia

The variable of primary interest from the Intermacs data set was the field cachexia/malnutrition. This binary variable is completed by the team member entering baseline clinical data at the implanting center. There are no diagnostic criteria or clinical guidance notes provided for this variable. If either the contributing comorbidity or contributing comorbidity‐2 field was positive for cachexia/malnutrition, the subject was designated as having baseline cachexia. Subjects were designated as having preimplantation cachexia present, cachexia absent, or cachexia status missing; for the analysis of outcomes related to cachexia status, only the first 2 subgroups were included, whereas patients with cachexia missing were eligible for inclusion in the separate analysis of weight change after LVAD implantation.

Statistical Analysis

We identified a priori the major demographics, baseline comorbidities, laboratory values, and hemodynamics that may be associated with a diagnosis of cachexia/malnutrition. Baseline data were extracted from the Intermacs patients and device files. Implausible extreme outlier values of baseline continuous variables were defined and exclusions applied accordingly. Clinical diagnoses including severe diabetes, frailty, and liver dysfunction are also defined without uniform Intermacs diagnostic criteria. These baseline characteristics were tabulated by cachexia status (present versus absent). When there was missingness for categorical values, the percentage for the numerator was calculated from the nonmissing denominator. Between‐group comparisons were made with pooled t tests for normally distributed continuous variables, Wilcoxon rank sum tests for nonnormally distributed continuous variables, and χ² tests for categorical data.

Primary and Secondary Outcomes

The primary outcome was mortality during LVAD support, obtained from the Intermacs events file. Secondary outcomes were in‐hospital mortality, postimplantation length of stay, discharge destination, Intermacs‐defined infectious events (composite of device and nondevice related), neurological events (stroke, intracranial hemorrhage, transient ischemic attack), gastrointestinal bleeding, pump thrombosis/malfunction, and right heart failure (RHF). The definitions of these clinical end points and the variables used as censoring indicators for each model are detailed in Tables S1–S3. For RHF, we used the Intermacs variable EventX, which is a composite of death due to RHF, subsequent right ventricular assist device implantation, rehospitalization for RHF, or documented hemodynamic RHF (central venous pressure >18 mm Hg, capillary wedge pressure ≤18 mm Hg, and cardiac index <2.3 L/min per m²). This variable was valid only until June 2014, and so we restricted our RHF analyses to this time frame. Because only the year and quarter of implant were available, to calculate the censoring time for patients who had not experienced RHF by June 2014, we used a factor of 3 months for each quarter to approximate the time to censoring.

We constructed Kaplan‐Meier plots and Cox proportional hazards models for the primary and secondary end points. The primary mortality models were censored on transplantation, LVAD explantation, or end of follow‐up period, whereas the secondary models were additionally censored on death as detailed in Tables S1–S3. For each of the secondary end points, we also calculated the unadjusted hazard ratios (HRs) with the alternate methodology of defining the end point as a composite of the secondary event plus mortality. To account for competing risks during LVAD support, the primary end point model was displayed as a Fine‐Gray cumulative incidence plot with outcomes of mortality, transplantation, or LVAD explantation. The primary and secondary end point models were adjusted for key demographics and clinical variables that were identified a priori as potential confounders of the relationship between cachexia and outcomes, or were unbalanced between the subgroups with and without cachexia. The adjustment covariates were age, sex, Intermacs profile, time since cardiac diagnosis, ischemic primary cardiac diagnosis, chronic renal disease, severe diabetes, history of alcohol abuse, history of pulmonary hypertension, and history of lung disease.

