The Association of the Pulmonary Artery Pulsatility Index and Right Ventricular Function after Cardiac Surgery

Background The pulmonary artery pulsatility index (PAPi) has been shown to correlate with right ventricular (RV) failure in patients with cardiac disease. However, the association of PAPi with right ventricular function following cardiac surgery is not yet established. Methods PAPi and other hemodynamic variables were obtained postoperatively for 959 adult patients undergoing cardiac surgery. The association of post-bypass right ventricular function and other clinical factors to PAPi was evaluated using linear regression. A propensity-score matched cohort for PAPi ≥ 2.00 was used to assess the association of PAPi with postoperative outcomes. Results 156 patients (16.3%) had post-bypass right ventricular dysfunction defined by visualization on transesophageal echocardiography. There was no difference in postoperative PAPi based on right ventricular function (2.12 vs. 2.00, p=0.21). In our matched cohort (n = 636), PAPi < 2.00 was associated with increased incidence of acute kidney injury (23.0% vs 13.2%, p < 0.01) and ventilator time (6.0 hours vs 5.6 hours, p=0.04) but not with 30-day mortality or intensive care unit length of stay. Conclusion In a general cohort of patients undergoing cardiac surgery, postoperative PAPi was not associated with postcardiopulmonary bypass right ventricular dysfunction. A postoperative PAPi < 2 may be associated with acute kidney injury.


Introduction
Right ventricular (RV) dysfunction is associated with poor perioperative outcomes in cardiac surgical patients, and early and accurate diagnosis of this pathology is essential for optimal management to alleviate these associated risks [1][2][3].However, continuous monitoring of RV function in the intensive care unit (ICU) presents unique challenges compared to the intraoperative setting, as conventional diagnostic and monitoring modalities have struggled to adequately balance the sensitivity and feasibility of use [4][5][6].
Te pulmonary artery pulsatility index (PAPi) is a novel hemodynamic parameter of right ventricular function gaining momentum in the assessment of cardiac failure.Defned as the ratio of the pulmonary arterial pulse pressure and the right atrial pressure, PAPi can ofer rapid insight regarding both the contractility of the right heart and its flling pressures, with a higher PAPi value typically indicating better RV function [7][8][9].Consequently, preoperative PAPi has been associated with RV failure following left ventricular assist device (LVAD) implantation and heart transplantation along with other perioperative morbidity such as acute kidney injury [10][11][12][13].Although an optimal range for PAPi has not yet been determined, values less than 2.0 have typically been associated with poor outcomes following cardiac surgery [9,[12][13][14].However, the prognostic utility of postoperative PAPi in general cardiac surgical patients as well as its relationship with postoperative RV function is presently unknown.
Tus, the primary aim of this study was to identify the relationship between postoperative PAPi with right ventricular function.Te secondary aim was to identify associations between postoperative PAPi and renal injury, mortality, ventilation time, and ICU length of stay.

Patients and Methods
2.1.Patient Data.Tis was a secondary analysis of a singlecenter retrospective observational cohort study of adult patients who underwent cardiac surgery between January 1, 2017, and December 31, 2019 [15].Our study was approved by the institutional review board, and patient consent was waived due to the retrospective nature of our study.Adult patients undergoing cardiac surgical procedures with cardiopulmonary bypass with general endotracheal anesthesia and a pulmonary artery (PA) catheter were included in the study.Exclusion criteria included missing PA hemodynamic data and echocardiographic evaluation of right ventricular function (Supplementary Figure 1).As the original study investigated the association between postoperative PAPi and renal injury, patients who were on hemodialysis preoperatively or those with missing serum creatinine data were also excluded [15].For duplicate patients, only the index procedure was included in the cohort.Te intraoperative anesthetic care and postoperative ICU management of patients were not protocolized but followed routine institutional standard practice [16].

