|Home | About | Journals | Submit | Contact Us | Français|
Despite improved outcomes and lower right ventricular failure (RVF) rates with continuous-flow left ventricular assist devices (LVADs), RVF still occurs in 20-40% of LVAD recipients and leads to worse clinical and patient-centred outcomes and higher utilization of healthcare resources. Preoperative quantification of RV function with echocardiography has only recently been considered for RVF prediction, and RV mechanics have not been prospectively evaluated.
In this single-centre prospective cohort study, we plan to enroll a total of 120 LVAD candidates to evaluate standard and mechanics-based echocardiographic measures of RV function, obtained within 7 days of planned LVAD surgery, for prediction of (i) RVF within 90 days; (ii) quality of life (QoL) at 90 days; and (iii) RV function recovery at 90 days post-LVAD. Our primary hypothesis is that an RV echocardiographic score will predict RVF with clinically relevant discrimination (C >0.85) and positive and negative predictive values (>80%). Our secondary hypothesis is that the RV score will predict QoL and RV recovery by 90 days. We expect that RV mechanics will provide incremental prognostic information for these outcomes. The preliminary results of an interim analysis are encouraging.
The results of this study may help improve LVAD outcomes and reduce resource utilization by facilitating shared decision-making and selection for LVAD implantation, provide insights into RV function recovery, and potentially inform reassessment of LVAD timing in patients at high risk for RVF.
The implantable left ventricular assist device (LVAD) is now a standard option as bridge-to-transplant (BTT) for patients with heart failure (HF) deteriorating while awaiting heart transplantation.1,2 Due to (i) donor organ shortage, (ii) increasing number of patients who are older or have comorbidities precluding transplantation, and (iii) improving technology, LVAD is gaining momentum as destination therapy (DT) also.3,4 Currently, 40% of the LVADs are implanted as DT,5 with 1-year survival exceeding 80% for continuous-flow devices,5 which currently represent ~100% of the implants.5 An estimated 150 000–250 000 patients are potential LVAD recipients in the USA.6 However, LVAD remains a high-risk, high-cost option, necessitating careful patient selection.
Outcomes after LVAD implantation are critically dependent on right ventricular (RV) function, which provides flow through the pulmonary vascular bed to fill the LVAD for optimal function.7 Despite improved outcomes and lower right ventricular failure (RVF) rates with continuous-flow over pulsatile-flow LVADs,3 RVF still occurs in 13–40% of the recipients.8 The aetiology of RVF, which is characterized by systemic hypoperfusion with elevated right-sided pressures, is multifactorial. Intrinsic RV dysfunction,9 ischaemia,10 ventricular interaction,11 worsening tricuspid regurgitation,12,13 intraoperative events,14,15 and LVAD management strategies16 have all been implicated. Patients who develop RVF have (i) reduced survival and progression to transplant17–24 and worse outcomes thereafter18,22; (ii) higher resources utilization with longer intensive care stays,15,21,23 more transfusions,23 use of inotropes,15,20 and need for right ventricular assist devices (RVADs)15,21; and (iii) long-term poor exercise capacity and quality of life (QoL).25
Although RVADs are available for short-term support, long-term mechanical circulatory support of the RV is still under development.26 Thus, identification of patients at high risk for RVF is of paramount importance, especially for DT implants. Elective, planned biventricular support is feasible and leads to better outcomes compared with ‘bailout’ RV support.27 Also, surgical series suggest that reduction of tricuspid regurgitation during LVAD surgery in patients with severe RV dysfunction may improve outcomes.28,29 Therefore, assessment of RVF risk has direct implications for patient selection and planning of surgical strategy.
