Dynamic Assessment of Pulmonary Artery Pulsatility Index Provides Incremental Risk Assessment for Early Right Ventricular Failure After Left Ventricular Assist Device

Abstract

Background:

The pulmonary artery pulsatility index (PAPi) has been studied to predict right ventricular failure (RVF) after left ventricular assist device (LVAD) implantation, but only as a single time point before LVAD implantation. Multiple clinical factors and therapies impact RV function in pre-LVAD patients. Thus, we hypothesized that serial PAPi measurements during cardiac intensive care unit (CICU) optimization before LVAD implantation would provide incremental risk stratification for early RVF after LVAD implantation.

Methods and Results:

Consecutive patients who underwent sequential pulmonary artery catherization with cardiac intensive care optimization before durable LVAD implantation were included. Serial hemodynamics were reviewed retrospectively across the optimization period. The optimal PAPi was defined by the initial PAPi + the PAPi at optimized hemodynamics. RVF was defined as need for a right ventricular assist device or prolonged inotrope use (>14 days postoperatively). Patients with early RVF had significantly lower mean optimal PAPi (3.5 vs 7.5, P < .001) compared with those who did not develop RVF. After adjusting for established risk factors of early RVF after LVAD implantation, the optimal PAPi was independently and incrementally associated with early RVF after LVAD implantation (odds ratio 0.64, 95% confidence interval 0.532−0.765, P < .0001).

Conclusions:

Optimal PAPi achieved during medical optimization before LVAD implantation provides independent and incremental risk stratification for early RVF, likely identifying dynamic RV reserve.

Durable continuous flow left ventricular assist devices (LVAD) are recommended to improve survival and quality of life for patients with advanced heart failure (HF) refractory to standard medical therapies.¹ Early right ventricular failure (RVF) remains a leading cause of premature morbidity and mortality after LVAD implantation.² However, the poor prognosis of early RVF after LVAD implantation can be attenuated in patients with planned placement of temporary right ventricle (RV) mechanical circulatory support (MCS) alongside durable LVAD.^3–9 Therefore, an accurate assessment of the risk of RVF in the early postoperative LVAD period is critical.

RVF after LVAD implantation is challenging to predict and current predictive tools continue to fall short of reliable risk stratification before LVAD implantation.^7,10,11 Several issues compound accurate RV function assessment in this population. First, RV function is highly dependent on loading conditions and left ventricular (LV) contractile function, which may fluctuate over the course of the assessment period. Second, after LVAD implantation RV function also depends on the RV having an adequate reserve to meet the significant intraoperative and perioperative demands associated with LVAD placement, including acute blood loss, hypoxemia, ischemia, and unfavorable ventricular geometry based on septal shift, in addition to increased workload demands from the mechanically supported LV.^10,12–17 Therefore, a significant limitation of existing static hemodynamic and imaging parameters used for risk prediction is that they are derived based on measurements at a single point in time.^2,18–27

The pulmonary artery pulsatility index (PAPi) is a validated hemodynamic parameter associated with early RVF in LVAD populations.^28,29 Although it has shown promise for predictive superiority over other commonly used hemodynamic parameters in initial studies, its clinical relevance to date has been limited by only being studied in static assessment study designs. It remains unclear whether a change in a hemodynamic parameter throughout a period of hemodynamic optimization has incremental predictive capabilities when compared with other parameters measured at single time points.^{14,19,20,25–27} We hypothesized that quantifying RV reserve through magnitude of change in serial invasive hemodynamics throughout specialist hemodynamic optimization would provide incremental risk stratification for post-LVAD RVF.

Methods

Study Population and Design

Consecutive patients from April 2008 to June 2017 at our single-center, Cleveland Clinic Foundation (Cleveland, Ohio), undergoing primary continuous flow LVAD implantation, either as destination therapy or a bridge to transplantation, were retrospectively included in the analysis. The institutional practice is to use pulmonary artery catheterization (PAC)-guided optimization for patients being considered for LVAD who have high-risk initial hemodynamics. This cohort was defined by patients with initial suboptimal and/or unstable invasive hemodynamics requiring admission to the cardiac intensive care unit (CICU) for PAC-guided medical optimization led by a HF specialist before durable LVAD implantation.

