Hemodynamic reserve predicts early right heart failure after LVAD implantation

Abstract

BACKGROUND:

Early right heart failure (RHF) remains a major source of morbidity and mortality after left ventricular assist device (LVAD) implantation, yet efforts to predict early RHF have proven only modestly successful. Pharmacologic unloading of the left ventricle may be a risk stratification approach allowing for assessment of right ventricular and hemodynamic reserve.

METHODS:

We performed a multicenter, retrospective analysis of patients who had undergone continuous-flow LVAD implantation from October 2011 to April 2020. Only those who underwent vasodilator testing with nitroprusside during their preimplant right heart catheterization were included (n = 70). Multivariable logistic regression was used to determine independent predictors of early RHF as defined by Mechanical Circulatory Support–Academic Research Consortium.

RESULTS:

Twenty-seven patients experienced post-LVAD early RHF (39%). Baseline clinical characteristics were similar between patients with and without RHF. Patients without RHF, however, achieved higher peak stroke volume index (SVI) (30.1 ± 8.8 vs 21.7 ± 7.4 mL/m²; p < 0.001; AUC: 0.78; optimal cut-point: 22.1 mL/m²) during nitroprusside administration. Multivariable analysis revealed that peak SVI was significantly associated with early RHF, demonstrating a 16% increase in risk of early RHF per 1 ml/m² decrease in SVI. A follow up cohort of 10 consecutive patients from July 2020 to October 2021 resulted in all patients being categorized appropriately in regards to early RHF versus no RHF according to peak SVI.

CONCLUSION:

Peak SVI with nitroprusside administration was independently associated with post-LVAD early RHF while resting hemodynamics were not. Vasodilator testing may prove to be a strong risk stratification tool when assessing LVAD candidacy though additional prospective validation is needed.

Left ventricular assist devices (LVADs) are increasingly utilized to improve survival and quality of life in patients with advanced heart failure.^1,2 Unfortunately, early right heart failure (RHF) after LVAD implantation is a common complication associated with significant morbidity and mortality.^3–5 Preimplant prediction of early RHF remains a challenge at least in part due to difficulties in properly assessing right ventricular (RV) function.^6–10 The gold standard to assess RV function relies on multibeat pressure volume relations to determine RV-pulmonary arterial (PA) coupling, yet these measures are expensive, technically difficult and not feasible for routine clinical practice.¹¹ Alternatively, measures of RV reserve present as an appealing option as it is technically easier to assess. Recent data suggest measurements of RV reserve are tightly associated with RV-PA coupling,¹² and therefore, assessing RV reserve may represent an opportunity for better prediction of post-LVAD early RHF compared with traditional resting evaluations.

Reserve measures are usually performed with exercise or inotropic agents like dobutamine. This can be problematic in potential LVAD recipients who may already require inotropic agents or have substantial exertional limitations. Administration of nitroprusside acutely lowers systemic afterload, lowers pulmonary vascular resistance (PVR), and increases left ventricular (LV) stroke volume thereby pharmacologically unloading the LV in a simulacrum of mechanical support devices.¹³ However, if the RV lacks “functional” reserve and is unable to maintain LV preload, stroke volume augmentation will be limited. We therefore, hypothesized that the assessment of hemodynamic reserve in response to vasodilator administration and the associated changes in stroke volume may be helpful to predict early RHF post LVAD implantation.

Methods

After IRB approval, we performed a multicenter retrospective analysis of patients who had undergone continuous-flow LVAD implantation from October 2011 to April 2020. The three centers included were the Medical University of South Carolina (2016–2020), the University of Pennsylvania (2017–2020) and the University Health Network in Toronto (2011–2020). Only patients who underwent clinically indicated vasodilator testing with nitroprusside within one year of LVAD implant as part of their advanced heart failure therapy evaluation were included (n = 70). Vasodilator testing was performed as part of an advanced therapy evaluation when pulmonary vascular resistance ≥2.5 WU (n = 62) or in the setting of significantly elevated pulmonary arterial wedge pressure (PAWP) and systolic pulmonary artery pressure (SPAP; n = 8).^14,15 Relevant demographics, clinical characteristics and hemodynamics at baseline and at peak vasodilation were recorded. Right heart failure was defined using the new 2020 Mechanical Circulatory Support––Academic Research Consortium (MCS-ARC) definition and included both early acute and early postimplant RHF (Supplemental Table 1). Stroke volume index was calculated as (cardiac output/heart rate)/body surface area. PA elastance (Ea) was calculated by systolic PA pressure divided by stroke volume. After determining the optimal cut-point for peak SVI to predict RHF, an additional 10 consecutive patients who received vasodilator testing prior to LVAD implant from July 2020 to October 2021 were also analyzed.

