Understanding longitudinal changes in pulmonary vascular resistance after left ventricular assist device implantation

Abstract

Background:

Elevated PVR is common in patients with advanced heart failure. PVR generally improves after LVAD implantation, but the rate of decrease has not been quantified and the patient characteristics most strongly associated with this improvement are unknown.

Methods and Results:

We analyzed 1581 patients from the INTERMACS registry that received a primary continuous-flow LVAD, had baseline PVR ≥ 3 Wood units (WU), and had PVR measured at least once postoperatively. Multivariable linear mixed effects modeling was used to evaluate independent associations between postoperative PVR and patient characteristics. PVR decreased by 1.53 WU (95% CI 1.27–1.79 WU) per month in the first 3 months postoperatively, and by 0.066 WU (95% CI 0.060–0.070 WU) per month thereafter. Severe mitral regurgitation (MR) at any time during follow up was associated with a 1.29 WU (95% CI 1.05–1.52 WU) higher PVR relative to absence of MR at that time. In a cross-sectional analysis, 15–25% of patients had persistently elevated PVR ≥ 3 WU at any given time within 36 months after LVAD implant.

Conclusion:

PVR tends to decrease rapidly early after implant, and only more gradually thereafter. Residual MR may be an important contributor to elevated postoperative PVR. Future research is needed to understand the implications of elevated PVR after LVAD implantation and the optimal strategies for prevention and treatment.

Introduction

Pulmonary hypertension (PH) due to left heart disease (World Health Organization [WHO] group II) is common in patients with heart failure¹ and is associated with increased morbidity and mortality.^2-4 For patients with end-stage systolic heart failure, PH is of particular concern, as it can limit availability of advanced heart failure therapies. PH increases right ventricular (RV) afterload and is one of the risk factors for right heart failure (RHF) after left ventricular assist device (LVAD) implantation.^5-8 Severe PH is also a contraindication to heart transplantation (HT).⁹

One marker of PH severity is pulmonary vascular resistance (PVR). Elevations in PVR have been suggested to correspond to progressive pulmonary vascular remodeling.¹⁰ While it is known that PVR generally improves after LVAD implantation,^11-13 the expected reduction over time has not been quantified because previous studies have been small and/or lacked longitudinal patient follow up. Furthermore, not all LVAD recipients experience an improvement in PVR,¹⁴ and the patient characteristics most strongly associated with favorable changes in PVR over time after LVAD implantation are not known. Understanding these associations could help frame patient and provider expectations prior to implantation and guide management of patients with persistently elevated PVR after LVAD implantation.

Therefore, the aims of this study were to describe and quantify the temporal changes in PVR after LVAD implantation among patients with elevated baseline PVR, and to identify important patient characteristics associated with these changes using a large, multicenter cohort of LVAD recipients.

Methods

Patient selection

The Interagency Registry for Mechanically Assisted Circulatory Support (INTERMACS) protocol was approved by the National Institutes of Health, the Institutional Review Board at the Data Coordinating Center at the University of Alabama at Birmingham, and each of the institutional review boards of the participating hospitals. Deidentified INTERMACS registry data through December 31, 2017 were obtained from the National Heart, Lung, and Blood Institute. We queried the registry for adult patients receiving their first durable continuous flow LVAD (CF-LVAD) who had complete hemodynamic data to calculate PVR at baseline and at least one follow up time point. Patients with PVR ≥ 3 Wood units (WU) were included in this study. Patients that received a right ventricular assist device (RVAD) implant before, coincident with, or after their LVAD implant; a total artificial heart; or a pulsatile-flow LVAD were excluded.

Statistical analyses

The primary outcome was the trend in postoperative PVR, calculated as the transpulmonary gradient divided by the cardiac output. For each continuous covariate, extreme outlier values were truncated at the 99^th percentile of the observations for that variable, to avoid implausible observations. Missing data for baseline covariates were multiply imputed using the “MICE” R package, version 3.6.0. Imputation was conducted using a multilevel imputation structure to preserve the longitudinal nature of the variable relationships. The most frequently missing variables were for RV function (26% missing) and preimplant phosphodiesterase-5 inhibitor (PDE5i) use (18% missing). Missingness was below 6% for all remaining variables. PVR at baseline or postoperatively was not imputed.