Laboratory and echocardiographic data were collected within 24 hours and 60 days before LVAD implant, respectively. The primary outcome was early RVF and was defined using components of the Interagency Registry for Mechanically Assisted Circulatory Support (INTERMACS) definition³⁰: postoperative inotrope use for more than 14 days after LVAD implantation and/or unplanned RVAD insertion within the first 30 days after LVAD implantation. A secondary outcome was death within the first 180 days after LVAD implantation, uncensored for transplant in bridge to transplant LVAD.

The Michigan RVF risk score (RVFRS) was calculated as previously described.²⁶ Patients were excluded if they had undergone a device exchange, were under 18 years of age or were undergoing HeartMate 3 (HMIII) implantation. HMIII patients were excluded given the time frame of this study was before clearance of HMIII for use by the US Food and Drug Administration.

Patients who did not have a CICU admission with at least 2 separate hemodynamics within 30 days of LVAD were excluded. All sequential hemodynamic data were obtained throughout the index hospitalization before LVAD placement. Hemodynamics were collected for analysis while a patient was on nondurable MCS devices that did not directly off load the RV; in our cohort, this included the Impella 5.0, Impella CP, and intra-aortic balloon pump devices, but not while on extracorporeal membrane oxygenation or Tandem Heart. The Cleveland Clinic Institutional Review Board approved the study.

Assessment of Serial Hemodynamics

Hemodynamic waveforms were validated at least every 24 hours by a HF cardiologist throughout the optimization period. All validated hemodynamic data available during PAC-guided optimization for the index hospitalization before LVAD placement were analyzed.

Wave forms were measured at end expiration. Mixed central venous blood gas was collected from the tip of the catheter in the pulmonary artery and cardiac output was estimated using Fick’s equation and indexed to body surface area to determine cardiac index.

Optimization of a patient’s clinical state was then performed under the guidance of an attending HF cardiologist using management, including diuretics, intravenous sodium nitroprusside and nitroglycerin, inotropes (milrinone and dobutamine), and nondurable MCS devices as deemed necessary to obtain an adequate cardiac output and optimal intracardiac filling pressures. Our HF CICU defines a cardiac index of more than 2.2, right atrial pressure of less than 10, and a pulmonary capillary wedge pressure of less than 20 optimal hemodynamics during PAC-guided optimization. In parallel with targeting these hemodynamic goals during optimization, there is careful attention to the prevention of hypotension (mean arterial pressure of <65 mm Hg) as well as worsening renal function. Therapy choice and medication titration based on hemodynamics and clinical parameters were ultimately at the discretion of the attending HF cardiologist.

Hemodynamic Calculations and Definitions

PAPi was measured as previously defined³¹: pulmonary artery systolic pressure minus pulmonary artery diastolic pressure, divided by the mean right atrial pressure or central venous pressure. The right atrial/pulmonary capillary wedge pressure was defined as the mean right atrial pressure or central venous pressure divided by the mean pulmonary capillary wedge pressure.

Hemodynamics were first recorded at the start of PAC-guided optimization before initiating CICU therapies. Serial hemodynamics were then analyzed for achievement of the patient’s most optimized hemodynamic state as evident by cardiac index and intracardiac filling pressure targets as defined elsewhere in this article or as close to these targets as possible if unable to achieve the hemodynamic goals. Hemodynamics were captured at this optimized state, including PAPi, termed “optimized PAPi.” The change in PAPi from initial hemodynamics to hemodynamics at their most optimized state as patients underwent PAC-guided optimization, was termed “delta PAPi.”

Statistical Methods

Continuous variables were expressed as mean ± standard deviation. The Student t test or Mann–Whitney U test were used to compare parametric and independent nonparametric continuous variables, respectively. Categorical variables were expressed as percentage (%) with comparisons via Fisher’s exact test or the χ² method. A clinical-based model was then constructed to best predict early RVF. Variables that were both statistically and clinically associated with the primary outcome of early RVF after LVAD insertion were considered for inclusion in the multivariable logistic regression model. Variables that reached significance at multivariate level (P < .05 based on Wald χ²) were then entered into the optimized model with the best R-square and C-statistic. To check for multicollinearity, the Pearson correlation coefficients between covariates were estimated and hemodynamic variables with a correlation coefficient or greater than 0.5 were each kept separate during the modeling process. Finally, the optimal cut off for optimal PAPi and delta PAPi that gave the highest combined sensitivity and specificity was determined by the maximum Youden Index. Primary and secondary outcomes were then further analyzed based off these optimal cut offs. Receiver operating characteristic curves were used to compare predictive capabilities of optimal PAPi vs delta PAPi.

