Rate of thromboembolic and bleeding events in patients undergoing concomitant aortic valve surgery with left ventricular assist device implantation ☆

Background: Significant aortic regurgitation at the time of left ventricular assist device (LVAD) implantation, requires concomitant aortic valve (AoV) replacement or repair. However, the impact of concomitant AoV surgery on morbidity remains unknown. Therefore, our aim is to determine the impact of concomitant AoV surgery on thromboembolic and bleeding events. Methods: A retrospective IMACS registry study, including patients implanted from 2013 until September 2017. Differences between different concomitant AoV surgery modalities were analyzed. Results: In total, 785 (5.1%) out of 15.267 patients (median age 58 IQR 49 – 66 years, 79% male) underwent concomitant AoV surgery (median age 63 IQR 54 – 69 years, 84% male); 386 (49%) patients received biological prostheses, 71 (9%) mechanical prostheses and 328 (42%) AoV repairs. In total, 54 (8%) patients with AoV surgery experienced a thromboembolic event and 1016 (9%) patients with no AoV surgery. Furthermore, concomitant AoV surgery was associated with an increased rate of all and nonsurgical bleedings. Following a multivariable Cox regression, concomitant AoV surgery remained an independent predictor for bleeding events. Conclusions: In LVAD patients undergoing concomitant AoV surgery, thromboembolic event rates were not higher, however both all and nonsurgical bleeding event rates were higher.


Introduction
During the last decade, the number of durable left ventricular assist device (LVAD) implantations has increased to unprecedented heights [1]. Valvular diseases including aortic regurgitation (AR) are associated with increased morbidity and mortality following LVAD implantation, due to a circulatory shortcut with the continuous flow [2,3]. Therefore, the International Society for Heart and Lung Transplantation [4] and the European Association for Cardio-Thoracic Surgery (EACTS) recommend that greater than mild AR, should prompt concomitant aortic valve replacement (AVR) or -repair during LVAD surgery [5,6]. However, concomitant aortic valve (AoV) surgery during LVAD implantation is not without risks as concomitant surgeries during LVAD surgery are associated with an increased mortality rate [7,8]. Moreover, reports on the impact of concomitant AoV surgery past the early period are scarcely available and were conducted with rather small number of patients. Therefore, the aim of the current study is to elucidate the impact of concomitant AoV surgery on, both early and late, thromboembolic (TE) ☆ All authors take responsibility for all aspects of the reliability and freedom from bias of the data presented and their discussed interpretation events and bleeding events in patients undergoing LVAD surgery.

Patients and methods
The IMACS registry is a multinational, multicenter database collecting prospective data, as has been described previously [9]. The goal of the IMACS registry is to gather data of patients treated with mechanical circulatory support (MCS) worldwide and consecutively conduct studies with the aim of improving outcomes. The registry receives data from the Interagency Registry for Mechanically Assisted Circulatory Support (INTERMACS) [10], European Registry for Patients with Mechanical Circulatory Support (EUROMACS), United Kingdom registry and the Japanese Mechanically Assisted Circulatory Support (JMACS) registries and various individual hospitals worldwide.

Ethical statement
This analysis was reviewed and approved by the IMACS Steering Committee. Informed consent was obtained by each of the participating registries and centers.