Statistical analysis

Statistical analyses were completed using SPSS version 25 (IBM corporation, Armonk, NY). Means and standard deviations were calculated for all continuous variables; frequencies and percentages for all categorical variables. Variables with continuous outcomes differences were assessed using independent samples t-tests with early postimplant RHF versus no early post-implant RHF serving as the independent groups. All continuous outcomes were examined using Levene’s Test for equality of variances. If significant variance differences were found between groups, results were reported based on adjusted degrees of freedom using the Welch-Satterthwaite method. Categorical variables were assessed using X² analyses or Fischer’s exact tests as appropriate.

A total of 50 patient covariates were screened for significant correlation with early RHF using univariable logistic regression models to ensure individual covariates had adequate statistical power for inclusion. Following univariable logistic regression models, 2 sets of multivariable logistic models were completed. In the first multivariable model (i.e., liberal model), any covariate with a univariable p value ≤0.10 was entered into a multivariable logistic regression model to determine independent predictors of early RHF. The exception to this criterion was to exclude cardiac index (CI) and PA Ea because of the strong significant correlation (collinearity) with SVI. SVI was selected over CI and PA Ea based on stronger Wald coefficient, odds ratio (OR), and p-value with the early RHF outcome. In the second multivarianble regression model (conservative model), any covariate with a univariable p value ≤0.05 was entered into a multivariable logistic regression model to determine independent predictors of early RHF. Logistic regression results are presented as OR with corresponding p values. P values ≤0.05 were considered statistically significant. Finally, area under the curve (AUC) was calculated with a receiver operating characteristic (ROC) curve using predicted probabilities from the logistic regression model with early post implant RHF status to demonstrate an aggregate of the true positive and false positive rates.

Results

Of the 70 patients included, 27 patients developed early RHF (39%) after LVAD implantation, including 14 patients who required right ventricular assist device (RVAD) support (20%). The mean age of the cohort was 49 ± 14 years with the majority of patients being men (77%) and with nonischemic cardiomyopathy (74%). Early RHF differed by centers, device type, and time between RHC and LVAD implant. There were no other significant difference in demographics, preimplant clinical characteristics, RHF Risk Scores,^16–18 relevant laboratory values, or echocardiographic assessment of LV, RV, or valvular function (Table 1). Institution specific data is shown in Supplemental Table 2.

Table 1.

Demographics and Clinical Characteristics.

