Background
Response to cardiac resynchronization therapy is variable. The aim of this study was to test the hypothesis that left ventricular (LV) lead position in proximity to myocardial scar is associated with less favorable outcomes.
Methods
A total of 149 patients were included in this substudy of the Speckle Tracking Assisted Resynchronization Therapy for Electrode Region trial using echocardiographic radial strain for basal and middle LV segments and transverse strain for apical segments to estimate scar. Sixty-four patients with high-dose rest 99m Tc sestamibi single-photon emission computed tomographic imaging were used for validation in 508 LV free-wall segments. The relationship of LV lead position to segments estimated as scar was defined as concordant, adjacent, or remote. The prespecified primary end point was heart failure hospitalization or death over 2 years.
Results
Wall thickening < 10% by radial strain in LV free wall segments was associated with absent 99m Tc sestamibi uptake indicative of scar with 77% sensitivity and 89% specificity and strain < 10% was accordingly used as the surrogate for scar. Event-free survival was most favorable in patients with nonischemic disease (79%), similarly favorable in patients with ischemic disease and LV leads remote from scar (74%), but significantly worse in patients with ischemic disease and LV leads adjacent to (61%) or within scar (41%) ( P = .004). Preserved wall thickening at the LV lead site was independently associated with favorable outcome and additive to pacing at the site of latest mechanical activation ( P = .001) and remained significant after adjusting for scar burden.
Conclusions
Echocardiographic speckle-tracking strain has the potential to estimate free wall scar in patients with ischemic cardiomyopathy and influence LV lead positioning away from scar to improve clinical outcomes after cardiac resynchronization therapy.
Cardiac resynchronization therapy (CRT) is an important therapy that alleviates symptoms and improves survival in the majority of patients with heart failure (HF). Factors associated with nonresponse, which occurs in approximately one third of CRT recipients, include lack of baseline dyssynchrony, global scar burden, and lead position according to the latest site of contraction. LV lead position within a scarred region has also been shown to have an unfavorable impact on LV reverse remodeling and clinical outcomes. However, clinical approaches to identify LV scar, including single-photon emission computed tomographic (SPECT) imaging and gadolinium cardiac magnetic resonance (CMR), require additional testing and may be expensive or not readily available.
Echocardiography is an accepted part of the workup of patients before undergoing CRT. Analysis of myocardial deformation by speckle-tracking radial strain applied to routine digital echocardiographic images has been useful to quantify mechanical LV dyssynchrony and the site of latest activation. Other reports have suggested that the absence of wall thickening by speckle-tracking strain may be a marker of scar. However, information regarding the effects of LV lead position to scar by speckle-tracking strain on patient outcomes has been limited. The Speckle Tracking Assisted Resynchronization Therapy for Electrode Region (STARTER) trial was a prospective, controlled, double-blinded, randomized clinical trial that demonstrated improved patient outcomes with LV positioning toward the site of latest mechanical thickening. As a substudy of STARTER, our objectives in this investigation were to test the hypotheses (1) that speckle-tracking radial strain can estimate regional scar in CRT patients using 99m Tc sestamibi SPECT imaging as a reference standard, (2) that LV lead positioning in regions remote from scar is associated with improved clinical outcomes, and (3) that LV lead positioning remote from scar may have an additive favorable effect to pacing at the site of latest mechanical activation on clinical outcomes.