Current risk prediction schemes for RVF take into account HF severity, end-organ damage, right-heart haemodynamics, and history of cardiac procedures.8 No scheme to date includes quantitative RV variables. Of note, these schemes were primarily developed in older, pulsatile-flow LVAD cohorts. Notably, although a C-statistic of 0.73–0.74 was reported for clinical scores in derivation cohorts,30,31 substantially lower values (C = 0.61–0.66) were observed in external cohorts.30,32
Quantitative RV function has only recently been studied for RVF prediction.8 However, most studies were retrospective and either included patients with pulsatile-flow devices or included pre-implant decision for biventricular support as RVF.8 RV mechanics have allowed for early detection of RV dysfunction in various settings. Retrospective studies in continuous-flow LVAD populations report that RV free wall strain predicted RVF independently of clinical risk scores and better than standard echocardiographic RV variables.32–34 However, prospective data are lacking.
The Right Ventricular Function After Left Ventricular Assist Device (RVF-LVAD) is a prospective, single-centre cohort study (www.clinicaltrials.gov NCT01999712) initiated in Q3 2012. In this study, we evaluate echocardiographic measures of RV function for prediction of (i) RVF, (ii) QoL, and (iii) RV function recovery at 90 days after LVAD implantation.
Our primary hypothesis is that an echocardiographic score will predict RVF (occurring within 90 days of implantation) with clinically relevant discrimination (C ≥ 0.85) and positive and negative predictive values (>80%). Our secondary hypotheses are that the RV echocardiographic score will predict (i) poor QoL at 90 days (C ≥ 0.85) and (ii) the degree of recovery of RV function at 30 and 90 days after LVAD. We expect longitudinal myocardial mechanics to have incremental value.
We plan to enrol 120 adults scheduled to receive an LVAD as BTT or DT at Emory Healthcare (Atlanta, GA, USA) with Interagency Registry for Mechanically Assisted Circulatory Support (INTERMACS) Profiles 2–7 (Box (Box1),1), without plans for mechanical RV support or unresponsive severe pulmonary hypertension (Table Table11). We opted not to include INTERMACS Profile 1 patients because these patients receive an LVAD in emergent situations, some as a bridge-to-decision after cardiogenic shock. These conditions are not conducive to complete evaluation for research purposes. Currently, Profiles 2–7 constitute 86% of the implants.5
Patients undergo complete echocardiography at baseline and then at 30 and 90 days after implant (Table Table22). QoL is assessed with the Kansas City Cardiomyopathy Questionnaire (KCCQ) and Euro-QoL 5D-3L at baseline and 90 days. Clinical outcomes are recorded through 90 days. Blood samples are drawn at baseline and 30 days for future analyses. Sonographers with extensive experience in research echocardiography have been designated to perform all echo studies. All data are entered into a REDCap® electronic data capture database securely hosted at Emory. The study has been approved by the Emory University Institutional Review Board and has been registered in clinicaltrials.gov with identifier NCT01999712.
Following the INTERMACS definition, RVF is defined as symptoms and signs of persistent RV dysfunction (defined as central venous pressure >18 mmHg with a cardiac index <2.0 L/min/m2 in the absence of pulmonary capillary wedge pressure >18 mmHg, tamponade, ventricular arrhythmias, or pneumothorax), requiring RVAD or inhaled nitric oxide (or other pulmonary vasodilator) or inotropic therapy for >7 days any time after LVAD implantation.
Health status and QoL are assessed with KCCQ, which has been established as a valid, reliable, and responsive health status tool in HF, including LVAD recipients.1,35 KCCQ scales are summarized into a single score (0–100), with higher scores reflecting better status. A KCCQ < 45 at 90 days will be considered to be indicative of poor QoL.36 Euro-QoL 5D-3L is used as an adjunct tool.37
We will report changes in (i) standard RV function parameters and (ii) RV mechanics from baseline to 30 and 90 days post-implant. For binary definition purposes, improvement of RV free wall strain by ≥5% units will be considered indicative of RV recovery.38
In previous studies, the definition of RVF has been variable, including need for RVAD alone or combined with need for prolonged use of inotropes.8 Need for pulmonary vasodilators following surgery has also been considered.8 All these components of RVF definition have been associated with worse outcomes.8 Although no time frame was specified in previous studies, most of the RVF develops immediately after implantation (<90 days). We have opted for the INTERMACS definition to facilitate comparison with INTERMACS registry data.39 However, we collect detailed data on post-LVAD haemodynamics and use of medications. Therefore, as ongoing work may redefine RVF, we will be able to update estimates for RVF and predictive parameters.