Study design, definitions, and endpoints
All patients who were scheduled for a continuous-flow LVAD implantation from January 2013 through September 2017 were selected. Supplemental fig. 1 shows the inclusion flowchart. Definitions of events were predetermined by the IMACS registry. The aim of the current study was to investigate the effect of concomitant aortic valve surgery on the primary endpoint. Subsequently, a sub-analysis, analyzing each aortic valve surgery modality separately was conducted. The primary endpoint of the study was the first occurrence of thromboembolic (TE) events. Thromboembolic events were defined as either early (during the first 90 days of follow-up) or late (up until 2 years of follow-up) ischemic strokes. Secondary endpoints included all major bleeding events (defined as mediastinal, pump pocket, pleural space, intra-abdominal, pulmonary, retroperitoneal, device anastomosis, urinary tract, and all gastrointestinal bleedings), nonsurgical bleeding events (defined as urinary tract or gastrointestinal bleeding), early and late pump thrombosis (defined as either suspected or confirmed cases), hemorrhagic stroke events, intensive care stay duration, total admission duration and mortality. Major bleeding events were predefined by the IMACS database: a suspected internal or external bleeding which resulted in one or more of the following things: death, re-operation, hospitalization, or transfusion with red blood cells. Furthermore, hemocompatibility related adverse events (HRAE's) were compared between the groups. The HRAEs, a composite endpoint, was defined as either a nonsurgical bleeding event, a neurologic event (i.e., hemorrhagic, or ischemic stroke) or pump thrombosis (suspected or confirmed). Lastly, independent predictors for bleeding events were evaluated.

Statistical analysis
Baseline characteristics are presented as mean, standard deviation (SD) or median with interquartile range (IQR) depending on the distribution of the continuous variables, and count and percentages (%) for categorical variables. Differences between patients' groups were compared with One-way ANOVA (Gaussian distribution) or Kruskal-Wallis (non-Gaussian distribution) for continuous variables. Categorical variables were compared with the Chi squared test. Kaplan-Meier curves were plotted for the occurrence of any of the primary or secondary endpoints. Differences in the rate of endpoints were compared with the Log-Rank test. Patients were censored at the time of transplantation, ventricular recovery, or death. The Fine-Gray method was applied for the competing outcomes analysis.
To determine the most optimal predictive model, a univariable cox hazard regression model was applied. Each individual baseline was tested for its predictive value. Following the univariable regression, a combined multivariable cox hazard regression model was built. The enter method was used to avoid the rather opportunistic nature of the forward and backward method. The multivariable Cox proportional hazards analysis was performed for the identification of covariates independently associated with bleeding events. Missing data were handled by performing multiple imputations, which was only performed for the missing variables used in the univariable and multivariable analysis (see Supplementary Table 1 for percentages missing). A maximum of 30% missing was deemed acceptable for inclusion to be imputed. Variables were only included in the multivariable models if their respective p was ≤0.10 in the univariable analysis. All multivariable models were constructed by using the enter method. A 2-tailed value of p < 0.05 was considered statistically significant. The analyses were performed using SPSS statistics version 26 for MacOS (IBM Corp, Armonk, NY) and environment for statistical computing (R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project. org/).

Results
In total, 15.267 patients were treated with a primary continuousflow LVAD implantation and were subsequently included. Concomitant AoV surgery was performed in 785 (5.1%) patients; 386 patients of them (49.2%) were treated with a biological prosthesis, 71 (9.0%) patients with a mechanical prosthesis and 328 (41.8%) patients were treated with concomitant AoV repair surgery. At the time of LVAD implantation, 38% of all patients undergoing concomitant AoV surgeries had a moderate-to-severe AR. The remainder of patients were reported to have only mild or even no AR. There was no specific information retrievable regarding the surgical indication for AoV surgery. Differences in baseline characteristics are displayed in Table 1. The differences in baseline characteristics for each individual modality are displayed in Supplementary table 2.

Thromboembolic events and pump thrombosis.
See Table 2 for a complete overview of the following clinical outcomes. Overall, TE rates were similar in all groups, regardless of concomitant AoV surgery. The rate of TE events (Supplemental Fig. 2a) was similar with (54 (8%)) or without (1016 (9%)) concomitant AoV surgery in LVAD patients (p = 0.66). Additionally, no difference in TE rate was observed between the different AoV surgical modalities. The rate of hemorrhagic strokes was comparable (Supplemental Fig. 2b). Furthermore, the rate of hemorrhagic stokes did not differ between the different AoV surgical modalities. To account for competing outcomes, a competing risk analysis (including stroke, death, or transplantation) was performed for both concomitant AVR and concomitant repair surgery, which revealed comparable outcomes between all groups (Supplemental The pump thrombosis rates, (73 (12%) vs 1326 (11%), p = 0.170, Supplemental Fig. 4) were comparable between patients with and without concomitant AoV surgery. However, pump thrombosis rates were significantly higher in the AoV repair group (Supplemental Fig. 5).
The nonsurgical bleeding rate (Supplemental Fig. 7), showed that concomitant AoV surgery patients had a higher rate (184 (28%) vs 2913 (25%), p = 0.002). The early nonsurgical bleeding rate was higher in the AoV repair group, whereas both concomitant biological and mechanical AVR were not associated with an increased rate. However, the late nonsurgical bleeding rate ( Fig. 1) was significantly higher in both the concomitant AoV repair and mechanical prosthesis group.
Lastly, we looked at the international normalized ratio (INR

