Body mass index and natriuretic peptides trends before and after left ventricular assist device

It is unknown to what degree of sarcopenia related to heart failure (HF) is reversible with resolution of the HF syndrome. We evaluated whether (1) weight loss prior to left ventricular assist device (LVAD) is associated with pre‐operative sarcopenia as quantified on pre‐operative chest CTs and (2) determine the relationship between weight recovery (increase) after LVAD implantation and reduction of NT‐proBNP levels.


Introduction
Left ventricular assist devices (LVADs) have been proven to improve survival in patients with refractory end-stage systolic heart failure (HF), and to date, over 25 000 LVADs have been implanted in the United States, and the number of new implants continues to rise each year. [1][2][3] As the number of patients supported with LVADs continues to rise, it is important to improve patient selection for best utilization of resources and best outcomes for patients being considered for this therapy. Skeletal muscle mass is known to decrease with the progression of HF before overt weight loss is apparent. [4][5][6][7][8] Acceleration of the age-related declined in skeletal muscle mass and strength, known as sarcopenia, is increasingly recognized as a risk factor for adverse outcomes in several disease states. [9][10][11][12] Psoas and more recently pectoralis muscle mass indexed to body surface area on pre-operative computed tomographic (CT) scans have been shown as predictors for adverse clinical outcomes while on LVAD support. [13][14][15] It is presently unknown to what degree the skeletal muscle wasting (both low mass and tissue quality) of advanced systolic HF might be reversible with correction of the HF syndrome, and there is a paucity of data about the patterns of weight change in patients both before and following LVAD implantation. [16][17][18][19][20][21][22] It is also unclear what metabolic factors or biomarkers might promote the prevention and reversal of skeletal muscle wasting. A recent prospective LVAD cohort analysis found that 52% of LVAD recipients were sarcopenic prior to implantation, with a gain of 2.3 kg in fat-free mass by dual X-ray absorptiometry in the first 3 months after LVAD implantation. 23 Understanding the potential for reversal of skeletal muscle wasting and the mechanism of this reversal has important applications among the general population of patients with HF beyond the LVAD subgroup.
In this analysis, we use a single-centre LVAD cohort and plotted pre-operative and post-operative body mass index (BMI) and natriuretic peptides to assess the relationship between these variables. We also assessed the association between pre-operative BMI trend and sarcopenia measures. We hypothesized that (1) weight loss prior to LVAD is associated with pre-operative sarcopenia as quantified on pre-operative chest CT and (2) weight recovery (increase) during LVAD support occurs after reductions in NT-proBNP levels. In an exploratory aim, we sought to determine if weight gain in the first-year post-LVAD is associated with subsequent survival during LVAD support. If these hypotheses are demonstrated to be true, we can infer that patients with advanced HF and sarcopenia may be expected to recover weight with control of the underlying HF syndrome and that HF neurohormonal activation is central to the pathogenesis of HF-associated skeletal muscle wasting.

Cohort
The Institutional Review Board of the University of Minnesota approved this study. A waiver of consent was obtained due to the nature of the study. The larger cohort included patients who underwent first-time continuous-flow LVAD implantation between 2007 and 2020 at the University of Minnesota (n = 502). Those with pre-operative and post-operative BMI and NT-proBNP data were included in the trend analysis (n = 403). A subset of patients with chest CTs within 3 months of LVAD implantation (n = 194) were included in the analysis to assess the relationship between pre-operative BMI trend and sarcopenia measures. The date of last follow-up was 1 August 2020. Patients were censored at the time of transplant. The following demographic and clinical covariate data are available in the database, which is updated through data extraction and manual chart review: age, gender, BMI, serum creatinine, serum albumin, pre-albumin, international normalized ratio (INR), Interagency Registry for Mechanical Circulatory Support (INTERMACS) profile, pre-operative haemodynamics, bridge to transplant (BTT) status, cardiomyopathy type, and presence of diabetes. Baseline haemodynamic variables were collected from the last formal right heart catheterization, and lab values were taken from the last value prior to implant. For the NT-proBNP and BMI trend analysis, all possible values were pulled via data extraction from the electronic medical record from 1 year prior to LVAD and 1-year post-LVAD but before transplantation (if applicable). Six-month haemodynamic variables were collected from the first right heart catheterization post-LVAD closest to the 6 months +/À 3 months.

