The clinical utility of stress echocardiography has been established over the more than three decades of its existence, with numerous studies documenting the diagnostic and prognostic value of the technique. Throughout this entire period of time, visual “eyeball” assessment of regional wall motion and thickening has been the standard method for detection of abnormalities in contractile function related to ischemia. The limitations of visual assessment were acknowledged 20 years ago, in two widely referenced studies showing reduced interinstitutional reproducibility and reduced accuracy for less than expert interpreters. With the expanding popularity of stress echocardiography in the 1990’s, emphasis was placed on training expert interpreters, which was shown to improve the accuracy and reproducibility of wall motion analysis. As the technique matured into an established modality, the emphasis on becoming an expert interpreter waned. There are currently two major approaches to improving our ability to assess abnormalities in regional function. The first, the use of contrast agents for endocardial border detection, has improved the clinical utility of stress echocardiography by increasing the number of studies that can be assessed by visual analysis. The second has been the development of quantitative techniques for the analysis of regional and global ventricular function, including automated endocardial border detection and pulsed and color tissue Doppler. Tissue Doppler was particularly appealing because it enabled the assessment of displacement, velocity, strain (shortening), and strain rate (rate of change of shortening) with stress. A modest number of studies showed that tissue Doppler–based methods provided quantitative information that could both increase the accuracy of nonexpert readers and yield similar accuracy to expert visual assessment of wall motion. However, tissue Doppler has technique-specific limitations, including labor- and time-intensive analysis, angle dependence, base-to-apex gradients in velocity and displacement, and noisy signals.
As the limitations of Doppler-derived methods of quantitation became apparent, assessment of strain based on two-dimensional speckle-tracking was introduced as another potentially useful quantitative method. The major advantages of speckle-tracking over tissue Doppler methodology include the lack of angle dependency, favorable signal-to-noise ratio, and less time-intensive analysis. An ever increasing number of studies have demonstrated the clinical value of speckle-tracking for the assessment of left ventricular longitudinal strain in the resting state. The enthusiasm generated by the results of these studies has raised our hopes that speckle-tracking will have equal clinical value when applied with stress testing. Abnormal longitudinal strain may be a sensitive marker of stress-induced ischemia because of the predominance of longitudinal fibers in the subendocardium. In an animal model of coronary stenosis, reduction in longitudinal strain was readily detected by speckle-tracking during dobutamine stress.
In this context, in their study reported in this issue of JASE , Joyce et al . compared the value of two-dimensional speckle-tracking and conventional visual assessment for the detection of significant (≥70% stenosis) coronary artery disease (CAD) in subjects with previous ST-segment elevation infarction treated with percutaneous intervention. After the exclusion of 20 subjects (15%) because of suboptimal image quality, the investigators enrolled 105 subjects who underwent dobutamine echocardiography 3 months after infarction and angiography within 1 year of their events. Angiography revealed significant CAD in 38 subjects. Those with significant disease had greater worsening of global longitudinal end-systolic strain from rest to peak dose (−16.8 ± 0.5% to −12.6 ± 0.5%) compared with subjects with no disease (−16.6 ± 0.4% to −14.3 ± 0.3%) ( P < .001). The optimal cutoff value for the rest-to-stress change in longitudinal strain for detection of disease was −1.9%; this cutoff value yielded sensitivity of 87%, specificity of 46%, and accuracy of 60%. Conventional wall motion analysis yielded sensitivity of 11%, specificity of 94%, and accuracy of 64%. Multivariate analysis showed that the change in longitudinal strain had incremental value for the prediction of significant disease compared with the combination of clinical variables and visual wall motion analysis.
Joyce et al . also investigated the utility of the −1.9% threshold value of the rest-to-stress change in longitudinal strain for the detection of disease in the three individual coronary artery territories. Using receiver operating characteristic analysis, areas under the curve ranged from 0.41 to 0.82 among the seven segments of the left anterior descending coronary artery territory. Areas under the curve ranged from 0.42 to 0.74 in the left circumflex coronary artery territory (five segments) and the right coronary artery territory (five segments). Using the segments in each territory with the optimal cutoff value identified by receiver operating characteristic analysis, accuracy for the detection of disease was 67%, 50%, and 42% in the left anterior descending, left circumflex, and right coronary artery territories, respectively. Interobserver variability was 16.7% for global longitudinal strain at peak stress, which translates to a 2.4% absolute difference in strain. Interobserver variability for regional strain was in a similar range. The investigators offer a mixed conclusion, stating that strain analysis is a promising new technique but acknowledging that its low specificity, segment-to-segment heterogeneity, and limitations of the technology currently prevent the routine clinical adoption of this method.