In the pivotal HeartMate II clinical trials, most QoL improvement after LVAD implantation has taken place by 90 days.40 A KCCQ summary score <45 has been associated with worse outcomes and health status in HF patients and has been used in previous studies in advanced HF41; therefore, we opted for this cut-off point for the binary definition of poor QoL.
The recovery of RV after LVAD has variable course, but most of the improvement is seen by 90 days.42 We quantify RV recovery at 30 and 90 days. However, we will also report RV recovery as a binary response, considering improvement in RV free wall strain ≥5% as evidence of RV recovery. In a serial study in pulmonary arterial hypertension,38 this degree of RV recovery was associated with lower mortality.
All studies are performed by registered sonographers on Vivid 7 and E9 systems (GE Healthcare, WI, USA) and transferred in uncompressed raw DICOM for offline analysis. Parameters of LV and RV functions and haemodynamics are recorded according to the American Society of Echocardiography (ASE) recommendations.
The ASE has suggested that RV function be quantified by two-dimensional and pulsed-wave tissue Doppler echocardiography.43 Based on previous experience and a thorough literature review,8 we have selected tricuspid annular plane systolic excursion (TAPSE), peak systolic velocity (s′) at the tricuspid annulus by tissue Doppler, RV myocardial performance index (MPI), and RV fractional area change (FAC) as the most representative standard measures of RV function (Table Table33). TAPSE is acquired in the apical four-chamber view by aligning an M-mode cursor with the tricuspid annulus and measuring the longitudinal motion of the annulus; colour M-mode can facilitate tracking of motion (Figure Figure11 A). In the same view, we obtain peak systolic velocity (s′) by placing a pulsed tissue Doppler sample volume in the middle of the RV free wall basal segment (Figure Figure11B). We use a carefully acquired tissue Doppler trace at the tricuspid annulus to measure MPI as this is the most reliable approach (Figure Figure11C). Finally, we measure the RV FAC from a dedicated RV apical four-chamber view, i.e. adjusting the transducer angle to focus on the RV chamber, with the goal of maximizing its chamber size (Figure Figure11D). On the basis of pathophysiology and previous data, we also consider estimated RV systolic pressure and tricuspid regurgitation44 as potential RVF predictors obtained with conventional echocardiography.
In addition to standard measures of RV function, we use speckle tracking to assess longitudinal RV mechanics (strain and strain rate), which have shown promise in RVF prediction (Table Table33). The RV is imaged from a dedicated apical view with an adequate field to image the entire RV throughout the cardiac cycle with high frame rate (>40 frames/s), as described previously.45,46 To assess RV longitudinal deformation, we adopt the six-segment RV model.45,46 Briefly, the endocardial border is manually outlined and the myocardium is then tracked by the algorithm and divided into six segments (Figure Figure22). The speckle-tracking algorithm detects the QRS onset from the electrocardiographic signal to define the point of zero strain, amenable to correction, and calculates segmental and global strain. By temporal derivation of strain, the corresponding strain rate is obtained. Global parameters refer to the total deformation of the chamber during the cardiac cycle in the selected view; the entire length of the tracked myocardium is considered as baseline length. We recorded global peak systolic strain (GS) and global systolic strain rate (GSRs) of the RV in the longitudinal direction as the main RV mechanics parameters of interest. We calculate GS and GSRs using the entire length of the RV myocardium, as shown in Figure Figure22A. The free wall RV strain and strain rate are calculated using the three free wall segments (Figure Figure22B). We have previously reported on the reproducibility of RV myocardial deformation parameters in our laboratory; the mean absolute percentage error for measurements performed in random order by two observers was 6.9 and 8.9% for GS and GSRs, respectively.47
To facilitate future biomarker studies, we draw blood samples during the pre-operative assessment (baseline) and at 30 days after implant. One serum and one plasma tube are centrifuged at 2500 g, 4°C for 10 min, and the supernatant is stored at −80°C in cryovials.
