Prediction of right ventricular failure after left ventricular assist device implantation in patients with heart failure: a meta-analysis comparing echocardiographic parameters

Abstract

 

OBJECTIVES

Between 10% and 40% of patients who receive a left ventricular assistance device (LVAD) suffer from right ventricular failure (RVF) shortly after the device is implanted. Patients with post-LVAD RVF tend to have poor outcomes. Only a few predictive factors concerning the right ventricle (RV) have been investigated. Our goal was to search for non-invasive variables that correlate with RV function, focusing on echocardiographic parameters of the RV.

METHODS

We selected 3 parameters: tricuspid annular plane systolic excursion, right ventricular fractional area change and right ventricular global longitudinal strain. We searched the literature and pooled relevant studies in a meta-analysis. Finally, we performed a statistical analysis to confirm whether each parameter was a reliable predictor of RVF after LVAD implantation.

RESULTS

We retained 19 articles involving a total of 1561 patients. We found a pooled standardized mean deviation of −0.13 cm for the tricuspid annular plane systolic excursion, with the lower and upper tails of −0.21 and −0.04 cm, respectively. Concerning the right ventricular fractional area change, the averaged standardized mean deviation was equal to −2.61%, with the lower and upper extremities of −4.12% and −1.09%, respectively. Finally, regarding the global longitudinal strain, the standardized mean deviation was equal to −2.06% with an uncertainty value between −3.23% and −0.88%.

CONCLUSIONS

The tricuspid annular plane systolic excursion could be a reliable parameter in RVF prediction. The right ventricular fractional area change and global longitudinal strain are likely to be stronger predictors of RVF after LVAD implantation. Prospective studies should be carried out to confirm this observation.

Heart failure (HF) is a clinical condition that occurs when the heart is no longer able to function sufficiently to supply the oxygen demands of the body [1].

INTRODUCTION

Heart failure (HF) is a clinical condition that occurs when the heart is no longer able to function sufficiently to supply the oxygen demands of the body [1]. Despite spectacular progress and innovation in the management of patients with HF the treatment of choice remains a heart transplant. While waiting for a heart transplant, it is possible for patients to benefit from a ventricular assistance device. These devices, which are surgically implanted, mainly in the left ventricle (LV) [left ventricular assistance device (LVAD)], are suitable for patients with optimal medical management and a reduced left ventricular ejection fraction [2, 3]. However, an LVAD is also associated with common complications, such as device infections or bleeding. One actual problem is that 10–40% of patients who benefit from an LVAD will later present with right ventricular failure (RVF). Patients with post-LVAD RVF tend to have worse outcomes than those without RVF, including multiple organ deficiency [4–6]. To avoid such poor outcomes, there is a need to classify patients as those for whom the right ventricular function is good enough to tolerate an LVAD and those for whom it might be more suitable to apply a biventricular assistance device (BiVAD), which assists both ventricles.

We searched for non-invasive variables that correlated with right ventricle (RV) function. We decided to focus on the RV echocardiographic parameters. Because of its particular anatomy, the RV has long been neglected, whereas the structure and function of the LV have been studied extensively [7–9]. The literature describes the predictive value of echocardiography as regards the RV in only a few individual studies. Only 1 meta-analysis, from Bellavia et al. [10], has been published thus far. Moreover, in the other individual studies, correlations with echocardiographic parameters were not well established: The studies described variations and contradictions concerning the reliability of the echocardiographic variables [10]. We decided to focus on 3 parameters that have been widely used to investigate RV function: tricuspid annular plane systolic excursion (TAPSE), RV fractional area change (RVFAC) and RV global longitudinal strain (GLS). The goal was to determine if significant difference in preoperative value before LVAD implantation exists between patients who presented with RVF and those who did not in order to draw conclusions about the predictive value of these parameters.

MATERIALS AND METHODS

Search strategy

We systematically searched the MEDLINE and EMBASE databases according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA). The MEDLINE database was investigated using OVID. A protocol was established before any research was begun. The period of interest was 1 January 1995–31 December 2020. We applied a cut-off before 1995 because of the dramatic evolution of LVAD working principles that has occurred in recent years (e.g. pneumatic pulsatile pump vs a continuous-flow device). Echocardiographic machines and parameters have gone under a comparable evolution (e.g. TAPSE and GLS). In the different databases, we combined terms related to the dysfunction of the RV in patients with an LVAD with terms referring to prediction using echocardiographic parameters. Our search strategy is available in the Supplementary Material.

