|Home | About | Journals | Submit | Contact Us | Français|
The acute adverse effects of left ventricular (LV) dyssynchrony on cardiac performance were first described in 1925 by Carl Wiggers1. In recent years, the accurate diagnosis of LV dyssynchrony has become the focus of a myriad of publications, driven by the advent of cardiac resynchronization therapy (CRT) to treat heart failure due to severe LV dysfunction in the setting of marked prolongation of the QRS interval2–4. In the initial large clinical trials of CRT, QRS duration was used as a measure of dyssynchrony to select patients for treatment2, 3. However, sensitivity5 and specificity6 of QRS duration to predict response to CRT were less than optimal. Subsequently, numerous “time-to-peak” parameters based directly on the motion of the LV walls were developed to diagnose LV mechanical dyssynchrony with echocardiography in an attempt to improve CRT selection criteria7.
Echocardiographic mechanical dyssynchrony parameters initially showed promise in predicting response to CRT in single-center studies8–12. However, the multicenter Predictors of Response to CRT (PROSPECT) study recently reported that no echocardiographic dyssynchrony parameter could be recommended to improve patient selection for CRT beyond current guidelines13. In addition, the Resynchronization Therapy in Narrow QRS (RETHINQ) trial recently reported that patients with narrow QRS and evidence of mechanical dyssynchrony do not benefit from CRT14.
So where do we go from here? Should selection of patients for CRT based on mechanical dyssynchrony be abandoned in the wake of the negative results from PROSPECT and RETHINQ? We believe that techniques to quantify LV mechanical dyssynchrony need to be refined, not forgotten, and will still play a role in improving CRT selection criteria in the future. This refinement of dyssynchrony quantification requires a paradigm shift. First, time-to-peak methods for quantifying dyssynchrony utilize only a single time point on the velocity or strain curves and should be replaced with more quantitatively sophisticated methods. Utilizing more data reduces variability and increases accuracy. Second, “response to CRT” should no longer be considered synonymous with “presence of left ventricular dyssynchrony”. There is no method based on LV dyssynchrony measures that by itself will predict response to CRT with high accuracy.
“Time-to-peak” parameters were developed primarily to diagnose mechanical LV dyssynchrony in an attempt to predict response to CRT better than QRS duration. The basic steps of time-to-peak analysis are as follows:
Numerous time-to-peak parameters initially showed promise in predicting response to CRT in single-center studies. These parameters included the septal-to-posterior wall motion delay (SPWMD) from M-mode echocardiography8 and multiple parameters based on Doppler tissue imaging (DTI) including: septal-to-lateral delay (SLD)9, peak velocity difference (PVD)10, 11, and the standard deviation of times to peak velocity (TsSD)12.
While fairly quick and easy to perform, time-to-peak analysis is quantitatively simplistic in that it utilizes only 1 point on a Doppler myocardial velocity curve that can consist of as many as 200 points per heart cycle. Thus, it is not surprising that recent studies have shown time-to-peak analysis to be highly operator dependent and non-reproducible, even in normal, healthy controls13, 15. Peaks in the Doppler velocity curves are often difficult to locate and highly susceptible to small changes in sample location, which leads to poor reproducibility16.
Diagnosis of mechanical dyssynchrony frequently differs depending on which time-to-peak parameter is utilized. Lafitte et al evaluated six published time-to-peak dyssynchrony parameters in a group of 65 patients prior to CRT17. These parameters agreed on the diagnosis of dyssynchrony in only 51% of patients.
Only recently, time-to-peak dyssynchrony parameters have been measured in large groups of normal controls. In a study of 120 subjects including 40 normal controls, Miyazaki et al found that 40 to 68% of healthy controls had dyssynchrony according to established threshold values18. Ng et al reported a mean TsSD of 37ms in a group of 122 healthy controls19, which is above the established 34ms threshold for diagnosing dyssynchrony20. Thus, time-to-peak methodology shows poor reproducibility and poor accuracy in healthy control subjects.