We plan to enrol 120 patients, based on a conservative 25% rate at 90 days. This sample provides 80% power to detect a difference at P = 0.05 between receiver operating characteristic (ROC) curve with C ≥ 0.85 (echocardiographic or combined clinical and echocardiographic score) vs. C = 0.70 (old clinical scores).
Because of relatively short follow-up, we will work in the logistic regression framework. For echocardiographic variable selection, we will use backwards elimination guided by clinical rationale. The final set of variables will be converted into a simple score. To assess the performance of the echocardiographic score, we will calculate (i) the C-statistic, (ii) sensitivity and specificity, and (iii) positive and negative predictive values for RVF. Because predictive models over-perform in the derivation cohort (‘optimism’),48 we will correct for optimism by bootstrapping.48 To assess the incremental value over previous models for RVF prediction, we will enter the score in multivariable models with (i) risk factors identified previously8 and (ii) existing RVF risk models.8
The association of the RV echocardiographic score with poor QoL (KCCQ Summary Score <45) will be assessed using logistic regression. To assess whether baseline RV parameters can predict the course of RV function over time (30 and 90 days), we will use multilevel mixed models. We will also assess RV response as a binary variable (ΔRV free wall strain by ≥5%).
As of Q2 2014, 41 patients have been enrolled, one-third of the total planned sample. At this point, we did an interim analysis for futility, evaluating the predictive value of echocardiography for the primary endpoint (RVF at 90 days). The pre-operative patient characteristics are summarized in Table Table44. Pre-operative echocardiograms were performed a median of 4 days (1, 14) prior to implantation. Complete RV assessment was possible in 38 patients (feasibility 92.7%). Three patients had inadequate echocardiographic views for complete RFV assessment and therefore were excluded from the analysis.
At 90 days, 15 of 38 patients (39.5%) had developed RVF (with 2 deaths) and 3 (11.5%) had died from sepsis. Median hospital stay was 15 days (25th and 75th percentiles: 12, 18), and intensive care stay was 7 days (6, 10). The median duration of inotrope use was 7 days (5, 12). Seven patients (18.4%) required inhaled or oral pulmonary vasodilators, and one patient required mechanical RV support.
Among pre-operative echocardiographic parameters, smaller LV diameters, lower RV global longitudinal strain, moderate or higher degree of semi-quantitatively assessed tricuspid regurgitation, and shorter pulmonary acceleration time and acceleration-to-ejection-time ratio were associated with RVF (Table Table55). TAPSE, tissue Doppler RV velocities, and RV FAC did not differ significantly between RVF groups.
Using stepwise elimination (P to enter 0.1 and P to leave 0.05) in a model with only echocardiographic parameters, the three parameters that remained in the final model were the semi-quantitative tricuspid regurgitation grade, LV systolic diameter, and RV global longitudinal strain. The echocardiographic score deriving from these three parameters had a C-statistic of 0.88 [95% confidence interval (CI): 0.75–1.00; P < 0.001] for RVF prediction (Figure Figure33). Sensitivity was 86.7% (95% CI: 59.5–98.3); specificity was 87.0% (95% CI: 66.4–97.2); and the positive and negative predictive values were 81.3% (95% CI: 54.4–96.0) and 90.9% (95% CI: 70.8–98.9), respectively. In comparison, the reference clinical score (Michigan) had C = 0.63 (95% CI: 0.47–0.78; P = NS), and P = 0.009 for the difference between the two curves. We did not include clinical covariates in the model, as the power was not sufficient at this point.
