Background
The reproducibility of the measurement of mechanical dyssynchrony by echocardiography including Doppler tissue imaging has recently been questioned. The aim of this study was to ascertain the role of a dedicated training program to improve skills and the reproducibility of dyssynchrony assessment.
Methods
In 70 patients with heart failure, color Doppler tissue images were acquired, and the time to peak systolic velocity of each segment and several dyssynchrony indices, including the standard deviation of time to peak systolic velocity, were measured by an expert to constitute a reference standard. The same images were then assessed by two beginners, who had only basic knowledge of dyssynchrony analysis after a 1-hour lecture, and two graduates, who had received a structured hands-on training program. Both sets of results were compared with the standard.
Results
For the standard deviation of time to peak systolic velocity, the linear correlations between the standard and beginner 1 ( r = 0.643) and beginner 2 ( r = 0.532) were only modest ( P < .001 for both). When referenced to the standard, interobserver variability was 18% for beginner 1 and 19% for beginner 2. Measurements with differences of ≥10 msec were found in 24% and 22% of cases by beginners 1 and 2, respectively. In contrast, the assessments made by graduates 1 and 2 were significantly improved. The correlation coefficients were 0.935 and 0.929 ( P < .001 for both), and interobserver variability values were 8% and 7%. The prevalence rates of measurements with differences ≥ 10 msec were 1.5% and 3%, respectively.
Conclusions
There is a learning curve for the measurement of systolic dyssynchrony using Doppler tissue imaging, but good reproducibility can be achieved by the use of a dedicated training program.
In the past few years, advanced echocardiographic technologies, including Doppler tissue imaging (DTI), have been used to evaluate mechanical systolic dyssynchrony. For the treatment of patients with heart failure by cardiac resynchronization therapy (CRT), it has been suggested that systolic dyssynchrony assessed by DTI may predict response to treatment. However, the role of DTI for dyssynchrony assessment in CRT candidates was challenged by the results of the Predictors of Response to Cardiac Resynchronization Therapy (PROSPECT) study, in which high variability for all tested parameters of dyssynchrony was reported, such that they were incapable of predicting response accurately. Therefore, since PROSPECT, there has been major concern as to whether DTI has sufficient reproducibility for the routine assessment of mechanical dyssynchrony. However, as with training for CRT device implantation, the application of this new imaging tool requires not only knowledge but also hands-on skills, for which a learning curve does exist. The aim of this study was to test if a dedicated hands-on training program would improve the reproducibility of dyssynchrony analysis using DTI.
Methods
Color Doppler tissue images of the three apical views (i.e., the apical four-chamber, two-chamber, and three-chamber views) from 70 study patients were acquired (Vivid 7; GE Vingmed Ultrasound AS, Horten, Norway) to assess longitudinal motion of the left ventricle, as previously described. Patients were in sinus rhythm with left ventricular (LV) ejection fractions < 50%. Doppler tissue images were optimized for pulse repetition frequency, color saturation, sector size, and depth, which allowed a maximum frame rate of >100 Hz. The alignment of the Doppler beam and LV walls was maximized in each view. At least three consecutive beats were digitally stored for offline analysis. Myocardial velocity curves were reconstituted with the aid of a customized software package (EchoPAC PC version 6.1.0; GE Vingmed Ultrasound AS), using the six basal and six mid segmental model, which consists of septal, lateral, anteroseptal, inferolateral, anterior, and inferior segments at both the basal and mid levels of the left ventricle. Referencing to the onset of the QRS complex, the time to peak myocardial systolic velocity during the ejection phase (Ts) was measured in each segment. The indices of mechanical systolic dyssynchrony derived from Ts were the standard deviation of Ts among the 12 LV segments (Ts-SD), the maximal difference of Ts among the 12 LV segments (Ts-DIF), and the maximal opposite wall delay among the 12 LV segments (Ts-OPW) as well as among the four basal segments (Ts-OPW-4).
The images were first interpreted by one cardiologist who was very experienced in dyssynchrony assessment by DTI (Q.Z.). To test the consistency of the readings, another very experienced cardiologist (Y.-J.L.) was asked to perform the same analysis, blinded to previous results. The Ts-SD values computed by these two cardiologists were then compared, indicating a close linear correlation ( r = 0.988, P < .001) and low interobserver variability (4 ± 3%). The correlation was also good for the calculated Ts-DIF ( r = 0.968), Ts-OPW ( r = 0.966), and Ts-OPW-4 ( r = 0.965) values ( P < .001 for all). Moreover, if the Ts-SD reading by Y.-J.L. had a difference of ≥5 msec compared with that by Q.Z. in any patient, the measurement was redone together to reach a final decision that would be adopted by Q.Z. Consequently, the readings by Q.Z., which were endorsed by an expert (C.-M.Y.), served as the “reference standard” of measures in the present study. In addition, the images from the 70 patients were labeled as level 1 (easy, 20 patients; Figure 1 a), level 2 (easy to difficult, 30 patients; Figure 1 b), or level 3 (difficult, 20 patients; Figure 1 c) according to difficulty in the analysis of Ts. This referred to the presence or absence of negative systolic velocity peak, postsystolic shortening peak, multiple systolic peaks, or artifacts.
We then recruited four cardiologists who specialized in echocardiography but had not been specifically trained for dyssynchrony analysis. Two of them were randomized to receive a 1-hour lecture followed by demonstration of the methods and practical tips on Ts measurement (beginners 1 and 2). The other two were randomized to receive a structured training program in the form of a 2-day workshop, which emphasized both online image acquisition and offline analysis skills (graduates 1 and 2). In addition to short lectures covering the fundamentals of DTI, clinical applications of DTI, performance of online scanning and offline analysis, as well as troubleshooting, this program in particular included several hands-on and self-testing sessions in which the participants were instructed by experienced tutors to practice image acquisition in two volunteers and Ts measurements in 50 training cases (different from the 70 cases included in the present study), with different levels of difficulty. All four cardiologists measured Ts in each segment of the 70 study patients at two different time points.