The variables of number of prior cardiac hospitalizations, history of atrial arrhythmias, and history of liver dysfunction were considered but not selected due to high proportions of missingness. To account for lesser degrees of missingness among the chosen covariates, we used multiple imputation methods (PROC MI) in all adjusted models but did not impute for the primary variable of interest (cachexia/malnutrition) or any of the end points. The proportional hazards assumption was examined using Martingale residuals, and variables that violated the assumption (P<0.05) were included in the mortality models as time‐varying covariates. For these variables, mortality was defined as early (<24 months postimplant) or late (≥24 months postimplant), because the data showed this to be a reasonable break point in mortality hazards. We additionally calculated the event rate, per 100 patient‐years, for each of the primary and secondary outcomes for the cachexia present versus cachexia absent groups.

Sensitivity Model

A sensitivity model was constructed to examine whether the relationship between cachexia and clinical outcomes would be consistent if the cachexia variable were redefined as a composite of the cachexia/malnutrition field and pre‐LVAD body mass index (BMI) <18.5 kg/m². The threshold of <18.5 kg/m² was selected to align with the Word Health Organization criteria for underweight. This alternate cachexia definition (titled cachexia/low BMI) is more inclusive, because it incorporates patients who were not recognized to be cachectic by the reporting center, as well as those who had a missing cachexia/malnutrition field but did have a baseline BMI. The Kaplan‐Meier plots and Cox proportional hazards models were performed as above, including the use of time‐varying covariates for variables that failed the proportional hazards assumption.

Weight Change Postimplantation

The Intermacs follow‐up file was used to determine the weight change after LVAD implantation from postimplant weight measurements in the full cohort. Weight was recorded at 1, 3, 6, and 12 months post‐LVAD implantation date (within 14 days of each time point) to define the early weight trajectory. If the percent weight change between these time points was >±25%, then the values were judged likely to be erroneous, and the weight for that visit was set to missing. Median values were displayed on a box plot, and the percent and absolute weight changes between baseline and postimplantation time points were calculated and displayed in a table.

Based upon the timeframe of metabolic recovery during LVAD support, ¹³ , ¹⁴ weight change between preimplantation and 3 months postimplantation was selected as the metric of interest. Only patients who survived to at least 3 months of LVAD support were included in this analysis of change in weight from implantation to 3 months. If a weight was not available at the 3‐month visit, and the percent change in weight from implantation to the 1‐month visit was ≥5%, then the weight change at the 1‐month visit was used instead. A histogram of percent weight change from preimplantation to 3 months postimplantation was constructed using this cohort. This cohort was then split into 2 groups by patients whose weight changed by +5% or greater at 3 months (≥5%) versus those who did not for survival analysis.

Survival was compared for patients with versus without ≥5% weight gain in the first 3 months of LVAD support using a Kaplan‐Meier plot and estimates. Unadjusted and adjusted Cox proportional hazards models were constructed for the end point of mortality, using the same censoring and adjustment considerations as detailed above. The at‐risk time for these weight change models began at 3 months post‐LVAD implantation. Patients who died, were transplanted or explanted, or lost to follow‐up within the first 3 months were not included in the survival analysis. We additionally constructed a restricted cubic spline curve to display the HR for mortality across the spectrum of percent weight change in the first 3 months of LVAD support to identify the range of early weight change that is associated with most favorable subsequent survival.

All analyses were conducted using SAS version 9.4 (SAS Institute, Cary, NC), except for the box plot, Kaplan‐Meier plots, and the cumulative incidence plots, which were constructed using R version 4.1.1 (R Foundation for Statistical Computing, Vienna, Austria). P values <0.05 were considered significant.

Results

Patient Cohort

Of the 22 458 primary LVAD implantations within the 2006 to 2017 inclusion window, there were a total of 1691 exclusions for implantation of a biventricular ventricular assist device, isolated right ventricular assist device, or total artificial heart, and 435 patients missing cachexia status (Figure 1). Therefore, the total cohort was 20 767, with 20 332 patients available for the analyses of baseline cachexia status. Overall, 78.9% of LVAD recipients were men, mean age was 56.7±12.9 years, mean BMI was 28.7±6.5 kg/m², and the most common strategy for LVAD implantation was bridge to transplantation (Table 1). The modal pre‐LVAD BMI category was 25.0 to 29.9 kg/m² (Figures S1–S3).

