Framework to Classify Reverse Cardiac Remodeling with Mechanical Circulatory Support: The Utah-Inova Stages

Abstract

BACKGROUND

Variable definitions and an incomplete understanding of the gradient of reverse cardiac remodeling following continuous flow left ventricular assist device (LVAD) implantation has limited the field of myocardial plasticity. We evaluated the continuum of left ventricular remodeling by serial echocardiographic imaging to define three stages of reverse cardiac remodeling following LVAD.

METHODS

The study enrolled consecutive LVAD patients across four study sites. A blinded echocardiographer evaluated the degree of structural (left ventricular internal diastolic diameter, LVIDd) and functional (left ventricular ejection fraction, LVEF) change after LVAD. Patients experiencing an improvement in LVEF ≥40% and LVIDd ≤6.0cm and were termed Responders, absolute change (Δ) in LVEF of ≥5% and LVEF <40%, were termed Partial Responders, and the remaining patients with no significant improvement in LVEF, Non-Responders.

RESULTS

Among 358 LVAD patients, 34 (10%) were Responders, 112 (31%) Partial Responders, and the remaining 212 (59%) were Non-Responders. The use of guideline-directed medical therapy for heart failure was higher in Partial Responders and Responders. Structural changes (LVIDd) followed a different pattern with significant improvements even in patients that had minimal LVEF improvement. With mechanical unloading the median reduction in LVIDd was: −0.6cm (IQR: −1.1 to −0.1cm, Non-Responders), −1.1cm (IQR: −1.8 to −0.4cm, Partial Responders), and −1.9cm (IQR: −2.9 to −1.1cm, Responders). Similarly, the median Δ in LVEF was: −2% (IQR: −6% to 1%), 9% (IQR: 6% to 14%), and 27% (IQR: 23% to 33%), respectively.

CONCLUSIONS

Reverse cardiac remodeling associated with durable LVAD support is not an all or none phenomenon and manifests in a continuous spectrum. Defining three stages across this continuum can inform clinical management, facilitate the field of myocardial plasticity, and improve the design of future investigations.

INTRODUCTION

Left ventricular assist device (LVAD) therapy reduces intraventricular filling pressures and decreases myocardial work by unloading the left ventricle (LV) which improves cardiac output, patient symptoms, and long-term survival. The improvement in LV function and structure with mechanical unloading was first described in 1994,¹ and the phenome of sufficient reverse remodeling that permits device explantation for myocardial remission or recovery has been recapitulated by many others.^1–13 In the Interagency Registry for Mechanically Assisted Circulatory Support (INTERMACS), significant reverse remodeling (LVEF ≥ 40%) was seen in ~10% of combined ischemic and non-ischemic cardiomyopathy (NICM) patients at one-year.^{2, 3} However, a critical limitation in the LVAD recovery field has been the dichotomous approach to assessing reverse remodeling, that defines a Responder by whether or not the patient achieves sufficient significant reverse remodeling to permit device explantation. Our current conceptual framework fails to describe the continuum of reverse remodeling that can occur after LVAD implant.^{7, 14–16} This limits the ability to study earlier stages of reverse remodeling and characterize the associated molecular processes and clinical management strategies that promote reverse remodeling.

Studies in heart failure with reduced ejection fraction (HFrEF), report reverse cardiac remodeling in 29–56% of patients treated with neurohormonal or cardiac resynchronization therapy.¹⁷ These studies also suggest that a partial improvement in LVEF is associated with reduced ventricular arrhythmias and improved patient survival.^18–20 Specifically, in the seminal study by Lowes and Bristow, changes in LVEF of ≥ 5% associated with beta-blocker therapy were found to define meaningful reverse cardiac remodeling and altered myocardial gene expression in a favorable manner.²¹ Further, a 5% change in LVEF reduces the odds of death 5-fold compared to patients with no change in LVEF.²²

We hypothesized that similar to HFrEF and many other biological processes, reverse cardiac remodeling with an LVAD does not manifest as an all or none phenomenon. Therefore, descriptors of the stages of recovery are needed to better characterize this continuum. Better descriptors of the stages of functional and structural changes associated with mechanical unloading will inform study design of future clinical and translational investigations that seek to understand and manipulate these events, as well as the use of adjunctive therapies to promote reverse remodeling.^23–31 Our research proposes a framework for staging reverse cardiac remodeling with mechanical unloading.