CT analysis
Methods of pectoralis muscle analysis were based on previous work conducted by Kinsey et al. 24 Measures of cross-sectional area in cm 2 , and mean Hounsfield units were obtained using SliceOMatic V5.0 software (Tomovision, Montreal, Canada) by a single reader who was blinded to patient outcomes. Once images were obtained in DICOM format, 5-10 min was required to measure each scan. Unilateral pectoralis muscle measurements were performed on a single axial slice directly superior to the aortic arch on the patient's right side. If a defibrillator was present on the right, the left pectoralis muscle was analysed instead. Muscles were manually shaded using a predefined attenuation range of À29-150 to obtain the mean Hounsfield units (PHU m ) and cross-sectional area (cm 2 ). Cross-sectional area measures were standardized for body size by dividing by heights in meters squared (m 2 ); this produced the measure of pectoralis muscle index (PMI, cm 2 /m 2 ). To determine intra-rater variability, a random subsample of scans (n = 30) were analysed a second time at least 2 weeks after completing initial measurements. To determine inter-rater variability, the same subsample of scans (n = 30) were analysed by a practicing cardiologist. Intraclass and interclass correlation coefficients values were generated using SAS Version 9.4 (SAS Institute, Cary, NC, USA) using the Shrout-Fleiss convention. 25 Intra-rater variability was found to be 0.99 (intraclass correlation coefficient). Inter-rater variability was found to be 0.97 (intraclass correlation coefficient).

Statistical analysis
Subsequent statistical analyses were performed with R version 4.0.2. A P-value of <0.05 was considered statistically significant. BMI trends before and after LVAD were computed and plotted using locally estimated scatterplot smoothing (LOESS). BMI plots were then repeated by pre-operative pectoralis muscle measure tertiles. BMI prior to LVAD was then regressed longitudinally. The association between BMI trend and pectoralis muscle measures were assessed via a linear mixed effects model by including fixed effects for time, the pectoralis muscle measures, and their interactions. BMI plots were then generated with log of NT-proBNP before and after LVAD. Log transformation was used because the variable of NT-proBNP was right-skewed. Linear mixed effects modelling was also performed to test the association between NT-proBNP and BMI.
As patients have varying numbers of BMI measurements in the first-year post-implant, we regressed the per cent BMI increase from baseline on time and took the regression coefficient to be the metric of BMI gain. This metric of BMI change between implant and 1 year was calculated for all subjects who survived at least a year post-implant. All BMI measurements between date of implant and 1-year post-implant were included in the calculation. A positive slope indicates an accelerating BMI growth rate, whereas a negative slope indicates a decelerating BMI growth rate. A slope of zero indicates a steady growth rate. We categorized patients into three tertiles based on this slope metric and plotted BMI and NT-proBNP before and after LVAD.
Baseline characteristics among the BMI gain tertiles plus a 4th category (death on LVAD within a year of implant) were compared with one-way analysis of variance (ANOVA) for normally distributed variables or the Kruskal-Wallis test for nonnormally distributed variables. Comparisons were repeated among the three BMI gain tertiles alone. Six-month NT-proBNP and haemodynamic variables were then compared between the groups. Chi-square tests were used to compare categorical variables. In the exploratory analysis to test the association between weight gain and subsequent survival modelled after 1 year of LVAD support, a Kaplan-Meier analysis was among the per cent BMI gain tertiles. We then compared the adjusted survival rate among the BMI gain tertiles via a Cox proportional hazards model adjusting for age at implant, race, sex, and baseline serum creatinine. Patients were censored upon receipt of a heart transplantation.

