Preoperative Three-Dimensional Strain Imaging Identifies Reduction in Left Ventricular Function and Predicts Outcomes After Cardiac Surgery Kimberly Howard-Quijano, MD, MS,* Ali Salem, MD,* Charles Barkulis, DO,* Einat Mazor, RDCS,* Jennifer C. Scovotti, MA,* Jonathan K. Ho, MD,* Richard J. Shemin, MD,† Tristan Grogan, MS,‡ David Elashoff, PhD,§ and Aman Mahajan, MD, PhD* BACKGROUND: Echocardiography-based speckle-tracking strain imaging is an emerging modality to assess left ventricular function. The aim of this study was to investigate the change in left ventricular systolic function after cardiac surgery with 3-dimensional (3D) speckle-tracking strain imaging and to determine whether preoperative 3D strain is an independent predictor of acute and long-term clinical outcomes after aortic valve, mitral valve, and coronary artery bypass grafting operations. METHODS: In total, 163 adult patients undergoing aortic valve, mitral valve, and coronary artery bypass surgeries were enrolled prospectively and had complete data sets. Demographic, operative, and outcome data were collected. 3D transthoracic echocardiograms were preformed preoperatively and on second to fourth postoperative day. Blinded off-line analysis was performed for left ventricular 2-dimensional (2D) ejection fraction (EF2D) and 3D ejection fraction (EF3D) and global peak systolic area, longitudinal, circumferential, and radial strain. RESULTS: 3D global strain correlated well with EF3D. Ventricular function as measured by strain imaging decreased significantly after all types of cardiac surgery. When preoperative EF3D was used, receiver operating characteristic curves identified reference values for 3D global strain corresponding to normal, mildly reduced, and severely reduced ventricular function. Normal ventricular function (EF3D ≥ 50%) corresponded to 3D global area strain −25%, with area under curve = 0.86 (0.81–0.89). Patients with reduced preoperative 3D global area strain had worse postoperative outcomes, including length of intensive care unit stay (4 vs 3 days, P = .001), major adverse events (27% vs 11%, P = .03), and decreased 1-year event-free survival (69% vs 88%, P = .005). After we controlled for baseline preoperative risk models including European System for Cardiac Operative Risk Evaluation score and surgery type, preoperative strain was an independent predictor of both short- and long-term outcomes, including length of intensive care unit stay, postoperative inotrope score, and 1-year event-free survival. CONCLUSIONS: This study shows that cardiac surgery was associated with an acute reduction in postoperative left ventricular function, when evaluated with 3D strain imaging. In addition, preoperative 3D strain was demonstrated to be an independent predictor of acute and long-term clinical outcomes after cardiac surgery. The use of noninvasive 3D transthoracic echocardiogram strain imaging before cardiac surgery may provide added information to aid in perioperative risk stratification and management for these high-risk patients. (Anesth Analg 2017;124:419–28)
M
ore than half a million cardiac surgeries are performed in the United States each year, and left ventricular (LV) systolic function is a strong predictor of their outcomes.1 Preoperative decrement in
From the Departments of *Anesthesiology and Perioperative Medicine, †Cardiothoracic Surgery, ‡Medicine Statistics Core, and §Biomathematics and Medicine, University of California, Los Angeles, Los Angeles, California. Accepted for publication May 6, 2016.
Funding: This study was funded by intramural department funds; Dr. Mahajan is supported by National Heart, Lung, and Blood Institute (NHLBI), National Institutes of Health (NIH) Research Project Grant (R01) HL084261, Bethesda, MD. The authors declare no conflicts of interest. Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal’s website (www.anesthesia-analgesia.org). Reprints will not be available from the authors. Address correspondence to Aman Mahajan, MD, PhD, Department of Anesthesiology and Perioperative Medicine, David Geffen School of Medicine at University of California, Los Angeles, 757 Westwood Blvd, Suite 3325 Los Angeles, CA 90095. Address e-mail to [email protected]. Copyright © 2016 International Anesthesia Research Society DOI: 10.1213/ANE.0000000000001440
February 2017 • Volume 124 • Number 2
ventricular function can cause hemodynamic instability in the acute postoperative period requiring escalating inotropic support, prolonged hospital stay, and result in longterm complications, including increased mortality. Systolic cardiac function routinely has been assessed through echocardiography by the use of 2-dimensional (2D) LV ejection fraction (EF2D) calculations. These traditional measurements are based on mathematical assumptions of LV geometry to calculate volume changes as an indirect measure of myocardial function.1,2 Strain is a direct measure of LV myocardial tissue deformation that may provide additional information regarding both regional and global myocardial functions.3–6 In recent studies that used magnetic resonance imaging (MRI) as a reference standard, strain was demonstrated to measure ventricular function with better accuracy and reliability than left ventricular ejection fraction.2,7–11 Furthermore, Macron et al12 showed that in patients with poor acoustic windows, strain values correlated more closely with MRI measures of ventricular function than traditional EF2D. www.anesthesia-analgesia.org
419
Copyright © 2016 International Anesthesia Research Society. Unauthorized reproduction of this article is prohibited.
Predictive Value of 3D Strain in Cardiac Surgery
Speckle-tracking strain can be measured by both 2D and 3-dimensional (3D) echocardiography; however, the newly developed 3D strain is an improvement over 2D strain measurements because speckle tracking is no longer limited to motion in a 2D imaging plane.8,10,13 The myocardial architecture of the LV leads to the rotational twisting and shortening of the left ventricle during contraction that may impede tracking in the 2D imaging plane.13,14 With 3D speckle-tracking echocardiography, the entire myocardium is acquired in 1 imaging loop so that speckles are tracked continuously without loss or tracking between frames. In addition, all strain measurements are taken from the same image acquisition compared with 2D speckle tracking, which requires 3 long-axis and 3 short-axis images collectively. This ability to measure all types of strain in 1 imaging loop with 3D speckle-tracking echocardiography improves accuracy and decreases the time for acquisition.15 The aim of this study was to investigate the change in LV systolic function after cardiac surgery with the use of 3D transthoracic echocardiogram (TTE) speckle-tracking strain imaging and to determine whether preoperative strain is an independent predictor of acute and long-term clinical outcomes after cardiac surgical operations. We hypothesized that ventricular function would be reduced in the acute postoperative period and that preoperative 3D strain measures of ventricular function would be an independent predictor of acute and long-term outcomes after cardiac surgery.
METHODS
After institutional review board approval and written informed consent, adult patients scheduled for elective cardiac surgery with cardiopulmonary bypass between October 2011 and July 2013 at the Ronald Reagan University of California, Los Angeles Medical Center were enrolled in this prospective study. Surgical procedures included aortic valve replacement for aortic stenosis (AVR), mitral valve repair or replacement for mitral regurgitation (MVR), coronary artery bypass graft (CABG), and combined multiple valve, CABG/valve, or AVR/aortic surgery for aortic insufficiency (combined). Exclusion criteria were cardiac arrhythmias (such as atrial fibrillation and paced rhythms), congenital heart disease, and poor TTE imaging windows.
Clinical Data Collection
Preoperative (age, gender, and European System for Cardiac Operative Risk Evaluation score [EuroSCORE]), intraoperative, and postoperative data were collected.16 Definitions of all data variables collected were per the Adult Cardiac Surgery Database Training Manual, v2.73.a Postoperative outcome variables were duration of tracheal intubation (from intensive care admission to extubation), length of intensive care unit (ICU) stay, length of hospital stay, inotrope requirement (calculated using the equation: dopamine [μk/kg/min] + dobutamine [μg/kg/min] + 100 × epinephrine [μg/kg/min] + 10 × milrinone dose [μk/kg/min]), major adverse cardiac events, or 1-year all-cause mortality.17–19 Major cardiac adverse events per http://www.sts.org/sites/default/files/documents/ACSDTraining Manual-3-14-14.pdf. Accessed March 20, 2013.
a
420 www.anesthesia-analgesia.org
the Society of Thoracic Surgeons database were cardiac arrhythmias (ventricular tachycardia, ventricular fibrillation, and asystole), multiple-organ failure, ventricular assist device (balloon pump or extracorporeal membrane oxygenation), and reoperation. Mortality was defined as all-cause 1-year mortality and was ascertained from electronic hospital records and telephone follow-up. One-year event-free survival was defined as percentage of patients free from any major cardiac adverse events or all-cause mortality up to 1 year after the date of study enrollment.20
Echocardiography
TTEs were performed preoperatively in awake spontaneously ventilating patients on the day of surgery and between 2 and 4 days postoperatively after surgery, once participants were extubated and chest tubes were discontinued. Echocardiographic studies were acquired by the same experienced sonographer (E.M.). Each patient underwent standardized 2D and 3D TTE for the determination of LV functional parameters. Ejection fraction was calculated by 2D echocardiography via Simpson’s methods of disks applied to 2- and 4-chamber apical images, according to the American Society of Echocardiography.21 Ejection fraction was calculated by 3D echocardiography by the use of full-volume data sets. End-diastolic and endsystolic frames were selected, and semiautomated LV border tracking software outlined the endocardial borders and calculated end-diastolic volume, end-systolic volume, and ejection fraction. All echo images were taken with Vivid 9 (GE Healthcare, Milwaukee, WI) with a 3V transducer; 2D images were acquired at frame rate of 55 to 90 frames/s, and 3D images were taken at frame rate of 25 to 55 frame/s.