The study by Joyce et al . raises an array of questions about the present value and potential future role of strain imaging in stress echocardiography. An honest appraisal of the technique might focus on answering the following questions.
First, is the feasibility of the strain imaging with pharmacologic and exercise stress high enough to convince clinicians to apply the technique on a routine basis? The success rates of obtaining adequate strain data during dobutamine stress have ranged from 77% to >90%, representing feasibility generally lower than the success rate of obtaining adequate images for visual interpretation of wall motion with contrast enhancement. Proposed reasons for the lower success rates of acquiring strain data include reduced image quality during stress, study-to-study variability in subject characteristics, experience in processing, and thresholds chosen for acceptable strain images. Reasonable-quality grayscale images are a prerequisite for applying strain technology. The excess cardiac motion that can occur at high levels of stress can result in speckles’ moving more quickly out of the two-dimensional imaging plane. Current technology uses semiautomated processing of strain data but also permits manual adjustment of regions of interest to improve tracking and the choice to accept or reject data from individual segments. Reprocessing of strain images by an expert operator may improve the yield of segments with usable data, at the expense of increasing processing times to several minutes per examination. The virtual absence of studies using speckle-tracking with exercise stress is a glaring deficiency in the literature attributable to the perceived difficulty of obtaining adequate data with lower quality exercise images. Strain data have been successfully acquired during low- to moderate-intensity bicycle exercise in non-CAD populations to assess responses in global ventricular function. An important step in determining the clinical value of strain imaging would be testing its feasibility during high-intensity bicycle exercise and after treadmill exercise in those with known or suspected CAD.
Second, what is the normal response of longitudinal strain to stress? In the study by Joyce et al ., there was a modest (−16.6% to −14.3%) but significant decline in global strain from rest to stress in those with no significant disease. This finding is counterintuitive to the expectation of most clinicians that in the absence of CAD, contractility should increase with stress. In previous studies, global strain has been shown to have a highly variable response to dobutamine stress in subjects without CAD. Several studies have shown that strain (shortening) slightly increases (larger negative values) from rest to peak stress in those without CAD, whereas one study (Yu et al .) showed a marked decline in strain in individuals without obstructive disease (−17.6% to −10.8%). The reasons for the wide disparity in the response of global strain in normal subjects may be both physiologic and technical. The normal hemodynamic response to dobutamine stress is usually a substantial reduction in preload and cavity volume. With lower preload and decreased diastolic myocardial fiber length, the increment in myocardial shortening may be less at peak stress than at rest, when cavity volume and diastolic fiber length are larger. Baseline fluid volume, cavity size, and individual hemodynamic responses to dobutamine are likely to affect the response of strain to dobutamine stress. Strain is a load-dependent parameter that has been shown in numerous studies to improve with reductions in afterload and decline to levels considered abnormal with modest increases in afterload. The response of blood pressure and afterload with dobutamine stress can be highly variable. The sampling rate limitations of current technology may be the major technical factor that contributes to lower than expected strain with the tachycardia and the increased out-of-plane motion that occur with peak stress. Rosner et al . found that 30 frames/cardiac cycle was the minimum sampling rate that provided accurate values of global longitudinal strain. Frame rates used for the acquisition of strain during stress have varied widely, ranging from 25 to 100 frames/sec. Heart rates achieved during stress routinely approach or exceed 150 beats/min, which would require frame rates of ≥75 frames/sec to achieve the desired 30 frames/cardiac cycle. Below 30 frames/cardiac cycle, strain values underestimate the true extent of shortening, indicating that the decline of global strain in subjects without CAD may be partially artifactual.