Predictors for bleeding events
An exploratory univariable Cox proportional hazard model was built to investigate potential predictors for bleeding events. See Table 3 for an overview of these predictors. Following a multivariable analysis, AoV surgery remained as an independent predictor for bleeding events.

Clinical outcomes
Overall, patients treated with concomitant AoV surgery had a longer intensive care unit [11] stay than those without. Furthermore, patients with concomitant AoV surgery were hospitalized longer. The outcomes of individual modalities of AoV surgery are listed in Supplemental  Table 3. The overall occurrence of HRAEs were additionally analyzed and while the early period favored no concomitant AoV surgery, at 2years of follow-up, no significant difference was observed between both groups. See Table 2 for an overview of the clinical outcomes.

Discussion
The main finding of this study was that concomitant AoV surgery during LVAD surgery was not associated with a higher rate of TE. However, concomitant AoV surgery was associated with an increased rate of bleeding events following LVAD implantation.

Thromboembolic events and pump thrombosis
This study was conducted with an initial hypothesis that concomitant AoV surgery could increase the rate of TE events following LVAD implantation. Prior, smaller studies revealed that aortic root thrombosis during LVAD support is not uncommon and that concomitant AoV surgery could play a role in the formation [12,13]. However, our current findings reveal similar rates of TE events between patients with and without concomitant AoV surgery.
The analysis revealed that concomitant AoV repair surgery was associated with a significantly increased rate of pump thrombosis following LVAD implantation. Previous studies have linked AoV closure at the time of LVAD implantation with an increased rate of pump thrombosis and decreased survival [7,14]. However, given the differences in patient characteristics (the AV repair patients were older, and were least likely to have a centrifugal device), this apparent increase in pump thrombosis is most likely explained due to these differences. Nonetheless, it remains possible that the AoV repair surgery is an instigator as well. The inherent risks of sutures tearing, and possible reoccurrence of AR could make patients more at risk for hematological complications such as pump thrombosis [15,16]. However, the underlying mechanisms are far from being elucidated.

Bleedings events
We found concomitant AoV surgery to be associated with an increased rate of bleeding events. To elucidate the underlying cause, we subsequently conducted a second analysis to only include nonsurgical bleeding. This revealed that, with prolonged support, both concomitant mechanical prostheses and AoV repair surgery were associated with an increased rate of nonsurgical bleeding. This is visible on the Kaplan Meier curves, showing an impact especially in the early period following surgery. Interestingly, a prior investigation stated that concomitant AoV surgery was not associated with an increased rate of bleedings in HeartMate II patients [17]. To validate our findings, we conducted a multivariable Cox regression analysis. After adjusting for multiple covariates, concomitant AoV surgery was found to be independent  predictor for bleedings events following LVAD implantation. We postulate that the initiation of an intensified regimen of platelet inhibitors and oral anticoagulation most likely plays a part, as both are associated with increased bleeding rates [18]. Moreover, concomitant AV surgery is more extensive, needs more suturing, aortic crossclamping and cardioplegia with dilution, and possibly more transfusion with possible impact on the blood homeostasis even in the days following the surgery. However, the increased shear stress and subsequent greater degree of acquired of von Willebrand factor deficiency could be contributors [19][20][21]. Lastly, the impact of intraoperative factors, such as the use of a cardiopulmonary bypass (CPB) machine, the duration of the CPB machine and the possible use of cardioplegia could all play a substation roll in the increased bleeding events, however these data were not available in this cohort.