Two-Dimensional Speckle-Tracking Analysis
Grayscale images obtained for 2D strain analysis included parasternal short axis at mitral valve, mid-papillary muscle, and apical level, as well as 4-chamber, 2-chamber, and apical long-axis, views. Patients were in the left lateral decubitus position or supine depending on the patients’ physical condition and adequate acoustic windows. Time to aortic valve closure was measured by continuous-wave Doppler in the LV outflow tract. 2D strain analysis was performed by 1 of 3 blinded echocardiographers experienced in regional and global strain analyses, using EchoPAC (version 110.1.1) system by GE Healthcare. Strain was measured with the use of automated tracing, with manual adjustments if needed. The integrity of tracking was visually confirmed and ascertained from the credibility shape of the strain curve. The region of interest was readjusted manually as necessary, in 12% of images analyzed. All measurements were obtained using the average of 3 cardiac cycles. Finally, 2D software output included the measurement of peak systolic global longitudinal (GLPS2D), circumferential (GCPS2D), and radial (GRPS2D) strain.
Three-Dimensional Speckle-Tracking Analysis
The 3D echocardiogram data acquisition was obtained in an adjustable imaging sector volume at the LV apex with breath hold lasting 4 heartbeats. The 3D data sets used for strain analysis were the same images used for EF3D analysis as well. Data analysis was preformed off-line by 1 of 3
anesthesia & analgesia
Copyright © 2016 International Anesthesia Research Society. Unauthorized reproduction of this article is prohibited.
experienced blinded echocardiographers using 4D Auto LVQ offline on EchoPAC system by GE Healthcare. The full-volume data set is displayed as orthogonal 4-chamber, 2-chamber, and short-axis images. Mitral annular and apical points are placed on these images in end-diastolic and end-systolic frames. Semiautomated endocardial surface detection software then outlines endocardial borders and tracks the endocardium in all frames and automatically calculates strain. If the endocardial border tracking appeared inadequate, then manual adjustments of automatic tracing were made and the sequence analysis was repeated. Manual adjustment was necessary in 10% of images analyzed. Software output includes calculation of peak systolic global longitudinal (GLS3D), circumferential (GCS3D), radial (GRS3D), and area (GAS3D) strain. Strain is a measure of myocardial deformation during systole; normal peak systolic longitudinal strain has a negative value because of LV shortening during systole and normal circumferential strain also has a negative value with reduction in LV circumference during systole, whereas radial strain has a positive value because of systolic radial thickening. Global area strain (GAS) is a new measurement unique to 3D strain and has a negative value. Representative 3D echocardiographic images and GAS3D output are shown in Figure 1.
Figure 1. Representative 3-dimensional (3D) left ventricular strain diagram depicting preoperative (A) and postoperative (B) global area strain results after mitral valve replacement. Each yellow line represents the area strain curve of an individual myocardial segment, and the white line represents cumulative left ventricular global area strain showing a reduction in postoperative global area strain.
February 2017 • Volume 124 • Number 2
Statistical Analysis
The number of enrolled patients was based on power calculation for 80% power at α = .05 to detect an odds ratio of 2.5 between those with normal and reduced ventricular function for the primary outcome of major adverse events, requiring sample size of 160 patients with additional 15% enrollment to account for possible study dropout.1 Histograms and quartile plots were examined in a bivariate analysis to determine whether distributions of continuous outcomes were approximately normally distributed. If not, the outcomes were logtransformed before analysis. Values were reported as mean (± standard deviation) for normally distributed outcomes and median (interquartile range) for skewed outcomes. The correlation between each strain parameter and 3D volume-based ejection fraction (EF3D) was assessed with Pearson correlation coefficients. EF3D has been shown to be closely correlated with MRI-based measures of ventricular function and was used in this study as the reference measurement to compare the other measures of ventricular function.10,22 Differences comparing degree of correlation between 2D and 3D strain with EF3D were measured with the William test. Receiver operating characteristic (ROC) curves were constructed for each type of strain to determine the cutoff value for normal (EF3D ≥ 50%), mild-moderate reduction (EF3D ≤ 45%), or severely reduced (EF3D ≤ 35%) ventricular function.7,23,24 Each strain cutoff value was chosen with the Wald statistic to maximize accuracy. For each strain parameter, the area under the curve (AUC), 95% confidence interval (CI), sensitivity, and specificity are reported. Mean differences in outcomes in patients with normal versus abnormal ventricular function, as defined by 3D strain ROC obtained values, were measured with the Wilcoxon test for continuous outcomes, Fisher exact test for dichotomous outcomes, and Kaplan–Meier survival curves with log-rank test for 1-year event-free survival. The differences between preoperative and postoperative ventricular function, as measured by EF and strain, were compared by the use of paired t tests. Differences between preoperative and postoperative measures, percent change in function preoperative to postoperative, and differences in outcomes between surgical categories were tested with repeated-measures analysis of variance and Tukey honestly significant difference test for pairwise comparisons. The relationship between outcomes (length of stay, duration of intubation, inotrope score) and strain variables were assessed by the use of linear multiple regression models to determine whether each strain measure was an independent predictor of outcomes. One-year event-free survival was assessed with Cox proportional hazards model with the outcome of time to event. A log transformation was computed for length of stay, duration of intubation, and inotrope score to satisfy the linear regression assumptions. We assessed linear model assumptions by examining residual plots (heteroscedasticity, pattern of residuals versus fitted values) and scatter plots (linearity). These assumptions were reasonably satisfied (Supplemental Digital Content 1, Appendix A, http://links.lww.com/AA/B449). Because of high colinearity between strain measures, each value of GAS, longitudinal strain, circumferential
www.anesthesia-analgesia.org
421
Copyright © 2016 International Anesthesia Research Society. Unauthorized reproduction of this article is prohibited.
Predictive Value of 3D Strain in Cardiac Surgery
Table 1. Demographic Data by Surgical Procedure Age (y) Sex, n (%) Weight (kg) Height (cm) EuroSCORE II CPB time (min)
Aortic Valve (n = 45) 65 (56–74) M = 33 (73) F = 12 (27) 77 (68–88) 172 (162–177) 1.5 (0.7–2.2) 124 (101–151)
Mitral Valve (n = 45) 62 (52–71) M = 28 (62) F = 17 (38) 74 (57–83) 170 (162–179) 1.4 (0.7–2.2) 178 (122–218)
Coronary Artery Bypass (n = 50) 64 (57–72) M = 41 (82) F = 9 (18) 76 (68–86) 168 (163–177) 1.9 (1.2–3.0) 109 (88–127)
Combined (n = 23) 63 (43–70) M = 18 (78) F = 5 (22) 77 (69–88) 167 (165–174) 3.6 (1.9–8.8) 180 (146–224)
Data are reported as median and interquartile range. Abbreviations: Combined, combined multiple valve, coronary artery bypass graft/valve, or aortic valve replacement for aortic stenosis/aortic surgery for aortic insufficiency; CPB, cardiopulmonary bypass; EuroSCORE II, European System for Cardiac Operative Risk Evaluation score; F, female; M, male.