Figure 1. Flowchart for cohort inclusions and exclusions.

*The cachexia missing subgroup is not included in the analysis of outcomes related to cachexia status but is included in the analysis of weight change postimplantation. BiVAD indicates biventricular assist device; LVAD, left ventricular assist device; RVAD, right ventricular assist device; TAH, total artificial heart; and VAD, ventricular assist device.

Table 1.

Baseline Characteristics by Cachexia Status Group

Recipient characteristics	Baseline cachexia status group
	Cachexia present (n=516)	Cachexia absent (n=19 816)	P value
Age, y	60.2±12.1	56.6±12.9	<0.0001
Female sex	113 (21.9)	4177 (21.1)	0.66
Ischemic heart failure	268 (51.9)	8995 (45.4)	0.003
Weight, kg	72.1±18.4	88.3±22.4	<0.0001
Height, cm	174.4±9.9	175.1±9.8	0.12
Body mass index, kg/m2	23.8±5.1	28.8±6.4	<0.0001
Body surface area, m2	1.9±0.2	2.0±0.3	<0.0001
Bridge to transplantation	269 (52.1)	11 468 (57.9)	0.009
Continuous flow device	505 (97.9)	19 235 (97.1)	0.29
Intermacs profile			<0.0001
1	116 (22.7)	3188 (16.1)
2	224 (43.8)	7113 (36.0)
3	109 (21.3)	6285 (31.8)
4	57 (11.2)	2462 (12.5)
5	5 (1.0)	695 (3.5)
Time from heart failure diagnosis			0.15
<1 mo	24 (4.8)	1080 (5.7)
1 mo–1 y	70 (14.1)	2084 (10.9)
1–2 y	35 (7.0)	1348 (7.1)
>2 y	368 (74.0)	14 598 (76.4)
No. of cardiac hospitalizations			0.23
0–1	120 (31.2)	4202 (35.5)
2–3	180 (46.9)	5193 (43.8)
≥4	84 (21.9)	2454 (20.7)
Implant era			<0.0001
2006–2011	78 (15.1)	4975 (25.1)
2012–2017	438 (84.9)	14 841 (74.9)
Albumin, g/dL, median [25th, 75th percentile]	3.1 [2.7, 3.5]	3.4 [3.0, 3.8]	0.70
Prealbumin, mg/dL	15.5±6.6	18.8±7.4	<0.0001
Total bilirubin, mg/dL	1.7±2.7	1.4±1.8	<0.0001
BUN, mg/dL	32.7±19.6	29.3±18.1	<0.0001
Cholesterol, mg/dL	115.9±36.4	129.3±42.6	<0.0001
Creatinine, mg/dL	1.3±0.7	1.4±0.7	0.006
Hemoglobin, d/dL	10.5±2.2	11.3±2.1	<0.0001
AST, U/L, median [25th, 75th percentile]	34.0 [24.0, 58.0]	29.0 [21.0, 44.0]	0.27
ALT, U/L, median [25th, 75th percentile]	32.0 [20.0, 59.0]	29.0 [19.0, 49.0]	0.67
Sodium, mmol/L	134.6±5.1	135.0±4.8	0.05
WBC, K/uL	9.1±4.5	8.7±4.2	0.05
INR	1.4±0.5	1.3±0.4	0.002
LVEDD, cm	6.7±1.1	6.8±1.1	0.005
LVEF			0.97
≥30%	21 (4.4)	840 (4.6)
20%–29%	116 (24.0)	4569 (25.2)
<20%	346 (71.6)	12 716 (70.2)
Pulmonary diastolic pressure, mm Hg	24.3±7.8	25.1±8.8	0.08
Pulmonary systolic pressure, mm Hg	49.0±13.9	50.0±14.8	0.15
Pulmonary wedge pressure, mm Hg	24.7±8.9	24.9±9.1	0.68
Right atrial pressure, mm Hg	13.6±8.2	13.1±8.2	0.27
Chronic renal disease	170 (33.1)	3651 (18.4)	<0.0001
History of atrial arrhythmia	151 (35.4)	2947 (20.5)	<0.0001
Severe diabetes	54 (10.5)	1559 (7.9)	0.03
Liver dysfunction	70 (16.4)	513 (3.6)	<0.0001
Peripheral vascular disease	35 (6.8)	799 (4.0)	0.002
History of solid organ cancer	24 (4.7)	868 (4.4)	0.75
Chronic infection	19 (3.7)	208 (1.1)	<0.0001
History of alcohol abuse	54 (10.5)	1241 (6.3)	<0.0001
History of drug use	41 (8.0)	1219 (6.2)	0.09
Current smoking	30 (5.9)	1086 (5.5)	0.72
Frailty	239 (46.5)	940 (4.7)	<0.0001
History of major stroke	26 (5.1)	593 (3.0)	0.007
History of other cerebrovascular disease	18 (4.2)	328 (2.3)	0.009
History of pulmonary hypertension	157 (30.7)	3628 (18.5)	<0.0001
History of lung disease	82 (16.0)	1483 (7.5)	<0.0001