	Entire cohort (N = 70)	No early RHF (N = 43)	Early RHF (N = 27)	p-value
Facility				0.006
MUSC(n)	28	18 [42]	10 [37]
Penn (n)	25	10 [23]	15 [56]
Toronto (n)	17	15 [35]	2 [7]
Gender				0.493
Male (n)	54	32 [74]	22 [81]
Race				0.63
Caucasian (n)	34	23 [53]	11 [41]
Black (n)	30	17 [40]	13 [48]
Asian (n)	1	1 [2]	0[0]
Hispanic (n)	2	1 [2]	1 [4]
Other(n)	3	1 [2]	2 [7]
BMI (kg/m2)	30 ± 7	29 ± 7	31 ± 7	0.374
Age (years)	49 ± 14	50 ± 15	49 ± 12	0.952
Prior Sternotomy (n)	6	4 [9]	2 [7]	1.00
Pulmonary Vasodilator during RHC (n)	7	3 [7]	4 [15]	0.417
Time between RHC and LVAD implant (days)	61 ± 87	80 ± 99	29 ± 99	0.007
Implant Strategy				0.654
Sternotomy (n)	66	40 [93]	26 [96]
Thoracotomy (n)	4	3 [7]	1 [4]
Device Type				0.025
HVAD (n)	14	8 [19]	6 [22]
HM2 (n)	14	13 [30]	1 [4]
HM3(n)	42	22 [51]	20 [74]
Heart Failure Etiology				0.596
Ischemic (n)	18	12 [28]	6 [22]
Non-Ischemic (n)	52	31 [72]	21 [78]
INTERMACS Profile				0.084
Profile 1 (n)	10	4 [9]	6 [22]
Profile 2 (n)	16	7 [16]	9 [33]
Profile 3 (n)	30	21 [49]	9 [33]
Profile 4–7 (n)	14	11 [26]	3 [11]
Temporary Mechanical Support at Implant				0.059
IABP (n)	9	5 [12]	4 [15]
Impella (n)	2	0[0]	2 [7]
ECMO at Implant (n)	4	1 [2]	3 [11]
Echo parameters
EDD LV (cm)	7.0 ± 0.9	6.9 ± 0.9	7.1 ± 1.0	0.606
LV Ejection Fraction (%)	18 ± 7	18 ± 6	17 ± 7	0.405
TAPSE (mm)	16 ± 5	16 ± 6	15 ± 5	0.388
TAPSE/PASP (mm/mmHg)	0.32 ± 0.19	0.33 ± 0.17	0.32 ± 0.22	0.901
Severe RV Dysfunction (n)	12	4 [9]	8 [30]	0.755
Severe Mitral Regurgitation (n)	24	12 [28]	12 [44]	0.199
Severe RV Dilation (n)	11	8 [19]	3 [11]	0.511
Severe Tricuspid Regurgitation (n)	7	6 [14]	1 [4]	0.238
Penn-Columbia Risks Score				0.503
Low Risk (n)	39	24 [56]	15 [56]
Medium Risk (n)	17	12 [28]	5 [19]
High Risk (n)	14	7 [16]	7 [26]
Euromacs RV Risk Score	3.2 ± 1.9	2.9 ± 1.9	3.7 ± 1.9	0.107
Kormos RHF Risk Score	1.1 ± 1.9	0.9 ± 1.7	1.6 ± 2.2	0.154
CRITT Score	1.6 ± 0.9	1.5 ± 1.0	1.8 ± 0.8	0.163
Lab Values
Hemoglobin (g/dL)	11.7 ± 2.2	11.7 ± 2.3	11.5 ± 2.2	0.671
Sodium (mmol/L)	135 ± 5	136 ± 4	134 ± 6	0.345
Creatinine (mg/dL)	1.5 ± 0.6	1.4 ± 0.5	1.7 ± 0.8	0.11
Total Bilirubin (mg/dL)	1.4 ± 1.5	1.5 ± 1.8	1.4 ± 0.2	0.871

Abbreviations: BMI, body mass index; Echo, echocardiography; ECMO, extracorpeal membrane oxygenation; EDD, end-diastolic diameter; HM2, Heartmate 2; HM3, Heartmate 3; HVAD, Heartware Ventricular Assist Device; IABP, Intra-aortic balloon pump; LV, left ventricle; LVAD, left ventricular assist device; MUSC, Medical University of South Carolina; PASP, pulmonary artery systolic pressure; Penn, University of Pennsylvania; RHC, right heart cathertization; RHF, right heart failure; RV, right ventricle; TAPSE, tricuspid annular plane systolic excursion; Toronto, University of Toronto.

Values are presented as mean ± standard deviation unless otherwise indicated; [ ] = percentage.

Hemodynamics

Table 2 summarizes preoperative baseline hemodynamics, hemodynamics measured at peak vasodilator infusion, and changes in hemodynamics from peak to baseline. At baseline, lower SVI, higher PA Ea, higher mean PA pressure, and higher heart rate were statistically associated with early RHF. However, absolute differences of these values between groups were relatively modest. During nitroprusside administration (mean 1.7 ± 1.6 mcg/kg/min), both groups realized similar reductions in systemic mean arterial pressure, pulmonary pressures, and PAWP. Compared to the group with early RHF, those without RHF achieved a higher peak SVI (30.1 ± 8.8 vs 21.7 ± 7.4 mL/m²; p < 0.001), higher peak CI (2.39 ± 0.62 vs 1.97 ± 0.36 L/min/m²; p = 0.001), lower PA Ea (0.83 ± 0.31 vs 1.19 ± 0.52 mmHg/mL; p = 0.002), and lower heart rate (82 ± 17 vs 96 ± 22; p = 0.004). The change in CI and SVI was also higher in the group without RHF, despite no significant difference in change in RV afterload (PVR or PA Ea).