Methods
STARTER was a prospective, double-blinded, randomized trial comparing echocardiography-guided LV lead positioning to a routine fluoroscopy-only approach in patients receiving CRT. The study received approval of the institutional review board at the University of Pittsburgh, and all patients gave written informed consent before enrollment. Briefly, 187 consecutive CRT patients who met routine criteria according to guidelines were recruited from three different affiliated hospitals and randomized to an echocardiographically guided ( n = 110) or a routine ( n = 77) approach to lead placement in a 3:2 ratio. Entry criteria were age > 18 years, LV ejection fraction (EF) ≤ 35%, and QRS width ≥ 120 msec. Symptomatic status was New York Heart Association class II (10%), III (70%), or IV (20%) on optimal medical therapy. All 187 patients were used in the intention-to-treat analysis for STARTER. Ninety-six of the 110 patients (87%) in the echocardiographically guided group had complete baseline data (six had poor echocardiographic images, seven had LV lead placement failure, and one had procedure cancellation). Sixty-nine of the 77 patients (90%) in the routine group had complete baseline data (two had poor echocardiographic images, three had LV lead placement failure, and three had procedure cancellation). The present STARTER substudy consisted of 149 of these 165 patients (90%) with complete data sets, including baseline echocardiographic speckle-tracking data suitable for strain analysis for timing and amplitude, matched fluoroscopic lead position, and follow-up outcomes data. The study design considered the existing recent independent study, Targeted Left Ventricular Lead Placement to Guide Cardiac Resynchronization Therapy (TARGET), which similarly randomized CRT patients to speckle-tracking-assisted LV lead positioning and also showed improved patient outcomes compared with routine control. TARGET used a speckle-tracking radial strain amplitude cutoff of <10% as a criterion to avoid scarred segments prospectively. STARTER did not prospectively use a criterion for scar and focused on directing the LV lead only to the site of latest mechanical activation. Accordingly, this STARTER substudy provided the opportunity to determine the impact of LV lead position in relation to scarred regions and the influence of the site of latest ventricular activation.
Echocardiography
All patients underwent echocardiography (Vivid 7; GE Vingmed Ultrasound AS, Horten, Norway), with digital data analyzed by a core lab at the University of Pittsburgh Presbyterian University Hospital by investigators blinded to all other imaging and clinical data (EchoPAC BT08-BT11; GE Vingmed Ultrasound AS). Digital grayscale two-dimensional cine loops were acquired at frame rates of 60 to 90 Hz. LV volumes were assessed using the biplane Simpson’s rule. Scar was first visually assessed as akinetic or dyskinetic segments with wall thicknesses ≤ 5 mm, usually with abnormally increased acoustic reflectance. Speckle-tracking analysis for radial strain was performed on LV free wall segments at the basal and middle LV levels and transverse strain for apical segments as previously described. Briefly, regions of interest were traced on endocardial and epicardial borders, with radial strain determined from short-axis images from the basal and middle LV levels, and transverse strain was determined for apical segments using images from apical windows. Peak amplitude (percentage wall thickening) was used as a surrogate for scar, and time to peak strain from the onset of QRS determined the site of latest mechanical activation. Mechanical dyssynchrony was defined as a time difference between anteroseptal and posterior peak strain of ≥130 msec.
SPECT Imaging
To test the association of strain data to identify scar, a validation cohort consisted of 64 patients enrolled in STARTER who also had high-dose rest 99m Tc sestamibi SPECT imaging ordered for routine clinical purposes. SPECT imaging was performed using a dual-headed system (Philips Medical Systems, Andover, MA). An 1,808 circular orbit (458 right anterior oblique to 458 left posterior oblique) and a step-and-shoot format with 30 seconds of imaging at each of 64 total stops was used. A scarred segment was predefined as having akinesis or dyskinesis and complete absence of isotope uptake consistent with transmural scar ( Figure 1 ). SPECT data were then compared with visual echocardiographic and speckle-tracking strain data in the same patients from blinded analysis by other investigators. We specifically focused on the LV free wall regions at the basal and middle LV levels with radial strain data for this validation study because these are clinically preferred for LV lead placement. Accordingly, the LV free wall (nonseptal) region was divided into four equal segments at the basal and midventricular levels. These same eight segments from the standard clinical 17-segment models for both SPECT and echocardiographic imaging were compared. A predefined cutoff of <10% wall thickening by peak radial strain was tested as representative of scar. Scar burden was predefined as significant if >50% of free wall segments had peak strain values <10% consistent with scar.