Continuous variables expressed as mean±SD, unless otherwise noted, and categorical variables expressed as number (%). ALT indicates alanine transaminase; AST, aspartate aminotransferase; BUN, blood urea nitrogen; INR, international normalized ratio; LVEDD, left ventricular end‐diastolic dimension; LVEF, left ventricular ejection fraction; and WBC, white blood cell.

Baseline Clinical Profile of Cachexia

Five hundred sixteen patients were identified as having baseline cachexia/malnutrition, representing 2.54% of patients with data for this field. Patients with cachexia were older, had more ischemic primary diagnoses, and had lower baseline hemoglobin, serum cholesterol, and prealbumin than the cohort without cachexia, and carried a higher burden of medical comorbidities. Patients with cachexia also were less likely to receive an LVAD as a bridge to transplantation (Table 1).

Primary Outcomes

Median survival in patients who were noncachectic with an LVAD was 47.8 months (95% CI, 46.4–49.2), and median survival in patients who were cachectic with an LVAD was 32.9 months (95% CI, 28.5–42.9; P<0.0001). Overall, 30.7% of patients died during LVAD support, 30.6% underwent heart transplantation, 3.6% were explanted without device replacement, and 35.1% were alive on LVAD support at the end of follow‐up. Mortality was significantly higher for LVAD recipients with baseline cachexia, with an HR of 1.36 (95% CI, 1.18–1.56; P<0.0001) for survival on LVAD support, censored for transplantation and LVAD explantation (Figure 2). After adjustment for key demographics and potential confounders, a significant association persisted between cachexia and mortality during LVAD support (adjusted HR, 1.28 [95% CI, 1.07–1.42]; Table S2).

Figure 2. Outcomes during LVAD support.

A, Kaplan‐Meier plot for primary end point of mortality during LVAD support. Blue line: cachexia absent. Red line: cachexia present. B, Black line: alive. Red line: dead. Orange line: LVAD explanted. Blue line: transplanted. C, Competing outcomes plot for patients without baseline cachexia during LVAD support. Black line: alive. Red line: dead. Orange line: LVAD explanted. Blue line: transplanted. LVAD indicates left ventricular assist device.

Sensitivity Model

Using the alternate definition of cachexia incorporating pre‐LVAD BMI <18.5 kg/m², 1022 of 20 458 patients met the category titled cachexia/low BMI (5.0% of cohort). Mortality during LVAD support was higher for patients with cachexia/low BMI than for those without (32.7% versus 30.5%). The unadjusted HR for mortality in the sensitivity model, censored for transplantation and LVAD explantation, was 1.17 (95% CI, 1.05–1.31; P=0.0057; Figure S2), with an adjusted HR of 1.12 (95% CI, 1.00–1.25; P=0.048).