In this study, we are prospectively enrolling 120 LVAD candidates to evaluate standard and mechanics-based echocardiographic measures of RV function for prediction of (i) RVF within 90 days; (ii) QoL at 90 days; and (iii) RV recovery at 90 days after LVAD implantation. For this purpose, we perform echocardiography within 7 days of planned LVAD surgery to evaluate the association of prespecified RV parameters with RVF and QoL at 90 days, in the belief that an RV echocardiographic score will provide clinically relevant discrimination (C ≥ 0.85) and positive and negative predictive values (>80%) for RVF. We also repeat echocardiograms at 30 and 90 days after LVAD surgery to examine the trajectory of RV recovery and how baseline RV function is associated with RV recovery. We expect RV mechanics to have incremental values for these outcomes. These results may help improve LVAD outcomes and reduce resource utilization by facilitating shared decision-making and selection for LVAD implantation vs. alternative option; provide insights into RV function recovery; and inform reassessment of LVAD timing.
In our preliminary analysis in the first 41 patients, in which only echocardiographic variables were considered, we identified three domains of potential importance for RVF prediction: tricuspid regurgitation, poor RV function, and smaller LV size. These parameters have been identified previously,8 but not in the context of a comprehensive assessment. Although both tricuspid regurgitation and RV function are ‘expected’ predictors of RVF, smaller LV size is perplexing. One potential explanation is that a smaller LV is prone to suction effects from the outflow cannula and therefore unfavourable ventricular interaction.
In contrast to tricuspid regurgitation severity and RV free wall strain, conventional RV function parameters were not predictors of RVF in our preliminary analysis. There are several potential explanations for this finding. In the USA, patients with INTERMACS Profiles 1–3 (corresponding to critical cardiogenic shock, progressively declining despite inotropic support, or stable but inotrope-dependent patients, respectively) represent currently the majority of LVAD recipients.49 This reflects also the current practice in our institution. Most of these really sick patients are imaged under severe haemodynamic stress a few days before planned surgery. As a result, all linear, load-dependent measures of RV function yield exaggerated values due to intense motion that does not actually represent true RV contraction. In our experience, TAPSE and RV tissue Doppler’s values are the most vulnerable parameters to this phenomenon (this was the case in the patient imaged in Figure Figure11). In fact, patients who developed RVF to date in our study (Table Table55) had slightly higher TAPSE and RV tissue Doppler’s values at baseline, highlighting this issue. RV FAC is less affected, but it becomes increasingly difficult to accurately assess RV FAC under severe haemodynamic stress and high heart rates, and it is therefore difficult to rely on RV FAC for patient classification and clinical decision-making.
Interestingly, RV global longitudinal strain was the most important predictor of RVF (among RV function parameters) in our preliminary analysis. In a retrospective study of 117 LVAD recipients, RV free wall longitudinal strain by velocity vector imaging (a variant of speckle tracking) predicted RVF with 76% specificity and 68% sensitivity at a cut-off of −9.6% and had incremental predictive value over a clinical score.32 In another study of 68 patients undergoing elective LVAD placement, RV longitudinal strain was significantly impaired pre-operatively (−12.6 ± 3.3 vs. −16.2 ± 4.3%, P < 0.001) in 24 patients (35.3%) who experienced RVF by 14 days.34 Thus, we need to prospectively validate the role of echocardiographic mechanics of RV function in post-LVAD RVF prediction.
Our study is conducted in a single centre and is therefore subject to biases associated with our institutional practice in terms of patient selection, surgical approaches, treatment of RVF (e.g. conservative approach vs. RVAD implantation), duration of inotrope use, and other potential considerations. Moreover, institutional practices may affect course of RV recovery and serial echocardiographic measurements. We believe, however, that the consistent definition of outcomes and the standardized echocardiographic approach will still provide a valid direction and strength of association of echocardiographic parameters of interest with RVF and course of RV function over time.
