Preexisting Right Ventricular Systolic Dysfunction May Be More Prognostic Than Left Ventricular Systolic Dysfunction for Mortality in High-Risk Patients Undergoing Non-Emergent Open Abdominal Surgery: A Retrospective Cohort Study

Background: The prognostic value of right ventricular systolic dysfunction in high-risk patients undergoing non-emergent open abdominal surgery is unknown. Here, we aim to evaluate whether presence of preexisting right ventricular systolic dysfunction in this surgical cohort is independently associated with higher incidence of postoperative major adverse cardiac events and all-cause in-hospital mortality. Methods: This is a single-centered retrospective study. Patients identied as American Society Anesthesiology Classication III and IV who had a preoperative echocardiogram within 1 year of undergoing non-emergent open abdominal surgery between January 2010 and May 2017 were included in the study. Incidence of postoperative major cardiac adverse events and all-cause in-hospital mortality were collected. Multivariable logistic regression was performed in a step-wise manner to identify independent association between preexisting right ventricular dysfunction with outcomes of interest. Results: Preexisting right ventricular systolic dysfunction was not associated with postoperative major adverse cardiac events (p=0.26). However, there was a strong association between preexisting right ventricular systolic dysfunction and all-cause in-hospital mortality (p=0.00094). After multivariate analysis, preexisting right ventricular systolic dysfunction continued to be an independent risk factor for all-cause in-hospital mortality with an odds ratio of 18.9 (95% CI: 1.8-201.7; p=0.015). Conclusion: In this retrospective study of high-risk patients undergoing nonemergent open abdominal surgery, preexisting right ventricular systolic dysfunction was found to have a strong association with all-cause in-hospital mortality.

surgery [10][11][12][13][14]. Moreover, knowledge about the predictive value of right ventricular (RV) dysfunction in noncardiac surgical cohort is largely unknown. Recently, we published the ndings that among high-risk patients undergoing major vascular surgery, preexisting RV systolic dysfunction was more predictive of postoperative MACE than LV systolic dysfunction. Indeed, our group found that while left ventricular ejection fraction (LVEF) was not an independent risk factor for MACE, presence of preexisting RV systolic dysfunction by itself was associated with six-folds increase in incidence of postoperative MACE [15].
Building on other authors' previous ndings in which the RCRI score was shown to have predictive value for morbidity and mortality in patients undergoing major abdominal procedures [16], and in which LVEF was not an independent risk factor for worse overall outcomes [17], we aimed to elucidate whether the effect of preexisting RV systolic dysfunction on postoperative cardiac morbidity and overall mortality would be similar to what we observed in our previous study. Our hypothesis was that preexisting RV systolic dysfunction would be more prognostic than LV systolic dysfunction for postoperative major adverse cardiac events. As a secondary outcome, we also hypothesized that preexisting RV systolic dysfunction would be associated with higher all-cause in-hospital mortality in major abdominal surgery. patients between the age of 18 and 89. We de ned high-risk patients as those who were identi ed as American Society of Anesthesiologist (ASA) Physical Status Classi cation of III or IV. Finally, we included only patients with a preoperative echocardiogram performed within one-year of the indexed surgery and for which the study report included evaluation and determination of the RV function. In patients who had multiple echo studies within one-year of the indexed surgery, we selected the study that was closest to the indexed surgery for nal review.

Methods
To collect patients' demographic and perioperative data, we performed manual chart review using the hospital's electronic record. While intraoperative variables were collected from Surgical Information Systems (Surgical Information Systems Corp, Alpharetta GA), demographic and postoperative variables were obtained from Quest (Allscripts Corporation, Alpharetta GA). For those who met the inclusion criteria, the following preoperative variables were collected: age, gender, body mass index (BMI), preoperative hemoglobin level, presence or absence of history of congestive heart failure (CHF), coronary artery disease (CAD), hypertension (HTN), cerebrovascular accident (CVA), diabetes (DM), chronic obstructive pulmonary disease (COPD), obstructive sleep apnea (OSA), and pulmonary hypertension. Since abdominal procedures can often be performed as part of a cancer treatment plan, presence of active cancer requiring the indexed abdominal surgery was also collected. Finally, an RCRI score was calculated and collected based on presence of its six clinical predictors [18].