Secondary Outcomes

Patients with cachexia were more likely to die during the implant hospitalization (8.1% versus 5.0%, P=0.013). Postimplantation hospital length of stay was nonsignificantly higher in patients with cachexia at median 21 days (14–34 days) versus 19 days (14–28 days) in patients without cachexia (P=0.48). Patients with cachexia were more likely to be discharged to a rehabilitation facility than patients without cachexia (39.5% versus 22.4%; Table S3). There were no significant differences in rates of infection, stroke, gastrointestinal bleeding, or RHF between LVAD recipients with and without baseline cachexia (Table 2). However, for the secondary outcome of pump thrombosis, recipients with baseline cachexia had a lower event rate (Figure S3), which remained significant after adjustment for potential confounders. When secondary event rates were alternatively expressed as a composite end point with mortality, there was a higher rate of stroke or mortality and bleeding or mortality for patients with cachexia (Table 2).

Table 2.

Primary and Secondary End Points by Baseline Cachexia Status

Events*	No. of events, cachexia present (n=516)	No. of events, cachexia absent (n=19 812)†	Event rate per 100 patient‐years, cachexia present (n=516)	Event rate per 100 patient‐years, cachexia absent (n=19 812)†	Unadjusted hazard ratio (95% CI)	Composite with mortality, unadjusted hazard ratio (95% CI)	Adjusted hazard ratio (95% CI)
Mortality	196 (38.0)	5983 (30.2)	25.71	18.90	1.36 (1.18–1.56)	N/A	1.23 (1.07–1.42)
Infection	235 (45.5)	9359 (47.2)	48.56	49.79	0.99 (0.87–1.12)	1.07 (0.96–1.19)	0.91 (0.8–1.04)
Thrombosis	103 (20.0)	4989 (25.2)	16.13	19.73	0.82 (0.67–0.99)	1.12 (1.00–1.27)	0.81 (0.67–0.99)
Stroke	112 (21.7)	4343 (21.9)	17.09	15.81	1.06 (0.88–1.28)	1.26 (1.11–1.43)	0.98 (0.81–1.19)
Bleeding	132 (25.6)	4794 (24.2)	23.03	19.60	1.16 (0.97–1.38)	1.26 (1.12–1.42)	1.03 (0.86–1.22)
Right heart failure‡	23 (9.0)	999 (9.2)	9.89	9.23	0.97 (0.64–1.47)	1.19 (0.96–1.47)	0.99 (0.65–1.50)

N/A indicates not applicable.

^*These percentages do not account for censoring and reflect only the proportion of outcomes that occurred during patient follow‐up.

^† Four patients were excluded from outcomes analyses because of time <0 days for mortality.

^‡ Variable was valid only up to June 2014, so right heart failure analyses were restricted to this time frame.

Weight Change Postimplantation

The cohort showed an increase in median body weight from 85.8 kg (25th, 75th percentile: 72.8, 101.3 kg) at implantation to 92.0 kg (25th, 75th percentile: 79.0, 107.4 kg) at 12 months of LVAD support (Figure 3). The mean percent weight change from implantation to 3 months of LVAD support was −1.2±8.4% (n=14 955 pairs; Table 3; Figure S4), with a mean change of 1.4±9.4% (n=361 pairs) in patients with cachexia and −1.2±8.5% (n=14 589 pairs) in patients without cachexia. Among patients with baseline cachexia and available data on weight change, 37.2% (n=132) achieved a 5% weight increase by 3 months. Conversely 21.6% (n=3094) of patients without baseline cachexia achieved a 5% weight increase by 3 months.

Figure 3. Change in weight over the first 12 months of LVAD support.