Table 2. Preoperative Ejection Fraction and Strain by Surgical Procedure Aortic Valve (n = 45) Ejection fraction (%) 47% (44 to 50) EF2D 43% (40 to 46) EF3D 2D peak systolic global strain (%) Longitudinal −15% (−14 to −16) Circumferential −14% (−13 to −15) Radial 36% (32 to 40) 3D peak systolic global strain (%) Area −22% (−20 to −24) Longitudinal −10% (−9 to −11) Circumferential −15% (−14 to −17) Radial 35% (31 to 39)
Mitral Valve (n = 45)
Coronary Artery Bypass (n = 50)
Combined (n = 23)
54%a (51 to 57) 53%a (50 to 56)
45% (42 to 48) 46% (44 to 49)
46% (41 to 51) 47% (44 to 50)
−19%a (−18 to −20) −14% (−13 to −15) 37% (32 to 42)
−16% (−15 to −17) −13% (−12 to −14) 35% (30 to 40)
−16% (−14 to −18) −15% (−13 to −17) 35% (28 to 42)
−30%a (−28 to −32) −15%a (−13 to −17) −19%a (−18 to −20) 51%a (46 to 56)
−24% (−22 to −26) −12% (−11 to −13) −15% (−14 to −16) 37% (33 to 44)
−27% (−24 to −30) −14% (−12 to −16) −17% (−15 to −19) 42% (37 to 47)
Data are reported as mean (95% confidence interval). Abbreviations: 2D, 2-dimensional; 3D, 3-dimensional; EF, ejection fraction. a All P < .01 for greater preoperative EF and strain values compared with aortic valve and coronary artery bypass.
strain, and radial strain was added individually to separate multivariate linear regression models to examine whether they could independently contribute to predicting outcome variables given the known covariates (EuroSCORE and surgical procedure). Models were examined in patients combined and subgrouped by patients with EF3D < 45% and EF3D > 45%. Statistical analyses were run using R version 3.0.2 (http://www.r-project.org) and SPSS 22 (IBM Corp., Armonk, NY). Statistical significance was defined as P ≤ .05.
Reproducibility Analysis
Twenty percent of patients were selected randomly and assessed for 3D strain intraobserver and interobserver reproducibility. Intraobserver reproducibility was assessed by the same echocardiographer, blinded to the previous results, analyzing each study again, >2 weeks from date of original study. Interobserver reproducibility was assessed by 1 of the 3 expert echocardiographers, who had not analyzed the case before and was blinded to the previous results, analyzing each study again. Agreement was measured using intraclass correlation coefficients (ICCs) ± 95% CIs.
RESULTS
In total, 182 patients were enrolled in this study. Fourteen patients had echocardiography studies unsuitable for speckle-tracking strain imaging (10 for postoperative arrhythmias and 4 technically difficult studies), 4 patients refused postoperative TTE, and 1 was lost to follow-up. Data on 163 patients were analyzed, and 115 (71%) patients were followed up for 1 year. Of patients undergoing aortic
422 www.anesthesia-analgesia.org
valve surgery, 38% had a bicuspid aortic valve. The median patient age was 63 years, weight was 77 kg, preoperative EuroSCORE II was 1.7, and cardiopulmonary bypass time was 134 minutes (Table 1).
Comparison of EF and Strain
Preoperative EF2D and EF 3D and strain measurements by surgical procedure are shown in Table 2. A stronger correlation was seen between 3D strain and reference EF3D, as compared with the 2D strain, for circumferential and radial strain (Table 3). No difference was observed between 2D and 3D longitudinal strain correlations.
Preoperative Versus Postoperative Ventricular Function: Changes in Ejection Fraction and Strain
Ejection fraction, as measured by 2D and 3D echocardiography, significantly decreased after surgery with collective cohort mean preoperative versus postoperative measurements of EF2D 48% vs 43% and EF3D 47% vs 40%. Speckletracking strain imaging measurements also decreased with preoperative versus postoperative mean values of GLPS2D = −17% vs −12%, GCPS2D = −14% vs −12%, and GRPS2D = 36% vs 29%; and GAS3D = −25% vs −18%, GLS3D = −13% vs −9%, GCS3D = −17% vs −11%, and GRS3D =41% vs 25%; all postoperative reductions in ventricular function were statistically significant (P < .005). By surgical subdivision (AVR, MVR, CABG, and combined), the only significant preoperative difference was increased measures of ventricular function (EF and strain)
anesthesia & analgesia
Copyright © 2016 International Anesthesia Research Society. Unauthorized reproduction of this article is prohibited.
Receiver Operating Characteristic Curves
Reference values for 2D strain have previously been published (normal GLPS2D ranges from −17% to −21%); however, there are few data on reference 3D strain values for varying degrees of LV function.25 Therefore, ROC curves were constructed for each measure of 3D preoperative strain (area, longitudinal, circumferential, and radial) to determine threshold values for ventricular function: normal (EF3D ≥ 50%), mild-to-moderately reduced ventricular function (EF3D ≤ 45%), and severely reduced function (EF3D ≤ 35%; Figure 2). GAS3D threshold value of −25% was associated with normal ventricular function, with an AUC of 0.86 (95% CI, 0.81–0.89), −21% with mild to moderate reduced function with an AUC of 0.85 (0.81–0.89), and −18% with severely reduced ventricular function with an AUC of 0.89 (0.85–0.93).
Postoperative Outcomes
Postoperative outcome measures for surgical cohorts can be seen in Table 5. The subgroup of combined surgical procedures had significantly worse outcomes with greater
February 2017 • Volume 124 • Number 2
Table 4. Ventricular Function Measurements by Surgical Procedure
in MVR compared with all other surgical groups, likely because of the retrograde ejection of volume into the low pressure left atrium with mitral regurgitation (Table 2). There was no difference seen in ventricular function measures across surgical subgroups after surgical repair (Table 4). A reduction in ventricular function (EF and strain) was seen after surgery across all surgical subgroups except EF and GCPS2D after CABG. The largest magnitude of reduction in EF3D was seen after MVR (23%) compared with AVR (9%) and CABG (7%), P = .004. Larger magnitudes of reduction in GRS3D were seen after MVR compared with AVR and CABG cohorts (P = .003). In addition, GCS3D decreased by a greater magnitude after MVR and Combined procedures as compared with AVR and CABG (P = .004).
Coronary Artery Bypass (n = 50)
Data are presented as Pearson correlation coefficient (95% confidence intervals). Abbreviations: 2D, 2-dimensional; 3D, 3-dimensional; EF, ejection fraction; TTE, transthoracic echocardiogram. a All P < .01 for significant correlation between strain and EF3D. b All P < .001 for degree of correlation between 3D strain and EF3D being greater than 2D strain and EF3D.
−17 ± 5 vs −12 ± 4c (29%) −14 ± 4 vs −12 ± 5c (14%) 36 ± 16 vs 29 ± 14c (19%) −26 ± 8 vs −18 ± 7c (31%) −13 ± 5 vs −9 ± 4c (31%) −17 ± 5 vs −11 ± 5c (35%) 41 ± 16 vs 25 ± 11c (39%)
Data are reported as mean ± standard deviation. Abbreviations: 2D, 2-dimensional; 3D, 3-dimensional; EF, ejection fraction; GCPS2D, global circumferential strain; GLPS2D, 2D global longitudinal strain; GRPS2D, global radial peak strain; GAS3D, global area strain; GCS3D, global circumferential strain; GLS3D, global longitudinal strain; GRS3D, global radial strain. a % Reduction in postoperative ventricular function greater than % reduction after aortic valve replacement for aortic stenosis (AVR) and coronary artery bypass graft. b % Reduction in postoperative ventricular function greater than % reduction after AVR. c All P < .005 for significant reduction in postoperative ventricular function compared with preoperative function.
Radial
−0.70a (−0.61 to −0.77) −0.54a (−0.42 to −0.64) −0.70a,b (−0.61 to −0.77) 0.70a,b (0.61 to 0.77)
−16 ± 4 vs −11 ± 4c (31%) −15 ± 4 vs −9 ± 4c (40%) 35 ± 18 vs 26 ± 15c (26%) −27 ± 7 vs −16 ± 6b,c (41%) −14 ± 4 vs −9 ± 4c (36%) −17 ± 5 vs −9 ± 5a,c (47%) 42 ± 13 vs 22 ± 10b,c (45%)
Circumferential
−0.76a (−0.69 to −0.82) −0.68a (−0.59 to −0.75) −0.76a,b (−0.69 to −0.82) 0.77a,b (0.70 to 0.83)
−16 ± 4 vs −12 ± 4c (25%) −13 ± 4 vs −13 ± 5 (0%) 35 ± 17 vs 28 ± 13c (20%) −24 ± 6 vs −17 ± 7c (29%) −12 ± 4 vs −9 ± 4c (25%) −15 ± 4 vs −11 ± 5c (33%) 37 ± 13 vs 24 ± 12c (35%)
Longitudinal
−0.63a (−0.53 to −0.71) −0.41a (−0.27 to −0.53) 0.37a (0.23 to 0.50)
46 ± 12 vs 39 ± 9 (15%) 47 ± 8 vs 37 ± 14c (21%)
3D strain Area
−0.69a (−0.60 to −0.76) −0.52a (−0.40 to −0.62) 0.46a (0.33 to 0.57)
45 ± 11 vs 45 ± 13 (0%) 46 ± 9 vs 43 ± 11 (7%)
Radial
Postoperative TTE
Combined (n = 23)
Circumferential
Preoperative TTE
Aortic Valve (n = 45) Mitral Valve (n = 45) Ejection fraction (%): pre versus post (% change) 47 ± 11 vs 44 ± 12c (6%) 54 ± 10 vs 43 ± 12c (21%) EF2D 43 ± 10 vs 39 ± 9c (9%) 53 ± 10 vs 41 ± 13a,c (23%) EF3D Peak systolic global strain (%): pre versus post (% change) −15 ± 4 vs −12 ± 3c (20%) −19 ± 4 vs −14 ± 4c (26%) GLPS2D −14 ± 4 vs −11 ± 4c (21%) −14 ± 4 vs −12 ± 5c (14%) GCPS2D 36 ± 15 vs 29 ± 13c (19%) 37 ± 17 vs 31 ± 14c (16%) GRPS2D −22 ± 7 vs −17 ± 6c (23%) −30 ± 7 vs −20 ± 7b,c (33%) GAS3D −10 ± 4 vs −9 ± 3c (20%) −15 ± 6 vs −11 ± 4b,c (33%) GLS3D −15 ± 5 vs −10 ± 4c (33%) −19 ± 4 vs −12 ± 5a,c (37%) GCS3D 35 ± 14 vs 24 ± 10c (29%) 51 ± 17 vs 29 ± 12a,c (43%) GRS3D
2D strain Longitudinal
All Patients (n = 163)
Table 3. Correlation Between EF3D and Strain
48 ± 11 vs 43 ± 12c (10%) 47 ± 10 vs 40 ± 12c (15%)
www.anesthesia-analgesia.org
423
Copyright © 2016 International Anesthesia Research Society. Unauthorized reproduction of this article is prohibited.