All statistical analyses were carried out using SAS version 9.3 and SigmaPlot version 11.0.

Results

A total of 397 consecutive patients underwent LVAD implantation at the Cleveland Clinic between 2008 and 2017. Most patients (80%) had their LVAD placed after 2010. After the exclusion of those who did not meet inclusion criteria (Supplementary Fig. 1), the final cohort comprised 315 patients (mean age 56 ± 12.4 years, 18.4% female, 85.7% INTERMACS profile I−III, 50.3% destination therapy).

Baseline and Perioperative Characteristics

Baseline demographic, clinical and echocardiographic characteristics of the study cohort stratified according to primary outcome of early RVF (n = 70) or no early RVF (n = 245) are provided in Tables 1 and 2.

Table 1.

Patient Characteristics at Baseline

	Early RVF
Characteristic	No (n = 245)	Yes (n = 70)	P Value
Age, years	56.4 ± 11.7	54.7 ± 14.5	.803
Male sex	201 (82.0)	52 (74.3)	.150
Body mass index, kg/m2	29.0 ± 5.3	29.2 ± 4.9	.670
INTERMACS			.005
I	27 (11.0)	18 (25.7)
II	103 (42.0)	31 (44.3)
III	73 (29.8)	18 (25.7)
IV–VII	42 (17.2)	3 (4.3)
Michigan risk score			<.001
≤3.0	217 (88.6)	51 (72.9)
3.0–5.5	16 (6.5)	5 (7.1)
≥5.5	12 (4.9)	14 (20.0)
LVAD type			.887
HeartMate II	187 (77.5)	54 (22.4)
Heartware	58 (78.3)	16 (21.6)
LVAD classification			.066
DT	116 (47.5)	42 (60.0)
BTT	128 (52.5)	28 (40.0)
Hypertension	124 (50.6)	31 (44.3)	.350
Diabetes mellitus	91 (37.1)	30 (42.9)	.386
Chronic kidney disease	87 (35.5)	19 (27.1)	.191
Atrial fibrillation	115 (46.9)	34 (48.6)	.809
Peripheral vascular disease	17 (6.9)	2 (2.9)	.265
Hyperlipidemia	113 (46.1)	30 (42.9)	.628
Cardiac resynchronization therapy	96 (39.2)	32 (45.7)	.327
Cerebrovascular accident	31 (12.7)	5 (7.1)	.201
Chronic lung disease	67 (27.3)	16 (22.9)	.452
Prior sternotomy	60 (26.4)	18 (25.7)	.87
Left ventricular ejection fraction, %	16.8 ± 5.9	16.2 ± 5.4	.337
LVIDd, cm	6.9 ± 1.2	7.0 ± 1.1	.443
RV function by ECHO parameters			.020
Normal	30 (12.2)	11 (15.7)
Mild dysfunction	65 (26.5)	16 (22.9)
Moderate dysfunction	89 (36.3)	16 (22.9)
Moderate to severe dysfunction	46 (18.8)	15 (21.4)
Severe dysfunction	15 (6.1)	12 (17.1)
Degree of tricuspid regurgitation			.031
None	35 (14.3)	8 (11.8)
Mild	89 (36.3)	20 (29.4)
Moderate	74 (30.2)	16 (23.5)
Moderate to severe	27 (11.0)	18 (26.5)
Severe	20 (8.2)	6 (8.8)

Values are mean ± standard deviation or number (%).

BTT, bridge to transplantation; DT, destination therapy; INTERMACS, Interagency Registry for Mechanically Assisted Circulatory Support; LVAD, left ventricular assist device; LVIDd, left ventricular end-diastolic diameter; RV, right ventricular; RVF, right ventricular failure.

Table 2.