METHODS

Data Sharing

The data, analytical methods, and study materials will be made available to other researchers for the purposes of reproducing results or replicating procedures. Please contact the corresponding author.

Population

Four study sites participated in this analysis and included all patients implanted with a continuous-flow LVAD from 2009 to 2017 and followed through 2018. Patients were prospectively consented and enrolled at sites 1–3, the Utah Transplantation Affiliated Hospitals (U.T.A.H.) Cardiac Transplant program (University of Utah Health Sciences Center, Intermountain Medical Center, Salt Lake VA Medical Center), and retrospectively reviewed at site 4, the Inova Heart and Vascular Institute (IHVI). Consecutive patients with HFrEF who required an LVAD were included. The study was approved by all involved institutional review boards as part of U.T.A.H. where subjects were prospectively consented, and a waiver of consent was granted at Inova where data was retrospectively reviewed.

Duration of HFrEF was defined as the time from symptom onset to LVAD implantation. To target patients with chronic, advanced cardiomyopathy we excluded patients with confirmed acute myocarditis or symptoms for less than 3 months (i.e., acute HF patients that may have been considered as more prone to cardiac recovery). Additional exclusion criteria were age < 18 years, infiltrative or restrictive cardiomyopathies, and heart failure with preserved ejection fraction. Demographics, comorbidities, laboratory data, and vital signs were collected within 24 hours prior to LVAD implantation.

LVAD and pharmacological management

Patients in this study were managed according to institutional protocols for blood pressure management and LVAD speed titration,^{7, 32} which were consistent with the standard HF and LVAD guideline directed medical therapy.^{33, 34} Neurohormonal therapy for heart failure, heart rate, and mean arterial pressure (i.e. Doppler opening pressure) for LVAD patients was recorded in the final 3 months of LVAD support.

Echocardiography

Echocardiographic examinations were performed by the echocardiography laboratories of the participating institutions and digitally stored. The echocardiograms were performed within 1-month preceding LVAD implantation and then 1, 2, 3, 6, 9, and 12 months subsequently based on institutional protocol. A requirement for study inclusion was the presence of a pre-LVAD and at least one echocardiogram ≥ 3-month post-LVAD implant. Each echocardiogram was reviewed by a blinded cardiologist at each institution (MAS and OWP) who was unaware of each patient’s responder status. Echocardiography was performed according to guidelines from the American Society of Echocardiography and the European Association of Cardiovascular Imaging.^{7, 35} To assess change in LV function and structure the last available echocardiographic LVEF and LVIDd was analyzed within a year of LVAD implant. The LVIDd was also indexed to the body-surface area. The left ventricular end-diastolic volume index (LVEDVi) and left ventricular end-systolic volume index (LVESVi) at follow-up echocardiogram were calculated when suitable images were available for volumetric analysis. The degree of mitral regurgitation was quantitated and classified as none, trivial, mild, moderate, and severe.

Responder Definitions

Following at least 3 months of mechanical patients achieving an LVEF to ≥ 40% and an LVIDd ≤ 6.0 cm were termed “Responders” based on prior publications.^{2–4, 7} The remaining patients were categorized by change in LVEF: “Partial Responders”: absolute LVEF improvement of ≥5% and LVEF <40%,^{19, 21} ,²² and “Non-Responders”: <5% change in LVEF or no improvement.