Table 2.

Hemodynamic Measurements

	Entire cohort (N = 70) [mean ± SD]	No early RHF (N = 43) [mean ± SD]	Early RHF (N=27) [mean ± SD]	p-value
Baseline hemodynamics
MAP (mmHg)	83 ± 9	82 ± 9	85 ± 8	0.185
PAPM (mmHg)	41 ± 7	40 ± 7	43 ± 7	0.044
CI (L/min/m2)	1.65 ± 0.36	1.66 ± 0.36	1.62 ± 0.36	0.669
RAP (mmHg)	11 ± 5	10 ± 4	12 ± 6	0.134
PAWP (mmHg)	25 ± 5	24 ± 5	25 ± 6	0.305
RAP:PAWP	0.45 ± 0.19	0.43 ± 0.18	0.49 ± 0.21	0.223
PVR (Wood units)	5.2 ± 2.2	4.9 ± 2.1	5.6 ± 2.3	0.248
SVI (mL/m2)	19 ± 6	21 ± 6	17 ± 5	0.01
Pulmonary Ea (mmHg/mL)	1.72 ± 0.64	1.58 ± 0.56	1.94 ± 0.70	0.021
Pca (mL/mmHg)	1.34 ± 0.68	1.40 ± 0.78	1.24 ± 0.50	0.342
PAPi(mmHg)	3.9 ± 3.0	4.3 ± 3.5	3.2 ± 1.9	0.155
SVR (dynes/sec/cm−5)	1807 ± 477	1796 ± 458	1824 ± 514	0.82
HR (bpm)	89 ± 20	83 ± 17	99 ± 22	0.001
Delta: Peak - Baseline
MAP (mmHg)	−14 ± 8	−14 ± 9	−14 ± 8	0.754
PAPM (mmHg)	−11 ± 7	−11 ± 7	−12 ± 8	0.472
CI (L/min/m2)	0.58 ± 0.50	0.73 ± 0.52	0.35 ± 0.37	0.001
PAWP (mmHg)	−7 ± 6	−8 ± 5	−7 ± 7	0.655
PVR (Wood units)	−2.0 ± 2.1	−2.1 ± 2.0	−1.8 ± 2.3	0.662
SVI (mL/m2)	7.5 ± 7.2	9.2 ± 7.5	4.7 ± 6.0	0.011
Pulmonary Ea (mmHg/mL)	−0.76 ± 0.48	−0.75 ± 0.41	−0.76 ± 0.59	0.98
HR (bpm)	−2 ± 10	−1 ± 7	−3 ± 13	0.51
Peak Vasodilator Hemodynamics
Nitroprusside dose (mcg/kg/min)	1.7 ± 1.6	1.8 ± 1.9	1.7 ± 1.2	0.816
MAP (mmHg)	69 ± 9	67 ± 10	71 ± 8	0.121
PAPM (mmHg)	30 ± 7	29 ± 6	31 ± 7	0.179
CI (L/min/m2)	2.23 ± 0.57	2.39 ± 0.62	1.97 ± 0.36	0.001
PAWP (mmHg)	17 ± 6	16 ± 6	18 ± 7	0.243
PVR (Wood units)	3.1 ± 1.4	2.9 ± 1.3	3.4 ± 1.6	0.152
SVI (mL/m2)	27 ± 9	30 ± 9	22 ± 7	<0.001
Pulmonary Ea (mmHg/mL)	0.97 ± 0.44	0.83 ± 0.31	1.19 ± 0.52	0.002
HR (bpm)	88 ± 20	82 ± 17	96 ±22	0.004

Abbreviations: CI, cardiac index; Ea, effective arterial elastance; HR, heart rate; MAP, mean arterial pressure; PAPi, pulmonary arterial pulsitility index; PAPM, mean pulmonary arterial pressure; PAWP, pulmonary arterial wedge pressure; Pca, pulmonary arterial compliance; PVR, pulmonary vascular resistance; RAP, right atrial pressure; SD, standard deviation; SVI, stroke volume index; SVR, systemic vascular resistance.

Values are presented as mean ± standard deviation unless otherwise indicated.

