Right Ventricular Longitudinal Strain In Left Ventricular Assist Device Surgery – A Retrospective Cohort Study

Abstract

Objectives

Right ventricular (RV) failure is common after left ventricular assist device (LVAD) surgery and is associated with higher mortality. Measurement of longitudinal RV strain using speckle-tracking technology is a novel approach to quantify RV function. We hypothesized that depressed peak longitudinal RV strain measured by intraoperative transesophageal echocardiography (TEE) examinations would be associated with adverse outcomes after LVAD surgery.

Design

Retrospective Cohort Study.

Setting

Tertiary academic medical center.

Participants

Following IRB approval, we retrospectively identified adult patients that underwent implantation of non-pulsatile LVAD. Exclusion criteria included: inadequate TEE images and device explantation within six months for heart transplantation.

Interventions

None.

Measurements and Main Results

The postoperative adverse event outcome was defined as a composite of one or more of death within six months, ≥14 days of inotropes, mechanical RV support, or device thrombosis. Intraoperative TEE images were analyzed for peak RV free wall longitudinal strain by two blinded investigators. Simple logistic regression was used to assess the relationship between adverse outcome and the mean of the strain measurements of the two raters. Agreement between the raters was assessed by intra-class correlation (0.62) and Pearson correlation coefficient (0.63). Of the 57 subjects, 21 (37%) had an adverse outcome. The logistic regression indicated no significant association between RV peak longitudinal strain and adverse events.

Conclusions

In this retrospective study of patients undergoing non-pulsatile LVAD implantation, peak longitudinal strain of the RV free wall was not associated with adverse outcomes within six months after surgery. Additional quantitative echocardiographic measures for intraoperative RV assessment should be explored.

Introduction

Heart failure is extremely common with an incidence of 10 per 1000 population older than 65 years of age and 1-year mortality of 20%.¹ Left ventricular assist devices (LVADs) are becoming more prevalent for the management of heart failure with the Interagency Registry for Mechanically Assisted Circulatory Support (INTERMACS) reporting 2,502 patients implanted in 2014, compared to only 1,010 patients in 2009.² Right ventricular (RV) failure occurs in 20–35% of patients following durable LVAD implantation, and it is associated with significantly worse survival, even if RV mechanical support is initiated.^{3, 4} Additionally, patients who require mechanical support for RV failure in conjunction with LVAD implantation are successfully weaned from RV mechanical support in only about half of all cases.⁵ Given that the proportion of patients who are unlikely to bridge to heart transplantation is rising, recognizing and treating RV dysfunction is critical at every stage of LVAD therapy. While intra-operative transesophageal echocardiography (TEE) provides many sophisticated methods for evaluation of left ventricular function, RV assessment is much less mature.⁶

While there exist multiple quantitative methods to assess RV function that have been validated including fractional area change (FAC), tissue doppler systolic velocity, strain via tissue doppler or speckle tracking, and 3-D imaging, a gold standard using transesophageal echocardiography in the operating room has yet to be established for patients undergoing cardiac surgery^{6, 7}.

The use of tricuspid annular plane systolic excursion (TAPSE) remains the most feasible from a quantitative standpoint in the operating room, with TAPSE being the most consistently obtained compared to FAC or speckle tracking strain values.⁸ When compared to TTE M-mode, speckle tracking TAPSE may more accurately compare than TEE M-mode derived TAPSE.⁹ The alignment of cardiac motion with the doppler interrogation angle can be a major limitation in addition to the far field position of the RV relative to the esophagus when using doppler modalities. The geometric shape of the RV also limits many of the modalities that have been useful for LV function analysis. The limitations in many speckle tracking and 3D modalities are still hampered by the need for offline analysis.

Current guidelines endorse the use of peak longitudinal strain via speckle tracking in an effort to advance RV assessment from a purely qualitative evaluation of normal to mild, moderate or severely reduced function, towards yielding a quantitative metric.⁷ Longitudinal strain analysis using transthoracic echocardiography has proven useful in determining RV function in LVAD patients and predict their likelihood of RV-associated morbidity and mortality.^10–12 Given that strain analysis is not dependent on the angle of incidence of the ultrasound beam, it may be an especially desirable technique when the ultrasound probe is limited by its position in the esophagus (i.e. during TEE). Here, we hypothesized that depressed free wall peak longitudinal RV strain determined from intraoperative TEE images would be associated with adverse outcomes within six months after implantation of durable, non-pulsatile LVADs (HeartMate II and Heartware devices).

Methods

The local Institutional Review Board (IRB) approved this study. The IRB waived the requirement for obtaining written consent. This manuscript was written in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement for the reporting of cohort studies.¹³

Selection and Description of Participants

All adult patients having had a non-pulsatile durable LVAD implanted at the study site were eligible for inclusion. Exclusion criteria were the presence of a previous non-pulsatile durable LVAD, the explantation of the primary device for heart transplantation within 180 days after the implant date, and inadequate or unavailable intraoperative TEE images.