Statistical Analysis

Data were summarized using descriptive statistics. Categorical variables were compared with the Pearson chi-square or Fisher exact test; since most continuous variables appear to be non-normally distributed based on the Shapiro-Wilk test, they were presented with the median (interquartile range, IQR) and compared with the Wilcoxon rank-sum test, or the Kruskal-Wallis test. The slope of the change in LVEF over the change in LVIDd (i.e., ΔLVEF (%)/ ΔLVIDd (cm)) was used to compare the response to mechanical unloading between responder groups. A Pearson correlation was performed to understand the relationship between the LVIDd and LVEDVi and LVESVi at the final echocardiographic follow-up. A two-sided p-value of < 0.05 was considered statistically significant. All analyses were performed with R (R Core Team, Vienna, Austria).³⁶

RESULTS

LVAD Patient Characteristics

After application of our exclusion criteria, there were 358 continuous-flow LVAD patients with baseline and repeat echocardiographic measurements 3 to 12 months following LVAD implantation. Of these, 231 (65%) were from sites 1–3 (UTAH), and 127 (35%) were from site 4 (INOVA). Overall, 1,349 echocardiograms were re-read by a blinded cardiologist as part of this analysis with an average of 4 echocardiograms per patient (1 pre-LVAD and 3 after LVAD implant). Baseline characteristics were similar between the study sites except that patients from Inova were more likely to have a non-ischemic rather than ischemic cardiomyopathy, be of African American race, and to be classified as INTERMACS Profile 3 rather than 4 (Table 1). The median time from HF diagnosis to LVAD implantation was 60 months (IQR: 24 to 120 months). Baseline echocardiographic measurements between the groups were similar as were the proportion of centrifugal-flow LVADs implanted (38%). The baseline LVEF for the entire cohort was 19% (IQR: 14% to 23%), with a LVIDd of 6.8 cm (IQR: 6.2 to 7.5 cm).

Table 1.

Patient Characteristics Prior to LVAD Implantation by Study Site

Attribute	Sites 1–3 (U.T.A.H.) (n=231, 65%)	Site 4 (INOVA) (n=127, 35%)	p-value
Age (years), median (IQR)	60 (51, 68)	57 (50, 64)	0.42
Male	196 (85%)	103 (84%)	0.78
Ischemic Cardiomyopathy	99 (43%)	38 (32%)	0.02
Hypertension	110 (48%)	72 (57%)	0.03
Diabetes Mellitus	81 (35%)	48 (38%)	0.61
Smoking History	116 (50%)	53 (42%)	0.12
Atrial Fibrillation	98 (42%)	52 (41%)	0.79
Pre-Implantation LVEF (%)	18 (15, 23)	19 (14, 24)	0.65
Pre-Implantation LVIDd (cm)	6.7 (6.1, 7.4)	6.9 (6.3, 7.6)	0.04
INTERMACS Profile			0.03
1	11 (5%)	14 (11%)
2	41 (18%)	29 (23%)
3	101 (44%)	65 (51%)
4	55 (24%)	12 (9%)
5–7	23 (10%)	7 (8%)
LVAD Type			0.01
HeartMate II	122 (53%)	81 (64%)
HVAD	86 (37%)	46 (36%)
Jarvik 2000	18 (8%)	0 (0%)
VentrAssist	3 (1%)	0 (0%)
Levacor	2 (1%)	0 (0%)

INTERMACS = Interagency Registry for Mechanically Assisted Circulatory Support; LVAD = left ventricular assist device; LVEF = left ventricular ejection fraction; LVIDd = left ventricular internal end-diastolic dimension. U.T.A.H.: Utah Transplant Affiliated Hospitals Cardiac Transplant Program: University of Utah Health & School of Medicine, Intermountain Medical Center & Salt Lake VA Medical Center

Of the 358 LVADs implanted, 203 (57%) were HeartMate II (Abbott Laboratories, Abbott Park IL), 132 (37%) were HVAD (Medtronic, Minneapolis MN), 18 (5%) were Jarvik 2000 (Jarvik Heart Inc., New York, NY), 3 (1%) were VentrAssist, and 2 (1%) were Levacor. There were no HeartMate 3 (Abbott Laboratories) devices included in this study.

