Active and passive cardiac indexes as complementary predictors of outcomes after left ventricular assist device implantation

Patients

This study was approved by the Mayo Clinic Institutional Review Board (#23–008424; approved August 31, 2023) with a waiver of informed consent. This was a retrospective, single-center study conducted at the Mayo Clinic in Rochester, Minn. We included 399 adult patients (aged 18 years or older) who underwent primary continuous-flow LVAD implants between January 2007 and September 2023. Patients were excluded if their index procedure was a LVAD exchange, received devices other than HeartMate 3 (Abbott), HeartMate II, or HeartWare (Medtronic), or did not have pre-operative right heart catheterization data available. Preoperative hemodynamic measurements obtained under temporary mechanical circulatory support (eg, extracorporeal membrane oxygenation and Impella [Abiomed] devices) were not used for this study. Data were collected and managed using Research Electronic Data Capture. Hemodynamic and clinical data definitions followed those established by the Society of Thoracic Surgeons Adult Cardiac Surgery Database.

Hemodynamic Index Calculation

Preoperative hemodynamic data were obtained from right heart catheterizations performed within 30 days before LVAD implantation, or from pulmonary artery catheter recordings acquired before induction of general anesthesia, without concurrent mechanical support. The ActCI and PasCI were calculated as previously described:²⁵

• $\mathrm{A}\mathrm{c}\mathrm{t}\mathrm{C}\mathrm{I}=\left[\right(\mathrm{m}\mathrm{P}\mathrm{A}\mathrm{P}-\mathrm{C}\mathrm{V}\mathrm{P})/\mathrm{m}\mathrm{P}\mathrm{A}\mathrm{P}]\times \mathrm{c}\mathrm{a}\mathrm{r}\mathrm{d}\mathrm{i}\mathrm{a}\mathrm{c}\mathrm{i}\mathrm{n}\mathrm{d}\mathrm{e}\mathrm{x}$

• $\mathrm{P}\mathrm{a}\mathrm{s}\mathrm{C}\mathrm{I}=(\mathrm{C}\mathrm{V}\mathrm{P}/\mathrm{m}\mathrm{P}\mathrm{A}\mathrm{P})\times \mathrm{c}\mathrm{a}\mathrm{r}\mathrm{d}\mathrm{i}\mathrm{a}\mathrm{c}\mathrm{i}\mathrm{n}\mathrm{d}\mathrm{e}\mathrm{x}$

The PAPi was calculated as previously described:²¹

• $\mathrm{P}\mathrm{A}\mathrm{P}\mathrm{i}=(\mathrm{s}\mathrm{y}\mathrm{s}\mathrm{t}\mathrm{o}\mathrm{l}\mathrm{i}\mathrm{c}\mathrm{P}\mathrm{A}\mathrm{P}-\mathrm{d}\mathrm{i}\mathrm{a}\mathrm{s}\mathrm{t}\mathrm{o}\mathrm{l}\mathrm{i}\mathrm{c}\mathrm{P}\mathrm{A}\mathrm{P})/\mathrm{C}\mathrm{V}\mathrm{P}$

The right ventricular stroke work index (RVSWi) was calculated using the standard formula:²⁶

• $\mathrm{R}\mathrm{V}\mathrm{S}\mathrm{W}\mathrm{i}=(\mathrm{m}\mathrm{P}\mathrm{A}\mathrm{P}-\mathrm{C}\mathrm{V}\mathrm{P})\times \mathrm{S}\mathrm{V}\mathrm{I}\times 0.0136$

where $\mathrm{S}\mathrm{V}\mathrm{I}=(\mathrm{c}\mathrm{a}\mathrm{r}\mathrm{d}\mathrm{i}\mathrm{a}\mathrm{c}\mathrm{i}\mathrm{n}\mathrm{d}\mathrm{e}\mathrm{x}\times 1000)/\mathrm{h}\mathrm{e}\mathrm{a}\mathrm{r}\mathrm{t}\mathrm{r}\mathrm{a}\mathrm{t}\mathrm{e}$ and CVP = central venous pressure, mPAP = mean pulmonary artery pressure, PAP = pulmonary artery pressure, and SVI = stroke volume index.

Outcomes and Definitions

The primary outcome was all-cause mortality following LVAD implantation. Secondary outcomes included postoperative RVF, postoperative renal failure, hepatic dysfunction, and respiratory failure. RVF was defined as a central venous pressure >18 mm Hg with a cardiac index <2.0 L/min/m² in the absence of tamponade, ventricular arrhythmias, or pneumothorax. This can manifest as requiring RVAD support, inhaled pulmonary vasodilator (eg, nitric oxide), and/or inotropic therapy for >1 week at any time following LVAD implantation.²⁷ Follow-up was censored at the time of heart transplantation or LVAD explant (without replacement) for myocardial recovery.

Statistical Analysis

Categorical variables are presented as numbers and percentages and were compared using Pearson χ² test. Continuous variables are expressed as medians with interquartile ranges and compared using the Kruskal-Wallis rank-sum test.