Continuous variables are expressed as mean ± SD. P values < .05 were considered statistically significant. Linear correlation analysis with Pearson’s correlation coefficient, Bland-Altman analysis by calculating the bias (mean difference) and the 95% limits of agreement (2 SDs from the mean difference), and interobserver variability tests were performed for comparison between the reference standard and the measurements performed by the four enrolled cardiologists, respectively. The intraobserver variability test was also used to compare the two readings of Ts-SD by the same operator.
Results
The mean LV ejection fraction of the 70 patients was 31.4 ± 8.6%, and the mean QRS duration was 125 ± 24 msec. The reference standard Ts-SD obtained by the expert reader (Q.Z.) was 45.7 ± 11.5 msec, and the mean Ts-DIF, Ts-OPW, and Ts-OPW-4 were 120 ± 32, 109 ± 30, and 95 ± 30 msec, respectively. Sixty patients (86%) had significant systolic dyssynchrony, defined as Ts-SD ≥ 33 msec, as previously validated. Ts-SD was closely correlated with Ts-DIF ( r = 0.872), Ts-OPW ( r = 0.876), and Ts-OPW-4 ( r = 0.787) ( P < .001 for all).
Compared with the standard, the readings by the beginners were unsatisfactory. As shown in Table 1 , the correlation coefficients ranged from 0.292 to 0.643 for the four dyssynchrony indices investigated. The interobserver variability values in measurement were large, ranging from 18% to 32%. The results of Bland-Altman analysis are summarized in Table 1 and Figures 2 to 5 , which display a wide range between the upper and lower 95% limits of agreement. When referenced to the expert data, Ts-SD measurements with differences ≥5 msec were found in 46% of cases done by beginner 1 and in 47% of cases by beginner 2. The figures for measurements of Ts-SD with differences ≥ 10 msec were 24% and 22%, respectively. Moreover, intraobserver variability in Ts-SD computation was 12 ± 17% for beginner 1 and 13 ± 18% for beginner 2.
| Beginner 1 | Beginner 2 | |||||||
|---|---|---|---|---|---|---|---|---|
| Ts-SD | Ts-DIF | Ts-OPW | Ts-OPW-4 | Ts-SD | Ts-DIF | Ts-OPW | Ts-OPW-4 | |
| Linear correlation ( r ) | 0.643 | 0.532 | 0.550 | 0.497 | 0.567 | 0.365 | 0.360 | 0.292 |
| Variability (%) | 18 ± 21 | 20 ± 20 | 19 ± 20 | 32 ± 37 | 19 ± 23 | 23 ± 24 | 21 ± 26 | 25 ± 34 |
| Bland-Altman | ||||||||
| Mean difference (msec) | 3.45 | 3.46 | 5.52 | 10.86 | 0.92 | −1.81 | 2.99 | 3.09 |
| 95% upper limit (msec) | 24.86 | 69.28 | 64.14 | 73.98 | 28.96 | 90.28 | 85.65 | 86.60 |
| 95% lower limit (msec) | −17.96 | −62.35 | −53.10 | −52.25 | −27.11 | −93.90 | −79.68 | −80.43 |
On the other hand, the dyssynchrony measurements by both graduates were significantly improved. As summarized in Table 2 , there were close relationships in the readings between the expert and the two graduates for all dyssynchrony indices, with correlation coefficients ranging from 0.800 to 0.937 ( P < .001 for all). Interobserver variability was about 10% in computing Ts-SD, Ts-DIF, and Ts-OPW and a little higher at 16% for Ts-OPW-4. Table 2 and Figures 2 to 5 show the results of Bland-Altman analysis for graduates 1 and 2. Not only did the mean differences diminish, but also the ranges between the upper and lower 95% limits of agreement became much narrower. Similarly, compared with the standard, measurements of Ts-SD with differences ≥ 5 msec were found in 24% of cases done by graduate 1 and in 19% of cases by graduate 2. The figures for measurements of Ts-SD with differences ≥ 10 msec were only 1.5% and 3%, respectively. Furthermore, intraobserver variability in Ts-SD measurement by graduate 1 was 5 ± 6%, while it was 4 ± 6% by graduate 2.
| Graduate 1 | Graduate 2 | |||||||
|---|---|---|---|---|---|---|---|---|
| Ts-SD | Ts-DIF | Ts-OPW | Ts-OPW-4 | Ts-SD | Ts-DIF | Ts-OPW | Ts-OPW-4 | |
| Linear correlation ( r ) | 0.935 | 0.837 | 0.814 | 0.856 | 0.929 | 0.840 | 0.807 | 0.800 |
| Variability (%) | 8 ± 7 | 12 ± 9 | 12 ± 10 | 15 ± 16 | 7 ± 7 | 11 ± 9 | 11 ± 11 | 16 ± 17 |
| Bland-Altman | ||||||||
| Mean difference (msec) | 0.97 | −3.56 | 0.67 | 1.28 | −0.63 | −6.34 | −2.55 | −0.85 |
| 95% upper limit (msec) | 9.43 | 34.60 | 38.26 | 32.17 | 7.85 | 29.72 | 33.64 | 34.53 |
| 95% lower limit (msec) | −7.49 | −41.73 | −36.91 | −29.60 | −9.12 | −42.41 | −38.74 | −36.23 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