Response to Mechanical Unloading and Effect of Turndown Studies

Individual response to mechanical unloading varied significantly between the baseline and follow-up echocardiograms, suggesting a gradient of reverse cardiac remodeling after LVAD (Figure 1). At sites 1–3, all patients were tested with prospective LVAD turndown studies for 30 minutes at the lowest manufacturer recommended LVAD speed setting. During the LVAD turndown study, the change from optimal LVAD support to minimal LVAD support in LVEF was a 0% (IQR: 0 to 2%) and LVIDd 0.13cm (IQR: −0.12 to 0.44cm). This was consistent with prior studies that also showed small changes between pre- and post-turndown LVAD studies.^{7, 37} For the purpose of these analyses we used values before the LVAD turndown study was performed as this was not routinely performed at site 4. Over one-year follow-up, the median overall change in LVIDd was a decrease of 0.7 cm (IQR: 0.2 to 1.6cm) with an increase in LVEF of 3% (IQR: −3% to 10%, Supplemental Figure I).

Figure 1. Echocardiographic Measurements at Baseline and After LVAD Implantation.

Visual depiction of change in each patient’s left ventricular cavity size and ejection fraction at baseline before LVAD and at follow-up. The blue circles represent the intersection of the patient LVEF and LVIDd at baseline before LVAD and the orange triangles represent the values at follow-up while on LVAD support. The LVIDd is graphed on the x-axis and LVEF on the y-axis. Dashed lines mark left ventricular end-diastolic internal diameter (LVIDd) of 6.0 cm and left ventricular ejection fraction (LVEF) of 40%.

Staging Reverse Cardiac Remodeling with Mechanical Unloading

A majority of LVAD patients experienced changes in structure as manifest by a reduction in the LVIDd within one-month of LVAD implant and this occurred even in patients with minimal to no improvement in systolic function (Figure 1). Change in ventricular function, LVEF, was quantified before and after LVAD, and patients were divided into subgroups based on their degree of functional response to mechanical unloading. Following at least 3 months of mechanical unloading 34 patients (10%) achieved an LVEF to ≥ 40% and an LVIDd ≤ 6.0 cm and were termed “Responders.” The remaining patients were categorized by change in LVEF: “Partial Responders”: absolute LVEF improvement of ≥5% and LVEF <40% (n=112, 31%), and “Non-Responders”: <5% change in LVEF or no improvement (n=212, 59%, Figure 2).

Figure 2. Development of Stages for Myocardial Reverse Remodeling with Mechanical Unloading.

Flow diagram to assist in categorizing the stages of response to mechanical unloading using baseline and follow-up left ventricular size and function. In the general LVAD population patients are classified as: Responders, LVEF ≥ 40% & LVIDd ≤ 6.0cm (10%),^2–4 Partial Responders (31%),^19,22 and Non-Responders (59%), based on change in ventricular function and structure with ventricular unloading.

We calculated the slope of the change in LVEF over change in LVIDd during LVAD support (ΔLVEF %/ ΔLVIDd cm), for each responder category. Specifically, we found that Responders had a steeper slope, indicating more significant remodeling (slope = −12.6, IQR: −19.6 to −9.1) compared to Partial Responders (slope = −8.3, IQR: −15.5 to −3.4), and Non-Responders (slope = 0, IQR: −2.9 to 7.7, Figure 3). Changes in ventricular structure and function were compared between the 3 responder groups; Responders had an improvement in LVIDd of 1.9cm (IQR: 1.1 to 2.9 cm) and Non-Responders by 0.6cm (IQR: 0.1 to 1.1 cm) (p < 0.001). The median change in ventricular function was an LVEF improvement in Responders of 27% (IQR: 23 to 33%) compared to a LVEF reduction in Non-Responders of −2% (IQR: −6 to −1%) (p < 0.001, Table 2). For each patient group, reverse remodeling in terms of structure and function is noted as early as the first month after LVAD implantation and continues to evolve throughout the first year (Figure 4).

Figure 3. Slope of Reverse Cardiac Remodeling based on Responder Stage.

Visual depiction of the change in each patient’s left ventricular cavity size on the x-axis (LVIDd) and ejection fraction on the y-axis (LVEF) at baseline before and after mechanical unloading with an LVAD. The circle indicates the baseline LVIDd and LVEF and the arrow tip represents the final LVIDd and LVEF. Each color group represents the three responder categories. The slope of the change in LVEF over the change in LVIDd is provided. Increasing negative numbers suggest more significant reverse remodeling.

Table 2.