For intraoperative variables, length of surgery, need for intraoperative transfusion of allogeneic blood products, intraoperative total uid balance, intraoperative hypotension, and intraoperative infusion of inotropic or vasopressor agents were collected. Furthermore, postoperative variables including postoperative development of respiratory complications and acute kidney injury, as well as, postoperative infections and need for subsequent surgeries during the same admission were also collected so they could be evaluated as confounding factors. Finally, diagnosis of sepsis made anytime during the entire hospital admission was also collected and evaluated as a confounder since presence of sepsis continues to be an important risk factor for morbidity and mortality in surgical patients [19].
For assessment of intraoperative hypotension, both noninvasive and invasive blood pressure measurements were extracted from the Surgical Information System (SIS). Data points without both a systolic and diastolic value were excluded. In addition, any systolic values outside of 20 -300 mmHg and diastolic values outside of 5 -200 mmHg were excluded as they were considered to be nonphysiological [20]. Blood pressure data from the noninvasive and invasive monitor were then combined in the following manner: if a systolic or diastolic value had another observation of the same type (systolic or diastolic) regardless of the source (noninvasive or invasive) and was within one minute of each other, the two values would be replaced with the average of the two. Mean arterial pressure (MAP) was calculated for each systolic/diastolic pair according to the following equation: 1/3 x systolic blood pressure + 2/3 x diastolic blood pressure. Intraoperative hypotension was de ned using MAP < 60mmHg as the threshold. This threshold was chosen because previous studies have been able to show an increased risk for myocardial injury and mortality when MAP is less than an absolute threshold of 60 for various duration during general surgery [20]. An episode of intraoperative hypotension was derived by calculating area under the threshold (AUT). AUT was calculated in the same manner as previously described by Vernooij et al. [21] Finally, total AUT was obtained by adding all AUTs for each surgical encounter. All blood pressure data processing described was performed via Python version 2.7 using SciPy and NumPy library of packages (Python Software Foundation, Wilmington, DE).
Postoperative MACE was de ned broadly as composite events including non-fatal cardiac arrest, myocardial infarction, development of congestive heart failure, cerebrovascular accident (Stroke), and cardiovascular mortality de ned as death attributable to any-or a combination of the adverse cardiovascular events just described [4][5][6]. Post-operative respiratory complication was de ned as prolonged intubation for more than 24 hours or need for re-intubation or tracheostomy. Post-operative acute kidney injury was de ned as patients with a post-operative rise of creatinine greater than 60% from the baseline [22]. Post-operative need for subsequent surgeries included all procedures that required anesthesia care. Post-operative infection was de ned as a composite event including wound or surgical site infection, urinary tract infection, pulmonary infection, and systemic infection. Finally, diagnosis of sepsis was made according to guidelines set by the International Sepsis De nitions Conference [23].
The preoperative echocardiogram obtained within one year of the index surgery was used to identify patients with RV systolic dysfunction. All of the echo studies were originally performed by the cardiology service at the study institution. The majority of the echo studies were performed via transthoracic echocardiogram (96%). All of the echo images were interpreted by the cardiologist from the study institution with the nal results reported and stored in the institution's cardiovascular imaging database (Syngo Dynamics -Siemens Healthcare, Tarrytown, NY). All study reports were reviewed and the following collected: LVEF, right ventricular systolic pressure (RVSP), any valvular pathology categorized as severe, presence of LV diastolic dysfunction, and RV function. RV function was reported as a binary variable (normal versus abnormal). RV function collected from the o cial report was determined based on visual estimation by the cardiologists. Visual estimation of the RV function was determined based on multiple acoustic windows including apical 4-chamber (lateral wall of the RV and RV apex), parasternal short-axis (anterior, lateral, and inferior wall of the RV), parasternal RV in ow (anterior and inferior wall of the RV), and subcostal 4-chamber (inferior wall of the RV).