LV Lead Position and Relation to Echocardiographic Segmentation
CRT was carried out using a routine transvenous approach with a right ventricular lead placed apically and an LV lead placed within an epicardial coronary vein. A routine 16-segment model was used for registration of echocardiographic segmentation with venographic-fluoroscopic data, described elsewhere in detail. Briefly, because the 17-segment model includes an apical cap that is not relevant to LV lead positioning, a 16-segment model was used. The LV free wall region was identified using coronary venography, with the great cardiac vein (in the anterior interventricular grove) defining the junction of the anterior septum and the middle cardiac vein (in the posterior interventricular groove) defining the junction of the posterior septum. This LV free wall arc of the coronary sinus was divided into four equal segments per level at basal and middle LV levels, similar to the echocardiographic segmentation. In the longitudinal plane, the left ventricle was divided into equal thirds. Accordingly, there were a total of 12 free wall LV segments, four each at the basal, middle, and apical levels. To facilitate communication with the device-implanting physician, the labeling was used to correspond to the coronary venous anatomy terminology as follows (corresponding speckle-tracking software labels listed first): anterior = anterolateral, lateral = lateral, posterior = posterolateral, and inferior = posterior. Biplane cine fluoroscopy documented the location of the LV lead. Using a standard 16-segment model, the relationship of LV lead to scar was classified as concordant when the lead was exactly within the scar segment, adjacent when the lead was in an immediate neighboring segment, or remote when the lead was at least one segment away.
Long-Term Clinical and Echocardiographic Follow-Up
The prespecified primary end point was a composite of first HF hospitalization or death over 2 years. The predefined secondary end points were (1) survival free from hard events, defined as death, LV assist device implantation, and transplantation over 2 years; (2) a ≥15% relative decrease in LV end-systolic volume (LVESV) at approximately 1 year after CRT; and (3) change in mechanical dyssynchrony after CRT. The association of the relation of LV lead position to scar on primary and secondary end points was prospectively evaluated. Clinical events were adjudicated independently by two investigators not involved in patient care.
Statistical Analysis
Continuous variables are presented as mean ± SD. Continuous variables from independent groups were compared using one-way analysis of variance followed by Bonferroni’s test for pairwise comparisons, and the χ 2 test was used for noncontinuous variables. Cutoff values, sensitivities, and specificities were determined using receiver operating characteristic curve analyses. Cross-tabulation κ analysis was performed to assess agreement of data from SPECT imaging and visual assessment of scar. Comparisons from baseline to treatment within groups were performed using paired-samples t tests, and the effect of lead localization on changes from baseline to treatment between the groups was assessed using an independent-samples t test. Kaplan-Meier analysis was used with a log-rank test to compare survival between patients grouped according to lead localization. A Cox proportional-hazards model was used to assess hazard ratios (HRs) and potential influence of covariates. Multivariate analysis was performed to examine the association of LV lead with scar when adjusted for scar burden. P values < .05 were considered significant.
Results
Echocardiographic Estimation of Scar
From the validation cohort of 512 free wall segments from 64 patients, matched SPECT, visual echocardiographic, and strain results were available in 508 segments. Transmural scar was determined in 56 segments with absent isotope uptake by SPECT imaging. Only 26 segments were determined as scar by visual assessment, of which 18 agreed with SPECT results, yielding low sensitivity (32%) for visual assessment. Although visual assessment had high specificity (98%), overall agreement between visual and SPECT analyses for scar was weak (κ = 0.38). By speckle-tracking, mean radial strain was 7.4 ± 4.9% in segments consistent with scar by SPECT imaging and 18.3 ± 9.1% in the remaining segments ( P < .0001). The predefined <10% peak free wall radial strain cutoff value was 77% sensitive and 89% specific for detecting scar determined by SPECT imaging, and this cutoff was also supported by receiver operating characteristic curve analysis ( Figure 2 ). Considering the 12 LV free wall segments used for LV lead positioning, a surrogate for scar was defined as radial strain < 10% in basal and middle LV segments and transverse strain < 10% in apical segments.