Predictive Value of 3D Strain in Cardiac Surgery
inotrope requirements than AVR, longer length of ICU stay and intubation than AVR and MVR, and longer hospital stay and major cardiac adverse events than all groups. There was no significant difference in 1-year postoperative
mortality by surgical subgroups: AVR (3/38), MVR (1/32), CABG (1/30), and combined (1/15) patients.
Association Between Clinical Outcomes and Preoperative Strain
Patients were dichotomized into normal versus mild-moderate abnormal ventricular function as defined by preoperative GAS3D cut-point of −21%, which corresponds to an EF3D of ≤ 45%, as per previously defined ROC curves, before surgical intervention. Patients with abnormal preoperative measures of GAS3D had longer length of ICU stay, total hospitalization, length of postoperative intubation, and greater rates of major adverse events (Table 6).
Preoperative Strain as a Predictor of Acute Postoperative Outcome
Multivariate regression analyses, in which we controlled for clinical predictors (EuroSCORE and surgery type), showed that 3D strain (area, longitudinal, circumferential, and radial) was associated independently with length of ICU stay in patients with EF3D > 45% (P < .001; Table 7), whereas for 2D strain, only radial strain was an independent predictor. In patients with normal ventricular EF preoperatively, strain was not an independent predictor for inotrope requirement or duration of intubation (Table 7). Strain was not found to be an independent predictor of 30-day mortality with this sample size. In patients with preoperative EF3D < 45%, preoperative GAS3D, GCS3D, and GRS3D as well as GLPS2D and GLCS2D were associated independently with increased length of ICU stay (P < .05), and all strain types were associated independently with inotrope requirements after surgery (P < .05; Table 7).
Event-Free Survival
Figure 2. Receiver operating characteristics curves (ROCs) demonstrating 3D global area (GAS), longitudinal (GLS), circumferential (GCS), and radial strain (GRS), as determinants of normal (ejection fraction [EF] ≥50%) ventricular function. Cutoff values, area under the curve, and 95% confidence intervals (CIs) are reported for normal ventricular function (EF ≥ 50%), mild to moderate reduced function (EF ≤ 45%), and severely reduced ventricular function (EF ≤ 35%). EF, ejection fraction.
When we used a preoperative GAS3D cut-point of −21%, which represents the strain value for normal versus mildmoderate abnormal ventricular function, Kaplan–Meir survival curves showed an increased probability of 1-year event-free survival in patients with normal preoperative 3D strain compared with abnormal preoperative strain 88% vs 69%, P = .005 (Figure 3). In addition, after we controlled for known preoperative risk predictors of EuroSCORE and surgery type, preoperative strain was found to be an independent predictor of 1-year event-free survival and added incremental predictive value (Table 8 and Supplemental Digital Content 2, Appendix B, http://links.lww.com/ AA/B450).
Table 5. Postoperative Outcomes by Surgical Procedure LOSICU (d) LOSHospital (d) LOI (h) Inotrope score Major adverse events
Aortic Valve (n = 45) 4 (2–5), P = .02 6 (5–8), P < .001 5 (3–8), P = .02 1 (0–5), P = .03 6 (13%), P = .002
Mitral Valve (n = 45) 3 (2–5), P = .02 7 (5–7) P < .001 5 (4–8), P = .04 0 (0–7), NS 2 (4%), P < .001
Coronary Artery Bypass (n = 50) 5 (3–5), NS 6 (5–9), P < .001 7 (5–10), NS 0 (0–7), NS 11 (22%), P = .03
Combined (n = 23) 5 (3–9)a 9 (7–16)b 10 (5–20)a 7 (3–13)c 11 (48%)a
Continuous data are reported as median (interquartile range). Abbreviations: AVR, aortic valve replacement for aortic stenosis; CABG, coronary artery bypass graft; ICU, intensive care unit; LOI, length of intubation; LOS, length of stay; MVR, mitral valve repair or replacement for mitral regurgitation. a Outcomes of Combined surgical cohort significantly > AVR, MVR, and/or CABG cohorts, P values listed for individual pairwise comparison with combined cohort. b Combined > AVR/MVR/CABG. c Combined > AVR.
424 www.anesthesia-analgesia.org
anesthesia & analgesia
Copyright © 2016 International Anesthesia Research Society. Unauthorized reproduction of this article is prohibited.
Reproducibility Analysis
Thirty-two patients were analyzed, and strong intrarater and interrater reproducibility was observed for 3D strain analysis, reported as ICCs ± 95% CI. There was no difference in preoperative versus postoperative ICC (Table 9).
DISCUSSION
This study demonstrates that in the acute period after cardiac surgery, there is a reduction in ventricular systolic function, as measured with 3D speckle-tracking strain imaging after mitral valve, aortic valve, and CABG operations. In addition, strain imaging was shown to be an independent predictor of acute and long-term postoperative outcomes and may provide additional predictive value to currently used clinical models of preoperative risk stratification. In this study, we show that all measures of 3D global strain correlate well with our reference EF3D and detect changes in ventricular function associated with cardiac surgery. GCS3D and GRS3D were observed to have a stronger correlation with EF3D compared with strain measured with 2D speckle tracking. Previous studies have demonstrated that 3D echocardiography, compared with 2D, allows faster, more reliable measures of strain, with increased accuracy and reproducibility.13,15,26,27 Schueler et al27 when comparing measurements of LV function in patients undergoing transcatheter aortic valve implantation reported findings
Table 6. Postoperative Outcomes in Patients With Reduced Preoperative 3D Global Area Strain ICU length of stay (d) Hospital length of stay (d) Length of intubation (h) Inotrope score Major adverse events (%)
Normal (n = 111) 3 (1 to 18) 6 (3 to 36) 5 (0 to 318) 2 (0 to 38) 16
Reduced (n = 52) 5 (2 to 19) 7 (4 to 27) 7 (2 to 191) 3 (0 to 46) 31
P Value .001 .008 .03 .20 .03
Data are reported as median (interquartile range) or percentages. Mild to moderately reduced 3D global area strain defined as ≤−21%. Abbreviations: 3D, 3-dimensional; ICU, intensive care unit.