Two investigators independently researched the relevant literature using the search strategy and sorted results according to the inclusion and exclusion criteria listed below. Data collection was performed in duplicate by 2 authors (L.-E.C. and P.T.) independently. References from articles were also reviewed. Disagreements were resolved by a discussion between the 2 investigators. If no consensus was reached, the study was removed from the pool.

Inclusion and exclusion criteria

We initially selected studies based only on their title and abstract. We first reviewed articles that reflected the topics RVF and/or ventricular assistance device. At this stage of the selection, we had no concerns with respect to the type of assistance (LVAD or BiVAD), failure type after surgery, or which echocardiographic investigations were performed.

The second step in our study selection was to apply more accurate criteria to include only published full-text articles. We retained studies that contained the required information, such as data on RVF and no-RVF patient groups, patient numbers in each group, data expressed as mean ± standard deviation (SD) or median and Q1–Q3 interval and echocardiographic preimplantation investigations for each group. Studies without any of these information areas were excluded. In addition, we excluded studies that did not use 2-dimensional transthoracic echocardiography, that were case reports or narrative reviews or that investigated RVF after a planned BiVAD implantation owing to the anticipated RVF risk.

Assessment of study quality

The final step before validating the pool of studies was to assess the quality of each individual article. The 22-point STROBE checklist was used to evaluate the validity and therefore the reliability of the articles. Only high-quality reports (a score over 15) were included in the review.

Outcome and definition of right ventricular failure

The goal of our systematic review was primarily to investigate whether there was an association between the preimplantation value of a parameter and the occurrence of RVF. In line with the definitions of the Interagency Registry for Mechanically Assisted Circulatory Support, we retained studies that defined the occurrence of RVF as (i) 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 elevated left atrial/pulmonary capillary wedge pressure >18 mmHg, tamponade, ventricular arrhythmia or pneumothorax); (ii) requirement for either right ventricular assistance device or BiVAD; and (iii) need for inhaled nitric oxide (or another pulmonary vasodilator) or inotropic therapy for >7 days any time after LVAD implantation [11].

Data collection

Once the definitive selection of publications had been completed, we carefully reviewed each article individually. The data of interest were the total sample size, number (n) of patients who benefitted from LVAD implantation, detailed number of patients in the RVF and no-RVF groups, study design, selection criteria and echocardiographic TAPSE (in cm or mm) values, RVFAC (in %) and right ventricular GLS (in %) for each group.

Statistical analyses

Our hypothesis was that any echocardiographic parameter that could have a potential predictive value of RVF would therefore display a significant difference in value between patients with RVF and patients without RVF. Because continuous variables were used, we defined the effect size of each study for a parameter by the difference between the mean values of the 2 groups. The result was divided by the SD of the study to obtain the standardized mean deviation (SMD). We used a random-effect model to aggregate the SMD of the different studies for each parameter. We displayed the results in forest plots for each echocardiographic parameter. If a study did not report values for a parameter, the study was removed from the pool for this parameter. For values that were displayed as median and Q1–Q3, we used the median as the mean, and the stated SD was equal to the interquartile range divided by 1.35. The conversion was based on the assumption of no other distribution occurring in the study aside from normal distribution.

We assessed the heterogeneity between studies by computing Cochrane’s Q statistic and the I2 value to stage it. We considered a P-value <0.05 statistically significant. Under this value, we considered that we could not exclude a lack of homogeneity in the study pool. Following the Cochrane recommendations, our reference for interpreting the I2 value was 0–30%: might not be important; 30–60%: may represent moderate heterogeneity; 60–75%: may represent substantial heterogeneity; 75–100%: considerable heterogeneity [12].