Association between clinical and hemodynamic characteristics and early RHF

Univariable and multivariable predictors of early RHF are shown in Table 3. In both liberal and conservative multivariable models, only baseline mean PA pressure and peak SVI remained significant, even when considering baseline SVI. ROC curves for SVI at baseline, delta, and peak are shown in Figure 1, with the AUC being the highest (0.780) for peak SVI. The optimal cut point was 22.1 mL/m², which had a 56% sensitivity, 84% specificity, 75% negative predictive value, and 68% positive predictive value for predicting postimplant early RHF (Figure 2). Multivariable analysis revealed that peak SVI was significantly associated with early RHF, demonstrating a 16% increase in risk of early RHF per 1 ml/m² decrease in SVI. Peak SVI, but not baseline or delta SVI, was also associated with early RHF when defined strictly by postoperative RVAD use (Supplemental Figure 1, p = 0.027; AUC: 0.693). Peak SVI also remained predictive of early RHF when limiting our cohort to only those with RHC within 6 months of VAD implant (n = 61, AUC: 0.757; p = 0.001), HeartMate 3 implants (n = 43; AUC: 0.832; p < 0.001), or those patients who did not require RVAD implant (n = 54; AUC: 0.800, p = 0.001). Cardiopulmonary bypass time (93 ± 39 vs 113 ± 50 min, p = 0.096) and intraoperative transfusions (1.1 ± 2.1 vs 2.2 ± 2.7, p = 0.089) were numerically higher in those who developed RHF, though neither reached statistical significance.

Table 3.

Univariable and Multivariable Predictors of Right Heart Failure

Characteristic	Univariable model OR|95% CI|p-value	Liberal model OR|95% CI|p-value	Conservative model OR|95°% CI|p-value
Facility
MUSC	Reference	Reference
Penn	(2.70)|0.888–8.21| p = 0.080	(1.35)|.264–6.93| p = 0.716
Toronto	(0.24)|0.045–1.27| p = 0.093	(.832)|.104–6.65| p = 0.863
Race
White	reference
Black	(1.60)|0.577–4.43| p = 0.367
Hispanic	(2.09)|0.119–36.63| p = 0.614
Asian	(0.00)|0.000–0.00| p = 1.00
Other	(4.18)|0.341–51.24| p = 0.263
Gender
Female	Reference
Male	(1.51)|0.461–4.96| p = 0.495
Age	(0.999)|0.966–1.03| p = 0.951
Device Type
HM3	Reference	Reference	Reference
HM2	(0.085)|0.010-.706| p = 0.023	(0.095)|0.007–1.26| p = 0.074	(0.081)|0.008–0.815| p = 0.033
HVAD	(0.825)|0.244–2.79| p = 0.757	Reference	Reference
Heart Failure Etiology
Non-Ischemic	reference
Ischemic	(0.738)|0.239–2.28| p = 0.597
INTERMACS Profile	(0.514)|0.298–0.885| p = 0.016	(0.745)|0.318–1.75| p = 0.499	(0.707)|0.339–1.48| p = 0.356
Temporary Mechanical Support at Implant
No	Reference	Reference
Yes	(3.2)|.921–11.12| p = 0.067	(1.03)|0.162–6.57| p = 0.972
Echo parameters
EDD LV (cm)	(1.15)|0.688 –1.91| p = 0.601
LV Ejection Fraction (%)	(0.968)|0.897 – 1.04| p = 0.401
TAPSE (mm)	(0.959)|0.872 –1.05| p = 0.384
Severe RV Dysfunction
No	Reference
Yes	(0.761)|0.205–2.82| p = 0.683
Severe RV Dilation
No	reference
Yes	(0.547)|0.132–2.27| p = 0.406
Severe Mitral Regurgitation
No	reference
Yes	(2.07)|0.753–5.68| p = 0.159
Severe Tricuspid Regurgitation
No	Reference
Yes	(0.240)|0.027 –2.12| p = 0.199
Baseline Hemodynamics
MAP (mmHg)	(1.04)|0.982–1.10| p = 0.186
PAPM (mmHg)	(1.08)|1.00–1.16| p = 0.051	(1.12)|1.00–1.24| p = 0.049	(1.12)|1.00–1.25| p = 0.046
CI (L/min/m2)	(.738)|.187–2.91| p = 0.664
RAP (mmHg)	(1.08)|.975–1.19| p = 0.143
PAWP (mmHg)	(1.05)|.957–1.15| p = 0.301
RAP:PAWP	(4.85)|0.380–61.89| p = 0.224
PVR (Wood units)	(1.15)|0.911–1.44| p = 0.245
SVI (mL/m2)	(0.882)|0.798–0.975| p = 0.014	(1.07)|0.916–1.25| p = 0.397	(1.06)|0.912–1.24| p = 0.436
Pulmonary Ea (mmHg/mL)a	(2.61)|1.11–6.12| p = 0.028
Pca (mL/mmHg)	(0.644)|0.254–1.64| p = 0.355
PAPi (mmHg)	(0.863)|0.699–1.06| p = 0.167
HR(BPM)a	(1.04)|1.01–1.07| p = 0.003
Peak Hemodynamics
MAP (mmHg)	(1.05)|0.987–1.12| p = 0.126
PAPM (mmHg)	(1.05)|0.977–1.14| p = 0.179
CI (L/min/m2)a	(0.169)|0.049-.579| p = 0.005
PAWP (mmHg)	(1.05)|0.972–1.14| p = 0.211
PVR (Wood units)	(1.29)|0.909–1.83| p = 0.154
SVI (mL/m2)	(0.864)|0.794-.939| p < 0.001	(0.852)|0.756–0.959| p = 0.008	(0.848)|0.754–0.955| p = 0.006
Pulmonary Ea (mmHg/mL)a Risk Score	(9.05)|2.19–37.41| p = 0.002
Penn Columbia Risk Score (Low Risk)	Reference
Penn Columbia Risk Score (Medium Risk)	(0.667)|0.196–2.27| p = 0.517
Penn Columbia Risk Score (High Risk)	(1.60)|0.468–5.48| p = 0.454
Euromacs RV Risk Score	(1.24)|0.952–1.61| p = 0.111
Kormos RHF Risk Score	(1.20)|0.928–1.56| p = 0.164
CRITT Score Lab Values	(1.47)|0.854–2.54| p = 0.164
Lab Values
Hemoglobin (g/dL)	(0.952)|0.764–1.19| p=.666
Sodium (mmol/L)	(0.953)|.862–1.05| p = 0.341
Creatinine (mg/dL)	(1.88)|.838–4.23| p = 0.126
Total Bilirubin (mg/dL)	(0.973)|0.704–1.34| p = 0.869
Days Between RHC and LVAD implant	(0.990)|0.981–0.999| p = 0.033	(0.999)|0.988–1.01| p = 0.815	(0.998)|0.987–1.01| p = 0.719