Baseline characteristics for continuous variables were summarized as means ± standard deviations or median (interquartile range) as appropriate based on distribution of the data. Categorical variables were summarized as number (percent). Temporal trends in PVR were visualized by plotting available data at each scheduled follow up time and summarized using locally weighted scatterplot smoothing (LOESS) curves.

Postoperative PVR was modeled using a multivariable linear mixed effects model. This approach allows for estimation of population-average covariate-outcome associations (fixed effects), while also allowing for per-individual deviations from those population average associations (random effects) to add uncertainty to their estimates. Random effects included a per-patient random intercept and time-PVR relationship (slope). Fixed effects were estimated for the following baseline variables: age, gender, height, weight, duration of heart failure, device strategy (bridge to transplant vs bridge to decision vs destination therapy), INTERMACS profile, sodium, creatinine, total bilirubin, international normalized ratio (INR), aspartate aminotransferase (AST), centrifugal vs axial flow pump, concomitant mitral valve surgery, concomitant tricuspid valve surgery, left ventricular ejection fraction (LVEF), left ventricular end-diastolic dimension (LVEDD), qualitative degree of RV dysfunction, qualitative severity aortic regurgitation, systolic (SBP) and diastolic blood pressures (DBP), right atrial pressure (RAP), pulmonary artery systolic (PASP) and diastolic (PADP) blood pressures, pulmonary capillary wedge pressure (PCWP), cardiac output (CO), and preimplant phosphodiesterase-5 inhibitor (PDE5i) use. Because valvular regurgitation severity changes over time and likely influences PVR, mitral regurgitation (MR) and tricuspid regurgitation (TR) severity were modeled as time-varying covariates. Baseline PVR was not included in the model because of collinearity with its components (PASP, PADP, PCWP, and CO). The fixed effect of duration of LVAD support was modeled as a piecewise linear relationship, with a breakpoint at 3 months. All variables were forced into the model without any selection procedures. Because clinical definitions of “high PVR” vary in the literature and in practice, we performed a sensitivity analysis using a cutoff of 4 WU to define the cohort before performing the imputation and modeling steps. As an exploratory analysis, we used the same modeling strategy to model temporal changes in two markers of pulsatile right ventricular load, pulmonary arterial compliance (PAC, calculated as [CO/heart rate]/[PASP-PADP]) and arterial elastance (Ea, calculated as PASP/[CO/heart rate]). All statistical analyses were performed in R statistical software, version 3.5.3 (R Foundation for Statistical Computing, Vienna, Austria). A two-sided α threshold of 0.05 was used to determine statistical significance.

Data Quality

For subjects with implausible hemodynamic data points, we applied a harmonization process we have used previously.¹⁵ In brief, this process was structured to ensure that all hemodynamic values were positive and plausible, and to ensure that neither RAP or PCWP could be greater than PADP.

Results

Study Cohort

There were 19669 patients in the INTERMACS registry that received a primary CF-LVAD and did not receive an RVAD at any time. Of these, 3741 (19%) had sufficient hemodynamic data recorded to calculate PVR at baseline and at least one postoperative time point (Figure 1). Compared with patients without postoperative hemodynamic data, notable differences include younger age, higher frequency of bridge to transplant device strategy, more recent LVAD implant, more favorable INTERMACS profile, higher frequency of centrifugal flow LVAD implantation, more severe RV dysfunction, and higher frequency of preoperative PDE5i use (Table S1).

Figure 1: Derivation of the cohort from the Interagency Registry for Mechanically Assisted Circulatory Support (INTERMACS).

LVAD, left ventricular assist device; PVR, pulmonary vascular resistance; RVAD, right ventricular assist device.

Baseline characteristics of the cohort of 1581 patients with baseline PVR ≥ 3 WU patients are presented in Table 1. Mean age was 55.4 years and 1207 (76.3%) patients were male. Nearly all patients (1317, 83.3%) had heart failure for more than 2 years prior to LVAD implantation. Severe MR was present in 434 (27.5%), and 98 (6%) underwent concomitant mitral valve surgery. Mean pulmonary artery systolic pressure at baseline was 59.8 mmHg and median PVR was 4.22 WU (IQR 3.52–5.50 WU). 215 (13.6%) patients were treated with PDE5i preoperatively. The frequency of PDE5i use was higher postoperatively, staying relatively constant at 20-25% over time (Figure S1).