We applied funnel plots to investigate potential publication bias with 95% confidence intervals (CIs). In line with Stern and Egger’s method, we plotted the mean difference on the x-axis and 1/SE on the y-axis. Interpretation of the funnel plots was performed visually by screening for any study asymmetry or suggestive lack in any side around the mean effect, which could indicate a publication bias.

The statistical computations and graphic generation were performed using Excel and STATA software (StataCorp LP, College Station, TX, USA).

RESULTS

Study selection

After removing the duplicates, 2 authors screened 3289 studies on the basis of their title and abstract. The process of inclusion and exclusion of the studies is described in Fig. 1. Overall, we retained 19 publications involving a total of 1561 patients and evaluated the parameters of interest. Of these patients, 468 developed RVF after LVAD implantation (29.9%). Of the publications retained, 12 [13–24] were retrospective reviews, and 7 [25–31] were prospective studies. In the study pool, 18 displayed TAPSE results [13–30], 15 reported values for RVFAC [13–15, 17, 18, 20, 22, 24–31] and 7 recorded GLS values [14, 15, 22, 25, 26, 29, 30]. Tables 1 and 2 display individual study characteristics and the values for the parameters of interest, respectively.

Figure 1:

PRISMA flow chart of included and excluded studies in the meta-analysis. BiVAD: biventricular assistance device; LVAD: left ventricular assistance device; RVF: right ventricular failure.

Table 1:

Study characteristics

First author, year of publication	Region	Study design	Start year	End year	N size	RVF definition
Aissaoui et al., 2015 [25]	EU	PC	2001	2011	42	RVAD or POI > 14 days
Bellavia et al., 2020 [13]	EU	RCC	2010	2020	74	RVAD or POI/iNO > 7 days
Boegershausen et al., 2017 [14]	EU	RCC	2014	2016	44	RVAD or POI > 7 days
Cameli et al., 2013 [26]	EU	PC	NR	NR	10	iNO > 48 h, POI > 14 days or RVAD
Charisopoulos et al., 2018 [15]	UK	RCC	NR	NR	70	RVAD
Cordtz et al., 2014 [16]	EU	RCC	2007	2012	33	iNO > 48 h POI > 14 days or RVAD
Dandel et al., 2013 [27]	EU	PC	2006	2012	205	RVAD, POI > 10 days or iNO > 10 days
Grant et al., 2012 [28]	USA	PC	2007	2011	117	RVAD or POI > 14 days
Kalogeropoulos et al., 2016 [29]	USA	PC	2012	2012	38	RVAD or POI > 7 days
Kato et al., 2012 [17]	USA	RCC	2007	2010	111	RVAD or POI > 14 days or iNO > 48 h
Kato et al., 2013 [30]	USA	PC	2010	2012	68	RVAD or POI > 14 days or iNO > 48 h
Kiernan et al., 2015 [18]	USA	RCC	2008	2011	26	BiVAD or POI > 14 days
Patil et al., 2015 [19]	UK	RCC	2003	2013	152	RVAD
Puwanant et al., 2008 [31]	USA	PC	2004	2007	33	Pulmonary vasodilators or POI > 14 days
Raina et al., 2013 [20]	USA	RCC	2008	2011	42	RVAD or POI > 14 days
Raymer et al., 2019 [21]	USA	RCC	2008	2014	216	RVAD or POI > 14 days
Ruiz-Cano et al., 2020 [22]	EU	RCC	2016	2018	80	RVAD or POI > 14 days or iNO
Sert et al., 2020 [23]	EU	RCC	2013	2016	71	RVAD or POI/iNO > 14 days
Vivo et al., 2013 [24]	USA	RCC	2004	2011	109	RVAD or POI > 14 days

BiVAD: biventricular assistance device; EU: European Union; iNO: inhaled nitric oxide; NR: not reported; PC: prospective cohort; POI: postoperative inotropes; RCC: retrospective case-control; RVAD: right ventricular assistance device; RVF: right ventricular failure.