Abbreviations: CI, cardiac index; Ea, effective arterial elastance; Exp(B), exponentiation of B coefficient; EDD, end-diastolic diameter; HM2, Heartmate 2; HM3, Heartmate 3; HR, heart rate; HVAD, Heartware Ventricular Assist Device; LV, left ventricle; LVAD, left ventricular assist device; MAP, mean arterial pressure; MCS, mechanical circulatory support; OR, odds ratio; PAPi, pulmonary arterial pulsatility index; PAPM, mean pulmonary artery pressure; PAWP, pulmonary arterial wedge pressure; Pca, pulmonary arterial compliance; Penn, University of Pennsylvania; PVR, pulmonary vascular resistance; RAP, right atrial pressure; RHC, right heart catheterization; RHF, right heart failure; RV, right ventricle; SVI, stroke volume index; TAPSE, tricuspid annular plane systolic excursion; Toronto, University of Toronto.

Liberal Model includes univariable covariates ≤ 0.10, Conservative model includes univariable covariates ≤ 0.05.

^aCI, Pulmonary Ea, HR, and SVI have significant collinearity so only SVI was included in the multivariable models.

Figure 1.

ROC curve of SVI association with early RHF. ROC curve for stroke volume index at baseline, delta, and peak values to predict MCS-ARC defined early right heart failure after LVAD implantation. *Abbreviations: AUC=area under the curve; LVAD=left ventricular assist device; RHF=right heart failure; ROC=receiver operating characteristic; SVI=stroke volume index.

Figure 2.

