Quantitative Gated SPECT–Derived Phase Analysis on Gated Myocardial Perfusion SPECT Detects Left Ventricular Dyssynchrony and Predicts Response to Cardiac Resynchronization Therapy

Study Population and Protocol

The study population consisted of 40 consecutive drug-refractory heart failure patients who were scheduled for implantation of a CRT device. Eligibility for CRT was based on advanced heart failure (New York Heart Association [NYHA] functional class III–IV), depressed systolic LV function (LVEF ≤ 35%), prolonged QRS duration (≥120 ms), and sinus rhythm. Patients with decompensated heart failure, a recent myocardial infarction (<3 mo), or a previously implanted pacemaker were excluded.

Before CRT implantation, all patients underwent evaluation of clinical status, GMPS with ^99mTc-tetrofosmin, and 2-dimensional (2D) echocardiography with TDI as part of the clinical assessment to determine therapeutic options.

The clinical status of the patient was assessed on the basis of NYHA functional class, exercise capacity (6-min-walk test), and QoL score (using the Minnesota QoL Questionnaire) (19).

Six months after CRT implantation, assessment of clinical status and 2D echocardiography with TDI were repeated. Patients with an improvement of 1 grade or more in NYHA class were classified as responders to CRT. Patients without improvement in NYHA class were considered nonresponders; patients who died of progressive heart failure were also considered nonresponders.

GMPS

2D Echocardiography

Transthoracic resting 2D echocardiography was performed with a commercially available system (Vivid Seven; General Electric-Vingmed), with the patient placed in a left lateral decubitus position. A 3.5-MHz transducer was used to obtain images at a depth of 16 cm in the parasternal and apical views (standard long- and short-axis, 2- and 4-chamber images). From apical 2- and 4-chamber images, LV end-systolic (LVESV) and LV end-diastolic (LVEDV) volumes and LVEF were calculated using the biplane Simpson's approach (24). In addition, TDI was performed to assess LV dyssynchrony. For TDI (triggered to the QRS complex), color Doppler frame rates exceeded 80 frames/s, and pulse repetition frequencies varied between 500 Hz and 1 kHz, yielding aliasing velocities between 16 cm/s and 32 cm/s. Tissue Doppler images and conventional echo images were saved in cine-loop format and analyzed by 2 experienced observers unaware of the GMPS and clinical patient data. All analyses were performed using commercially available software (Echopac 6.0.1; General Electric-Vingmed). To determine LV dyssynchrony, the sample volume was placed in the basal portions of the septum and lateral wall; peak systolic velocities and time-to-peak systolic velocities were obtained, and the delay in peak systolic velocity between the septum and the lateral wall was calculated as an indicator of LV dyssynchrony (25). Reliable tissue Doppler curves for the assessment of septal-to-lateral delay were obtained in 39 patients.

The peak systolic velocity analysis was limited to the LV ejection phase (marked by the opening and closing of the aortic valve). The inter- and intraobserver agreements for assessment of LV dyssynchrony were 90% and 96%, respectively (26). Substantial LV dyssynchrony on TDI was defined as a delay in peak systolic velocity greater than or equal to 65 ms between the basal septum and the lateral wall (7).

CRT Implantation

The LV pacing lead was inserted transvenously via the subclavian route as reported previously (25). A coronary sinus venogram was obtained during occlusion of the coronary sinus using a balloon catheter. The LV pacing lead was inserted into the coronary sinus with an 8-French guiding catheter and positioned as far as possible in the venous system, preferably in the posterolateral vein. The right atrial and right ventricular leads were positioned conventionally. For each patient, the atrioventricular interval was adjusted to maximize the mitral inflow duration using pulsed-wave Doppler echocardiography. No adjustments were made to the V-V interval during the first 6 mo of CRT. The CRT-device and LV-lead implantations were successful in all patients, and there were no major complications. All patients received a CRT-D device (Contak Renewal [Guidant]; InSync III, InSync Marquis, or InSync Sentry [Medtronic Inc.]; Atlas HF [St. Jude Medical]), and 4 types of pacemaker leads were used (Acuity 4554 or Easytrak 4517–4524 [Guidant]; Attain 4193–4194 [Medtronic]; Quicksite 1056T [St. Jude Medical]).

Statistical Analysis

Continuous data are expressed as mean ± SD, and categoric data are presented as percentages. When appropriate, the 2-tailed paired or unpaired Student t test and the χ² test were used for numeric or categoric data.