Preoperative Characteristics and Laboratory Parameters Within 24 Hours Before LVAD Implantation

	Early RVF		P Value
Characteristic	No (n = 245)	Yes (n = 70)
Inotropes	193 (78.8)	63 (90.0)	.034
Vasopressors	23 (9.4)	16 (22.9)	.003
Intubation	24 (9.8)	12 (17.1)	.088
MCS*	79 (32.2)	28 (40.0)	.227
Cardiac arrest during index stay	10 (4.1)	4 (5.7)	.521
Creatinine, mg/dL	1.3 ± 0.57	1.5 ± 0.5	.001
Blood sodium, mEq/L	133.6 ± 4.8	132.0 ± 5.7	.020
BUN, mg/dL	26.3 ± 15.1	34.7 ± 19.2	<.001
ALT, U/L	53.3 ± 118.0	86.4 ± 208.0	.009
AST, U/L	41.3 ± 58.1	64.6 ± 115.9	<.001
Total bilirubin, mg/dL	1.2 ± 0.9	2.2 ± 2.2	<.001
INR	1.1 ± 0.1	1.2 ± 0.1	.007
Albumin, g/dL	3.5 ± 0.5	3.4 ± 0.5	.310
Hemoglobin, g/dL	11.1 ± 1.9	10.9 ± 2.2	.429
White blood cell count, 109/L	8.4 ± 3.4	9.4 ± 4.3	.022

^*Within the 107 patients with MCS, there were 98 intra-aortic balloon pumps and 9 Impella CP or 5.0 devices. Values are number (%) or mean ± standard deviation. ALT, alanine aminotransferase; AST, aspartate aminotransferase; BUN, blood urea nitrogen; CICU, cardiac intensive care unit; INR, international normalized ratio; MCS, mechanical circulatory support. Other abbreviations as in Table 1.

There were no significant differences in age, cardiovascular comorbidities, implant indication (destination therapy vs bridge to transplant), or number of prior sternotomies. Patients who developed RVF had a significantly lower INTERMACS profile at the time of LVAD (53% INTERMACS profile I−II in the RVF group vs 70% in the RVF group, P = .005) and were treated with preoperative inotropes (90% vs 78.8%, P = .034) or vasopressors (22.9% vs 9.4%, P = .003) significantly more often. For each group, the majority (>90%) of nondurable MCS was with an intra-aortic balloon pump. There was no difference in the nondurable MCS use between each group. RVF patients also more frequently displayed severe RV dysfunction grade by visual assessment with a higher tricuspid regurgitation grade on preimplant transthoracic echocardiography. Patients that developed RVF also had significantly worse renal and hepatic laboratory profiles (Table 2). Those who developed RVF had a significantly higher total RVFRS (2.5 vs 1.0, P < .001) as well as a larger percentage with a RVFRS of 5.5 or higher (20.0 vs 4.9, P < .001) compared with those who did not develop RVF. There was not a significant difference in RVF between centrifugal and axial flow pumps (Supplementary Fig. 2).

Pre-LVAD Implantation Hemodynamics

Table 3 compares most optimized invasive hemodynamics before LVAD implantation across the RVF and no RVF groups. Patients who developed RVF had significantly higher biventricular filling pressures and significantly lower cardiac indices after optimization compared with those without RVF. This outcome was despite a significantly greater use of inotropes for optimization in the RVF group and no difference in nondurable MCS use between the groups.

Table 3.

Preoperative Hemodynamic Parameters at the Time of Most Optimized Hemodynamics, Stratified by Early RVF Outcomes

	Early RVF
Hemodynamics	No (n = 245)	Yes (n = 70)	P Value
RA pressure, mm Hg	5.3 ± 3.39	9.1 ± 5.18	<.001
PA systolic pressure, mm Hg	47.6 ± 13.47	46.4 ± 11.21	.533
PA diastolic pressure, mm Hg	20.1 ± 6.61	22.3 ± 6.25	.018
Mean PA pressure	30.6 ± 8.57	31.4 ± 7.67	.640
PCWP, mm Hg	17.4 ± 5.48	19.6 ± 6.88	.073
Cardiac index, L min−1 m−2	2.5 ± 0.57	2.2 ± 0.48	.006
RA/PCWP	0.3 ± 0.17	0.5 ± 0.22	<.001
Optimal PAPi	7.5 ± 5.46	3.5 ± 2.31	<.001
Initial PAPi	2.0 ± 1.14	1.4 ± 0.77	<.001
Delta PAPi	5.5 ± 5.20	2.1 ± 2.09	<.001
Delta RA/PCWP	0.2 ± 0.24	0.16 ± 0.21	.054

Values are mean ± standard deviation.

The PA systolic pressure, PA diastolic pressure, mean PA pressure, PCWP, and cardiac index were measured at the time of most optimized hemodynamics.