SVI after nitroprusside predicts early RHF. Population pyramid dichotomized at SVI of 22.1 mL/m². The x-axis details the number of patients whose SVI is above or below 22.1. The y axis details if the patient did or did not develop RHF. Percentages for each quadrant is depicted with the total percentage representing the entire cohort. The majority with SVI <22.1 mL/m² developed early RHF, with a specificity of 84%, sensitivity of 56%, negative predictive value of 75%, and positive predictive value of 68%. *Abbreviations: LVAD=left ventricular assist device; RHF=right heart failure; SVI=stroke volume index.

Follow up cohort

Three of the 10 patients included in the follow up cohort developed RHF. Clinical and hemodynamic characteristics of this cohort are found in Supplemental Table 3. Only peak SVI was different between those who did and did not develop RHF (p = 0.017). All of the patients were categorized appropriately according to the previously established peak SVI cut-point of 22.1 mL/m².

LVAD implants without vasodilator testing

Supplemental table 4 outlines the demographics and clinical characteristics of the 191 excluded patients who received LVAD therapy over the study period but did not receive pre-implant vasodilator testing. Many characteristics were similar. This group was slightly older, had a higher proportion of INTERMACS 1 profile, and less commonly had severe mitral regurgitation. In terms of preimplant hemodynamics, the non-vasodilator group had similar degrees of right and left heart dysfunction but lower, albeit still elevated, RV afterload. One hundred forty-seven of the 191 patients (77%) met at least one of the criteria for vasodilator testing.

Discussion

The major findings of our study can be summarized as follows: (1) Early postimplant RHF was common in our cohort, (2) baseline clinical characteristics, echocardiographic and hemodynamic measures were not strongly and independently predictive of early RHF, and (3) hemodynamics at the peak of nitroprusside vasodilator administration, particularly peak SVI, was a good discriminator of those who did and did not develop early postimplant RHF. These findings suggest that vasodilator testing and acute pharmacologic LV unloading may be useful in risk stratification prior to LVAD implantation.

RV functional reserve response to LV unloading

Prediction of early RHF after LVAD remains challenging with most risk scores, and prediction tools only show modest discrimination in validation cohorts.¹⁹ Indeed, 4 published risk models included in our analysis were not predictive of RHF nor were echocardiographic or previously identified hemodynamic values like pulmonary artery pulsatility index and RAP:PAWP ratio. Both perioperative and postoperative factors clearly contribute to postoperative RHF, ones that are either not present or predictable preoperatively.^20,21 However, limitations related to preoperative assessment of RV function also play a role. Recent evidence suggests measures of RV reserve may be more closely associated with the gold standard of RV functional assessment (RV-PA coupling) than resting measures. In our study, nitroprusside, a potent systemic afterload reducer, predictably unloaded the LV and lowered LV filling pressures which in turn reduced pulmonary pressures and PVR. Stroke volume increased in the majority of patients, with a disparate response between those who ultimately did and did not develop early RHF after LVAD implantation. This disparate response may relate to a lack of RV functional reserve, and specifically, the inability to maintain sufficient LV preload as systemic afterload is reduced. It is important to note that in our cohort PA pressures, PVR, and pulmonary Ea had nearly identical reductions in those who developed early RHF and those who did not. This suggests that the disparate changes in SVI were not merely a reflection of differences in RV afterload but rather differences in RV contractility to handle increasing venous return. Indeed, a recent study by Kremer and colleagues suggested RV volume loading may be able to uncover impaired RV-PA coupling and lack of RV reserve.²² Importantly, pharmacologic unloading of the LV and the subsequent RV response are hemodynamically similar to what would occur with mechanical unloading with LVAD implantation. Whether testing the response of the RV to temporary LV mechanical support would also be predictive of early RHF is unknown.

The definition of early RHF in our study was based on the new MCS-ARC adverse events definitions.²³ This definition is similar to the 2013 INTERMACS criteria for severe and severe-acute RHF which has been associated with poor outcomes after LVAD.²⁴ In our study, all subjects declared to have early RHF met both definitions. Peak SVI remained predictive of early RHF even if post-implant RHF was defined more narrowly as the need for RVAD. RHF was common with 39% of subjects meeting the broader criteria for early RHF and 20% requiring RVAD support within the first 30 days. While relatively high rates of early RHF are similar to recent reports of centrifugal devices,^25–27 center specific thresholds for RVAD implantation likely also played a role with 8 of the 14 RVAD implantations planned at time of LVAD implant. Importantly, and given the retrospective nature of this study, the decision to use a planned RVAD was likely not related to vasodilator testing results. A subgroup analysis excluding all patients with RVAD implant also exhibited consistent results. Additionally, most but not all subjects had combined post- and pre-capillary PH; thus, higher RV afterload may have contributed to the high rates of early RHF. However, PVR has not been found to be a strong predictor of early RHF in other studies.^11,16