Table 1: Baseline characteristics of the patients.

Continues variables presented as means (standard deviation) or median [interquartile range]. Categorical variables presented as number (%). ALT, alanine aminotransferase; AST, aspartate aminotransferase; BTD, bridge to decision; BTT, bridge to transplant; CO, cardiac output; DBP, diastolic blood pressure; INR, international normalized ratio; INTERMACS, Interagency Registry for Mechanically Assisted Circulatory Support; LVAD, left ventricular assist device; LVEF, left ventricular ejection fraction; MV, mitral valve; PADP, pulmonary artery diastolic blood pressure; PASP, pulmonary artery systolic blood pressure; PCWP, pulmonary capillary wedge pressure; PDE5i, phosphodisesterase-5 inhibitor; RV, right ventricle; SBP, systolic blood pressure; TV, tricuspid valve, WU, Wood units.

	N=1581
Age (years)	55.4 (12.0)
Male gender	1207 (76.3)
Height (cm)	174.29 (9.57)
Weight (kg)	85.07 (21.81)
Duration of heart failure
<1 month	31 (2.0)
1-12 months	137 (8.7)
1-2 years	96 (6.1)
≥ 2 years	1317 (83.3)
Device strategy
Bridge to recovery	3 ( 0.2)
BTT (currently listed)	503 (31.8)
BTD	526 (33.3)
DT	547 (34.6)
Rescue	2 ( 0.1)
INTERMACS profile
1	171 (10.8)
2	583 (36.9)
3	591 (37.4)
4–7	236 (14.9)
Sodium (mEq/L)	134.55 (4.69)
Creatinine (mg/dL)	1.37 (0.53)
Total bilirubin (mg/dL)	1.10 [0.70, 1.60]
INR (mean (sd))	1.30 (0.32)
AST (U/L)	29.00 [21.00, 41.00]
ALT (U/L)	28.00 [19.00, 45.00]
LVEF (%)
>50	1 ( 0.1)
40-49	3 ( 0.2)
30-39	56 ( 3.5)
20-29	413 (26.1)
>20	1108 (70.1)
RV dysfunction
Normal	313 (19.8)
Mild	454 (28.7)
Moderate	518 (32.8)
Severe	296 (18.7)
Mitral regurgitation
None	67 ( 4.2)
Mild	520 (32.9)
Moderate	560 (35.4)
Severe	434 (27.5)
Tricuspid regurgitation
None	107 ( 6.8)
Mild	755 (47.8)
Moderate	546 (34.5)
Severe	173 (10.9)
Aortic regurgitation
None	983 (62.2)
Mild	531 (33.6)
Moderate	61 ( 3.9)
Severe	6 ( 0.4)
SBP (mmHg)	106.0 (15.7)
DBP (mmHg)	66.1 (11.3)
RAP (mmHg)	13.0 (6.8)
PASP (mmHg)	59.8 (13.3)
PADP (mmHg)	29.1 (8.3)
PCWP (mmHg)	23.5 (8.0)
CO (L/min)	3.43 (1.08)
PVR (WU)	4.22 [3.52, 5.50]
Preimplant PDE5i use	215 (13.6)
Surgical details
Centrifugal LVAD	468 (29.6)
Concomitant MV surgery	98 ( 6.2)
Concomitant TV surgery	254 (16.1)

Postoperative hemodynamic assessment frequency

Among the 1581 patients with baseline PVR ≥ 3 WU, 1078 (68%) had one postoperative hemodynamic assessment reported in INTERMACS, 346 (22%) had 2, and 157 (10%) had more than 2. The median time of postoperative assessment was 6 months (IQR 3–18 months).

Postoperative trajectory of RV afterload

Among patients with baseline PVR ≥ 3 WU, PVR appeared to decrease sharply in the first 3 months after LVAD implantation, followed by a modest rebound towards pre-implant values before decreasing more gradually over time (Figure 2). The rate of decline in the early postoperative period was 1.53 WU (95% CI 1.27–1.79 WU, p < 0.001) per month for the first 3 months, followed by a slower decline of 0.066 WU (95% CI 0.060 – 0.070 WU, p < 0.001) for each month thereafter (Table 2). However, in a cross-sectional analysis, at any given time over the first 3 years after LVAD implantation, 15–25% of patients assessed at that time had PVR ≥ 3 WU (Figure 3). Transpulmonary gradient (calculated as mPAP–PCWP) showed a similar trajectory, with a steep decline in the early postoperative period, followed by a more gradual decline thereafter (Figure S2).