Weight (kilograms) at LVAD implantation (0 months) and 1, 3, 6, and 12 months of LVAD support. The horizontal line denotes the median value, with the lower and upper borders of the box showing the 25th and 75th percentile boundaries. The whiskers extend 1.5 times the interquartile range from the box, and the data points show the minimum and maximum values. Numbers of patients are: 0 months, n=17 359; 1 month, n=12 065; 3 months, n=14 955; 6 months n=12 806; 12 months, n=8554. LVAD indicates left ventricular assist device.

Table 3.

Changes in Weight Over the First 12 Months of LVAD Support

Time from LVAD implant, mo	No. of subjects	Median weight, kg	25th, 75th percentile, kg	Mean percent weight change from implantation±SD (no. of pairs)	Mean absolute weight change from implantation, kg±SD (no. of pairs)
0	17 359	85.8	72.8, 101.3	N/A	N/A
1	12 065	83.9	71.8, 98.0	−2.6±7.6% (n=12 065)	−2.6±6.8 (n=12 065)
3	14 955	85.0	73.0, 99.0	−1.2±8.4% (n=14 955)	−1.6±7.6 (n=14 955)
6	12 806	88.0	76.0, 102.7	1.9±9.6% (n=12 806)	1.1±8.7 (n=12 806)
12	8554	92.0	79.0, 107.4	5.0±10.3% (n=8554)	3.9±9.4 (n=8554)

LVAD indicates left ventricular assist device; and N/A, not applicable.

Twenty‐three percent (n=3483) of patients with available data met the threshold of ≥5% weight gain between implantation and 3 months of LVAD support (Figure S4). When comparing the groups with versus without a ≥5% weight gain at 3 months, there was an independent association between ≥5% weight gain and subsequent survival on LVAD support (unadjusted HR, 0.90 [95% CI, 0.84–0.98]; P=0.012; adjusted HR, 0.89 [95% CI, 0.82–0.97]; P=0.006). The most favorable adjusted mortality hazard on the restricted spline curve was seen within the range of 5% to 10% weight gain during the first 3 months of LVAD support (Figure 4).

Figure 4. Restricted cubic spline curve for percent weight change from baseline LVAD implant to 3 months of LVAD support and subsequent mortality (adjusted model).

Number of patients eligible for the analyses of change in weight over the first 3 months of LVAD support is 14 641. LVAD indicates left ventricular assist device. Blue line represents the spline curve. Dashed lines represent the 95% confidence interval. Vertical dotted line represents the reference value of 0% weight change.

Discussion

Within the Intermacs cohort, 2.54% of 20 332 primary LVAD recipients were locally identified to have cachexia/malnutrition at baseline and had a higher‐risk clinical profile than patients without cachexia. Preimplant cachexia was independently associated with lower survival during LVAD support, with the association remaining robust on a sensitivity analysis that incorporated patients with BMI <18.5 kg/m². The rates of Intermacs‐defined LVAD adverse events expressed separately from mortality did not appear to be elevated in patients with cachexia. Across the cohort, weight gain ≥5% in the first 3 months of LVAD support was independently associated with subsequent survival, with an optimal range of 5% to 10% weight gain for future survival.