The relationship between histogram bandwidth and phase SD derived from GMPS and LV dyssynchrony derived from TDI was evaluated by Pearson correlation analysis. Next, univariate and multivariate analyses were performed to determine predictors of response to CRT. Receiver-operating-characteristic (ROC) curves were analyzed to determine the optimal cutoff values of histogram bandwidth and phase SD to predict response to CRT. The highest sum of sensitivity and specificity was considered as the optimal cutoff value for the prediction of response to CRT.

For all analyses, a P value of less than 0.05 was considered statistically significant.

Study Population

Baseline characteristics of the 40 patients (32 men; mean age, 63 ± 8 y) are summarized in Table 1. The etiology of heart failure was ischemic cardiomyopathy in 70% of patients and idiopathic dilated cardiomyopathy in 30%. Medication consisted of diuretics (90% of patients), angiotensin-converting enzyme inhibitors (90% of patients), and β-blockers (75% of patients) in maximum tolerated dosages. Before CRT implantation, the mean NYHA functional class was 3.1 ± 0.4 and the mean QoL score was 38 ± 19. On the surface electrocardiogram, the mean QRS interval was 156 ± 34 ms, and the 6-min-walk distance was 323 ± 96 m. On 2D echocardiography, the mean LVESV was 193 ± 79 mL, with a mean LVEDV of 252 ± 84 mL and a mean LVEF of 25% ± 8%. Mean heart rate during GMPS was 68.6 ± 12.9 beats/min, and the mean heart rate during TDI was 68.5 ± 13.2 beats/min (P = not significant). The mean number of cardiac segments with a perfusion defect was 6 ± 3.

LV Dyssynchrony: GMPS Versus TDI

Phase analysis on GMPS yielded a mean histogram bandwidth of 84° ± 26° (range, 40°−152°) and a mean phase SD of 23° ± 7° (range, 12°−41°). On TDI, the mean LV dyssynchrony was 53 ± 35 ms (range, 0–130 ms). Pearson correlation showed a significant correlation between LV dyssynchrony on TDI and histogram bandwidth on GMPS (r = 0.69, r² = 0.48, SEE = 25.4, n = 39, P < 0.01, Fig. 1A). Also, a significant correlation between LV dyssynchrony on TDI and phase SD on GMPS was demonstrated (r = 0.65, r² = 0.42, SEE = 26.8, n = 39, P < 0.01, Fig. 1B). Finally, a good correlation between histogram bandwidth and phase SD was observed (r = 0.95, r² = 0.91, SEE = 8.0, P < 0.01).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1. 

Agreement between LV dyssynchrony on TDI and histogram bandwidth (A) or phase SD (B) derived from GMPS. Significant correlation between LV dyssynchrony on TDI and histogram bandwidth or phase SD as assessed with GMPS was observed.

Patients with LV dyssynchrony greater than or equal to 65 ms on TDI (indicating substantial LV dyssynchrony) had significantly larger histogram bandwidth and phase SD than did patients with LV dyssynchrony below 65 ms on TDI (Fig. 2).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 2. 

(A) Histogram bandwidth was significantly larger in patients with substantial LV dyssynchrony on TDI (defined as mechanical delay between basal septum and lateral wall ≥ 65 ms) than that in patients without LV dyssynchrony on TDI (delay < 65 ms) (105 ± 23 vs. 71 ± 17, P < 0.05). (B) Phase SD was significantly larger in patients with substantial LV dyssynchrony on TDI (defined as mechanical delay between basal septum and lateral wall ≥ 65 ms) than that in patients without LV dyssynchrony on TDI (delay < 65 ms) (28 ± 6 vs. 20 ± 5, P < 0.05).

Clinical Responders and Nonresponders

At the 6-mo follow-up, 24 (60%) patients showed an improvement of 1 or more grades in NYHA functional class and were considered responders to CRT. Among the 16 (40%) nonresponders, 5 patients died within 6 mo of CRT because of progressive heart failure.

In Table 2, baseline clinical and imaging variables for responders and nonresponders are summarized. Responders showed a significantly longer 6-min-walk distance and larger LVEDV. Furthermore, responders had significantly more LV dyssynchrony on TDI (Fig. 3).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 3. 

Extent of LV dyssynchrony on TDI at baseline was significantly larger in responders to CRT than that in nonresponders to CRT (62 ± 37 vs. 40 ± 26, P < 0.05).

For the GMPS data, responders showed significantly larger histogram bandwidth (94° ± 23° vs. 68° ± 21°, P < 0.01) and phase SD (26° ± 6° vs. 18° ± 5°, P < 0.01) than did nonresponders. Responders to CRT showed a significantly lower number of cardiac segments with perfusion defects than did nonresponders to CRT (5 ± 2 vs. 8 ± 2, P < 0.01). Examples of GMPS data in patients with and without response to CRT are shown in Figure 4.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 4. 