RA, right atrium; RA pressure; PA, pulmonary artery; PAPi, pulmonary artery pulsatility index; PCWP, pulmonary capillary wedge pressure. Other abbreviations as in Tables 1 and 2.

The entire cohort started PAC-guided therapy an average 9.5 days before LVAD implantation, when the initial hemodynamics were taken and had an initial mean PAPi of 1.87 (Supplementary Fig. 3 and Supplementary Table 1). PAC-guided therapy achieved hemodynamic optimization, on average 5 days from the initial hemodynamics (Supplementary Table 2). There was significant improvement in hemodynamics in both the no RVF and early RVF groups. For the entire cohort, the mean optimized PAPi was 6.6 ± 5.20.

Patients in the no RVF group were able to increase their PAPi with optimization on average more than twice as much as the RVF group (delta PAPi 5.5 vs 2.1, P < .001), giving them an optimal PAPi over twice as large (7.5 vs 3.5, P < .001). This magnitude of improvement occurred despite the no RVF group starting with a higher initial PAPi then the RVF group (2.0 vs 1.4, P < .001).

There was no difference between the number of hemodynamics analyzed between each group and no difference in the number of days between the initial and optimized hemodynamics for each group (Supplementary Fig. 3). Finally, there was no significant correlation between the number of hemodynamic data points observed and the level of either the change in PAPi (delta PAPi) or the PAPi at optimization. (Supplementary Fig. 3).

Change in PAPi According to Specific Pre-LVAD Optimization Intervention

Assessment of change in PAPi was analyzed by specific intervention to improve hemodynamics (Fig. 1). The delta PAPi was significantly after the following initiation of inotropes and nondurable MCS in the non-early RVF group compared with the early RVF group. Finally, when analyzing only the patients who started with what is currently considered a high-risk PAPi (<2.0),²⁹ the no RVF group had a significantly higher delta PAPi than the early RVF group. This outcome was despite starting with a higher initial PAPi.

Fig. 1.

Delta PAPi by intervention type. *One hundred ninety-one patients had an original PAPi without inotropes and went on to have optimized hemodynamics while on inotropes. Significantly higher delta PAPi in the no RVF group (6 ± 5.47) vs 1.7 ± 1.65) in the RVF group. **P < .0001. ƚ Eighty-eight patients had an original PAPi without nondurable MCS and went on to have optimized hemodynamics while on MCS. There was a significantly higher mean delta PAPi in the no RVF group (5.67 ± 4.43 vs 1.8 ± 1.9 in the RVF group). ǂ Two hundred one patients had an original PAPi of less than 2.0 and underwent optimization. The mean delta PAPi was significantly higher in the no RVF group than the RVF group (5.2 ± 4.78 vs 2.0 ± 1.8). MCS, mechanical circulatory support; PAPi, pulmonary artery pulsatility index; RVF, right ventricular failure.

Primary and Secondary Outcomes

The primary outcome of RVF occurred in 22.2% of the patient population. The majority (n = 52, 74% of patients with RVF (17% of the overall cohort) met the criteria for RVF based on prolonged inotrope use (Supplementary Fig. 4). For the entire cohort, the 180-day mortality rate was 13%. Those who developed RVF had a significantly higher 180-day post-LVAD mortality rate compared with those who did not develop RVF (30% vs 8.2%, P < .0001) (Supplementary Fig. 4) When the primary outcome was analyzed by determinant of RVF, those with prolonged inotrope use had a similar mortality rate as those who underwent an unplanned RVAD implantation. (32.7% vs 22.2%, P = .55) (Supplementary Fig. 4).

Serial PAPi Assessment and Outcomes

Table 4 illustrates the optimal multivariate logistic regression model for prediction of RVF after LVAD implantation, with a total of 12 variables meeting both the statistical and the clinical criteria considered for inclusion. The delta PAPi and optimized PAPi had significant multicollinearity based on a Pearson correlation coefficient of more than 0.5 and were therefore entered separately into the clinically based model. The delta PAPi and the optimized PAPi each were found to be independently associated with the primary outcome. The optimized PAPi was more predictive than the delta PAPi (area under the curve of 0.83 optimal PAPi vs area under the curve of 0.81 delta PAPi, P = .028) (Supplementary Fig. 5). We, therefore, used the optimal PAPi in our final model.

Table 4.