Statistical Analysis
All statistical analysis was performed using SPSS for windows version 24 (SPSS Inc, Chicago, IL). The cohort was divided into 2 groups: those with and without RV systolic dysfunction. For comparative analysis, Fisher's exact test was used for dichotomous variables while Student t-test or Mann-Whitney U test were used for continuous variables with normal and non-normal distribution respectively. For test of normality, Shapiro-Wilk test was employed. Dichotomous variables are reported as counts and percentages while continuous variables are described as either mean and standard deviation for normal distribution or median with interquartile range for non-normal distribution.
Logistic regression analysis was performed to estimate odds ratio (OR) and 95% con dence interval (CI) for effect of RV systolic dysfunction on binary outcomes. The selection of variables to include in the univariable logistic model was based on both group differences and a priori predictors. Variables that were individually associated with outcome of interest with p-value <0.1 in univariable analysis were further included into multivariable analysis in a step-wise manner. Since RV systolic dysfunction, CHF, and RCRI are highly correlated with each other and are expected to exhibit multicollinearity, they were not included in the same regression models during the step-wise multivariable analysis. For goodness-of-t of the regression model, Hosmer and Lemeshow test was employed. For all tests, a p-value <0.05 was considered statistically signi cant.

Results
A total of 122 patients met nal inclusion criteria and were included for data analysis (Fig 1). 5.7% (N=7) of the patients in this cohort had preexisting RV systolic dysfunction evident on preoperative echocardiogram at the time of surgery. A comparison of demographic data in patients with and without RV systolic dysfunction showed that there was no difference in gender, age, and BMI between the groups (Table 1). For other preoperative covariates, a higher percentage of patients with RV systolic dysfunction had a history of CAD and CHF (p=0.0098 and p=0.00052 respectively). In addition, a higher percentage of patients with RV systolic dysfunction had an RCRI score >3 (p=0.00072). For the remaining preoperative covariates such as history of COPD, OSA, hypertension, pulmonary hypertension, and preoperative hemoglobin level, no differences were found between the groups (Table 1). Similarly, there was no statistical differences between the groups when evaluating other echo parameters such as diastolic dysfunction, LVEF, and right ventricular systolic pressure (RVSP). In regards to valvular pathology, only 2 patients had severe valvular pathology (mitral regurgitation and aortic stenosis). Both patients had normal RV function.
For intraoperative covariates, all patients underwent general anesthesia with or without epidural catheter. A higher proportion of patients with RV systolic dysfunction received inotropic or vasopressor infusion during surgery compare to those without RV systolic dysfunction (p=0.014). There was no difference in area under the threshold (AUT) for MAP < 60mmHg in the analysis of intraoperative hypotension (p=0.41). Moreover, no differences were found between the groups for both intraoperative uid balance, need for transfusion of blood products, and total surgical time ( Table 2) Neither RV systolic dysfunction nor LVEF were found to be associated with higher incidence of MACE in univariable analysis. An RCRI score > 3 on the other hand was found to be associated with higher incidence of MACE (p=0.0030). In addition, CHF and CAD were also found to be associated with higher incidence of MACE (p=0.012 and p=0.021 respectively). Multivariable analysis was not carried out due to expected multicollinearity among MACE, CHF, and CAD (Table 3).
A total of 7 (5.7%) patients had expired during the hospital stay: 3.5% among patients with normal RV function and 42.9% among patients with abnormal RV function (p=0.0037). 1 patient died from sepsis resulting from ischemic bowel, 2 patients died from multi-organ failure relating to sepsis, 1 patient died from protracted course relating to organ rejection after renal transplant, 1 patient died from protracted hospital course relating to severe gastrointestinal bleed, 1 patient died from decompensated heart failure, and 1 patient died from multi-organ failure relating to metastatic cancer. There were no intraoperative deaths; all deaths occurred postoperatively during hospital admission.