Hypothesis 1.
Weight loss prior to LVAD is associated with pre-operative sarcopenia as quantified on pre-operative chest CT.
BMI decreased prior to LVAD, decreased further shortly after LVAD, and then steadily increased (Figure 1). Across pectoralis muscle tertiles, the same pattern was observed. At 100 days prior to LVAD, both pectoralis muscle measures were significantly associated with BMI (P < 0.05). The trend of BMI prior to LVAD was significantly associated with PHU m (P < 0.05), with a more severe decline in BMI being associated with lower tissue attenuation (PHU m ).

Hypothesis 2.
Weight recovery during LVAD support occurs after the reductions in NT-proBNP levels.
The plots of BMI and log NT-proBNP and for the entire cohort before and after LVAD are displayed in Figure 2. BMI declined prior to LVAD implantation (Day 0), with further decline early post-implant, followed by a recovery (increase) in BMI between 100 and 300 days post-implant. The average per cent increase in BMI in Year 1 was 7.6%, 95% CI: 6.3-8.8%. NT-proBNP levels decreased during the first 100 post-LVAD implantation (À5.4%, 95% CI: À6.6 to À4.2%). Patients in the highest BMI gain tertile had the lowest NT-proBNP post-implantation, which stayed lower throughout the follow-up period. Post-LVAD NT-proBNP and BMI trends were significantly associated, with a decrease of 1 unit log NT-proBNP associated with an increase in 0.81 BMI (CI: 0.53-1.09, P < .001).

Assessment of patient characteristics by per cent BMI gain tertile
The baseline characteristics of the study cohort by per cent BMI gain tertile is displayed in Table 1. Among the four groups (death on LVAD within a year of implant, per cent BMI gain tertiles by 1 year), patients who died within the first-year post-implant had the lowest pre-operative pectoralis muscle measures, both by mass and mean Hounsfield units. Patients who had the steepest per cent BMI gain over the first year had the second lowest pectoralis muscle mass. There was no difference in NT-proBNP pre-LVAD between the groups. The highest weight gain group was younger, more likely to be bridge to transplant, more likely to have non-ischaemic cardiomyopathy, and less likely to be diabetic. Those in the steepest weight gain tertile had the lowest creatinine prior to LVAD. Those who died within the first year of LVAD support had the lowest pectoralis muscle measures, had the highest pre-operative creatinine, were older, and were more likely to be ischaemic in aetiology and diabetic. The right atrial pressure was highest in the group who died within a year of LVAD implantation. Patients who failed to gain weight post LVAD had the highest 6-month post-LVAD natriuretic peptides (lowest per cent BMI gain tertile NT-proBNP: 2208 vs. highest 1635 pg/mL, P < 0.001). This

Figure 2
Body mass index and log NT-proBNP trend before and during LVAD support. Time zero is LVAD implantation. The red arrow represents the phase of heart failure improvement (decrease of NT-proBNP) after LVAD implantation. The phase of heart failure improvement precedes body mass index recovery (green arrow) towards pre-implant levels.  is visually demonstrated in Figure 3, where those in the steepest weight gain tertile have the lowest NT-proBNP.

Exploratory aim to assess the impact of weight gain on LVAD and subsequent survival
After 1 year of LVAD support, patients with the steepest weight gain in the year post-LVAD had the highest survival ( Figure 4; highest vs. lowest % BMI gain tertile hazard ratio of 0.59, 95% CI: 0.342-1.021, P = 0.0592). The direction of the relationship persisted in the adjusted model with the steepest weight gain tertile had a lowest mortality (highest vs. lowest tertile HR 0.84, 95% CI 0.45-1.56, P = 0.57); however, this was not statistically significant.