of reduced correlation between 2D and 3D measures, likely seen because 2D echocardiography-based speckle tracking cannot track the complex 3D motion of the LV myocardium throughout systole.28 The results of our study show that 3D strain demonstrates a reduction in ventricular function immediately after cardiac surgery in all surgical subgroups. To our knowledge, this is the first study using 3D strain to comprehensively characterize acute postoperative changes in ventricular function associated with cardiac surgery. Studies that used 2D strain have shown confounding results and were limited by small sample size and restrictive enrollment criteria.29,30 Although an acute decrease in function after MVR may be expected, the decrease in function after AVR was unanticipated, given the reduction in LV afterload after surgery. Previous studies at 1 year and 6 months have shown that myocardial function improves after AVR for aortic stenosis and remains mildly depressed after MVR for mitral regurgitation.28,29,31,32 The difference between the acute versus long-term changes in observed ventricular function in AVR patients is likely a reflection of the immediate reduction in myocardial performance because of the effects of cardiac surgery and cardiopulmonary bypass.33 There are no studies to date defining reference values for 3D strain in patients with normal, mild-moderate, and severely reduced ventricular function. An important contribution of this study is to provide possible references values for each 3D strain measure corresponding to different levels of ventricular function (Figure 2). These reference values could allow investigation of outcome differences in patients with normal versus abnormal ventricular function, as defined by 3D strain measures. Area strain is a novel measurement unique to 3D imaging that reflects the changes in endocardial surface area with LV contraction. Area strain incorporates changes in both subendocardial longitudinal and circumferential strain and therefore has been shown to be a more accurate measure of overall ventricular function and early myocardial dysfunction.11,34 In this study, patients with reduced GAS3D, compared with those with normal preoperative GAS3D, had the worse postoperative outcomes
Table 7. Preoperative 3D Strain as Predictors of Postoperative Outcomes LV ejection fraction ≥45% (n = 105) Clinical predictors Preoperative strain Area Longitudinal Circumferential Radial LV ejection fraction 60%. J Am Coll Cardiol. 2010;55:671–679. 36. Witkowski TG, Thomas JD, Debonnaire PJ, et al. Global longitudinal strain predicts left ventricular dysfunction after mitral valve repair. Eur Heart J Cardiovasc Imaging. 2013;14:69–76.
anesthesia & analgesia
Copyright © 2016 International Anesthesia Research Society. Unauthorized reproduction of this article is prohibited.
M
ore than half a million cardiac surgeries are performed in the United States each year, and left ventricular (LV) systolic function is a strong predictor of their outcomes.1 Preoperative decrement in
From the Departments of *Anesthesiology and Perioperative Medicine, †Cardiothoracic Surgery, ‡Medicine Statistics Core, and §Biomathematics and Medicine, University of California, Los Angeles, Los Angeles, California. Accepted for publication May 6, 2016.
Funding: This study was funded by intramural department funds; Dr. Mahajan is supported by National Heart, Lung, and Blood Institute (NHLBI), National Institutes of Health (NIH) Research Project Grant (R01) HL084261, Bethesda, MD. The authors declare no conflicts of interest. Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal’s website (www.anesthesia-analgesia.org). Reprints will not be available from the authors. Address correspondence to Aman Mahajan, MD, PhD, Department of Anesthesiology and Perioperative Medicine, David Geffen School of Medicine at University of California, Los Angeles, 757 Westwood Blvd, Suite 3325 Los Angeles, CA 90095. Address e-mail to [email protected]. Copyright © 2016 International Anesthesia Research Society DOI: 10.1213/ANE.0000000000001440
February 2017 • Volume 124 • Number 2
ventricular function can cause hemodynamic instability in the acute postoperative period requiring escalating inotropic support, prolonged hospital stay, and result in longterm complications, including increased mortality. Systolic cardiac function routinely has been assessed through echocardiography by the use of 2-dimensional (2D) LV ejection fraction (EF2D) calculations. These traditional measurements are based on mathematical assumptions of LV geometry to calculate volume changes as an indirect measure of myocardial function.1,2 Strain is a direct measure of LV myocardial tissue deformation that may provide additional information regarding both regional and global myocardial functions.3–6 In recent studies that used magnetic resonance imaging (MRI) as a reference standard, strain was demonstrated to measure ventricular function with better accuracy and reliability than left ventricular ejection fraction.2,7–11 Furthermore, Macron et al12 showed that in patients with poor acoustic windows, strain values correlated more closely with MRI measures of ventricular function than traditional EF2D. www.anesthesia-analgesia.org
419
Copyright © 2016 International Anesthesia Research Society. Unauthorized reproduction of this article is prohibited.
Predictive Value of 3D Strain in Cardiac Surgery
Speckle-tracking strain can be measured by both 2D and 3-dimensional (3D) echocardiography; however, the newly developed 3D strain is an improvement over 2D strain measurements because speckle tracking is no longer limited to motion in a 2D imaging plane.8,10,13 The myocardial architecture of the LV leads to the rotational twisting and shortening of the left ventricle during contraction that may impede tracking in the 2D imaging plane.13,14 With 3D speckle-tracking echocardiography, the entire myocardium is acquired in 1 imaging loop so that speckles are tracked continuously without loss or tracking between frames. In addition, all strain measurements are taken from the same image acquisition compared with 2D speckle tracking, which requires 3 long-axis and 3 short-axis images collectively. This ability to measure all types of strain in 1 imaging loop with 3D speckle-tracking echocardiography improves accuracy and decreases the time for acquisition.15 The aim of this study was to investigate the change in LV systolic function after cardiac surgery with the use of 3D transthoracic echocardiogram (TTE) speckle-tracking strain imaging and to determine whether preoperative strain is an independent predictor of acute and long-term clinical outcomes after cardiac surgical operations. We hypothesized that ventricular function would be reduced in the acute postoperative period and that preoperative 3D strain measures of ventricular function would be an independent predictor of acute and long-term outcomes after cardiac surgery.
METHODS
After institutional review board approval and written informed consent, adult patients scheduled for elective cardiac surgery with cardiopulmonary bypass between October 2011 and July 2013 at the Ronald Reagan University of California, Los Angeles Medical Center were enrolled in this prospective study. Surgical procedures included aortic valve replacement for aortic stenosis (AVR), mitral valve repair or replacement for mitral regurgitation (MVR), coronary artery bypass graft (CABG), and combined multiple valve, CABG/valve, or AVR/aortic surgery for aortic insufficiency (combined). Exclusion criteria were cardiac arrhythmias (such as atrial fibrillation and paced rhythms), congenital heart disease, and poor TTE imaging windows.
Clinical Data Collection
Preoperative (age, gender, and European System for Cardiac Operative Risk Evaluation score [EuroSCORE]), intraoperative, and postoperative data were collected.16 Definitions of all data variables collected were per the Adult Cardiac Surgery Database Training Manual, v2.73.a Postoperative outcome variables were duration of tracheal intubation (from intensive care admission to extubation), length of intensive care unit (ICU) stay, length of hospital stay, inotrope requirement (calculated using the equation: dopamine [μk/kg/min] + dobutamine [μg/kg/min] + 100 × epinephrine [μg/kg/min] + 10 × milrinone dose [μk/kg/min]), major adverse cardiac events, or 1-year all-cause mortality.17–19 Major cardiac adverse events per http://www.sts.org/sites/default/files/documents/ACSDTraining Manual-3-14-14.pdf. Accessed March 20, 2013.
a
420 www.anesthesia-analgesia.org
the Society of Thoracic Surgeons database were cardiac arrhythmias (ventricular tachycardia, ventricular fibrillation, and asystole), multiple-organ failure, ventricular assist device (balloon pump or extracorporeal membrane oxygenation), and reoperation. Mortality was defined as all-cause 1-year mortality and was ascertained from electronic hospital records and telephone follow-up. One-year event-free survival was defined as percentage of patients free from any major cardiac adverse events or all-cause mortality up to 1 year after the date of study enrollment.20
Echocardiography
TTEs were performed preoperatively in awake spontaneously ventilating patients on the day of surgery and between 2 and 4 days postoperatively after surgery, once participants were extubated and chest tubes were discontinued. Echocardiographic studies were acquired by the same experienced sonographer (E.M.). Each patient underwent standardized 2D and 3D TTE for the determination of LV functional parameters. Ejection fraction was calculated by 2D echocardiography via Simpson’s methods of disks applied to 2- and 4-chamber apical images, according to the American Society of Echocardiography.21 Ejection fraction was calculated by 3D echocardiography by the use of full-volume data sets. End-diastolic and endsystolic frames were selected, and semiautomated LV border tracking software outlined the endocardial borders and calculated end-diastolic volume, end-systolic volume, and ejection fraction. All echo images were taken with Vivid 9 (GE Healthcare, Milwaukee, WI) with a 3V transducer; 2D images were acquired at frame rate of 55 to 90 frames/s, and 3D images were taken at frame rate of 25 to 55 frame/s.
Two-Dimensional Speckle-Tracking Analysis
Grayscale images obtained for 2D strain analysis included parasternal short axis at mitral valve, mid-papillary muscle, and apical level, as well as 4-chamber, 2-chamber, and apical long-axis, views. Patients were in the left lateral decubitus position or supine depending on the patients’ physical condition and adequate acoustic windows. Time to aortic valve closure was measured by continuous-wave Doppler in the LV outflow tract. 2D strain analysis was performed by 1 of 3 blinded echocardiographers experienced in regional and global strain analyses, using EchoPAC (version 110.1.1) system by GE Healthcare. Strain was measured with the use of automated tracing, with manual adjustments if needed. The integrity of tracking was visually confirmed and ascertained from the credibility shape of the strain curve. The region of interest was readjusted manually as necessary, in 12% of images analyzed. All measurements were obtained using the average of 3 cardiac cycles. Finally, 2D software output included the measurement of peak systolic global longitudinal (GLPS2D), circumferential (GCPS2D), and radial (GRPS2D) strain.