Table 2:

Study population

First authors, year of publication	Number of men, n (%)	Mean age (SD)	LVAD therapy (DT/BTT/BTR) (%)	HF ischaemic aetiology (%)
Aissaoui et al., 2013 [25]	44 (90)	56 (15.6)	33/67/–	47
Bellavia et al., 2020 [13]	67 (91)	59 (10.5)	22/58/20	NR
Boegershausen et al., 2017 [14]	46 (85)	61.2 (7.5)	33/37/30	83
Cameli et al., 2013 [26]	9 (90)	66.2 (5.0)	NR	60
Charisopoulos et al., 2018 [15]	59 (84)	47.4 (11.8)	0/100/0	25
Cordtz et al., 2014 [16]	29 (88)	53.6 (21.83)	NR	30
Dandel et al., 2013 [27]	182 (89)	52 (27.4)	NR	42
Grant et al., 2012 [28]	92 (79)	58 (12.96)	33/67/–	40
Kalogeropoulos et al., 2016 [29]	25 (61)	51.8 (13.5)	58.5/41.5/–	34
Kato et al., 2012 [17]	87 (78)	55.7 (15.2)	NR	32
Kato et al., 2013 [30]	61 (90)	62.6 (11.8)	NR	44
Kiernan et al., 2015 [18]	19 (73)	46 (15.9)	NR	58
Patil et al., 2015 [19]	121 (80)	44.3 (13.3)	–/100/–	10
Puwanant et al., 2008 [31]	NR	53.7 (12.7)	21.2/66.8/12	67
Raina et al., 2013 [20]	30 (71)	54.5 (17.4)	45/55/–	41
Raymer et al., 2019 [21]	173 (80)	58.1 (10.9)	36.5/63.5/–	45
Ruiz-Cano et al., 2020 [22]	72 (90)	61 (57.7)	37.5/62.5/–	34
Sert et al., 2020 [23]	59 (83)	42.9 (14.6)	NR	35
Vivo et al., 2013 [24]	84 (77)	54 (13)	49/49/2	56

BTR: bridge to recovery; BTT: bridge to transplant; DT: destination therapy; HF: heart failure; LVAD: left ventricular assistance device; SD: standard deviation; NR: not reported.

Patient characteristics

The population of interest comprised 1561 patients. The mean age was 54.1 years. The predominant gender was male (79%). The goal of the LVAD was described for 1006 patients: it was a bridge to a transplant for 684 (67.9%), destination therapy for 276 (27.4%) and a bridge to recovery for 46 (4.7%). HF was primarily due to non-ischaemic events (59.6%).

Heterogeneity evaluation and funnel plots

We assessed the pool’s heterogeneity for each parameter. Computations with the pool of publications for TAPSE reported I2 = 65.22%, so we could not exclude a moderate lack of homogeneity.

For studies reporting the RVFAC, we obtained I2 = 21.86% and a P-value of 0.21. No lack of homogeneity was suspected. Finally, for the GLS, we found I2 = 57.34% and P = 0.03; and so could not exclude a moderate lack of homogeneity.

The funnel plots (see Supplementary Material) assessing the publication bias did not show any obvious bias in the study pool. In addition, since no to a moderate level of heterogeneity was assessed, no meta-regression or subgroup analysis was performed.

Echocardiographic parameters

We found a pooled SMD of −0.13 cm for TAPSE with lower and upper tails ranging from −0.21 to −0.04 cm; thus, we found significance for predicting RVF. The TAPSE results are illustrated in Fig. 2.

Figure 2:

Forest plot for tricuspid annular plane systolic excursion parameter. CI: confidence interval; N: number of patients; RVF: right ventricular failure; SD: standard deviation.

Concerning RVFAC, with a summary SMD of −2.61% and lower and upper extremities of −4.12% and −1.09%, respectively, we established that the RVFAC was lower in patients who developed RVF compared to those without complications and that the differences were statistically significant. The RVFAC results are displayed in Fig. 3.

Figure 3:

Forest plot for right ventricle fractional area change parameter. CI: confidence interval; N: number of patients; RVF: right ventricular failure; SD: standard deviation.

Finally, we found a significantly higher rate of RVF post-LVAD implantation in patients exhibiting a lower GLS. Indeed, as shown in Fig. 4, the summary SMD for GLS was equal to −2.06% with an uncertainty value between −3.23% and −0.88%.

Figure 4:

Forest plot for global longitudinal strain parameter. CI: confidence interval; N: number of patients; RVF: right ventricular failure; SD: standard deviation.