To identify patients who had received HeartMate II or Heartware LVAD at the University of Colorado, we obtained the registry data for our site from INTERMACS. In the period from 07/10/2006 until 03/31/14 we identified 89 records of patients (n=81 HeartMate II, and n=8 for Heartware). One record was a replacement device for a patient with a pre-existing durable LVAD device. Of the remaining 88 patients, 21 patients had their primary device explanted for transplant within 180 days. Of the remaining 67 patients, ten patients did not have adequate intraoperative midesophageal four-chamber view images for analysis. Determination of inadequate image quality was based on consensus of the two echocardiographers. For the final analysis, 57 patients were included (Figure 2).

Figure 2.

Study enrollment flow diagram.

Demographics and Michigan RV failure predictive score

Retrospective chart review was performed to obtain pre-operative data used in the calculation of the Michigan RV failure prediction score.¹⁴ Baseline demographics, pre-operative care and lab values (AST, bilirubin, creatinine, and vasopressor requirements) were collected in a manner consistent with prior studies, for the period of 24 hours before LVAD placement. Information from the operative record was collected for LVAD device type and cardiopulmonary bypass times.

Technical Information

Longitudinal strain analysis was performed offline using commercially available software (Syngo US Workplace, Siemans Medical Solutions, Inc, Ultrasound Division, Mountain View, CA). Pre-cardiopulmonary bypass (CPB) images from the mid-esophageal four-chamber or modified four-chamber view were used to trace the RV endocardial border, and strain curves were generated (Figure 1). Peak longitudinal strain was generated from the three segments corresponding to the RV free wall and averaged to generate the mean peak longitudinal strain value. Two cardiac anesthesiologists certified in advanced perioperative TEE performed the strain analysis independently. Both were blinded to the clinical outcomes of the patients analyzed.

Figure 1.

Example of longitudinal strain analysis of the TV free wall, using speckle tracking SYNGO software. In this example, the RV apex is incompletely depicted to optimize imaging of the RV free wall.

Statistics

Summary statistics (Table 1 and 2) were reported as mean (standard deviation) for continuous data and counts (proportions) for categorical data. Tests for associations between demographic factors, CPB time, and Michigan Score with adverse RV-related events were performed with t-tests except for gender and LVAD device type (Fisher’s exact tests). Laboratory values and other clinical variables are reported with t-tests other than AST (Wilcoxon test), vasopressor use, and renal replacement therapy (Pearson Chi-square). P-values reported in Table 1 are for 2-sided tests.

Table 1.

Pre-operative demographic information. Vasopressor requirement was defined as receiving phenylephrine, norepinephrine, or vasopressin

Patient and Procedure Characteristics	Total (n=57)	>=1 Adverse Outcome (n=21)	No Adverse Outcome (n=36)	P-value
Female n (%)	11 (19)	7 (33)	4 (11)	0.0783
Age on Implant Date MEAN (STD)	53.4(14.1)	57.3(10.5)	51.0(15.5)	0.1032
Device Type
Heartware n (%)	7 (12)	2 (10)	5 (14)	1.0000
Heartmate II n (%)	50 (88)	19 (90)	31 (86)
CPB Length (min) MEAN (STD)	96.1(42.9)	103.2(47.9)	91.9(39.8)	0.3394
Michigan Score MEAN (STD)	4.0(2.6)	4.6(2.8)	3.7(2.4)	0.1768
Vasopressor requirement	41 (72)	16 (76)	25 (69)	0.5845
AST (IU/liter) MEDIAN (IQR)	34(25,45)	36(28,49)	30(22,44)	0.0928
Bilirubin total (mg/dl) MEAN (STD)	1.7(1.0)	2.2(1.3)	1.5(0.7)	0.0108
Creatinine (mg/dl) MEAN (STD)	1.3(0.4)	1.4(0.4)	1.3(0.4)	0.7538
Renal replacement therapy n (%)	6 (11)	3 (14)	3 (8)	0.4800

Table 2.

Mean RV longitudinal strain measurements

	Overall (n=57)	>=1 Adverse Outcome (n=21)	No Adverse Outcome (n=36)
RV strain (STD)	−9.7 (4.3)	−9.2 (4.2)	−10.0 (4.4)

Inter observer agreement was assessed with Pearson correlation analysis and intra-class correlation (ICC). Common cutoffs used for ICC follow: ICC values less than 0.4 indicate poor agreement between raters, 0.40–0.59 indicate fair agreement, 0.60–0.74 indicate good agreement, and 0.75–1.0 indicate excellent agreement¹⁵.