Three-Dimensional Speckle-Tracking Analysis
The 3D echocardiogram data acquisition was obtained in an adjustable imaging sector volume at the LV apex with breath hold lasting 4 heartbeats. The 3D data sets used for strain analysis were the same images used for EF3D analysis as well. Data analysis was preformed off-line by 1 of 3
anesthesia & analgesia
Copyright © 2016 International Anesthesia Research Society. Unauthorized reproduction of this article is prohibited.
experienced blinded echocardiographers using 4D Auto LVQ offline on EchoPAC system by GE Healthcare. The full-volume data set is displayed as orthogonal 4-chamber, 2-chamber, and short-axis images. Mitral annular and apical points are placed on these images in end-diastolic and end-systolic frames. Semiautomated endocardial surface detection software then outlines endocardial borders and tracks the endocardium in all frames and automatically calculates strain. If the endocardial border tracking appeared inadequate, then manual adjustments of automatic tracing were made and the sequence analysis was repeated. Manual adjustment was necessary in 10% of images analyzed. Software output includes calculation of peak systolic global longitudinal (GLS3D), circumferential (GCS3D), radial (GRS3D), and area (GAS3D) strain. Strain is a measure of myocardial deformation during systole; normal peak systolic longitudinal strain has a negative value because of LV shortening during systole and normal circumferential strain also has a negative value with reduction in LV circumference during systole, whereas radial strain has a positive value because of systolic radial thickening. Global area strain (GAS) is a new measurement unique to 3D strain and has a negative value. Representative 3D echocardiographic images and GAS3D output are shown in Figure 1.
Figure 1. Representative 3-dimensional (3D) left ventricular strain diagram depicting preoperative (A) and postoperative (B) global area strain results after mitral valve replacement. Each yellow line represents the area strain curve of an individual myocardial segment, and the white line represents cumulative left ventricular global area strain showing a reduction in postoperative global area strain.
February 2017 • Volume 124 • Number 2
Statistical Analysis
The number of enrolled patients was based on power calculation for 80% power at α = .05 to detect an odds ratio of 2.5 between those with normal and reduced ventricular function for the primary outcome of major adverse events, requiring sample size of 160 patients with additional 15% enrollment to account for possible study dropout.1 Histograms and quartile plots were examined in a bivariate analysis to determine whether distributions of continuous outcomes were approximately normally distributed. If not, the outcomes were logtransformed before analysis. Values were reported as mean (± standard deviation) for normally distributed outcomes and median (interquartile range) for skewed outcomes. The correlation between each strain parameter and 3D volume-based ejection fraction (EF3D) was assessed with Pearson correlation coefficients. EF3D has been shown to be closely correlated with MRI-based measures of ventricular function and was used in this study as the reference measurement to compare the other measures of ventricular function.10,22 Differences comparing degree of correlation between 2D and 3D strain with EF3D were measured with the William test. Receiver operating characteristic (ROC) curves were constructed for each type of strain to determine the cutoff value for normal (EF3D ≥ 50%), mild-moderate reduction (EF3D ≤ 45%), or severely reduced (EF3D ≤ 35%) ventricular function.7,23,24 Each strain cutoff value was chosen with the Wald statistic to maximize accuracy. For each strain parameter, the area under the curve (AUC), 95% confidence interval (CI), sensitivity, and specificity are reported. Mean differences in outcomes in patients with normal versus abnormal ventricular function, as defined by 3D strain ROC obtained values, were measured with the Wilcoxon test for continuous outcomes, Fisher exact test for dichotomous outcomes, and Kaplan–Meier survival curves with log-rank test for 1-year event-free survival. The differences between preoperative and postoperative ventricular function, as measured by EF and strain, were compared by the use of paired t tests. Differences between preoperative and postoperative measures, percent change in function preoperative to postoperative, and differences in outcomes between surgical categories were tested with repeated-measures analysis of variance and Tukey honestly significant difference test for pairwise comparisons. The relationship between outcomes (length of stay, duration of intubation, inotrope score) and strain variables were assessed by the use of linear multiple regression models to determine whether each strain measure was an independent predictor of outcomes. One-year event-free survival was assessed with Cox proportional hazards model with the outcome of time to event. A log transformation was computed for length of stay, duration of intubation, and inotrope score to satisfy the linear regression assumptions. We assessed linear model assumptions by examining residual plots (heteroscedasticity, pattern of residuals versus fitted values) and scatter plots (linearity). These assumptions were reasonably satisfied (Supplemental Digital Content 1, Appendix A, http://links.lww.com/AA/B449). Because of high colinearity between strain measures, each value of GAS, longitudinal strain, circumferential
www.anesthesia-analgesia.org
421
Copyright © 2016 International Anesthesia Research Society. Unauthorized reproduction of this article is prohibited.
Predictive Value of 3D Strain in Cardiac Surgery
Table 1. Demographic Data by Surgical Procedure Age (y) Sex, n (%) Weight (kg) Height (cm) EuroSCORE II CPB time (min)
Aortic Valve (n = 45) 65 (56–74) M = 33 (73) F = 12 (27) 77 (68–88) 172 (162–177) 1.5 (0.7–2.2) 124 (101–151)
Mitral Valve (n = 45) 62 (52–71) M = 28 (62) F = 17 (38) 74 (57–83) 170 (162–179) 1.4 (0.7–2.2) 178 (122–218)
Coronary Artery Bypass (n = 50) 64 (57–72) M = 41 (82) F = 9 (18) 76 (68–86) 168 (163–177) 1.9 (1.2–3.0) 109 (88–127)
Combined (n = 23) 63 (43–70) M = 18 (78) F = 5 (22) 77 (69–88) 167 (165–174) 3.6 (1.9–8.8) 180 (146–224)
Data are reported as median and interquartile range. Abbreviations: Combined, combined multiple valve, coronary artery bypass graft/valve, or aortic valve replacement for aortic stenosis/aortic surgery for aortic insufficiency; CPB, cardiopulmonary bypass; EuroSCORE II, European System for Cardiac Operative Risk Evaluation score; F, female; M, male.
Table 2. Preoperative Ejection Fraction and Strain by Surgical Procedure Aortic Valve (n = 45) Ejection fraction (%) 47% (44 to 50) EF2D 43% (40 to 46) EF3D 2D peak systolic global strain (%) Longitudinal −15% (−14 to −16) Circumferential −14% (−13 to −15) Radial 36% (32 to 40) 3D peak systolic global strain (%) Area −22% (−20 to −24) Longitudinal −10% (−9 to −11) Circumferential −15% (−14 to −17) Radial 35% (31 to 39)
Mitral Valve (n = 45)
Coronary Artery Bypass (n = 50)
Combined (n = 23)
54%a (51 to 57) 53%a (50 to 56)
45% (42 to 48) 46% (44 to 49)
46% (41 to 51) 47% (44 to 50)
−19%a (−18 to −20) −14% (−13 to −15) 37% (32 to 42)
−16% (−15 to −17) −13% (−12 to −14) 35% (30 to 40)
−16% (−14 to −18) −15% (−13 to −17) 35% (28 to 42)
−30%a (−28 to −32) −15%a (−13 to −17) −19%a (−18 to −20) 51%a (46 to 56)
−24% (−22 to −26) −12% (−11 to −13) −15% (−14 to −16) 37% (33 to 44)
−27% (−24 to −30) −14% (−12 to −16) −17% (−15 to −19) 42% (37 to 47)
Data are reported as mean (95% confidence interval). Abbreviations: 2D, 2-dimensional; 3D, 3-dimensional; EF, ejection fraction. a All P < .01 for greater preoperative EF and strain values compared with aortic valve and coronary artery bypass.
strain, and radial strain was added individually to separate multivariate linear regression models to examine whether they could independently contribute to predicting outcome variables given the known covariates (EuroSCORE and surgical procedure). Models were examined in patients combined and subgrouped by patients with EF3D < 45% and EF3D > 45%. Statistical analyses were run using R version 3.0.2 (http://www.r-project.org) and SPSS 22 (IBM Corp., Armonk, NY). Statistical significance was defined as P ≤ .05.
Reproducibility Analysis
Twenty percent of patients were selected randomly and assessed for 3D strain intraobserver and interobserver reproducibility. Intraobserver reproducibility was assessed by the same echocardiographer, blinded to the previous results, analyzing each study again, >2 weeks from date of original study. Interobserver reproducibility was assessed by 1 of the 3 expert echocardiographers, who had not analyzed the case before and was blinded to the previous results, analyzing each study again. Agreement was measured using intraclass correlation coefficients (ICCs) ± 95% CIs.