All the intervals corresponded to 95% CI.

DISCUSSION

RVF is a frequent complication arising after LVAD implantation. The mechanistic explanation for this is that a patient with left-sided HF that requires an LVAD will most likely exhibit general HF, thus displaying an RV that is weaker than predicted. Therefore, upon LVAD implantation, the rise in cardiac output due to the assistance cannot be well tolerated by the RV, which leads to RVF [6]. Another proposed explanation is that the LVAD, when set, will drive ventricular suction on the septum, which will thus move into the LV. Such a move could significantly impact RV function [7]. Despite its postoperative incidence of 10–40% [5, 32], it remains difficult to predict RVF occurrence for a patient who undergoes LVAD. Predictive scores of RV function exist. However, these scores appear to be unable to predict the need for RV support after an LVAD implant [33]. Only 1 meta-analysis has been published concerning predictive factors for RVF during the preimplantation assessment [10]. This research adds several clinical characteristics with good specificity, such as the need for continuous renal replacement therapy or mechanical ventilation, as well as other biological and haemodynamic parameters like white blood cell counts or central venous pressure, to the field of promising RVF predictors.

We searched for non-invasive variables that correlated with RV function. We decided to focus on the RV echocardiographic parameters. In echocardiography, the RV is difficult to visualize, and it is thus challenging to obtain reliable data. This difficulty is mainly due to its particular anatomy, i.e. its smaller size and triangular shape. Therefore, quantifying RV function appears challenging, because it requires many assumptions. For those reasons, the RV has been neglected for a long time, whereas the structure and function of the LV has been studied extensively. In contrast to the LV, no exact geometric model exists for the RV in the echocardiography field [7–9].

Bellavia et al. [10] also investigated echocardiographic parameters as reliable predictors for post-LVAD RVF. It has been suggested that the RV/LV ratio and right ventricular free-wall longitudinal strain should be routinely assessed using echocardiography because of their predictive impact. Indeed, the higher the ratio, the more likely is the patient to develop RVF once cardiac output has been restored due to the insertion of an LVAD. Another echocardiographic parameter used to assess RV function is TAPSE, which is commonly discussed in the literature in this indication [8, 9]. In our study, TAPSE seems to be a reliable parameter for RVF prediction. Indeed, we observed a significant SMD representing the difference of TAPSE values between the groups with and without RVF. In addition to a CI close to 0 for the SMD, this observation may be interpreted with caution following other reports because TAPSE is likely not a strong indicator of RV function in patients with severe RV dysfunction, such as those with regional motion abnormalities or severe tricuspid regurgitation [34, 35]. This lack of reliability may be explained by the way TAPSE is used to investigate RV function: It is a 2-dimensional measurement assessing the motion of a 3-dimensional structure exposed to a load dependency. Moreover, this parameter only investigates the longitudinal component of the RV contraction, thereby limiting its reliability [10]. With a more pronounced difference between the RVF and the non-RVF groups, RVFAC was found to be a likely indicator of RVF occurrence after insertion of an LVAD. Indeed, the RVF group presented a significantly lower RVFAC compared to the group without RVF. This result was in line with the European guidelines on RV imaging: RVFAC offers a reliable indication of RV systolic function evaluation, owing to the global evaluation it provides. The guidelines also state that a 35% cut-off value is suggestive of systolic dysfunction [8, 36]. However, in our study, most patients already exhibited an RVFAC value <35%. We concluded that RVFAC is an interesting parameter in predicting RVF but that further research is needed for a better understanding and for its application in practice, such as setting a cut-off value for the preimplantation LVAD setting.

Because of the RV shape, longitudinal strain is a suitable measurement of systolic function. This fact can be explained as follows: RV contraction is predominantly generated by longitudinal shortening reduction rather than by cavity diameter reduction [37], which can be accounted for by the composition of the myocardial tissue wall of the RV. Muscle layers are divided into superficial and deep layers, with differing fibre organizational patterns: The superficial layer is arranged circumferentially, whereas the deeper muscles are longitudinally aligned from base to apex and account for ∼80% of RV contraction [38]. In our study, this parameter was identified as a potential candidate to evaluate the probability of RVF: The between-group difference in GLS values was significant, with a lower GLS in the RVF versus the non-RVF group. However, the cut-off values displayed in the literature for this parameter did not apply to the specific preimplantation population; therefore, more in-depth investigations are required [8, 38].