The adverse event outcome was defined as a composite of one or more of the following events occurring after LVAD implantation: death within six months, more than 14 days on inotropes, need for mechanical RV support, or device thrombosis requiring explant. Associations between the individual adverse outcome types comprising the composite outcome and peak RV longitudinal strain measurements were also ascertained. Simple logistic regression was used to assess the relationship between adverse event outcome and the mean of the peak RV longitudinal strain measurements of the two raters. A one-sided p value was used to test the null hypothesis that the regression coefficient was less or equal to zero versus the alternative hypothesis that the regression coefficient was greater than zero.

The power calculation was based on a previously reported study using transthoracic echocardiography (TTE)-derived strain parameters to monitor RV function before and after LVAD implantation.¹¹ Our sample of 57, would have provided 94% power to detect an odds ratio as small as 1.33 for a 1% decrease in RV strain, assuming RV strain is normally distributed (mean=−16%, SD=4%) and a 0.025 significance level for a one-sided test. All statistical tests were performed using SAS 9.4. (SAS, Cary NC) and Prism 6.0 (GraphPad Software Inc., La Jolla, CA).

Results

From the study population of 57 LVAD patients, 21 (37 %) met criteria for at least one adverse event. Basic demographic data such as gender (p=0.08), age (p=0.10), device type (Heartware vs. Heartmate II) and cardiopulmonary bypass times (CPB) (103.2 vs. 91.9 minutes, p=0.34) were not found to be significant factors to be associated with adverse events. Also, previously published pre-operative RV failure prediction scores (Michigan Score) for those who experienced adverse events were mean (SD) of 4.6 (2.8) and 3.7 (2.4) for those without adverse events, not statistically different for our composite adverse event rates (p=0.18) (Table 1).

The strain measurements of the two raters were significantly (p<0.0001) associated. The Pearson correlation coefficient was found to be 0.63. The ICC was calculated to be 0.62, which indicates good agreement between the two raters (Figure 3). The strain rates are displayed in Table 2. The mean peak RV longitudinal strain of both raters for all patients receiving LVAD implantation was −9.7 (4.3). Those with an adverse outcome had a mean peak strain of −9.2 (4.2) and those without adverse outcomes of −10.0 (4.4), (p=0.53).

Figure 3.

Inter-rater correlation of RV strain measurements. A Pearson correlation coefficient of 0.63 and Intra-class correlation (ICC) of 0.62 were obtained for the inter-rater correlation between two investigators blinded to each other’s results.

The results of the logistic regression are depicted in Table 3. On average, for every 1% increase in average strain (less negative strain indicative of reduced systolic function), the odds of an adverse event increased 4% (95% CI of −10%, +21%) which is not significantly greater than zero (p=0.26). There was no significant association between individual rater measurements and adverse events.

Table 3.

Regression analysis for composite outcomes vs strain rate. Averages and individual rater’s analysis. Hosmer and Lemeshow Tests (p=0.74) indicate no significant lack of fit for the logistic model. However, the c-index (0.54) indicates poor ability of the model to discriminate between low risk and high risk subjects.

	Parameter Estimate	Odds Ratio	(95% CI)	One sided p value
Average	0.0415	1.042	(0.90,1.21)	0.26
Rater A	−0.005	0.995	(0.88,1.13)	0.47
Rater B	0.063	1.065	(0.95,1.19)	0.13

From the adverse event cohort, 8 (38%) patients required RV mechanical support and 8 (38%) patients were on prolonged inotropic therapy. Five (24%) patients had devices explanted without recovery or transplantation and 6 (29%) died within six months of surgery (Table 4). When individual adverse events were analyzed by cause from the composite index, there was also no significant relationship discovered for mechanical support (p= 0.77), prolonged inotropic support (p= 0.40), death within six months (0.24), with a non-significant trend for device explantation. (Table 5).

Table 4.

Specific adverse events reflected in the composite postoperative outcome.

Adverse Outcome	>=1 Adverse Outcome (n=21)
RV mechanical support	8 (38%)
Inotropes>14 days	8 (38%)
Device explanted other than for transplant or recovery	5 (24%)
Died within 6 months of implant	6 (29%)

Table 5.

Mean LVAD Strain measurements within subgroups of adverse outcomes. Device explanted is for reasons other than recovery or heart transplantation.

	RV Strain Mean for those with Outcome	RV Strain Mean for those without Outcome	P-value
Device Explanted	−6.2(4.3)	−10.0(4.2)	0.0534
Died with 6 months	−7.7(3.3)	−9.9(4.4)	0.2329
Inotropes > 14 days	−10.9(3.4)	−9.5(4.4)	0.4032
RV Support	−10.1(5.2)	−9.6(4.2)	0.7722

Discussion

Our hypothesis that peak intraoperative longitudinal RV free wall strain would be associated with adverse outcomes after LVAD implantation was rejected.