RESULTS
In total, 182 patients were enrolled in this study. Fourteen patients had echocardiography studies unsuitable for speckle-tracking strain imaging (10 for postoperative arrhythmias and 4 technically difficult studies), 4 patients refused postoperative TTE, and 1 was lost to follow-up. Data on 163 patients were analyzed, and 115 (71%) patients were followed up for 1 year. Of patients undergoing aortic
422 www.anesthesia-analgesia.org
valve surgery, 38% had a bicuspid aortic valve. The median patient age was 63 years, weight was 77 kg, preoperative EuroSCORE II was 1.7, and cardiopulmonary bypass time was 134 minutes (Table 1).
Comparison of EF and Strain
Preoperative EF2D and EF 3D and strain measurements by surgical procedure are shown in Table 2. A stronger correlation was seen between 3D strain and reference EF3D, as compared with the 2D strain, for circumferential and radial strain (Table 3). No difference was observed between 2D and 3D longitudinal strain correlations.
Preoperative Versus Postoperative Ventricular Function: Changes in Ejection Fraction and Strain
Ejection fraction, as measured by 2D and 3D echocardiography, significantly decreased after surgery with collective cohort mean preoperative versus postoperative measurements of EF2D 48% vs 43% and EF3D 47% vs 40%. Speckletracking strain imaging measurements also decreased with preoperative versus postoperative mean values of GLPS2D = −17% vs −12%, GCPS2D = −14% vs −12%, and GRPS2D = 36% vs 29%; and GAS3D = −25% vs −18%, GLS3D = −13% vs −9%, GCS3D = −17% vs −11%, and GRS3D =41% vs 25%; all postoperative reductions in ventricular function were statistically significant (P < .005). By surgical subdivision (AVR, MVR, CABG, and combined), the only significant preoperative difference was increased measures of ventricular function (EF and strain)
anesthesia & analgesia
Copyright © 2016 International Anesthesia Research Society. Unauthorized reproduction of this article is prohibited.
Receiver Operating Characteristic Curves
Reference values for 2D strain have previously been published (normal GLPS2D ranges from −17% to −21%); however, there are few data on reference 3D strain values for varying degrees of LV function.25 Therefore, ROC curves were constructed for each measure of 3D preoperative strain (area, longitudinal, circumferential, and radial) to determine threshold values for ventricular function: normal (EF3D ≥ 50%), mild-to-moderately reduced ventricular function (EF3D ≤ 45%), and severely reduced function (EF3D ≤ 35%; Figure 2). GAS3D threshold value of −25% was associated with normal ventricular function, with an AUC of 0.86 (95% CI, 0.81–0.89), −21% with mild to moderate reduced function with an AUC of 0.85 (0.81–0.89), and −18% with severely reduced ventricular function with an AUC of 0.89 (0.85–0.93).
Postoperative Outcomes
Postoperative outcome measures for surgical cohorts can be seen in Table 5. The subgroup of combined surgical procedures had significantly worse outcomes with greater
February 2017 • Volume 124 • Number 2
Table 4. Ventricular Function Measurements by Surgical Procedure
in MVR compared with all other surgical groups, likely because of the retrograde ejection of volume into the low pressure left atrium with mitral regurgitation (Table 2). There was no difference seen in ventricular function measures across surgical subgroups after surgical repair (Table 4). A reduction in ventricular function (EF and strain) was seen after surgery across all surgical subgroups except EF and GCPS2D after CABG. The largest magnitude of reduction in EF3D was seen after MVR (23%) compared with AVR (9%) and CABG (7%), P = .004. Larger magnitudes of reduction in GRS3D were seen after MVR compared with AVR and CABG cohorts (P = .003). In addition, GCS3D decreased by a greater magnitude after MVR and Combined procedures as compared with AVR and CABG (P = .004).
Coronary Artery Bypass (n = 50)
Data are presented as Pearson correlation coefficient (95% confidence intervals). Abbreviations: 2D, 2-dimensional; 3D, 3-dimensional; EF, ejection fraction; TTE, transthoracic echocardiogram. a All P < .01 for significant correlation between strain and EF3D. b All P < .001 for degree of correlation between 3D strain and EF3D being greater than 2D strain and EF3D.
−17 ± 5 vs −12 ± 4c (29%) −14 ± 4 vs −12 ± 5c (14%) 36 ± 16 vs 29 ± 14c (19%) −26 ± 8 vs −18 ± 7c (31%) −13 ± 5 vs −9 ± 4c (31%) −17 ± 5 vs −11 ± 5c (35%) 41 ± 16 vs 25 ± 11c (39%)
Data are reported as mean ± standard deviation. Abbreviations: 2D, 2-dimensional; 3D, 3-dimensional; EF, ejection fraction; GCPS2D, global circumferential strain; GLPS2D, 2D global longitudinal strain; GRPS2D, global radial peak strain; GAS3D, global area strain; GCS3D, global circumferential strain; GLS3D, global longitudinal strain; GRS3D, global radial strain. a % Reduction in postoperative ventricular function greater than % reduction after aortic valve replacement for aortic stenosis (AVR) and coronary artery bypass graft. b % Reduction in postoperative ventricular function greater than % reduction after AVR. c All P < .005 for significant reduction in postoperative ventricular function compared with preoperative function.
Radial
−0.70a (−0.61 to −0.77) −0.54a (−0.42 to −0.64) −0.70a,b (−0.61 to −0.77) 0.70a,b (0.61 to 0.77)
−16 ± 4 vs −11 ± 4c (31%) −15 ± 4 vs −9 ± 4c (40%) 35 ± 18 vs 26 ± 15c (26%) −27 ± 7 vs −16 ± 6b,c (41%) −14 ± 4 vs −9 ± 4c (36%) −17 ± 5 vs −9 ± 5a,c (47%) 42 ± 13 vs 22 ± 10b,c (45%)
Circumferential
−0.76a (−0.69 to −0.82) −0.68a (−0.59 to −0.75) −0.76a,b (−0.69 to −0.82) 0.77a,b (0.70 to 0.83)
−16 ± 4 vs −12 ± 4c (25%) −13 ± 4 vs −13 ± 5 (0%) 35 ± 17 vs 28 ± 13c (20%) −24 ± 6 vs −17 ± 7c (29%) −12 ± 4 vs −9 ± 4c (25%) −15 ± 4 vs −11 ± 5c (33%) 37 ± 13 vs 24 ± 12c (35%)
Longitudinal
−0.63a (−0.53 to −0.71) −0.41a (−0.27 to −0.53) 0.37a (0.23 to 0.50)
46 ± 12 vs 39 ± 9 (15%) 47 ± 8 vs 37 ± 14c (21%)
3D strain Area
−0.69a (−0.60 to −0.76) −0.52a (−0.40 to −0.62) 0.46a (0.33 to 0.57)
45 ± 11 vs 45 ± 13 (0%) 46 ± 9 vs 43 ± 11 (7%)
Radial
Postoperative TTE
Combined (n = 23)
Circumferential
Preoperative TTE
Aortic Valve (n = 45) Mitral Valve (n = 45) Ejection fraction (%): pre versus post (% change) 47 ± 11 vs 44 ± 12c (6%) 54 ± 10 vs 43 ± 12c (21%) EF2D 43 ± 10 vs 39 ± 9c (9%) 53 ± 10 vs 41 ± 13a,c (23%) EF3D Peak systolic global strain (%): pre versus post (% change) −15 ± 4 vs −12 ± 3c (20%) −19 ± 4 vs −14 ± 4c (26%) GLPS2D −14 ± 4 vs −11 ± 4c (21%) −14 ± 4 vs −12 ± 5c (14%) GCPS2D 36 ± 15 vs 29 ± 13c (19%) 37 ± 17 vs 31 ± 14c (16%) GRPS2D −22 ± 7 vs −17 ± 6c (23%) −30 ± 7 vs −20 ± 7b,c (33%) GAS3D −10 ± 4 vs −9 ± 3c (20%) −15 ± 6 vs −11 ± 4b,c (33%) GLS3D −15 ± 5 vs −10 ± 4c (33%) −19 ± 4 vs −12 ± 5a,c (37%) GCS3D 35 ± 14 vs 24 ± 10c (29%) 51 ± 17 vs 29 ± 12a,c (43%) GRS3D
2D strain Longitudinal
All Patients (n = 163)
Table 3. Correlation Between EF3D and Strain
48 ± 11 vs 43 ± 12c (10%) 47 ± 10 vs 40 ± 12c (15%)
www.anesthesia-analgesia.org
423
Copyright © 2016 International Anesthesia Research Society. Unauthorized reproduction of this article is prohibited.