Restricted cubic spline Cox proportional hazards models were used to examine the relationship between ActCI, PasCI, PAPi, and RVSWi and postoperative mortality, and to identify optimal spline-derived thresholds for each index. Survival was assessed using Kaplan-Meier analysis with log-rank testing and Cox proportional hazards regression. Both continuous and threshold-based models were evaluated.

Multivariable Cox models were adjusted for pre-specified clinical covariates (detailed in Table E1). Separate models were constructed for long-term survival and RVF. Univariate hazard ratio plots were generated to illustrate the continuous relationship between ActCI/PasCI and mortality risk without adjustment for covariates.

C-statistics with 95% CI were calculated for each index to assess comparative predictive performance. Subgroup analysis was performed in patients implanted with HeartMate 3 devices; however, the small sample size precluded meaningful hypothesis testing. Combined ActCI/PasCI risk stratification was examined by categorizing patients into 4 groups based on high/low values of both indices. All statistical analyses were performed using R software (version 4.4.2) via RStudio (version 2024.12.0 + 467) (R Foundation for Statistical Computing).

Study Population Characteristics

A total of 399 adult patients underwent continuous-flow LVAD implantation at Mayo Clinic between January 2007 and September 2023. The cohort included patients who received HeartMate 3 (9.8%), HeartMate II (69.9%), or Heartware (20.3%) devices. The median follow-up duration was 24.3 months (interquartile range, 9.6–53.1 months).

ActCI and PasCI Thresholds

Spline-based survival analysis identified a survival threshold for ActCI at 0.903 L/min/m² (Figure 1, A), stratifying the cohort into 133 patients with ActCI ≤0.903 L/min/m², and 266 with ActCI >0.903 L/min/m². For PasCI, the survival threshold was identified at 0.778 L/min/m² (Figure 1, B), with 266 patients with PasCI ≤0.778 L/min/m² and 133 patients with PasCI >0.778 L/min/m². These thresholds were used to stratify baseline characteristics, postoperative complications, and survival.

Baseline Characteristics for ActCI Group

Patients with higher ActCI (>0.093) demonstrated overall more favorable clinical and hemodynamic profiles compared with those with lower ActCI (≤0.903) (Table 1). The high ActCI group exhibited better preserved end-organ function, with significantly lower total bilirubin level (1.0 vs 1.3 mg/dL; P < .001), lower aspartate aminotransferase level (31.0 vs 36.0 U/L; P = .006), and higher albumin level (3.9 vs 3.6 g/dL; P < .001) as well as a trend toward lower creatinine level (1.3 vs 1.4 mg/dL; P = .053).

Higher ActCI also reflected superior RV function, including lower rates of moderate-to-severe RV dysfunction (48.5% vs 69.8%; P < .001) and moderate-to-severe tricuspid regurgitation (46.7% vs 66.9%; P < .001). Hemodynamically, higher ActCI was associated with lower right atrial pressure (11.0 vs 20.0 mm Hg; P < .001) and lower pulmonary capillary wedge pressures (22.0 vs 25.0 mm Hg; P < .001) as well as higher cardiac output (3.9 vs 2.9 L/min; P < .001) and cardiac index (2.0 vs 1.4 L/min/m²; P <.001) . Consistent with this, fewer patients in the high ActCI group required preoperative intra-aortic balloon pump support (39.1% vs 57.1%; P < .001), and inotropic support was also less common (63.8% vs 75.8%; P = .016). These findings reinforce that higher ActCI is associated with improved preoperative RV function, reduced venous congestion, and better end-organ perfusion.

Baseline Characteristics Stratified by PasCI

Compared with patients with lower PasCI (≤0.778), those with higher PasCI (>0.778) were significantly younger (median age, 61 vs 64 years; P = .026) and had a higher body mass index (30.0 vs 28.0; P = .026) (Table 2). The high PasCI group also demonstrated signs of worsened end-organ function, including higher creatinine (1.4 vs 1.3 mg/dL; P = .001), higher total bilirubin (1.2 vs 1.0 mg/dL; P = .016), and lower albumin levels (3.7 vs 3.9 g/dL; P = .004).

Higher PasCI was associated with a higher incidence of moderate-to-severe tricuspid regurgitation (65.9% vs 47.3%; P < .001) and moderate-severe mitral regurgitation as well as a trend toward greater moderate/severe RV dysfunction (62.5 vs 52.1%; P = .053) Patients with higher PasCI also had significantly elevated right atrial pressure (19.0 vs 11.0 mm Hg; P < .001), greater mean pulmonary capillary wedge pressure (24.5 vs 22.0 mm Hg; P = .004), and a higher cardiac index (2.0 vs 1.7 L/min/m²; P < .001).

Operative Characteristics Stratified by ActCI and PasCI

LVAD indication status did not significantly differ between groups stratified by ActCI or PasCI (all P values > .05), suggesting that observed outcome differences were unlikely to be driven by initial treatment intent. ActCI groups (Table 3) did not differ in LVAD type or bypass time. However, the low ActCI group had more tricuspid valve surgery (57.9% vs 41.4%; P = .002) and RVAD placement (10.5 vs 3.4%; P = .004) consistent with poorer RV function. Higher ActCI was associated with more aortic valve surgery (18.4 vs 10.5%; P = .041).