Comparison of Change in Left Ventricular Structure and Function with Mechanical Unloading among Responder Categories

Echocardiographic Parameter	Non-Responder	Partial Responder	Responder	p-value (Kruskal-Wallis Rank Sum Test)
Baseline LVIDd (cm), median (IQR)	6.8 (6.3, 7.6)	6.8 (6.2, 7.3)	6.5 (5.7, 7.0)	0.02
Follow Up LVIDd (cm)	6.3 (5.6, 7.0)	5.7 (4.8, 6.5)	4.6 (3.9, 5.0)	<0.001
Δ in LVIDd (cm)	−0.6 (−1.1, −0.1)	−1.1 (−1.8, −0.4)	−1.9 (−2.9, −1.1)	< 0.001
Baseline LVEF (%)	20 (15, 25)	15 (10, 20)	17 (15, 25)	<0.001
Follow Up LVEF (%)	16 (13, 20)	26 (20, 30)	47 (42, 52)	<0.001
Δ in LVEF (%)	−2 (−6, −1)	9 (6, 14)	27 (23, 33)	< 0.001

LVIDd = left ventricular internal end-diastolic dimension; LVEF = left ventricular ejection fraction.

Figure 4. Time Course for Change in Left Ventricular Function and Structure after LVAD Based on Responder Stage.

By comparing the baseline and last follow-up echocardiogram, changes in LVEF and LVIDd were used to categorize LVAD patients into three distinct groups: Responders (blue), Partial Responders (green), and Non-Responders (red). Using serial echocardiography, the change in function (LVEF) and structure (LVIDd) after LVAD implant are depicted by responder category.

In a subset of our population, we had adequate echocardiographic images to perform volumetric assessments of the LVEDVi and LVESVi at baseline and last follow-up (n=172, Supplemental Table I). Although most patients had a reduction in the LVEDVi with mechanical unloading, the magnitude was greatest in the Responder group (−87; IQR: −93 to −58 ml/m2) compared to Partial Responders (−43; IQR: −71 to −27 ml/m2), and Non-Responders (−29; IQR: −55 to −9 ml/m2, p<0.001). Similar changes were also noted for the LVESVi. Volumetric LV measurements at final follow-up were compared to the LVIDd, as well as the LVIDd indexed to the patient body-surface area. There was a high correlation between the final LVIDd and LVIDd indexed to the LVEDVi and LVESVi (R = 0.53–0.60, p < 0.001, Supplemental Figure II). The results are similar after adjusting for baseline measurements of LV size and function (R =0.48 – 0.52, p < 0.001, data not presented).

Clinical Characteristics of Responders

Clinical characteristics of patients by responder stage were compared. In comparing Non-Responders, Partial Responders and Responders, we see key statistical differences or non-significant trends across groups which are consistent with prior myocardial recovery publications. As the degree of cardiac reverse remodeling increases (i.e., Non-Responder to Partial Responder to Responder), we see that patients are younger, more likely to be female, have a shorter duration of heart failure before LVAD, a smaller baseline LVIDd, received an axial-flow device, and the proportion of NICM patients increases (Table 3). In addition, we looked at medical therapy between patients at final follow-up. Although final heart rate and blood pressure did not vary across responder categories, the proportion of patients treated with an angiotensin-converting enzyme inhibitor (ACE-I) or angiotensin II receptor blocker (ARB) was higher in Partial Responders and Responders (p=0.04) and the use of β-blockers also tended to be higher (p=0.11, Table 4).

Table 3.