Data Collection.
All necessary demographic, clinical, imaging, pharmacologic, and hemodynamic data were acquired from the electronic medical record system (Epic Systems, Verona, WI).Invasive central venous pressure (CVP), PA pressure, mean arterial pressure (MAP), and cardiac index measurements were collected hourly for either the frst 48 hours of postoperative care or until PA catheter removal, whichever was shorter, and an average value was obtained.PAPi was calculated as [(PA systolic pressure -PA diastolic pressure)/CVP], where CVP was used as a surrogate for right atrial pressure [7].RV evaluation from preoperative and intraoperative echocardiograms was recorded and categorized as with or without the presence of dysfunction.Postoperative RV function was defned as postcardiopulmonary bypass transesophageal (TEE) documentation of RV function.
Selective medications administered in the ICU for the duration of their ICU stay were identifed and categorized as vasopressors (norepinephrine, vasopressin, and phenylephrine), inotropes (epinephrine, dobutamine, and milrinone), vasodilators (nitroglycerin, nicardipine, and nitroprusside), and diuretics (furosemide, bumetanide).Patients were identifed as having received a medication class if it was present on the medication record at any point while they had a PA catheter in place in the ICU.

2.3.
Outcome.Te primary outcome was the association of postoperative PAPi in the ICU and the presence of RV dysfunction defned by post-bypass TEE visualization.Secondary outcomes were association of postoperative PAPi with acute renal injury (defned as a rise in serum creatinine by 50% or more of baseline value) [17], 30-day mortality, total postoperative ventilator hours, and ICU length of stay.

Missing Data.
Missing hemodynamic data were imputed using multiple imputations with predictive mean matching in order to maximize statistical inference for our primary outcome.[18] We performed 25 imputations of 10 iterations each.Imputed data were pooled, and a sensitivity analysis was performed by adjusting imputed estimates by 10-20% and repeating the primary outcome analyses to evaluate the robustness of models.Outcome data were not imputed.
2.5.Statistical Analysis.Cohort characteristics were described by median and interquartile range for continuous and by frequency for categorical variables.Comparisons between groups were carried out by Student's t-test, Wilcoxon rank sum tests, or Fisher's exact test when applicable.A multivariable linear regression model was performed to assess the association between select variables of interest with postoperative PAPi.A natural log transformation for PAPi was performed to improve model performance.Predictor variables with a variance infation factor of 4 or greater were excluded from the model to adjust for multicollinearity.A Bonferroni-Holm correction was applied to control for multiple comparisons.
For secondary outcomes, we generated a propensity score based on the probability of a postoperative PAPi of 2.0 or greater, estimated by multivariable logistic regression.Variables used to generate propensity scores were based on hypothesized association or confounding with postoperative PAPi and included age, sex, postoperative RV dysfunction, pulmonary disease, ejection fraction, bypass time, surgical procedure, serum creatinine, BMI, and postoperative vasoactive agents (Supplementary Figure 2).Greedy nearest neighbor propensity-score matching without replacement using a prespecifed caliper width of 0.1 was used to generate a matched cohort [19].Baseline diferences and outcomes between groups were tested using Student's t-test, Wilcoxon rank sum test, or Chi-square when applicable.
All analyses were conducted using R Statistical Software (v4.2.1, R Core Team 2022).A p value of <0.05 was used throughout as the threshold for statistical signifcance.

Secondary
Outcome.Clinical characteristics of our propensity-matched cohort are given in Table 3.After matching, the cohort included 318 patients in each group.Standardized mean diferences of covariates were 0.1 or less, indicating appropriate covariate balancing (Supplementary Figure 2).Patients in the PAPi <2.00 group had a median postoperative PAPi of 1.48 (IQR 1.20-1.71)and postoperative CVP of 10.9 mmHg (IQR 9.4-12.8).Patients in the PAPi < 2.00 group also had a higher incidence of AKI (23.0%vs 13.2%, p < 0.01) and marginally increased hours on mechanical ventilation postoperatively (6.0 vs 5.6, p � 0.04) (Table 4).