Discussion
In this analysis, we demonstrate the temporal relationships between neurohormonal recovery and body mass recovery in patients with advanced systolic HF before and during LVAD support. For our two hypothesis, we found the following: (1) Weight loss prior to LVAD was associated with pre-operative sarcopenia as quantified on pre-operative chest CT, and (2) weight recovery during LVAD support occurred after the recovery of NT-proBNP levels. In addition, the highest tertiles of weight gain in the first year after LVAD had the highest unadjusted survival after that first year of LVAD support com-pared with the other groups, suggesting that weight gain was important. These data collectively suggest that recovery of body mass in patients with HF-associated skeletal muscle wasting may be dependent on resolution of the underlying HF neurohormonal activation ( Figure S1). Given the highest mortality rate in the lowest pectoralis muscle mass measures category, there may be a threshold upon which skeletal muscle mass wasting is not recoverable by treatment of the advanced HF syndrome. The observation that these patients within this muscle mass category had higher RA pressures and higher creatinine prior to LVAD implantation may suggest a greater severity of HF-especially right HF-that LVAD support alone may be unable correct. Furthermore, the combination of a higher creatinine and lowest muscle mass in this patient category highlights that renal function may have been overestimated by serum creatinine measures prior to LVAD implantation. Although it is conceivable that post-implantation renal dysfunction may be associated with persistence of a higher NT-proBNP level after LVAD, our results suggest that the persistence of the elevated NT-proBNP levels associates with a lack of BMI recovery after LVAD implantation.
Patients with the second lowest pectoralis muscle mass had the highest per cent BMI gain during LVAD support. These patients were younger, less likely to have diabetes or ischaemic cardiomyopathy and displayed a highly recoverable muscle wasting phenotype. An unresolved question is whether this phenotype of LVAD-responsive skeletal muscle wasting could be identified with more sophisticated muscle imaging pre-implantation, for example, with a muscle quality signal. The ability to identify patients with recoverable muscle wasting would be a substantial advancement in our patient selection risk stratification prior to LVAD implantation.
This work is consistent with previous demonstrating that weight gain, and more recently specifically muscle gain, occurs post LVAD. 23 It is also important to note that patients can have substantial muscle wasting and yet maintain a normal or high BMI, suggesting that it is muscle recovery after LVAD implantation that is probably most closely linked to survival. This also adds to the literature that demonstrates the importance of normalizing left sided filling pressures post-LVAD implantation. The association between persistence of elevated natriuretic peptides and a failure to gain weight during LVAD support may in part explain why patients with right HF after LVAD implantation are at high risk of mortality. 26 Next steps will include determining the pathophysiological pathways between decompensated HF and muscle wasting, for example, through biomarker analysis of patients who recover muscle mass after LVAD implantation. It is currently unclear whether improvements in nutrition and activity levels contribute towards post-LVAD weight gains and whether pre-operative efforts to improve strength or nutrition would have the potential to improve post-implantation outcomes.

Limitations
These analyses were performed using retrospective data. Important variables such as nutritional intake, cardiac rehabilitation participation, and pre-LVAD functional assessments are missing in this analysis. It is important to note, however, that patients who are being evaluated for LVAD and cardiac transplant are often too sick to perform formal functional assessments, especially when pre-LVAD temporary mechanical circulatory support is required. 27,28 There was no body composition assessment after LVAD implantation, and the use of BMI is an imperfect surrogate for muscle mass recovery. It is also acknowledged that obesity is associated with a negative bias on NT-proBNP levels in patients with HF, although the average BMI remained below the threshold of obesity (30 kg/m 2 ) during weight recovery. 29 Given the mortality rate is lower after the first year of LVAD support, the exploratory aim to assess the impact of weight gain on mortality was underpowered. Despite this, the unadjusted analysis demonstrated improved survival among those with the highest weight gain. It is important to acknowledge the susceptibility of any CT measures of quantity and attenuation to changes in X-ray techniques and presence of contrast.
Finally, it is recognized that these findings are hypothesis-generating and that the temporal sequence of NT-proBNP and BMI changes do not indicate causality or illustrate the mechanisms of weight recovery.

Conclusions
Weight recovery during LVAD support occurred after the recovery in NT-proBNP levels. Failure to gain weight after LVAD implantation was associated with persistently elevated NT-proBNP levels. These data collectively suggest that recovery of body mass in patients with advanced HF may be dependent on resolution of underlying HF neurohormonal activation.