Optimized Multivariate Logistic Model to Predict Early RVF After LVAD Implantation

Variables In This Model* (R-Square: 0.2829 And C-Statistic: 0.871)	Wald Chi-Square	Odds Ratio (95% CI)	P value
Sex	5.7	3.071 (1.227–7.686)	.0166
INTERMACS level	3.9	0.579 (0.338–0.990)	.0458
Total Michigan Score	4.8	1.214 (1.021–1.443)	.0045
Fick CI ƚ	8.1	0.250 (0.096–0.650)	.00788
Optimal PAPi	23.4	0.638 (0.532–0.765)	<.0001

CI, cardiac index. Other abbreviations as in Tables 1 and 3.

^*Considered for multivariate logistic model that did not meet significance: RA pressure, initial PAPi, RA/PCWP, myocardial oxygen consumption, inotrope at time of optimized PAPi, MCS at time of optimized PAPi, echo severity of RV, and tricuspid regurgitation degree on echo.ƚ The Fick CI is at the time of most optimized hemodynamics.

The final optimized model included the optimal PAPi, sex, INTERMACS profile, RVFRS, and Fick cardiac index at time of optimized hemodynamics as variables that are independently associated with RVF after LVAD implantation (Table 4). Based on this model, for each unit increment increase in the optimal PAPi after following optimization, the risk of RVF post LVAD decreased by 36%.

Fig. 2 provides a graphical summary of the results of the study, using the likelihood ratio test for a comparison of the models to illustrate the incremental value of incorporating serial measurement of right heart hemodynamics (optimal PAPi) into an overall risk prediction model for the development of early RVF after LVAD implantation.

Fig. 2.

Incremental value of optimal PAPi in likelihood ratio test. Likelihood ratio test for the prediction of early RVF after LVAD. Optimal PAPi shows incremental value to predict early RVF when assessed on top of conventional clinical, echocardiographic and standard hemodynamic parameters. Standard echo parameters include RV function grade and severity of TR. The diagram shows a proposed flow using the incremental value of predictors of early RVF based on the Likelihood ratio test.

CI, cardiac index; INTERMACS, Interagency Registry for Mechanically Assisted Circulatory Support; LVAD, left ventricular assist device; PAPi, pulmonary artery pulsatility index; RV, right ventricle; RVF, right ventricular failure; TR, tricuspid regurgitation.

Incorporating the optimal PAPi, on top of the standard clinical, echocardiographic, and hemodynamic variables currently integrated in standard risk assessment, provides an incremental risk assessment for this important outcome of RVF after LVAD implantation.

The optimal cut off value for the absolute value of PAPi after incremental PAC-guided therapy (optimal PAPi) was greater than 3.33. This value was determined by using the intersection of the curves methods that provided the greatest sensitivity and specificity for developing early RVF. Using the same methods, we provided a further analysis of the predictive capabilities of an improvement in RV hemodynamics with PAC-guided optimization, and established that the optimal delta PAPi was greater than 2.08. Thus, in patients with initial poor hemodynamics, increasing one’s PAPi by more than 2 with PAC optimization in the CICU, to a level of greater than 3.3, provides predictive capabilities of both early RVF (optimal PAPi and delta PAPi) in addition to 180-day mortality after LVAD implantation (delta PAPi). (Table 5, Fig. 3, and Supplementary Table 3).

Table 5.

Primary and Secondary Outcomes by Optimal PAPi Groups

		Optimal PAPi Group*
Outcomes, n (%)	Total (N = 315)	>3.33 (n = 236)	≤3.33 (n = 79)	P-value
The primary outcome: Early RVF after LVAD implantation	70 (22.2)	26 (11)	44 (55.7)	<.001
The secondary outcome: Death after LVAD implantation	41 (13.0)	26 (11)	15 (19.0)	.068

Abbreviations as in Tables 2 and 3.

^*The optimal cut off determined with the maximum Youden index and gave a sensitivity of 0.66 and a specificity of 0.85, respectively.

Fig. 3.

Six-month mortality based on optimal cut off for delta PAPi. Time to event analysis at 180 days, outcome is death with cumulative incidences estimated with the Kaplan–Meier method. The P value is generated from the log-rank test. PAPi, pulmonary artery pulsatility index

Fig. 3 shows the Kaplan−Meier curve highlighting the significant difference in death based on the optimal cut off of the delta PAPi 2.08.