Figure 2: Temporal trends in right ventricular afterload.

A locally weighted scatterplot smoothing (LOESS) line is fit to the observed data points, and the inset highlights the first 30 months of follow up.

Table 2: Factors significantly associated with changes in pulmonary vascular resistance (PVR) over time after left ventricular assist device implantation from a multivariable model among patients with baseline PVR ≥ 3 WU and had at least one postoperative PVR.

Only statistically significant associations are shown. For a full description of model covariates, see Methods. CI, confidence interval. Other abbreviations as in table 1.

	N=1581
	beta	95% CI	p-value
Age (per 10 years)	0.06	(0.01, 0.11)	0.013
Height (per 10 cm)	−0.10	(−0.18, −0.03)	0.007
Weight (per kg)	0.003	(0, 0.006)	0.023
PASP (per 5 mmHg)	0.31	(0.28, 0.34)	<0.001
PADP (per 5 mmHg)	0.55	(0.49, 0.60)	<0.001
PCWP (per 5 mmHg)	−0.89	(−0.96, −0.83)	<0.001
CO (per L/min)	−0.78	(−0.84, −0.72)	<0.001
Tricuspid regurgitation at any time
None	ref
Mild	−0.27	(−0.48, 0.06)	0.012
Moderate	−0.01	(−0.24, 0.23)	0.94
Severe	−0.07	(−0.37, 0.23)	0.64
Mitral regurgitation at any time
None	Ref
Mild	0.1	(−0.07, 0.28)	0.24
Moderate	0.81	(0.60, 1.01)	<0.001
Severe	1.29	(1.05, 1.52)	<0.001
Concomitant mitral valve surgery
At median LVEDD (6.9 cm)	−0.18		0.014
Per 1 cm decrease in LVEDD	−0.27	(−0.49, −0.04)	0.020
Duration of LVAD support (per month)
0–3 months	−1.53	(−1.79, −1.27)	<0.001
≥ 3 months	−0.066	(−0.07, −0.06)	<0.001

Figure 3: Proportion of patients with persistently elevated pulmonary vascular resistance (PVR) at a given time after left ventricular assist device implantation.

Error bars correspond to 95% confidence intervals. WU, Wood units.

Factors associated with change in RV afterload over time

In patients with elevated baseline PVR, there were several factors that were independently associated with changes in PVR over time (significant associations shown in Table 2, all model parameters shown in Table S2). In addition to the duration of LVAD support, the most notable of these was severity of MR. PVR was higher with each incremental increase in MR severity, with severe MR at a given follow-up time associated with a 1.29 WU increase in PVR compared with absence of MR at that time (95% CI 1.05–1.52 WU, p < 0.001). Higher pulmonary artery pressures at baseline were associated with higher post-LVAD PVR across follow up and higher baseline CO was associated with lower post-LVAD PVR over time. Of note, pump type (axial vs centrifugal flow) was not associated with postoperative PVR.

Preoperative LV end-diastolic dimension (LVEDD) was higher in patients with more severe MR (7.2 ± 1.0 cm vs 6.6 ± 0.9 cm for severe vs mild MR, p < 0.001). Among patients with severe MR, there was no difference in the association with postoperative PVR based on preoperative LVEDD (interaction p=0.37). 98 (6.2%) patients had mitral valve surgery at the time of LVAD implantation. These patients had slightly higher LVEDD compared with those who did not (7.1 ± 1.0 cm vs 6.9 ± 1.1 cm, p=0.032). For the median LVEDD of 6.9cm, mitral valve surgery was associated with a 0.18 WU reduction in PVR at any time (p=0.014) compared with no mitral valve surgery. This decrease in PVR was more pronounced in patients with smaller LV size: for every 1 cm decrease in LV size, mitral valve surgery was associated with an additional 0.27 WU reduction in PVR (interaction p=0.020) compared with no surgery.