Predictive Value of 3D Strain in Cardiac Surgery
inotrope requirements than AVR, longer length of ICU stay and intubation than AVR and MVR, and longer hospital stay and major cardiac adverse events than all groups. There was no significant difference in 1-year postoperative
mortality by surgical subgroups: AVR (3/38), MVR (1/32), CABG (1/30), and combined (1/15) patients.
Association Between Clinical Outcomes and Preoperative Strain
Patients were dichotomized into normal versus mild-moderate abnormal ventricular function as defined by preoperative GAS3D cut-point of −21%, which corresponds to an EF3D of ≤ 45%, as per previously defined ROC curves, before surgical intervention. Patients with abnormal preoperative measures of GAS3D had longer length of ICU stay, total hospitalization, length of postoperative intubation, and greater rates of major adverse events (Table 6).
Preoperative Strain as a Predictor of Acute Postoperative Outcome
Multivariate regression analyses, in which we controlled for clinical predictors (EuroSCORE and surgery type), showed that 3D strain (area, longitudinal, circumferential, and radial) was associated independently with length of ICU stay in patients with EF3D > 45% (P < .001; Table 7), whereas for 2D strain, only radial strain was an independent predictor. In patients with normal ventricular EF preoperatively, strain was not an independent predictor for inotrope requirement or duration of intubation (Table 7). Strain was not found to be an independent predictor of 30-day mortality with this sample size. In patients with preoperative EF3D < 45%, preoperative GAS3D, GCS3D, and GRS3D as well as GLPS2D and GLCS2D were associated independently with increased length of ICU stay (P < .05), and all strain types were associated independently with inotrope requirements after surgery (P < .05; Table 7).
Event-Free Survival
Figure 2. Receiver operating characteristics curves (ROCs) demonstrating 3D global area (GAS), longitudinal (GLS), circumferential (GCS), and radial strain (GRS), as determinants of normal (ejection fraction [EF] ≥50%) ventricular function. Cutoff values, area under the curve, and 95% confidence intervals (CIs) are reported for normal ventricular function (EF ≥ 50%), mild to moderate reduced function (EF ≤ 45%), and severely reduced ventricular function (EF ≤ 35%). EF, ejection fraction.
When we used a preoperative GAS3D cut-point of −21%, which represents the strain value for normal versus mildmoderate abnormal ventricular function, Kaplan–Meir survival curves showed an increased probability of 1-year event-free survival in patients with normal preoperative 3D strain compared with abnormal preoperative strain 88% vs 69%, P = .005 (Figure 3). In addition, after we controlled for known preoperative risk predictors of EuroSCORE and surgery type, preoperative strain was found to be an independent predictor of 1-year event-free survival and added incremental predictive value (Table 8 and Supplemental Digital Content 2, Appendix B, http://links.lww.com/ AA/B450).
Table 5. Postoperative Outcomes by Surgical Procedure LOSICU (d) LOSHospital (d) LOI (h) Inotrope score Major adverse events
Aortic Valve (n = 45) 4 (2–5), P = .02 6 (5–8), P < .001 5 (3–8), P = .02 1 (0–5), P = .03 6 (13%), P = .002
Mitral Valve (n = 45) 3 (2–5), P = .02 7 (5–7) P < .001 5 (4–8), P = .04 0 (0–7), NS 2 (4%), P < .001
Coronary Artery Bypass (n = 50) 5 (3–5), NS 6 (5–9), P < .001 7 (5–10), NS 0 (0–7), NS 11 (22%), P = .03
Combined (n = 23) 5 (3–9)a 9 (7–16)b 10 (5–20)a 7 (3–13)c 11 (48%)a
Continuous data are reported as median (interquartile range). Abbreviations: AVR, aortic valve replacement for aortic stenosis; CABG, coronary artery bypass graft; ICU, intensive care unit; LOI, length of intubation; LOS, length of stay; MVR, mitral valve repair or replacement for mitral regurgitation. a Outcomes of Combined surgical cohort significantly > AVR, MVR, and/or CABG cohorts, P values listed for individual pairwise comparison with combined cohort. b Combined > AVR/MVR/CABG. c Combined > AVR.
424 www.anesthesia-analgesia.org
anesthesia & analgesia
Copyright © 2016 International Anesthesia Research Society. Unauthorized reproduction of this article is prohibited.
Reproducibility Analysis
Thirty-two patients were analyzed, and strong intrarater and interrater reproducibility was observed for 3D strain analysis, reported as ICCs ± 95% CI. There was no difference in preoperative versus postoperative ICC (Table 9).
DISCUSSION
This study demonstrates that in the acute period after cardiac surgery, there is a reduction in ventricular systolic function, as measured with 3D speckle-tracking strain imaging after mitral valve, aortic valve, and CABG operations. In addition, strain imaging was shown to be an independent predictor of acute and long-term postoperative outcomes and may provide additional predictive value to currently used clinical models of preoperative risk stratification. In this study, we show that all measures of 3D global strain correlate well with our reference EF3D and detect changes in ventricular function associated with cardiac surgery. GCS3D and GRS3D were observed to have a stronger correlation with EF3D compared with strain measured with 2D speckle tracking. Previous studies have demonstrated that 3D echocardiography, compared with 2D, allows faster, more reliable measures of strain, with increased accuracy and reproducibility.13,15,26,27 Schueler et al27 when comparing measurements of LV function in patients undergoing transcatheter aortic valve implantation reported findings
Table 6. Postoperative Outcomes in Patients With Reduced Preoperative 3D Global Area Strain ICU length of stay (d) Hospital length of stay (d) Length of intubation (h) Inotrope score Major adverse events (%)
Normal (n = 111) 3 (1 to 18) 6 (3 to 36) 5 (0 to 318) 2 (0 to 38) 16
Reduced (n = 52) 5 (2 to 19) 7 (4 to 27) 7 (2 to 191) 3 (0 to 46) 31
P Value .001 .008 .03 .20 .03
Data are reported as median (interquartile range) or percentages. Mild to moderately reduced 3D global area strain defined as ≤−21%. Abbreviations: 3D, 3-dimensional; ICU, intensive care unit.
of reduced correlation between 2D and 3D measures, likely seen because 2D echocardiography-based speckle tracking cannot track the complex 3D motion of the LV myocardium throughout systole.28 The results of our study show that 3D strain demonstrates a reduction in ventricular function immediately after cardiac surgery in all surgical subgroups. To our knowledge, this is the first study using 3D strain to comprehensively characterize acute postoperative changes in ventricular function associated with cardiac surgery. Studies that used 2D strain have shown confounding results and were limited by small sample size and restrictive enrollment criteria.29,30 Although an acute decrease in function after MVR may be expected, the decrease in function after AVR was unanticipated, given the reduction in LV afterload after surgery. Previous studies at 1 year and 6 months have shown that myocardial function improves after AVR for aortic stenosis and remains mildly depressed after MVR for mitral regurgitation.28,29,31,32 The difference between the acute versus long-term changes in observed ventricular function in AVR patients is likely a reflection of the immediate reduction in myocardial performance because of the effects of cardiac surgery and cardiopulmonary bypass.33 There are no studies to date defining reference values for 3D strain in patients with normal, mild-moderate, and severely reduced ventricular function. An important contribution of this study is to provide possible references values for each 3D strain measure corresponding to different levels of ventricular function (Figure 2). These reference values could allow investigation of outcome differences in patients with normal versus abnormal ventricular function, as defined by 3D strain measures. Area strain is a novel measurement unique to 3D imaging that reflects the changes in endocardial surface area with LV contraction. Area strain incorporates changes in both subendocardial longitudinal and circumferential strain and therefore has been shown to be a more accurate measure of overall ventricular function and early myocardial dysfunction.11,34 In this study, patients with reduced GAS3D, compared with those with normal preoperative GAS3D, had the worse postoperative outcomes
Table 7. Preoperative 3D Strain as Predictors of Postoperative Outcomes LV ejection fraction ≥45% (n = 105) Clinical predictors Preoperative strain Area Longitudinal Circumferential Radial LV ejection fraction 60%. J Am Coll Cardiol. 2010;55:671–679. 36. Witkowski TG, Thomas JD, Debonnaire PJ, et al. Global longitudinal strain predicts left ventricular dysfunction after mitral valve repair. Eur Heart J Cardiovasc Imaging. 2013;14:69–76.
anesthesia & analgesia
Copyright © 2016 International Anesthesia Research Society. Unauthorized reproduction of this article is prohibited.