Patient Characteristics at LVAD Implant by Reverse Remodeling Stage

Characteristic	Non-Responder (n=212, 59%)	Partial Responder (n=112, 31%)	Responder (n=34, 10%)	p-value
Age (years), median (IQR)	59 (51, 66)	59 (50, 67)	56 (44, 65)	0.23
Male	186 (89%)	91 (80%)	24 (71%)	0.01
Heart Failure Duration (months)	72 (25, 132)	60 (23, 108)	16 (6, 44)	<0.001
Ischemic Cardiomyopathy	90 (43%)	38 (33%)	9 (27%)	0.08
Hypertension	104 (50%)	62 (54%)	16 (47%)	0.63
Diabetes Mellitus	84 (40%)	36 (32%)	9 (27%)	0.15
Smoking History	100 (48%)	54 (47%)	15 (44%)	0.93
Atrial Fibrillation	85 (41%)	53 (47%)	12 (35%)	0.41
Axial-flow device	117 (56%)	78 (68%)	26 (77%)	0.01
Study Site 4 Patient	71 (34%)	42 (37%)	14 (41%)	0.66
INTERMACS Profile				0.84
1	16 (8%)	7 (6%)	2 (6%)
2	39 (19%)	22 (19%)	9 (27%)
3	101 (48%)	52 (46%)	13 (38%)
4	39 (19%)	20 (18%)	8 (24%)
5–7	15 (7%)	13 (11%)	2 (6%)

INTERMACS = Interagency Registry for Mechanically Assisted Circulatory Support; LVAD = left ventricular assist device; LVIDd = left ventricular internal end-diastolic dimension; LVEF = left ventricular ejection fraction.

Table 4.

Medical Therapy, Vital Signs, and Echocardiographic Parameters at Final Follow-up

Characteristic	Non-Responder	Partial Responder	Responder	P value
Vital Signs
Heart Rate, bpm, median (IQR)	82 (75, 90)	80 (73, 90)	81 (73, 90)	0.47
Mean Arterial Pressure, mmHg	84 (77, 90)	84 (77, 92)	88 (76, 94)	0.95
Guideline-Directed Medical Therapy
ACE-I or ARB	95 (45%)	67 (59%)	20 (59%)	0.04
β-blocker	129 (61%)	80 (70%)	26 (76%)	0.11
MRA	83 (40%)	46 (40%)	13 (38%)	0.97
Digoxin	26 (12%)	20 (18%)	7 (21%)	0.28
Echocardiographic Parameters
LVEF, %	16 (13, 20)	26 (20, 30)	47 (42, 52)	<0.001
LVIDd, cm	6.3 (5.6, 7.0)	5.7 (4.8, 6.5)	4.6 (3.9, 5.0)	<0.001
LVIDd indexed, cm/m2	3.0 (2.6, 3.4)	2.7 (2.4, 3.1)	2.2 (2.1, 2.4)	<0.001
Mitral Regurgitation				0.16
None-Mild	146 (90%)	84 (90%)	24 (89%)
Moderate	16 (10%)	6 (7%)	2 (7%)
Severe	0 (0%)	3 (3%)	1 (4%)

ACE-I: angiotensin-converting enzyme inhibitor, ARB: angiotensin II receptor blocker, MRA: mineralocorticoid receptor antagonist, LVIDd indexed: LVIDd / body surface area.

DISCUSSION

Our work describes the continuum of LV myocardial functional and structural change following LVAD-induced mechanical unloading and circulatory support using a large, cohort of advanced HFrEF patients. Reverse cardiac remodeling is a continuous phenomenon³⁸, and we propose a framework for staging the functional and structural changes after LVAD support into three distinct categories that can be used to characterize reverse remodeling: Responder (LVEF ≥ 40%, LVIDd ≤ 6.0cm), Partial Responder (LVEF Δ ≥ 5%, LVEF<40%), and Non-Responder (LVEF Δ <5% or no improvement).

Reverse Remodeling in HFrEF

Reverse cardiac remodeling in HFrEF patients treated with neurohormonal blockade and cardiac resynchronization therapy, is associated with improvements in survival compared to patients without reverse remodeling.^17–20 A meta-analysis suggests that even a 5% improvement in LVEF corresponds to a 14% reduction in mortality.²² This suggests that improvements in LVEF with a therapeutic intervention may serve as surrogate for clinical outcomes.²² The concept of staging reverse remodeling in HFrEF by absolute changes in LVEF is well-established, but has not been applied to LVAD patients. Whether an LVAD patient who is classified as a Partial Responder has improved clinical outcomes (e.g., less gastrointestinal bleeding, heart failure), similar to other HFrEF populations, will need to be validated in future studies with larger patient populations. A recent INTERMACS analysis strongly suggests that standard heart failure drug therapy/neurohormonal blockade and the associated reverse remodeling are beneficial to LVAD patients.³⁹ In our analyses we found that the use of guideline-directed medical therapy was higher in Partial Responders and Responders than Non-Responders.