Insufficiency of the tricuspid valve is also a matter of concern in patients with an LVAD. Incompetence of the tricuspid valve is common among patients with HF. However, current practices for patients undergoing surgery for advanced cardiomyopathy do not recommend treatment of tricuspid valve regurgitation [39]. Indeed, after restoring the coaptation of the leaflets, an increase in RV afterload could occur, favouring an RV failure. When one considers the haemodynamic modifications brought about by an LVAD implant, such as reshaping the RV or reducing the LV filling pressure, the value of preoperative tricuspid insufficiency observed by echocardiography is subject to discussion. Researchers have concluded that an LVAD implant could reduce tricuspid regurgitation whereas significant residual tricuspid incompetence following the operation is also frequent and associated with poorer outcomes [40, 41]. As a solution to deal with these considerations, Veen et al. [42] performed a meta-analysis showing that concomitant tricuspid surgery with an LVAD implant did not worsen outcomes and, according to some data, could even be beneficial.

No severe lack of homogeneity was observed based on the values of the I2 statistics in the pool of the 3 parameters. However, the P-values of the TAPSE pool and the GLS pool were significant; thus, we cannot exclude moderate heterogeneity among these studies. No publication bias was suspected after looking at the funnel plots, none of which displayed asymmetry among the studies.

Limitations

The main limitations of this meta-analysis are the inherent limitations of the retrospective and prospective studies included in the analysis. Regarding the pooled articles, we cannot exclude several biases, such as selection bias. Concerning GLS, even if we did not identify any bias or lack of homogeneity among the publications, a statistical evaluation was performed on a small number of studies, which could decrease the viability of the investigation when taking into account its heterogeneity. In addition, experience of the operator play a significant role in the measures of the echocardiographic parameters and thus could differ among the studies. Finally, the aggregated nature of this meta-analysis is another major limitation because it introduces higher variability and confounding factors specific to the studies but also to the patients.

CONCLUSION

It is possible to predict the occurrence of RVF using non-invasive measures. We focused on echocardiographic assessments and found that TAPSE is likely to be a reliable parameter of RVF occurrence after LVAD implantation, but other parameters, such as RVFAC and GLS, may also be interesting and worth investigating. During our data collection, we observed that only a few studies reported echocardiographic parameters, especially GLS. It might be profitable to systematically report these parameters when studying RV and to conduct prospective studies to increase our understanding of their potential use as RVF predictors in order to provide better care for patients with HF.

SUPPLEMENTARY MATERIAL

Supplementary material is available at ICVTS online.

Funding

This work was supported by the University of Lausanne and the Department of Cardiovascular Surgery of Lausanne University Hospital.

Conflict of interest: none declared.

Author contributions

Louis-Emmanuel Chriqui: Conceptualization; Investigation; Methodology; Writing—original draft. Pierre Monney: Formal analysis; Supervision; Validation. Matthias Kirsch: Formal analysis; Funding acquisition; Supervision; Validation. Piergiorgio Tozzi: Conceptualization; Funding acquisition; Methodology; Project administration; Writing—review & editing.

Reviewer information

Interactive CardioVascular and Thoracic Surgery thanks Eric J. Lehr and the other, anonymous reviewer(s) for their contribution to the peer review process of this article.

ABBREVIATIONS

BiVAD

Biventricular assistance device

CI

Confidence interval

GLS

Global longitudinal strain

HF

Heart failure

LV

Left ventricle

LVAD

Left ventricular assistance device

RV

Right ventricle

RVF

Right ventricular failure

RVFAC

Right ventricle fractional area change

SD

Standard deviation

SMD

Standardized mean deviation

TAPSE

Tricuspid annular plane systolic excursion

 

Presented at the 34th Annual Meeting of the European Association for Cardio-Thoracic Surgery, Barcelona, Spain, 8–10 October 2020.