In univariate logistic regression, the odds ratio (OR) for all-cause in-hospital mortality in presence of preexisting RV systolic dysfunction was 20.8 (95% CI, 3.4-125.8; p=0.00094). Other covariates found to have signi cant association with all-cause in-hospital mortality included sepsis, MACE, CHF, RCRI>3, and postoperative development of respiratory complications (Table 3). Accounting for all other covariates, RV systolic dysfunction remained independently associated with higher incidence of all-cause in-hospital mortality with an OR=18.9 (95% CI, 1.8-201.7; p=0.015) (Figure 2). RCRI>3 was not included in the same regression model as RV systolic dysfunction because of expected multicollinearity. Evaluating RCRI>3 in a separate multivariable regression model, unlike RV dysfunction, it was not an independent risk factor for all-cause in-hospital mortality ( Table 3).

Discussion
In this retrospective single-centered study, we evaluated the association of preexisting RV systolic dysfunction with postoperative major adverse cardiac event (MACE) and all-cause in-hospital mortality in high-risk patients undergoing open abdominal surgery. We found that while preexisting RV systolic dysfunction was not shown to be associated with postoperative MACE in this surgical cohort, it was shown to be associated with higher all-cause in-hospital mortality with a nearly 20-fold increase in the odds ratio for risk. Similar to previous ndings by Gundes et al. [17], we did not nd LV systolic dysfunction de ned by LVEF to be associated with either outcomes of interest.
In regards to postoperative MACE, its low incidence rate (4.1%) together with the study's small sample size likely result in the study being underpowered to detect the true effect of preexisting RV systolic dysfunction on postoperative MACE. In regards to all-cause in-hospital mortality, despite the low sample size in the RV dysfunction group, the nding of a strong association between preexisting RV systolic dysfunction and mortality may have clinical signi cance. First, it has been shown that among patients in the intensive care unit with diagnosis of sepsis, 47% has isolated RV dysfunction and 53% has combined biventricular dysfunction evident on echocardiogram [24]. In our surgical cohort, 3 of the 7 patients ultimately died from sepsis. Among these three patients, one had preexisting RV dysfunction. It is possible that development of sepsis in this patient exacerbated the ventricular function in an already compromised RV. Since the RV plays a critical role in delivering deoxygenated blood to the lungs, maintaining forward ow from venous return thereby preventing organ congestion, as well as global systemic circulatory hemostasis [25], it is reasonable to assume that a preexisting RV dysfunction in a septic patient may contribute to increased mortality.
Second, from clinical experience, management of critically-ill patients with concomitant RV systolic dysfunction can be quite di cult. Aside from addressing the underlying problems, optimization of RV function in a patient with preexisting RV dysfunction is challenging. Part of this challenge results from the fact that an already compromised RV, unlike that of a healthy RV, is not only sensitive to pressure overload, but also exquisitely sensitive to volume loading. In other words, while preload may be important for optimal RV function, the safety margin in which the rate of uid replacement and total volume given without further exacerbating ventricular function can be quite narrow in a dysfunctional RV. Combining this challenge with the frequent hemodynamic derangement of hypotension resulting from postsurgical bleed, uid shift, or in setting of sepsis, the clinical problem and management often becomes much more complex [26].
In addition to the challenge of when and how to optimally replace uid in a patient with RV systolic dysfunction, avoiding factors that can worsen pulmonary vascular resistance may be just as di cult in a critically-ill patient. This is because a postsurgical and critically-ill patient is often faced with multitude of clinical problems that have negative effect on pulmonary vascular resistance and therefore RV function. These problems may include physiological changes relating to surgically induced stress [27], pulmonary complication resulting in hypoxemia or hypercarbia, mechanical ventilatory setting that increases intrathoracic pressure, or systemic hypotension that requires use of vasoconstrictor agents [28]. In a critically-ill patient with preexisting RV systolic dysfunction, any one or a combination of the above processes would undoubtedly add complexity to the clinical care of the patient. In summary, sepsis may play a role in worsening RV or biventricular function; in a critically-ill patient with preexisting RV systolic dysfunction, development of sepsis may therefore further exacerbate ventricular function and contribute to mortality. Moreover, preexisting RV systolic dysfunction in a postsurgical and critically-ill patient can be particularly challenging and may also contribute to overall in-hospital mortality.