Reverse Remodeling with LVADs

One of the reasons for the reported clinical variability in reverse remodeling with LVAD studies, includes the lack of consensus definitions that can inform clinical and translational investigations. Myocardial recovery leading to device explantation is reported to occur in about 1–2% of patients based on INTERMACS analyses and other large registries.^{2, 3, 40} Perhaps, more importantly, close to 10% of LVAD patients demonstrate significant improvements in myocardial function and achieve an LVEF ≥ 40% on follow-up echocardiography.^{2, 3, 40} What proportion of patients have milder degrees (i.e. Partial Responder) of reverse cardiac remodeling has not been analyzed. These earlier responder categories may benefit from further mechanical unloading or titration of guideline-direct medical therapy (GDMT) to improve to LV structure and function and ultimately become a Responder.

The distinction between change in LV structure (LVIDd) and function (LVEF) provides a more nuanced expression of reverse remodeling that can encourage future identification of the molecular, cellular, and histological signatures that underlie each of these two components of the observed change. In our analysis we saw that all patients experienced a reduction in ventricular size with mechanical unloading and this occurred within one-month after the initiation of LVAD support (Supplemental Figure I). This can likely be attributed to reductions in left ventricular pressures that reduce LV “distension” with mechanical unloading. Changes in ventricular function on the other hand were delayed and occurred throughout the first year (Figure 4). INTERMACS and single center analyses support the assertion that improvements in LVEF trail temporally behind reductions in LV dimension.^{3, 37} Importantly, to contrast cardiac reverse remodeling with an LVAD to how it is conceptualized within broader HF research (e.g. medical therapy, cardiac resynchronization therapy), we performed volumetric LV measurements (i.e. LVEDVi, LVESVi) in patients who had adequate image quality. The volumetric measures were highly correlated with the LVIDd, suggesting that in LVAD patients, LVIDd can reliably be used to characterize changes in LV structure with cardiac reverse remodeling.

Response to Ventricular Unloading and Clinical Implications

We found that the prevalence of certain clinical characteristics was more commonly seen with enhanced reverse cardiac remodeling (i.e., Non-Responder to Responder). This includes a higher proportion of younger patients with a NICM, shorter duration of HF, female sex, receipt of an axial-flow device, and a smaller pre-implant LVIDd. These clinical predictors of response to mechanical unloading have previously been reported in LVAD-induced myocardial recovery studies and strengthen our proposed staging of reverse cardiac remodeling.^{2, 3, 5, 40, 41} Critical to our staging of recovery is the classification of Non-Responders to mechanical unloading. This patient population represents ~60% of patients receiving an LVAD and is a HFrEF population with severe adverse remodeling that seems to gradually progress despite mechanical unloading and/or adjuvant therapy. These patients will likely remain on long-term LVAD support as destination therapy (DT) and if eligible could be considered for cardiac transplantation.

Myocardial recovery with mechanical unloading is seen as a clinically rare phenomenon, that can only be achieved in a minority of patients receiving an LVAD (1–2%).² Our multicenter data suggests that in fact 10% of all LVAD patients may be classified as Responders. Responders have sufficient phenotypic evidence of left ventricular reverse remodeling and can proceed to formal cardiac recovery testing.^{5, 6, 11, 37, 42–44} An additional 30% of all LVAD patients can be classified as Partial Responders. These patients demonstrate reverse left ventricular remodeling and may achieve Responder status with appropriate therapeutic interventions. By dichotomizing our definitions however, we remove this critical nuance and ignore the significant remodeling changes observed in a large part of the LVAD patient population, which could inform the myocardial plasticity and reverse remodeling field. For example, in Partial Responders with an improved LVEF, but persistently dilated LV, an LVAD-ramp study could be considered to maximize ventricular unloading and reduce mitral regurgitation in order to facilitate a further decrease in LV size.^{32, 45} Partial Responders with an LVIDd ≤6.0cm could benefit from continued optimization of guideline-directed medical therapy and regular echocardiographic re-testing (i.e. 3 months). These recommendations are based on prior studies and will need to be validated in large-scale prospective studies (Figure 5).