Indications
We want to preface this section with the notion that the decision for concomitant AoV surgery is multifactorial and driven by more than the degree of AR. Nonetheless, to our surprise, this study revealed that the majority (48%) of concomitant AoV surgeries were performed in patients with a preoperative diagnosis of mild AR. This was unanticipated, since the current guidelines recommend concomitant AoV surgery only in patients with moderate-to-severe AR. Plausible explanations for deferring form this recommendation include the caution of heart teams in this obscure clinical entity in the era of LVAD surgery. Furthermore, during the perioperative period, surgeons can defer, or schedule concomitant surgeries based on the current perioperative condition of patients. The introduction of improved left ventricular flow could reveal a higher severity of aortic regurgitation, which might have been concealed previously due to poor left ventricular function [22]. Secondly, the data of this current study are provided from 2013 until 2017 and are subject to evolving indications and experience. Moreover, there seems to be evidence that concomitant AoV surgery is perhaps warranted in select LVAD candidates with mild AR. Previous studies found that mild preoperative AR is a significant predictor for worsening AR during LVAD support [3,23]. Nonetheless, these findings highlight the disparities between contributing centers.
Of note, inclusion of mechanical prosthesis as a treatment modality was highly unanticipated. The ISHLT guidelines and the EACTS expert consensus recommend replacement of mechanical prostheses with biological prosthesis at the time of LVAD surgery and recommends the use of a biological prosthesis in case of a scheduled AoV surgery [6,24]. These findings underline the far from crystallized indications for concomitant aortic valve surgery in patients undergoing LVAD implantation.

Clinical implications
Concomitant AoV surgery is associated with significantly higher mortality, longer need of ICU stay and hospital stay and bleedings rates, most noticeably in the early period following LVAD implantation. Furthermore, patients who had not undergone concomitant aortic valve surgery were more likely to be transplanted. This most likely derives from their baseline differences as patients who had not undergone concomitant aortic surgery were younger and more frequently implanted as bridge-to-transplantation. Therefore, the decision for concomitant aortic valve surgery should warrant thorough evaluation of bleeding risks prior to surgery. Lastly, the deployment of less invasive intervention such as transcatheter aortic valve implantation (TAVI) can provide a solution in some patient who are prone to complication when undergoing concomitant aortic valve surgery. Rather than extending the operation time, increase CPB time and aortic cross-clamping, the TAVI procedure can be scheduled either prior to surgery or following surgery [25].

Limitations
The current study is the performed with data from registries allowing for the inclusion of a large number of concomitant AoV surgeries. However, it's important to note that the data supplied by every center is subject to erroneous and missing data. Furthermore, the retrospective nature of this study does not allow for establishing causality. Secondly, no detailed information was available on the indications for aortic valve surgery, implanted device type, and therefore no further analyses could be conducted between the differences regarding device types. Furthermore, the use and the duration of a cardiopulmonary bypass machine, and possible off-pump implantations were not captured in the database. Therefore, no separate analyses could be conducted to account for the effect of CPB in this cohort. Additionally, no information was available on the decision-making process during concomitant aortic valve surgery. Third, some data were missing for end points used in this study and therefore could have altered the results of the study. Furthermore, no data on prior aortic valve surgery or aortic root dilation were provided. Fourth, missing data were imputed. Nonetheless, the percentage of missing data was limited, with most covariates missing less than 10%. Of note, the imputed data was solely used in the Cox regression models. Lastly, no data was available on the reasoning for concomitant AoV surgery modality and the registry did not distinguish a separate AoV closure group.

Conclusion
In LVAD patients, concomitant AoV surgery was not associated with an increased rate of TE events. However, concomitant AoV surgery was associated with an increased rate of bleeding events, especially in the early postoperative period. Prospective, randomized trials are required to determine the appropriate indications and management of concomitant AoV surgery at the time of LVAD implantation.

Funding
None.

Declaration of Competing Interest
None.