Relation of LV Lead Position to Scar on Primary Outcomes
We grouped patients with ischemic and nonischemic disease, because of the importance of scar in ischemic disease. Patients were classified as having ischemic disease by having at least one major epicardial coronary artery with >50% stenosis demonstrated by previous coronary angiography. Cardiomyopathy type was ischemic in 96 patients (64%) and nonischemic in 53 patients (36%) ( Table 1 ). The overall distribution of LV leads was basal anterolateral in 20 patients (13%), basal lateral in 20 (13%), basal posterolateral in 10 (7%), basal posterior in five (3%), middle anterolateral in four (3%), middle lateral in 47 (32%), middle posterolateral in 14 (9%), middle posterior in two (1%), apical anterolateral in one (1%), apical lateral in 14 (7%), apical posterolateral in eight (5%), and apical posterior in four (3%). Over a 2-year follow-up interval, 11 patients with nonischemic cardiomyopathy (21%) and 37 patients with ischemic cardiomyopathy (39%) reached the primary end point ( P = .026): there were two deaths and nine HF hospitalizations in the nonischemic group and 13 deaths and 24 HF hospitalizations in the ischemic group. In patients with ischemic disease, the relation of LV lead position to scar by speckle-tracking strain was classified as remote from scar in 38 patients, adjacent to scar in 36 patients, and within scar in 22 patients. HF hospitalization or death of ischemic patients with LV lead positioned remote from scar was comparable with that of patients with nonischemic cardiomyopathy, with event rates of 26% (10 of 38 patients) and 21% (11 of 53 patients), respectively, compared with patients with LV lead adjacent to scar (39% [14 of 36 patients]) or within scar (59% [13 of 22 patients]) ( P = .008). By Kaplan-Meier analysis, 2-year event-free survival was 79% in patients with nonischemic disease, 74% in patients with ischemic disease with leads remote from scar, 61% in those with leads adjacent to scar (log-rank P = .004 vs both nonischemic disease and ischemic disease with LV lead remote from scar), and 41% in those with leads within scar (log-rank P = .004 vs both nonischemic disease and ischemic disease with LV lead remote from scar) ( Figure 3 ). LV lead position remote from scar versus within or adjacent to scar was associated with a 55% risk reduction in the primary end point (HR, 0.46; 95% confidence interval [CI], 0.23–0.96; P = .04), similar to outcomes in patients with nonischemic disease. In the ischemic subgroup, LV lead position within scar was independently associated with poor outcomes (HR, 2.34; 95% CI, 1.78–4.65; P = .015), and this association remained significant after adjustment for potential covariates, including ischemic disease, age, gender, QRS duration < 150 msec, and radial dyssynchrony < 130 msec, in the multivariate model ( Table 2 ). Because total LV scar burden has been shown to be associated with outcomes after CRT, the primary end point was assessed by patients grouped according to ≥50% versus <50% free wall segments classified as scar. The 32 patients with scar burden ≥ 50% tended to have worse outcomes than the 117 patients with <50% total scar burden, but this did not reach statistical significance ( P = .131). Using multivariate analysis adjustment for scar burden, regional LV lead concordance with scar remained significantly associated with the primary end point of event-free survival ( P = .031).
| Characteristic | Nonischemic disease ( n = 53) | Ischemic disease | |
|---|---|---|---|
| LV lead within or adjacent to scar ( n = 58) | LV lead remote from scar ( n = 38) | ||
| Age (y) | 60.1 ± 13.1 | 70.7 ± 10.0 ∗ | 68.9 ± 9.9 ∗ |
| Men | 53% | 88% † | 84% ‡ |
| QRS duration (msec) | 166 ± 26 | 157 ± 29 | 159 ± 29 |
| NYHA class | |||
| II | 13% | 5% | 16% |
| III | 74% | 60% | 79% |
| IV | 13% | 35% | 5% |
| EF (%) | 25.5 ± 6.8 | 24.2 ± 5.6 | 28.4 ± 5.6 ‡,‖ |
| LVESV (mL) | 145 ± 78 | 156 ± 56 | 126 ± 19 |
| LVEDV (mL) | 191 ± 88 | 201 ± 71 | 174 ± 59 |
| RV pacing | 25% | 21% | 26% |
| LV pacing threshold (V) | 1.8 ± 1.2 | 1.6 ± 0.9 | 1.7 ± 1.2 |
| Radial dyssynchrony (msec) | 241 ± 135 | 188 ± 115 | 212 ± 112 |
| Radial strain at the LV pacing site (%) | 18 ± 11 | 13 ± 8 ‡,§ | 22 ± 11 |
| Diabetes mellitus | 26% | 40% | 39% |
| Serum creatinine (mg/dL) | 1.14 ± 0.6 | 1.32 ± 0.4 | 1.40 ± 0.5 |
| ACE inhibitors | 83% (44) | 88% (51) | 82% (31) |
| β-blockers | 87% (46) | 86% (50) | 95% (36) |
∗ P < .001 versus nonischemic.
§ P < .0001 versus LV lead remote from scar.
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