Figure 5: Conceptual Framework for Clinical Trial Design Incorporating the U-NOVA Reverse Remodeling Stages.

Reverse left ventricular remodeling is not an all or none phenomenon and as many other biological phenomena manifests as a continuous spectrum. We describe a new framework to characterize stages of reverse remodeling with mechanical unloading. We identified three distinct HF sub-populations and their cardiac functional and structural response to durable mechanical circulatory support. This framework for grading reverse remodeling allows for heart failure clinicians and investigators to readily recognize the stages of reverse remodeling and identify subpopulations for personalized management. This functional/structural reverse cardiac remodeling staging may also facilitate advancing the field of myocardial plasticity by informing the study design of future clinical and translational investigations. GDMT: Guideline directed medical therapy.

Study Limitations

Limitations of this study include the relatively small sample size and inclusion of data from only four study sites. Nonetheless, there was little difference in LV myocardial response between the sites suggesting generalizability of the results despite practice differences in clinical management after LVAD implantation (e.g., turn-down studies, speed titration, GDMT utilization). Additional limitations include the lack of centralized echocardiographic interpretation, the use of only one modality to assess mechanical unloading and its limitations to assess right ventricular structure and function. The cardiologists, who reviewed the echocardiograms, were however blinded to final responder status. We note that patients left ventricular function may fluctuate (up or down) during LVAD support in between the proposed stages of response, and our analysis was limited to the last available echocardiographic time point. Further, a 5% change in LVEF may be hard to detect with echocardiography, but we used the biplane Simpson’s method which has been shown to have the highest correlation to cardiac magnetic resonance imaging,⁴⁶ and has been validated in previous HFrEF studies as a meaningful improvement in LVEF.²²

Conclusion

Reverse cardiac remodeling with mechanical circulatory support is not an all or none phenomenon and as many other biological phenomena manifests as a continuous spectrum. We describe a framework to characterize the stages of reverse remodeling during mechanical circulatory support with an LVAD. We identified three distinct groups and characterize their left ventricular functional and structural response to durable mechanical circulatory support. This framework for grading reverse remodeling allows for clinicians and investigators to readily recognize and communicate changes in reverse cardiac remodeling and forms the basis for future clinical and translational investigations.

Clinical Perspective.

What Is New?

• We describe the continuum of left ventricular myocardial functional and structural change following left ventricular assist device (LVAD) induced mechanical unloading and circulatory support in a large, multicenter, cohort of advanced heart failure patients receiving LVAD support.

• Reverse cardiac remodeling is a continuous phenomenon, and we identify 3 distinct stages of response to mechanical unloading: Responder (LVEF ≥ 40%, LVIDd ≤ 6.0cm), Partial Responder (LVEF Δ ≥ 5%, LVEF<40%), and Non-Responder (LVEF Δ <5% or no improvement).

• The proportion of unselected, LVAD patients achieving each responder category were: Responder (10%), Partial Responder (31%), and Non-Responder (59%).

What Are the Clinical Implications?

• Use of the U-NOVA (Utah-INOVA) stages of reverse cardiac remodeling can improve delineation of the gradient of recovery seen with mechanical unloading and neurohormonal therapy for LVAD patients.

• These stages can be incorporated into the design of future clinical trials investigating how reverse cardiac remodeling improves clinical response to LVAD therapy or how novel therapeutic strategies can enhance myocardial recovery rates.

ABBREVIATIONS

GDMT

guideline-directed medical therapy

HFrEF

heart failure with reduced ejection fraction

INTERMACS

Interagency Registry for Mechanically Assisted Circulatory Support

LV

left ventricle

LVAD

left ventricular assist device (continuous flow)

LVEF

left ventricular rejection fraction

LVEDVi

left ventricular end-diastolic volume index

LVESVi

left ventricular end-systolic volume index

LVIDd

left ventricular internal end-diastolic diameter

NICM

non-ischemic cardiomyopathy