Discussion

The present analysis establishes several key findings. This report is the first on degree of change in serial invasive hemodynamics providing risk stratification for the development of early RVF after LVAD implantation. The novel concept of optimal PAPi provided independent and incremental risk stratification to previously validated predictors of early RVF. The optimal PAPi has established the importance of quantifying hemodynamics serially through PAC-guided therapy in patients deemed at high risk for early RVF based on current parameters including static hemodynamic assessment. On average, patients reached their optimal PAPi after only 5 days of PAC-guided optimization, highlighting the clinical reproducibility of this novel parameter. Finally, we have provided insight into the interpretation and usefulness of sequential hemodynamic assessments in an intensive care setting in significantly decompensated patients before LVAD implantation, with initial hemodynamics consistent with cardiogenic shock, and a sizeable burden of INTERMACS profile 1 and 2 patients relative to INTERMACS registry averages.²² This patient groups is one of the most challenging to predict early RVF.^33–35 This study provides previously lacking reference values for such high-risk patients to aid the interpretation of advanced hemodynamics such as PAPi while being supported by advanced medical and temporary MCS therapies.

Static interpretation of PAPi has been associated with superior predictive capabilities to predict early RVF over other commonly used static hemodynamic parameters. However, these studies derived their suggested optimal hemodynamic cut off values in patient populations that had relatively high filling pressures, had a minority of patients on inotropic or nondurable MCS support at time of hemodynamic monitoring, and either did not report INTERMACS profile or had a relatively higher INTERMACS profile going into LVAD.^28,29,37 Morine et al has established a static PAPi value of 1.85 as constituting a cut off for having a “good PAPi,” which has been published into guidelines and reviews.¹⁰ Notably, in our higher risk, sicker cohort, 82% of patients who developed early RVF had an optimized PAPi that was greater than 1.85, with an average value of 3.5. This observation questions the predictive value of the currently accepted PAPi cut off to patients requiring inpatient hemodynamic optimization before LVAD implantation. The discongruency in PAPi values between our study and prior studies is intriguing and important. Despite our relatively higher risk cohort for the development of early RVF, we had significantly higher optimal PAPi values than previously reported static PAPi values.^28,29,37

It needs to be emphasized that the optimal PAPi is acquired sequentially while undergoing PAC-guided optimization in a CICU. We believe the relatively high PAPi values across our cohort to be a function of the nature in which the optimal PAPi is obtained. The elevated absolute value of the optimized PAPi level in our cohort, relative to prior publications, is a testament to the novelty of optimal PAPi. We are the first to capture sequential hemodynamic changes in a cohort of patients with decompensated advanced HF as they undergo intensive CICU optimization before LVAD implantation. The application of the magnitude of these changes to risk of early RVF after LVAD placement has high clinical relevance to this cohort at significant risk for early RVF. The initial hemodynamics of most of our cohort had poor biventricular hemodynamics. HF centers caring for such high-risk patients are unlikely to have the luxury of time and a decision on a patient’s candidacy for definitive advanced therapies is often required during that same hospitalization. An imperative part of this decision will be an accurate assessment of the likelihood of developing early RVF after LVAD implantation in these high-risk patients. For the first time we have provided a reference for what constitutes an appropriate increase in PAPi, to a goal PAPi after hemodynamic optimization. Patients who increase their PAPi to a level greater than 3.33 with intensive, specialist-driven, hemodynamic optimization seem to exhibit adequate RV reserve and have a significantly lower risk of developing early RVF. Thus, the inclusion of optimal PAPi in the current arsenal of established predictive variables of early RVF after LVAD placement will arm HF providers with a stronger risk/benefit analysis during the LVAD candidacy process. This development has the potential to improve clinical decision-making on when to pursue temporary RV support upfront with durable LVAD or when to declare the patient at prohibitive risk for an LVAD, thereby leaving heart transplantation or palliative-focused care as the remaining options. Of course, further studies ideally using multicenter and prospective designs will be required to validate the use of serial hemodynamics including optimal PAPi to guide the advanced HF therapies decision tree.

The optimal PAPi was superior to delta PAPi in predicting the primary outcome. When considering changes in hemodynamics and predicting outcomes, optimal PAPi captures the best hemodynamic achieved and intrinsically controls for patients with initial favorable RV hemodynamics. Such patients are likely to have less room to improve their hemodynamics with optimization, potentially leading to a relatively low delta PAPi that does not as accurately capture the true assessment of their hemodynamic RV function. However, the independent and incremental predictive capabilities of delta PAPi in this study suggests it still has significant prognostic value when applied to high-risk cohorts with poor starting hemodynamics. During CICU optimization, the inability to increase one’s PAPi by more than 2 was associated with an increased risk in mortality and is likely a signal for lack of RV reserve.