Third, are global and regional strain values reproducible? Global longitudinal strain values obtained at rest in subjects evaluated by expert observers, using a single vendor’s machine, have been shown to have acceptable reproducibility. As expected, interobserver variability of global strain is higher when evaluated with dobutamine stress. In the study by Joyce et al ., interobserver variability in global strain at peak stress averaged 16.7%, which translates to a 2.4% difference in strain, which is larger than the 1.9% cutoff value for the change in strain that best distinguished between those with and without significant CAD. Substantial interobserver variability for global strain has also been reported in the majority of previous dobutamine stress studies, lowering one’s confidence about the reliability of using published threshold values of global strain—or the rest-to-stress change in global strain—for the diagnosis of CAD. Regional strain may have similar or greater interobserver variability than global strain. Interobserver variability for regional strain during dobutamine stress ranged from 11.3% to 18.8% in the present study. Similarly wide interobserver variability (22%) was reported by Yamada et al . for regional strain. The reasons for less than desirable reproducibility of global and regional strain values include many of the aforementioned factors that lower the feasibility of strain imaging during stress.
In addition to the challenge of wide interobserver variability in strain, the present study showed that there can be substantial variability of strain values during stress among adjacent segments in the same coronary distribution. Because of this variability, the investigators selected one segment for each coronary territory for analysis to determine the presence or absence of significant disease. Even with this approach, accuracy for the identification of disease in the posterior circulation was low. Hanekom et al . also showed that localization of disease using segmental strain values was challenging (areas under the curve ranging from 0.59 to 0.85). The reasons for wide adjacent-segment variability in strain are more likely to be technical rather than physiologic. Algorithms used to smooth data, and region-of-interest size, may play a role. Both interobserver variability and segment-to-segment variability in strain appear highest in the posterior circulation, which has a preponderance of basal segments. The increased variability of basal segment strain may be attributed to the effects of artifacts on basal segment image quality and greater difficulty in tracking basal versus apical segments during stress because of the increased rotation and translation of basal segments.
Intervendor variability of global and regional strain adds an additional challenge when one tries to apply strain results from published studies in clinical practice. In the recent past, the reproducibility of global strain values has been suboptimal when compared among different vendors and among different software versions from a single vendor. The use of vendor-independent software for analysis, and the incorporation of agreed-upon guidelines to reduce variability of global strain in the newest generation of proprietary software, has improved intervendor variability to the range of 6% to 7% for global longitudinal systolic left ventricular strain. However, even with improvement in intervendor reproducibility of global strain, the variability may remain high enough that diagnostic values of global strain determined using a particular machine and software may not be applicable when another machine or software package is used.
An ideal quantitative method would be primarily automated, requiring limited operator input, and be usable with both pharmacologic and exercise stress. More important, the method must provide accurate and reproducible information that can be readily understood by the interpreting physician. Pooled results of studies comparing the accuracy of assessment of longitudinal end-systolic strain by speckle-tracking versus visual analysis for the detection of CAD have not shown a clear-cut advantage of this quantitative parameter. The optimal quantitative parameter for the detection of CAD, end-systolic strain, change in strain with stress, end-systolic strain rate, or postsystolic strain, has yet to be determined. Strain rate (rate of change in shortening) rather than strain may be the preferred parameter for use with dobutamine stress because of the greater increase in strain rate versus strain with catecholamine stimulation. However, the requirement of very high frame rates for the assessment of strain rate during stress might potentially compromise spatial resolution and tracking accuracy. The ideal quantitative method should also provide information that has at least supplemental value to what can be obtained from visual analysis using current state-of-the-art technology. The study of Joyce et al . demonstrated that the change in strain with stress had incremental value over visual assessment. However, this was in a setting in which the sensitivity of visual analysis was 11%. It is unlikely that stress echocardiography would be used at all for the detection of disease in the postinfarction setting if the sensitivity were truly that low. The real value of strain assessment will be known if or when it improves on the accuracy of visual assessment performed with high–frame rate harmonic imaging with contrast enhancement. Nagy et al . recently reported successful acquisition of strain data in 75% of subjects with contrast-enhanced images undergoing dobutamine stress. Strain assessment and visual analysis had comparable accuracy.
We seem quite removed from the day when a clinician will be able to take a 30-sec look at polar maps of strain at rest and with stress and, in conjunction with visual analysis, identify the absence or presence of ischemia and its location with confidence. However, the combined and concerted effort of scientists and vendors to advance the technology for the purposes of improving the acquisition and reproducibility of strain data is ongoing. It should be the role of talented investigators and institutions, like the Leiden University group, to perform additional clinical studies aimed at addressing the many physiologic and technical issues that currently make the application of speckle-tracking to stress echocardiography a challenging exercise.
Dr Sawada is supported by the Strategic Research Initiative, Indiana University School of Medicine.