In a sensitivity analysis using a cutoff value of 4 WU to define high PVR, we observed similar temporal changes in PVR (−2.07 WU per month for the first 3 months, −0.09 WU per month thereafter), and the association between MR (1.56 WU increase with severe MR relative to no MR) and PVR was similar in magnitude as in the primary analysis (Table S3). MR was also significantly and independently associated with improvements in PAC and Ea, markers of pulsatile RV afterload. Severe MR at a given follow-up time was associated with 0.99 mL/mmHg (95% CI 0.51–1.47 mL/mmHg) lower PAC and a 0.56 mmHg/mL (95% CI 0.40–0.73 mmHg/mL) higher Ea compared with no MR at that time (Table S4).

Discussion

In this study, we have shown in a large, multicenter cohort that for LVAD recipients with elevated PVR, PVR decreases most significantly in the first 3 months after LVAD implantation, and continues to decline at a slower rate thereafter. These findings confirm what has been observed in smaller cohorts and extend those findings by quantifying the expected early and late changes in PVR. They also highlight the independent association of mitral regurgitation severity as a potential cause of persistently elevated RV afterload postoperatively. This is the largest study of changes in RV afterload after LVAD implantation and more broadly, provides important insights into the timing, frequency, and potential mechanisms of the reversibility of WHO group II PH after treatment of left heart disease.

The current findings suggest that even in a large population with longstanding heart failure and hemodynamic markers of “mixed” PH, unloading the left heart can improve PVR in some patients with group II PH within a relatively short time. While previous studies have shown similar early decreases in PVR, duration of follow up in these studies was limited and it has remained unclear whether continued LVAD support would result in further decreases in PVR. One of the key strengths of this analysis is the use of a large cohort with follow-up data available over several years that can overcome some of these limitations. While this cohort does not contain sufficient data to determine detailed postoperative PVR trajectories for individual patients, leveraging a large population registry like INTERMACS and using mixed-effects modeling allows for estimation of population-average trajectories whose confidence intervals account for the uncertainty introduced from individual patient deviations. Our results show that, other factors being equal, the additional PVR reduction expected with LVAD support beyond the first 3 months is small. A similar small decrease in PVR after the immediate perioperative period was seen in a United Network for Organ Sharing analysis of patients with elevated PVR listed for HT after LVAD implantation, which identified a 0.055 WU per month decrease in PVR while on LVAD support.¹⁶ These findings can help to frame expectations for patients with persistently elevated PVR after LVAD implantation who are ineligible for HT because of PH.

Our cross-sectional analysis (Figure 3) suggests that a significant proportion of patients with elevated baseline PVR may have persistently elevated PVR (≥ 3 WU) as many as 36 months after LVAD implantation. This finding has important implications for the selection of patients with PH for LVAD therapy, both for patients anticipating potential HT candidacy as well as for destination therapy patients, for whom increased RV afterload is a contributor to late RHF, decreased quality of life, and increased mortality.^17,18 Better identification of patients with PH who are unlikely to experience significant reductions in PVR following prolonged LVAD support would help to frame patient expectations and inform management decisions in advance of surgery. The impact of persistent PH on long-term outcomes in LVAD recipients is also an important topic for future study.

The early steep decline in PVR is the expected hemodynamic effect of acute reduction of left atrial pressure (LAP). Elevated LAP passively raises mPAP and shifts the pulmonary vascular resistance-compliance relationship down and to the left,¹⁹ both of which contribute to increased RV load. Visual analysis of PVR trajectory demonstrated a possible rebound in PVR towards preoperative values between 1 and 3 months. Similar results have been observed in the early postoperative trends of other hemodynamic parameters by Fujino et al in a cohort of 150 patients.²⁰ The basis for these findings is unclear, and may be attributable to resolution of early postoperative physiologic changes or changes in vasoactive therapy that commonly occur around this time. However, we cannot exclude that because hemodynamic measurements in INTERMACS are clustered in the early postoperative period, the LOESS line may show more random variation in the first 3 months than beyond.

The more modest decrease in PVR seen with prolonged LVAD support beyond the early postoperative period may reflect that reduction of left atrial pressure in WHO group II PH can induce favorable structural remodeling in the pulmonary vasculature. While studies have identified various molecular pathways that lead to adverse pulmonary vascular remodeling in the setting of chronically elevated left atrial pressure,^21,22 there are fewer data available investigating pathways that regulate reverse remodeling once left atrial pressure is decreased. Animal models suggest that hemodynamic unloading of the pulmonary vasculature can cause histologic regression of lesions characteristic of PH,²³ though human data are lacking.