Different from the study performed by Jakobson et al.[16], in which RCRI>3 was evaluated for in-hospital mortality, as well as both short-and long-term mortality, our group only investigated the association between RCRI>3 and all-cause in-hospital mortality. In agreement with this prior study[16], our multivariable analysis did not nd RCRI>3 to have independent association with all-cause in-hospital mortality. This nding is consistent with the understanding that while RCRI predicts perioperative cardiac events well, it does not reliably predict death in noncardiac surgical cohort [29].

Limitations
While the ndings in this study may have clinical implications and were strongly statistically signi cant, the sample size was too small. Finding agreement with these results in a larger surgical cohort would be reassuring that they are not a uke. In addition, the study has other important limitations. First, the retrospective nature of the study means that the quality of the study ndings is dependent on accuracy of medical charting. Since our nding of a mortality rate of 5.7% is comparable with prior studies, this suggests that quality of the study was not likely compromised [30][31][32]. In addition, the retrospective nature of the study allowed evaluation of well-documented covariates only. As such, there are important a priori variables relevant to this surgical cohort that the study could not account for [33,34].
Second, including preoperative echocardiogram as far as 1 year prior to the indexed surgery may not accurately capture changes in ventricular function. However, the majority of the patients (89%) had preoperative echocardiogram (echo) performed within 6 months of the surgery. Moreover, it is the practice of the institution in which repeat echo is obtained during preoperative evaluation for nonemergent surgery when patient either demonstrate or endorse clinical changes that may have cardiac relevance. With this, we can reasonably assume that the remaining 11% of the patients in this surgical cohort did not have clinically signi cant cardiac changes at the time of the surgery.
Lastly, unlike our previous nding in the vascular cohort in which patients with RV systolic dysfunction had lower LVEF and higher RVSP [15], we did not nd a signi cant difference in LVEF and RVSP between patients with and without RV systolic dysfunction in the current surgical cohort. While this may be a re ection of an inherent difference between the vascular versus the abdominal surgical cohort, it may represent a skewed sampling of the population resulting from the study's small sample size. If the latter is true, the study would have additional limitation in that it does not re ect the true epidemiology of patients with RV systolic dysfunction undergoing open abdominal surgery.

Conclusion
The present study demonstrated that 5.7% of the patients classi ed as ASA III and above undergoing nonemergent open abdominal surgery had preexisting RV systolic dysfunction. The presence of RV systolic dysfunction was independently associated with all-cause in-hospital mortality with an almost 20fold increase in odds. LVEF on the other hand, was not associated with overall in-hospital mortality.
Based on the study nding, preexisting RV systolic dysfunction appears to be more prognostic for mortality than LV systolic dysfunction in high-risk patients undergoing non-emergent open abdominal surgery. Further studies with larger sample size that includes additional relevant covariates are required to validate current ndings.  Abbreviation: RVD = right ventricular (systolic) dysfunction, CHF = congestive heart failure, CAD = coronary artery disease, EF = ejection fraction, RCRI = revised cardiac risk index, MACE = major adverse cardiac events, AKI = acute kidney injury, OR = odds ratio, CI = confidence interval *Multivariate logistic regression model was not performed for MACE due to non-significant finding for RVD and multicollinearity among RCRI, CAD, & CHF **Only results from the final models are shown for the multivariate logistic regression analysis for all-cause in-hospital mortality. RVD and RCRI>3 were analyzed in separate models due to multicollinearity. Specific step-wise models are provided in the supplemental.
*** CHF was not included in the multivariate logistic regression due to expected multicollinearity with RVD and RCRI. Figure 1 Inclusion and Exclusion Diagram Multivariate Logistic Regression on all-cause in-hospital mortality Abbreviation: RVD = right ventricular (systolic) dysfunction, MACE = major adverse cardiac events, RCRI= revised cardiac risk index *p-value<0.05