Commensurate with the single-centre design, our planned sample size of 120 patients is relatively small for multivariable analyses. On the basis of the observed enrolment rates during the pilot period of the study, we have powered the study to demonstrate prognostic and decision-making superiority of quantitative echocardiography over clinical scores for RVF. However, the number of events will be relatively small to obtain reliable estimates for individual risk factors. Therefore, we expect our findings to demonstrate the direction and strength of association for important domains of echocardiographic and clinical parameters with RVF, but not necessarily provide the ultimate estimates for each parameter. This would require a multicentre approach. In contrast, our study is overpowered for detection of changes in echocardiographic parameters with serial echocardiography and for clinically relevant correlations with QoL.
Regardless of our ability to accurately quantify RV function, RVF prediction is a moving target. Intraoperative events, concomitant surgical procedures, and post-operative changes in pulmonary haemodynamics and device settings are important (and unpredictable) confounders. Additional challenges include evolving mechanical circulatory support technology and shifts in the target population. However, the full potential of echocardiography has not been fully used in the pre-operative assessment of candidates with LVAD.
Our imaging approach to the RV is incomplete compared with other clinical methods of RV imaging (e.g. cardiac magnetic resonance) and reflects our attempt to strike a balance between practical considerations for clinical applicability in this patient population and scientific rigour. For example, our imaging approach does not include the RV outflow tract. Although the outflow tract plays a lesser role in RV output for adults without congenital heart disease, as the infundibulum accounts for <15% of RV stroke volume,50 it is still an important limitation. Also, we did not attempt to incorporate multilayered strain analysis in our study. There are pathophysiological and practical considerations when applying this approach to the RV. From a pathophysiology perspective, the fibres in the RV free wall lack the multilayered orientation of the LV, and therefore, the incremental benefit would be unclear. From a practical perspective, unless there is hypertrophy due to congenital heart disease, for example, the RV free wall is too thin to track into three layers. However, such an approach could have provided insights for the interventricular septum. Finally, we did not consider short-axis RV mechanics in our study. However, unless there is continuous overload during development, as in transposition of the great arteries,51 for example, the RV relies predominantly on longitudinal function to produce output,52 whereas twisting and rotational movements do not contribute significantly to RV contraction (in contrast to the LV).53 In experimental animal models, longitudinal strain seems to best quantify RV contractile function.54 An additional concern is the difficulty in acquiring reliable short-axis images in enlarged and displaced RVs, as is frequently the case in patients with advanced HF.
By identifying LVAD candidates at high risk for RVF and poor QoL, our ultimate goal is to (i) facilitate shared decision-making for LVAD vs. planned alternatives (e.g. biventricular support or concomitant procedures),28,29 potentially improving outcomes and QoL, and (ii) reduce resource utilization and costs. Finally, although improvement of RV function after LVAD implantation has been reported,42 recovery may not be feasible when baseline RV dysfunction exceeds a certain threshold. This might be a signal for earlier LVAD implantation in some patients when RV is still ‘viable’. By investigating RV function with serial echocardiography, we will provide mechanistic insights into RV recovery and, in conjunction with pre-operative evaluation, may inform LVAD timing for patients with similar degrees of LV dysfunction but varying degrees of RV dysfunction. In this direction, our preliminary results from the interim analysis are encouraging.
Conflict of interest: none declared.
This work was supported by (i) the American Heart Association (13SDG15960001); (ii) an Emory University Research Committee and Atlanta Clinical and Translational Science Institute Grant (supported by NIH, National Center for Advancing Translational Sciences under Award No. UL1TR000454); and (iii) PHS Grant UL1 RR025008 from the Clinical and Translational Science Award program, NIH, National Center for Research Resources (Research and Woodruff Health Sciences IT Division—REDCap).