For the first aim, the proportion of patients recognized as having cachexia/malnutrition at the time of LVAD implantation, in the absence of any specific criteria for coding this Intermacs registry variable, was much lower than anticipated from prior studies. The incorporation of pre‐LVAD BMI <18.5% into the criteria for the sensitivity analysis raised the proportion of patients screening in for cachexia up to 5.0%. These analyses suggest underrecognition of cardiac cachexia in routine care within the advanced HFrEF population, given that other studies using more robust diagnostic definitions have identified a higher prevalence of wasting. ¹⁵ For example, a 10.5% to 16% cachexia prevalence has been reported per >5% unintentional edema‐free weight loss criteria (n=238, mean age 71 years; n=11 712 clinical trial participants, mean age 65 years). ² , ¹⁶ A recent study of 200 inpatients and outpatients with New York Heart Association class III–IV heart failure in the United Kingdom used consensus cachexia criteria incorporating weight loss, anthropometrics, strength, symptoms, and biochemistry, and found a 15% prevalence, ¹⁷ whereas a US analysis of heart failure inpatient diagnostic codes yielded 7% cachexia. ¹⁸ Rates ranging from 19.5% to 52% are reported using dual x‐ray absorptiometry appendicular lean mass criteria (n=200, mean age 70 years; n=38, mean age 75 years). ⁷ , ¹⁹ Vest et al prospectively evaluated the prevalence of skeletal muscle wasting at the time of LVAD implantation using dual x‐ray absorptiometry appendicular lean mass criteria and found a 51.9% baseline prevalence (n=27 with dual x‐ray absorptiometry scans, mean age 56 years), ¹² suggesting clinical underrecognition of skeletal muscle wasting in advanced HFrEF.

Clinical assessment for cardiac cachexia at the time of LVAD implantation is currently challenging due to the lack of validated diagnostic criteria. Reliance on the physical examination or percent weight loss thresholds may underrecognize cachexia onset, because overt wasting on physical examination is considered a late manifestation, weight loss criteria are often insensitive to muscle wasting in the presence of fluid retention, and both are common as HFrEF becomes advanced. The related, but distinct, condition of malnutrition also has multiple diagnostic criteria, with limited validation among patients with advanced HFrEF. ¹ , ²⁰ The prevalence of malnutrition at LVAD/transplant evaluation was 24.7% in one cohort, as assessed by the screening tool Mini Nutritional Assessment Short Form. ²¹ Thus, we conclude that the Intermacs data set substantially underestimates wasting prevalence due to a lack of validated and easily applied diagnostic criteria for cardiac cachexia. It is anticipated that LVAD recipients who were designated with preimplantation cachexia in the Intermacs registry had severe skeletal muscle and/or subcutaneous adipose wasting upon physical examination that was readily recognizable to clinicians.

Patients designated as having cachexia were older and had a higher prevalence of baseline comorbidities including chronic infection, alcohol abuse, stroke, lung disease, peripheral vascular disease, liver dysfunction, and severe diabetes, each of which could cause or exacerbate the inflammatory and catabolic pathophysiology of cachexia. The laboratory findings for the subgroup with cachexia were aligned with prior literature on the metabolic dysfunction associated with inflammation and catabolism, including lower serum cholesterol, lower prealbumin (but not albumin), and lower hemoglobin. Highlighting the importance of work that has questioned the appropriateness of serum creatinine for estimation of glomerular filtration rate in the advanced heart failure population, ²² , ²³ the subgroup with cachexia had higher blood urea nitrogen and a higher prevalence of chronic kidney disease, as compared with those without cachexia, and yet a lower serum creatinine. This likely reflects low production of serum creatinine secondary to low skeletal muscle mass among patients with severe cachexia.

For the second aim, baseline cachexia was independently associated with mortality after LVAD implantation. This finding aligns with observations in the general ambulatory heart failure population, where cardiac cachexia and skeletal muscle wasting has been recognized as a strong independent predictor of mortality. ¹ , ⁴ , ¹⁸ , ²⁴ Specific to LVAD recipients, lower computed tomography pectoralis muscle mass and quality preimplantation has shown association with higher mortality during LVAD support. ²⁵ Similarly, abdominal skeletal muscle mass was associated with unadjusted mortality after LVAD implantation in a cohort from Japan. ²⁶ This Intermacs analysis supports the current practice of considering severe cachexia to be at least a relative contraindication to LVAD implantation, although it should be remembered that the majority of recipients with cachexia left the hospital alive (91.9%), and their median survival was 32.9 months.