(A) Example of a patient without LV dyssynchrony on GMPS. Synchronous contraction pattern is reflected by homogeneous phase angle distribution of polar map (left) and narrow highly peaked histogram (right). At 6-mo follow-up, no improvement in NYHA functional class was observed and LVEF remained unchanged (32% at baseline vs. 33% at 6-mo follow-up). (B) Example of a patient with extensive LV dyssynchrony on GMPS. LV dyssynchrony is indicated by heterogeneous phase angle distribution of polar map (left) and wide histogram (right). At 6-mo follow-up, NYHA functional class improved from 3 to 2, with increase in LVEF from 21% to 33%.

Baseline and 6-Mo Follow-up Data

At the 6-mo follow-up, responders showed a significant improvement in NYHA functional class, QoL score, and 6-min-walk distance, whereas nonresponders did not show significant improvement in any of the clinical parameters (Table 3). Furthermore, at the 6-mo follow-up only responders demonstrated a significant decrease in LVEDV and LVESV and a significant increase in LVEF.

Univariate analysis was performed on clinical, echocardiographic, and GMPS variables and demonstrated that 6-min-walk distance, LVEDV, and LV dyssynchrony on TDI were predictors of response to CRT at the 6-mo follow-up. In addition, histogram bandwidth and phase SD were predictive of CRT response at the 6-mo follow-up. Multivariate analysis showed that 6-min-walk distance (P = 0.042) and phase SD (P < 0.01) were the only independent predictors for response to CRT.

ROC curve analysis was used to define the optimal threshold for prediction of response to CRT. For histogram bandwidth, optimal cutoff value was defined at 72.5°, which yielded a sensitivity of 83% and a specificity of 81%; the area under the curve (AUC) was 0.83 (Fig. 5A). For phase SD, optimal cutoff value was defined at 19.6°, yielding a sensitivity of 83% and a specificity of 81%; the AUC was 0.85 (Fig. 5B).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 5. 

ROC curve analysis for histogram bandwidth (A) and phase SD (B) for prediction of response to CRT. ROC curve analysis for histogram bandwidth showed good predictive value (AUC, 0.83) of response to CRT. Optimal cutoff value was defined at 72.5°, yielding sensitivity of 83% and specificity of 81%. For phase SD, ROC curve analysis demonstrated good predictive value (AUC, 0.85) of response to CRT. Optimal cutoff value was defined at 19.6°, yielding sensitivity of 83% and specificity of 81%.

LV Dyssynchrony Derived from GMPS

Patients are selected for CRT on the basis of the current American Heart Association/American College of Cardiology/European Society of Cardiology guidelines, which recommend CRT in patients with NYHA class III or IV heart failure, an LVEF of no more than 35%, and a prolonged QRS interval greater than or equal to 120 ms (6–10).

However, a substantial number of the selected patients do not show response to CRT (2,5–8). Therefore, efforts have been made to improve current selection criteria for CRT, and previous studies have demonstrated that LV dyssynchrony is related to CRT response and may have additional value in the selection of potential candidates for CRT (6–8). One of the potential explanations for nonresponse in patients meeting contemporary guidelines is that a wide QRS complex may reflect interventricular rather than intraventricular dyssynchrony (26).

To detect LV dyssynchrony, a variety of imaging techniques can be used, including echocardiography with TDI (6–8) or myocardial strain (rate) imaging (11), MRI (12), and nuclear imaging with GMPS (14–18). Most studies on CRT have reported on echocardiographic methods to assess LV dyssynchrony; in particular, echocardiography using TDI has been used as a tool for the assessment of LV dyssynchrony (6-8). Recently, the feasibility of phase analysis on GMPS for detection of LV dyssynchrony has been shown (27,28). This approach provides automatically quantitative parameters of LV dyssynchrony using count-based analysis and has been developed to extract amplitude and phase angle from changes in counts per region during the cardiac cycle (14–17). Amplitude of the phase curve indicates maximal systolic wall thickening, whereas phase angle indicates the beginning of mechanical contraction per region, providing information on the degree of hetero- or homogeneity of the mechanical activation pattern within the ventricle. A histogram was constructed that represents the distribution of the phase angle for all sample points of the left ventricle, providing quantitative parameters for LV dyssynchrony including histogram bandwidth (includes 95% of the elements of the phase histogram) and phase SD (SD of phase distribution) (18).