RV functional reserve in prediction of early RHF

Two prior studies have hinted that measures of RV reserve may be useful in predicting early RHF after LVAD implantation. In one small study of 6 patients with resting RV dysfunction by echocardiography, the RV response to dobutamine infusion (changes in estimated SPAP and tricuspid annular plane systolic excursion) was modestly different between those subjects who did and did not develop early RHF.²⁸ Although certainly dobutamine or another inotropic agent could be used to assess RV functional reserve, many higher risk patients are already on inotropic agents at the time of LVAD evaluation. In this cohort, 56% of subjects were already on an inotrope at time of RHC. Inotropic agents may also not unload the LV to the same extent as nitroprusside. These two factors may therefore limit the utility of such assessments. Another more recent manuscript by Gonzalez et al. compared the predictive ability of baseline hemodynamics to those when the patient’s heart failure was “optimized” prior to implant.²⁹ Optimization included a combination of diuretic therapy, inotropic agents and vasodilators. The authors concluded that the optimized hemodynamics, specifically the optimized PA pulsatility index, was additive to prediction models of post-LVAD early RHF. Similar to our study, the most optimized (peak) hemodynamic values, rather than the change from baseline to optimized, were more predictive of early RHF.

The AUC for peak stroke volume index in our study fell into the acceptable to good range (0.78), with 2 subgroup analyses (HM3 only; RHF excluding RVAD use) being slightly higher. As a comparison the AUC for the Euromacs RHF risk score was 0.7 in the derivation and 0.67 in the validation cohort.¹⁶ There is clearly still progress to make in risk prediction of RHF. Given this persistent uncertainly, using these risk stratification tools in shared-decision making discussions and perhaps to identify high risk individuals who may benefit from additional perioperative support may be more appropriate than restricting LVAD candidacy altogether.

Limitations

There are several limitations of this work which merit further discussion. First, this was a retrospective analysis, and only subjects who had a clinical indication for vasodilator testing were included. Therefore, the cohort is somewhat biased in that only patients with elevated PVR ≥2.5 woods or significantly elevated PAWP and SPAP underwent vasodilator testing. A comparison cohort of patients who underwent LVAD implant without vasodilator testing had generally similar characteristics, but lower, albiet still elevated, RV afterload. It remains unknown if this strategy would be helpful in predicting early RHF in those without a pre-capillary component to their PH or in those with a more modest baseline PAWP and SPAP. However, the latter hemodynamic profile would be uncommon in LVAD candidates as those with normal or near normal PAWP may not be optimal candidates for LVAD therapy as additional LV unloading would be unlikely to improve heart failure symptoms.³⁰ While we believe the multicenter nature of our study is an overall strength, the incidence of early RHF was lower in one center compared to the other two centers. This difference was not maintained in multivariable modeling suggesting that patient characteristics rather than center specific differences played a more significant role. Further, we completed both a liberal and conservative multivariable model approach to examine the tradeoffs between statistical power (i.e., increased in the liberal model but this also increased chances of a type 1 error) and a conservative model (i.e., more parsimonious and less likely to be over fit); across both models peak SVI remained significant. However, we acknowledge the limitation of sample size and need for additional external validation with future prospective studies. Finally, we did not have measures of right atrial pressure during nitroprusside infusion which limits our ability to calculate changes in right heart hemodynamics.

Conclusions

In summary, peak SVI during nitroprusside infusion was significantly associated with post-LVAD early RHF. These findings suggest vasodilator testing and acute pharmacologic LV unloading may be useful in risk stratification prior to LVAD implantation in patients with elevated pulmonary pressures and should be prospectively studied.

Abbreviations:

CI

cardiac index

Ea

elastance

MCS-ARC

Mechanical Circulatory Support–Academic Research Consortium

PAWP

pulmonary artery wedge pressure

RHC

right heart catheterization

RHF

right heart failure

RVAD

right ventricular assist device

SVI

stroke volume index