Patients with higher PasCI (Table 4) had a greater proportion of HeartMate II implants (81.2 vs 64.3%; P < .001), and a lower incidence of HeartMate 3 implants (1.5 vs 13.9%; P < .001). The high PasCI group also had significantly longer cardiopulmonary bypass times (112.0 vs 96.5 minutes; P = .003). Tricuspid valve surgery was also more frequently performed in the high PasCI group (62.4% vs 39.1%; P < .001).

Postoperative Outcomes Stratified by ActCI and PasCI

Patients with lower ActCI (≤0.903) experienced more postoperative complications (Table 3), including postoperative RVF (11.3% vs 4.9%; P = .018), temporary RVAD support (10.5% vs 4.9%; P = .035), dialysis (27.8% vs 11.7%; P < .001), respiratory failure (33.1% vs 17.4%; P <.001) , hepatic dysfunction (21.1% vs 8.3%; P <.001) , 30-day mortality (9.0% vs 4.1%; P = .048), and in-hospital mortality (14.3% vs 5.6%; P = .004).

Conversely, patients with higher PasCI (>0.778) had more complications (Table 4) such as increased rates of dialysis (28.6% vs 11.3%; P < .001), respiratory failure (28.6% vs 19.5%; P = .042), hepatic dysfunction (21.8% vs 7.9%; P < .001), and gastrointestinal bleeding (44.4% vs 32.3%; P = .019) as well as a trend toward higher in-hospital mortality (12.0 vs 6.8%; P = .076).

Long-term Survival Analysis

Patients with high ActCI had improved survival compared with those with low ActCI (P = .0068) (Figure 2, A). Multivariable analysis confirmed that higher ActCI was associated with a 29% reduction in mortality (hazard ratio, 0.71; 95% CI, 0.53–0.95; P = .019).

In contrast, high PasCI was associated with worse survival compared with low PasCI (P = .001) (Figure 2, B). On multivariable analysis, higher PasCI remained an independent predictor of increased mortality (hazard ratio, 1.39; 95% CI, 1.05–1.85; P = .022).

We further evaluated the combined prognostic utility of ActCI and PasCI by stratifying patients into 4 groups based on spline-derived high/low thresholds for both indices. Kaplan-Meier analysis (Figure E1) showed that Group A (high ActCI + low PasCI) had the most favorable survival, whereas Group C (low ActCI + low PasCI) had the worst. Groups B and D had intermediate outcomes, with convergence of survival curves by 5 years. These findings support the added value of combined ActCI/PasCI stratification for early and midterm risk assessment. Univariate hazard ratio plots demonstrated that higher ActCI was associated with progressively lower mortality risk, whereas higher PasCI was associated with progressively higher risk (Figure E2). The discriminative performance of each index, assessed by C-statistics from multivariable models, including clinical covariates, was similar: ActCI, 0.652; PasCI, 0.656; PAPi, 0.647; and RVSWi, 0.648 (Table E2).

Comparison With Alternative RV Function Indices

Alternative right heart indices, including PAPi and RVSWi, did not demonstrate significant associations with survival. PAPi was available for all 399 patients (mean, 2.4 ± 2.1), with a spline-identified inflection point at 2.5, but this was not predictive of survival (P = .073) (Figure 3, A). Kaplan-Meier curves stratified by this cutoff showed no difference in survival (P = .47) (Figure 3, B), and multivariable Cox analysis showed no significant association (hazard ratio, 1.06; 95% CI, 0.98–1.13; P = .125). Similarly, RVSWi did not show a significant association with survival in univariable or multivariable analyses (Figure E3). In contrast, both ActCI and PasCI remained independently associated with long-term outcomes across continuous and categorical models (Tables E3 and E4).

Study Limitations

Our study is a single-institution, retrospective study with inherent biases such as a small study population. It is also not possible to know the outcomes of alternative management decisions nor can we exclude the influence of patient management changes over the study duration and between individual surgeons. Echocardiographic evaluation of RV contractility was not quantitative and therefore has aspects of subjective visual judgment. Objective measures of RV function (eg, tricuspid annular plane systolic excursion, RV dimension, and RV ejection fraction) were not available. Given the relatively small study population, we were also unable to meaningfully determine the influence of LVAD device type on outcomes. Our PasCI and ActCI thresholds are derived from the durable LVAD candidate population and its utility and threshold for other clinical settings needs to be further defined. Advancements in LVAD technology are also likely to have a major influence on expected outcomes. Intrinsic myocardial biology is expected to influence ventricle function but is not considered in this study examining patient demographic characteristics, comorbidities, and hemodynamics. Further experimental hemodynamic studies in animal models and/or human patients are also needed to clearly define the cardiac output contributed by central venous pressure versus right-sided myocardial contraction. Although we have validated much of the results pertaining to PasCI from the University of Michigan study, these result support further examination of a national and/or international cohort to interrogate the value of the PasCI and ActCI metrics.