What remains unknown is whether nutritional interventions such as protein‐calorie supplementation, for example via tube feeding, can meaningfully change outcomes for an LVAD candidate with baseline cachexia or malnutrition. Although preimplantation nutritional optimization is an appealing option, clinical experience suggests the neurohumoral metabolic state of advanced HFrEF that drives inflammation and catabolism cannot be adequately overcome simply by enhancing macronutrient intake. Future prospective studies are needed to determine what degrees of cachexia are recoverable by LVAD implantation alone and whether nutritional supplementation or other metabolic interventions are effective in enhancing the improvements in body composition that occur after LVAD implantation.

The observation that cachexia is associated with a lower rate of pump thrombosis was somewhat unexpected but has been previously reported when analyzing LVAD events by baseline BMI in an overlapping cohort, ²⁷ probably reflecting confounding by obesity and associated hypercoagulability. Additional analyses from other LVAD cohorts also corroborate a higher rate of pump thrombosis for patients with BMI >30 kg/m². ²⁸ Such reports have also noted a higher risk of infections, particularly driveline site infection, in patients with baseline obesity, ²⁷ , ²⁸ , ²⁹ , ³⁰ and the converse of a lower risk of infections with cachexia was not seen in the current study.

For the third aim, we determined that an early ≥5% weight gain is associated with superior survival during subsequent LVAD support. This is an important addition to the existing literature about weight and body composition in durable LVAD recipients. Prior publications have noted an association between lower baseline BMI and subsequent weight gain during LVAD support, ¹¹ with recent confirmation that skeletal muscle mass increases within 6 months of LVAD support, ¹² but the present analysis advances these observations by demonstrating the survival benefits of an early weight gain. The tendency for patients with baseline obesity to gain weight after surgical LVAD implantation is well established, which can have negative consequences on transplant candidacy and obesity‐associated complications, ³¹ whereas this new observation of a survival advantage after a 5% to 10% weight gain in the first 3 months of LVAD support casts weight gain in a more beneficial light. It is hypothesized that this early weight gain accompanies the systematic metabolic and inflammatory recovery already reported over the same period of LVAD support, which are also associated with greater survival. ¹³ , ³² , ³³

The principal limitations of this Intermacs analysis are the retrospective nature of data collection, particularly the subjective definition and nonstandardized data entry for cachexia/malnutrition. There was also significant missing data in the cachexia/malnutrition field, which we chose not to impute for given the importance of accuracy for this field, as well as for other important variables such as baseline liver function. There is potential for other unmeasured confounders, possibly including LVAD brand, which is not available within the Biologic Specimen and Data Repository Information Coordinating Center data set. Furthermore, the accuracy of clinically indicated weight measurements in the Intermacs registry is uncertain, and there are likely data entry errors, which we partly addressed by setting implausible value limits. Despite the limitations in data granularity and completeness, the findings from a national registry are far more likely to be generalizable than from a single‐center study. Additionally, our conservative statistical methods were designed to minimize the impact of missing or erroneous data entry and the incomplete recording of outcomes such as RHF. Another inherent limitation of this analysis is that only patients clinically deemed acceptable for LVAD surgery are included in the Intermacs registry; inclusion of more severely cachectic patients would be expected to strengthen the mortality relationship.

In conclusion, recognition of cachexia at the time of LVAD implantation is low within the Intermacs cohort at only 2.5%, which is substantially lower than published assessments of skeletal muscle wasting in similar populations. Cachexia was independently associated with higher mortality during LVAD support. Early weight gain was associated with lower subsequent mortality, suggesting that clinicians should promote strategies to facilitate early weight regain in the range of 5% to 10% after LVAD implantation. Future clinical trials evaluating interventions that augment recovery from cardiac cachexia may further improve outcomes for LVAD recipients.

Sources of Funding

None.

Disclosures

Dr Kiernan reports consultancy for Medtronic. Dr Couper reports consultancy for Abbott and Medtronic. The remaining authors have no disclosures to report.

Supporting information

Tables S1–S3

Figures S1–S4