The association between MR severity and higher PVR following LVAD implantation suggests a potential therapeutic target for patients with persistently elevated PVR postoperatively. Just as mitral valve surgery can cause resolution of PH in patients with primary MR,^24,25 MR improvement as a result of LVAD implantation, even without concomitant mitral valve intervention, can improve PH. In a retrospective analysis of 69 patients, Kassis et al. found that patients with residual moderate or severe MR after LVAD implantation had higher pulmonary arterial pressures and worse RV function compared to patients with no or mild MR.²⁶ PVR, however, was not reported in this study. In a separate observational cohort of 270 patients with significant valvular regurgitation of any valve (a majority with MR), patients without residual valvular regurgitation had lower PVR compared with those with residual disease.²⁷ In both of these studies, patients with significant persistent valvular regurgitation, despite LVAD support, had increased risk of hospitalization or death, emphasizing the clinical relevance of residual MR in LVAD recipients. While the question of whether this association with adverse clinical outcomes is mediated through elevated RV afterload remains, the present findings further emphasize that therapies aimed at reducing residual MR, such as hemodynamically guided LVAD speed augmentation, may be an effective strategy to manage persistently elevated PVR in LVAD recipients.²⁸ Our finding that MR severity was also associated with increased RV pulsatile afterload (lower PAC and higher Ea) further highlights the importance of postoperative MR in LVAD recipients as a potential driver of RV load.

As expected, our data show larger LV size is associated with more severe MR, suggesting that MR is most likely functional in etiology. We found that concomitant mitral valve surgery was associated with a greater reduction in postoperative PVR in patients with smaller ventricles. While we were not able to quantify LV volumes with the data available, this finding raises the possibility that concomitant mitral valve surgery at the time of LVAD implantation is more effective for patients with out-of-proportion MR than in patients with purely functional MR, in whom LV unloading with LVAD alone may be sufficient to improve mitral valve coaptation. These results should be interpreted with caution, however, as the number of patients in this cohort that underwent such procedures was small. Further research is needed to understand which patients are at highest risk for residual MR to better select patients for mitral valve intervention at the time of LVAD implantation. This is particularly important given that concomitant procedures have been associated with increased risk of adverse events at the time of LVAD implantation.⁸

Limitations

There are several limitations to this study. By design, only patients that had postoperative hemodynamic data recorded in the INTERMACS registry were included. These patients were slightly different than patients who did not have follow up data available in the registry (Table S1), presumably because there was a clinical indication for repeat hemodynamic assessment. As such, the results may not be generalizable to the entire population of LVAD recipients. However, given that these data are derived from over 1500 LVAD recipients, these observations remain relevant despite potential limitations in generalizability. Analysis of hemodynamic data from large registries is also limited by the fact that details regarding management strategies that can impact pulmonary physiology, like LVAD speed or vasoactive medication use, are incompletely recorded. Similarly, reasons for obtaining a postoperative hemodynamic assessment are not captured in the registry. While some centers may record postoperative hemodynamic data in a protocolized fashion, it is likely that in many patients, postoperative hemodynamic data was obtained at a time of clinical decompensation and may not reflect the hemodynamic profile of a stable patient. It is also possible that the temporal relationships we observed may be due in part to regression to the mean, or the loss of patients due to death, who may have had higher PVR than survivors. Alternatively, if individual patients are transplanted or are less likely to have repeat hemodynamic measurements recorded after a significant PVR reduction has been observed, the observed rate of decline would appear to attenuate over time as these patients are selected out of the cohort.

Conclusion

While PVR improves after LVAD implantation, the most significant reduction in PVR occurs in the early postoperative period, with a more modest rate of reduction after 3 months, and a significant proportion of patients may have persistently elevated PVR as many as 3 years after implantation. The presence of moderate to severe postoperative MR is strongly associated with higher postoperative PVR and suggests a potential target for therapy for patients with persistent PH. More research is needed to understand the clinical implications of elevated PVR after LVAD implantation and the optimal treatment strategies to prevent it.