Discussion
Te present study found that in patients undergoing cardiac surgery with cardiopulmonary bypass, there were no  With known challenges in identifying RV dysfunction, PAPi was conceptualized to ofer a simplifed but unique assessment of right ventricular function [8][9][10]20].PAPi not only refects RV contractility but may be more sensitive to changing RV loading conditions than other hemodynamic measurements [8].Following its initial use in patients with right ventricular myocardial infarction, it has been associated with clinical outcomes in diverse patient populations, most notably in patients with heart failure [10][11][12][20][21][22].However, current studies on PAPi emphasize the association of preoperative PAPi and postoperative clinical outcomes rather than identifying validated postoperative PAPi thresholds.
Tis study illustrates that PAPi is not a robust surrogate for RV function in a diverse postoperative cardiac surgical patient cohort.Tis could be because PAPi is not as sensitive of a marker of RV dysfunction in the setting of rapid hemodynamic and volume changes following cardiac surgery [15,23,24].Indeed, RV failure in the perioperative cardiac surgical period is multifactorial and afected by preload, afterload, and stunning from perioperative insults such as cardiopulmonary bypass [25][26][27][28].Alternatively, our fndings could be confounded by awareness of the TEE assessment of RV dysfunction which could have led the ICU team to modify management in order to initiate or enhance RV support [25,28,29].Tis would lead to improved PAPi, theoretically, and negate the association of post-bypass RV dysfunction with lower PAPi values.However, there was no independent association identifed between vasoactive or diuretic medication exposure in the ICU and postoperative PAPi.Tis could simply be because the majority of patients in this study received RV support medications, as is common in postoperative management in the present era [5,25,28,29].Tus, the impact of vasoactive medications on RV dysfunction and correlation to postoperative PAPi cannot be fully deduced by the present study.
Additionally, this study and those investigating postoperative PAPi illustrate that PAPi values need to be contextualized for various patient populations.We selected a postoperative PAPi threshold of 2.00 based on available current literature that identifed similar values as a threshold for perioperative outcomes [9,[11][12][13][14]22].Although the present study found an association of this threshold with perioperative morbidity such as AKI, this fnding may be driven by CVP [15,30,31].A doubling of CVP will halve the PAPi value, which would have major implications if postoperative management is based on PAPi values alone.Ultimately, studies on postoperative PAPi values of "normalcy" need to be individualized for various patient populations, especially in the setting of complex physiology such as pulmonary hypertension, heart transplantation, or mechanical circulatory support [8,23,24].

Limitations.
Tere are a number of important limitations of our study.We did not evaluate prospective development of RV dysfunction in our patients and rather relied on intraoperative post-bypass evaluation of RV function.Tis could have led to the initiation of therapeutics that signifcantly afect postoperative PAPi values.However, this mimics current practice in ICU management of RV dysfunction where there is limited accurate monitoring of this morbidity.Additionally, data acquisition of medications did not account for the frequency or total dosing of each class of medication, but only for the initiation and use of the medication.Lastly, discrepancies in qualitative TEE assessment of RV function in cardiac surgical patients may be present, especially when most of these assessments rely upon sonographer profciency and judgement [32].However, the defnition of RV dysfunction as ascertained from the ofcial TEE report mirrors real-world practice in how RV failure is identifed in cardiac surgical patients and is generally thought to be the gold standard for RV failure identifcation in these patients [25].

Conclusion
Post-bypass RV dysfunction was not associated with postoperative PAPi in the ICU in a mixed cardiac surgical population.When adjusted for confounders, a postoperative

Table 2 :
Multivariable linear regression analysis.Linear regression table performed ln(PAPi) as the dependent variable.Beta coefcient for each predictor was exponentiated to geometric mean diference (odds ratio) per unit change in predictor.Geometric mean adjusted to refect percent change in PAPi per unit change in covariate.Adjusted p values are based on Bonferroni-Holm correction.

Table 4 :
6econdary outcomes in the propensity-matched cohort for PAPi ≥ 2.00.Continuous data presented as median (interquartile range), and categorical data presented as count (percentage of total).6CriticalCare Research and Practice PAPi less than 2.00 was associated with an increased incidence of AKI and a less clinically signifcant increase in ventilator time but was not associated with 30-day mortality and ICU length of stay.Overall, our study emphasizes the importance of developing an accurate and continuous assessment of RV function following cardiac surgery and further validating postoperative PAPi values for individual patient populations.