Multiple studies have been performed to predict RV failure after LVAD insertion including using assessment of organ function and hemodynamic data. ^{3, 14, 16} Certainly, the advancement in technology over the last decade, specifically the preference of continuous flow over pulsatile LVADs, complicates interpretation of older studies to current clinical settings. Patients having adverse events in our population had no differences in the Michigan RV failure risk score. Although we did use a composite endpoint that included mortality and device explant as opposed to just isolated RV failure, this is agreement with another follow-up study showing conflicting results to this scoring system’s ability to predict adverse outcomes.¹⁶

Attempts to include echocardiography to enhance the predictive ability of pre-operative evaluation using these models have also been performed. Assessment of RV longitudinal strain has shown shows an additive benefit to RV failure risk scores obtained via TTE.¹² Again using serial TTE examinations, others have also shown a divided population among LVAD recipients, with those having the lowest (less negative) RV strain continuing to decline and having poorer outcomes, while those with higher (more negative) RV strain showing recovery over time.¹⁰

There exists wide variation in reported normal strain values, with additional variation based on modality of the ultrasound imaging (TTE vs TEE). Defined as a percent of deformation, Kowalski et al reported RV free wall strain values of –19 to –32% in healthy subjects, increasing from base to apex.¹⁷ Amongst those patients presenting for cardiac surgery, RV free wall values ranged from –23 to –31%, but decreasing from base to apex.⁸ It is known that pre-existing pulmonary hypertension also reduces strain values compared to control, with ranges from –17 to –23% as assessed by TTE.¹⁸ Patients presenting for LVAD implantation show consistently lower values, with a cutoff of −9.6% predicting RV failure with 76% specificity, and 68% sensitivity.¹² These values were consistent with what we observed in our population for free wall baseline RV strain rates, but the failure of our study to show differences may be the result of an underpowered study as compared to Grant et al.

Although the use of TEE as a means to guide treatment has become a standard in LVAD cases, intraoperative echocardiography may have additional challenges when compared to TTE in assessing RV function and especially when applying strain-imaging technology. Compared to TTE, several other echocardiographic analysis techniques have proven inconsistent for TEE assessment of RV function. Myocardial performance index performed in awake patients by TTE showed poor correlation to TEE derived values when under general anesthesia.¹⁹ No correlation was found between tissue Doppler-derived strain when TEE was compared to TTE values both performed under general anesthesia.²⁰ While the use of TEE-derived strain for RV assessment during cardiac surgery does seem possible using speckle tracking⁸ and comparisons of TEE to TTE may be similar in non-sedated patients,²¹ others have shown only moderate correlation with TTE assessments under general anesthesia.²² Furthermore, recent publications suggested the use of non-strain applications of speckle tracking RV analysis to be more useful when correlating to the TTE standard evaluation of RV function.⁹ Overall, while speckle tracking and strain remain hopeful new candidates for utility in intra-operative assessment, considerable validation work for normalized values and comparisons with other echocardiographic modalities remains.²³

In a recent report by Starling and colleagues’ rates of LVAD thrombosis were found to have increased from 2.2% prior to 2011 to 8.4% until 2013.²⁴ Thrombosis is driven by Virchow’s triad of hypercoagulability, endothelial injury and stasis.²⁵ Stasis is facilitated amongst other factors by decreased flow from impaired RV function.²⁶ With a trend towards significance for patients with poor RV strain who had LVAD explanation in our analysis, this may reflect a subgroup of patients whose RV failure contributed to device explanation.

Limitations of the current study include the retrospective nature of the analysis and single center design. Strain analysis is also dependent on high-quality 2D images, and almost 15% (10/67) of patients eligible for inclusion were excluded due to poor image quality. While a standard perioperative TEE exam was performed, optimization of imaging for RV free wall measurements was not done, which may reduce the quality of speckle tracking and introduce variability. Agreement between investigators was good according to published standards using ICC and consistent with other studies looking at intraoperative TEE strain values using tissue Doppler.²⁰ Furthermore, strain is a load-dependent analysis,²⁷ and we cannot comment on other hemodynamic parameters at the time of TEE that may have affected these measurements. Given the retrospective nature of the study, it also remains a challenge to comment on the decision making process for inotrope therapy or criteria for RVAD placement at the time of clinical care.

In summary, this report suggests a lack of association between intraoperative RV longitudinal strain using intraoperative TEE and adverse outcomes after LVAD implantation. Intraoperative RV strain imaging as a means to guide clinical care is still a developing field that will require further investigation. Our study provides further evidence that there may be distinct differences between transthoracic and intraoperative transesophageal echocardiography in the context of RV strain analysis. Current risk models used to predict RV failure after LVAD surgery may not be directly translatable to intraoperatively derived TEE imaging data.