Various studies have shown that this approach can detect LV dyssynchrony in patients with heart failure and a wide QRS complex (27,28). Subsequent studies have shown the relationship between phase analysis on GMPS and echocardiographic techniques for the assessment of LV dyssynchrony (29,30). Henneman et al. (29) evaluated 75 heart failure patients with a wide QRS complex in a head-to-head comparison between phase analysis on GMPS and 2D echocardiography with TDI. The authors demonstrated that histogram bandwidth and phase SD derived from GMPS related well to LV dyssynchrony derived from TDI. Marsan et al. (30) used real-time 3-dimensional echocardiography and also demonstrated a good relationship with phase analysis on GMPS for the detection of LV dyssynchrony in 40 patients with heart failure. In the current study, GMPS data were processed with the QGS algorithm, which was expanded to provide quantitative parameters for the assessment of LV dyssynchrony. It was shown that LV dyssynchrony derived from TDI related significantly to histogram bandwidth and phase SD obtained with the QGS algorithm.

GMPS for Prediction of Response to CRT

Many echocardiographic studies have shown that LV dyssynchrony is predictive for response to CRT (6–8). Only 1 study has reported on the use of GMPS for prediction of response to CRT (31); in that study, 42 patients with heart failure (NYHA class III–IV), an LVEF of no more than 35%, and a QRS greater than 120 ms underwent GMPS before CRT implantation. In the patients, reflectors of LV dyssynchrony (phase SD and histogram bandwidth) were identified as predictors of response to CRT over a 6-mo follow-up period. Particularly, a cutoff value of 43° for phase SD yielded a sensitivity and specificity of 74% for prediction of response to CRT; a cutoff value of 135° for histogram bandwidth had a sensitivity and specificity of 70%.

In the current study, a different algorithm derived from the QGS program was used for the assessment of phase SD and histogram bandwidth. In line with previous work (31), histogram bandwidth and phase SD were significantly larger in CRT responders than in CRT nonresponders (Table 2). When ROC curve analysis was applied, a cutoff value of 19.6° for phase SD yielded a sensitivity of 83% and a specificity of 81% (AUC, 0.85) to predict response to CRT at the 6-mo follow-up. For histogram bandwidth, a similar sensitivity and specificity were obtained at a cutoff value of 72.5° (AUC, 0.83). The accuracy to predict response to CRT in the current study is in line with earlier work using different software to analyze GMPS (31). The optimal cutoff values for phase SD and histogram bandwidth in the present study are not identical to the values obtained in the study by Henneman et al. (31), and this may be related to differences in software packages or to differences in study populations. One of the major differences between these software packages is that QGS uses an ellipsoidal sampling system (21), whereas the Emory Cardiac Toolbox uses the hybrid cylindric–spheric coordinate system (32).

In both the study by Henneman et al. (31) and the current study, the accuracy of GMPS to predict response to CRT was not perfect. Compared with TDI, however, the advantages of GMPS include providing global measurements and having measurements that are automatically derived. Given these advantages and the accuracy demonstrated by this and previous work, GMPS may provide a reasonable alternative for clinical use to TDI in predicting CRT response.

It is not realistic to expect a perfect accuracy for prediction of CRT response based solely on LV dyssynchrony. Indeed, recent work has indicated that issues other than LV dyssynchrony are important for response to CRT. Scar tissue has been identified as an important factor limiting the response to CRT (33,34). It has been shown that the extent of scar tissue was inversely related to the response to CRT; also, patients with scar tissue in the region where the LV lead was positioned did not respond well to CRT and the extent (the total scar burden) of scar tissue has been shown to prevent response to CRT. Therefore, the preferred method would be to assess not only the presence of LV dyssynchrony but also the extent and location of scar tissue in the left ventricle. In this perspective, GMPS has the potential to provide both these parameters from the same dataset.

Finally, it has been shown that patients with the LV lead positioned outside the region of latest mechanical activation benefitted less from CRT than did patients with the lead positioned in the area of latest activation (35). More recently, Ypenburg et al. (36) reported on the outcome of patients undergoing CRT with the LV lead positioned inside or outside the region of latest mechanical activation and demonstrated that long-term survival was better when the LV pacing lead was positioned at the site of latest mechanical activation.

Study Limitations

The current work should be considered a feasibility study, validating a new algorithm on phase analysis with QGS. Patients with extensive scar tissue in the left ventricle benefit less from CRT. In addition, patients with the LV pacing lead located outside the region of latest mechanical activation will have a less favorable response to CRT. In the current study, only the number of cardiac segments with a perfusion defect was integrated. To provide a meaningful analysis on all these issues, however, a much larger study population is needed. Whether integration of all this information will result in a superior prediction of response to CRT needs further testing in future studies.