Notably, the average initial PAPi of the entire cohort was 1.87. Although the initial PAPi was predictive and should still be considered useful, the concept of change in PAPi through CICU PAC optimization to achieve optimal PAPi captures the dynamic nature of RV reserve and yields better predictive value for RVF than traditional PAPi alone. Significantly, our subanalysis in patients starting with a conventional low PAPi value of less than 2.0 showed patients who were able to augment their PAPi with optimization were in fact not at high risk of developing post-LVAD RVF. This suggests that original high-risk hemodynamics require reassessment for true risk of early RVF after LVAD implantation. This study reveals that PAPi levels are sensitive to both intravascular congestion and depressed LV function. Thus, only through serial hemodynamic assessment can one ensure that both filling pressures and LV cardiac output are truly optimized. This is likely a major reason that serial hemodynamics have a superior predictive capability to static hemodynamics. PAC-guided optimization essentially unmasks PAPi values that are low owing to congestion or significant LV dysfunction rather than true RV compromise. Those PAPi values that failed to increase significantly declared themselves as truly decompensated RV’s independent of volume status, afterload, or LV function. It needs to be emphasized that our results are pertinent to relatively high-risk patients before LVAD implantation. In addition, owing to our institution’s practice of favoring early nondurable MCS in cardiogenic shock, our cohort had a relatively large portion of patients in each group receiving this support for optimization before LVAD implantation. Other centers may not have the same institutional practice and should consider this factor when interpreting our results to their patients before LVAD implantation.

The predictive capabilities of optimal PAPi were not studied in low-risk patients not requiring CICU optimization. This finding is in part why the RVFRS was selected over other validated RVF risk scores, for comparison with the optimal PAPi. The RVFRS was derived from relatively sick patients going into LVAD placement.²⁶ Furthermore, novel to other risk scores, Matthews et al²⁶ found clinical consequences of RV dysfunction rather than direct measurements of RV dysfunction, to be the most predictive of early RVF after LVAD implantation. The RVFRS therefore captures how sick a patient is going into LVAD therapy in a more granular manner than the INTERMACS profile. Given our relatively high-risk cohort, it was of high clinical relevance to assess for optimal PAPi to have incremental prediction of early RVF over the RVFRS.

Limitations

Several limitations require acknowledgement. This was a retrospective study from a single center that included a preselected cohort of patients deemed eligible for LVAD implantation. Although the presence of vasoactive infusion is noted, the specific types of vasoactive infusions or their doses at the time of hemodynamic measurements are not available. Second, these results have yet to be validated in another pre-LVAD cohort. Third, these results do not include patients implanted with the newest generation centrifugal device based on full magnetic levitation technology. However, the 2-year results reported with this device show a persistent significant incidence of post-device RVF similar to earlier generation device cohorts.^32,36 Also, our cohort had no difference in the rate of early RVF between centrifugal vs axial flow devices. In addition, all currently used hemodynamic parameters to predict early RVF after LVAD placement were derived in cohorts either before HMIII clinical use or in cohorts with a very small percentage of HMIII devices.^{20,21,23,28,29,37} We therefore believe the results of this study, remain applicable to HMIII devices in clinical practice. Finally, although more than 80% of patients had LVAD implants after 2010, our study still spans a relatively large time frame it pertains to device and management strategies evolution. We cannot exclude inevitable advancement in knowledge of both pre- and postmanagement of LVADs within both the advanced HF field at large and within our institution. However, the 2 principal devices in use during this period were both represented, and the study inclusion period took place across a time frame when use of both was well-established potentially minimizing any era effect.

Conclusions

This study provides significant new insight into the risk prediction of early RVF after LVAD implantation. In a large continuous flow durable LVAD population, serial hemodynamic assessment using PAPi provided incremental risk prediction to other validated clinical and hemodynamic parameters, likely through its more sensitive identification of dynamic RV reserve. These findings will enhance the pre-LVAD risk assessment particularly in higher risk patient populations and should be validated in a prospective, ideal multicenter clinical trial setting.