In light of these limitations with time-to-peak parameters, it is not surprising that the Predictors of Response to CRT (PROSPECT) study recently reported that no echocardiographic dyssynchrony parameter could be recommended to improve patient selection for CRT beyond current guidelines13. Eight of the twelve dyssynchrony parameters that were measured in PROSPECT were time-to-peak parameters and the other four parameters were cardiac timing intervals based on a single study21. All parameters tested had an area under the receiver-operating characteristic curve of 0.62 or less for predicting clinical or volume response to CRT. In addition, inter-observer coefficients of variation were unacceptably high (32 – 72%) with kappa coefficients showing poor inter-observer agreement (0.15 – 0.35).
Similarly, the Resynchronization Therapy in Narrow QRS (RETHINQ) trial recently reported that patients with narrow QRS and evidence of mechanical dyssynchrony do not benefit from CRT14. The RETHINQ trial utilized SPWMD and a version of the peak velocity difference (PVD) to diagnose mechanical dyssynchrony22. To our knowledge, at the inception of the trial, SPWMD and PVD had never been measured in a large group of healthy controls and the inter-center reproducibility of these parameters had not been quantified. It is now clear from recent studies that these time-to-peak parameters show dyssynchrony in approximately half of normal controls and have poor reproducibility15, 18. Thus, in retrospect, the criteria used to define mechanical dyssynchrony in the RETHINQ trial were sub-optimal, and may have contributed to the negative results.
In the wake of the negative results of the PROSPECT and RETHINQ trials, new parameters for diagnosing ventricular mechanical dyssynchrony have been described. Two such studies are published in this issue of JASE23, 24. Olsen et al developed a new method that utilizes the entire systolic portion of the DTI velocity curve (not just peak values) to measure dyssynchrony with a cross-correlation function24. The investigators compared the ability of their new cross-correlation of acceleration (XCA) method to the time-to-peak parameters SLD and TsSD to 1) discriminate between healthy controls and responders to CRT (their “positive control” group with assumed dyssynchrony) and 2) predict response to CRT. XCA discriminated between healthy controls and responders to CRT with a significantly higher area under the receiver-operating characteristic curve (AUC) than SLD and TsSD (0.95 vs 0.59 and 0.75, respectively). XCA also showed a significant ability to predict response to CRT (AUC = 0.66), but this was not significantly different, statistically, from the AUC for SLD (0.55) and TsSD (0.58).
Olsen et al measured XCA from acceleration curves that were derived from Doppler tissue velocity curves by taking a derivative24. Acceleration is theoretically more closely related to myocardial contraction than velocity, but it is important to note that differentiation amplifies noise25. The authors report that cross-correlation analysis of systolic velocity curves26 showed no difference from XCA in the ability to predict response to CRT (AUC = 0.69 vs 0.66, respectively). Thus, differentiation to acceleration may not be required, and the salient point is that cross-correlation analysis of velocity or acceleration curves limited to systole are superior to time-to-peak parameters in diagnosing LV mechanical dyssynchrony.
Also published in this issue of JASE is a study by Bertola et al in which the authors calculated the temporal uniformity of strain (TUS) index from speckle tracking radial strain data23. Similar to XCA, TUS also utilizes all points on the strain curves (not just peak values) and was originally developed for strain curves derived from tagged MRI27. The authors compared TUS to the standard deviation of times to peak strain, which was recently shown to be superior to velocity-based TsSD in predicting response to CRT28. Bertola et al found that 1) TUS showed better reproducibility than TsSDstrain 2) TUS predicted response to CRT with an AUC of 0.65 while TsSDstrain did not (AUC = 0.54) and 3) TUS was the only variable in a multivariate model which predicted post-CRT improvement in ejection fraction.
Bertola et al concluded that there were no limitations in applying TUS to CRT patients23. However, it may be prudent to note several words of caution. First, TUS was not very reproducible based on the Bland-Altman plots shown in Figure 4: the 95% limits of agreement are nearly as large as the mean values of TUS. Second, the authors reported a low CRT responder rate of 37%, which is significantly lower than the response rates of 60–70% typically seen in large CRT trials. This may have been due to their more stringent definition of response: a reduction in LV end-systolic volume of ≥15% and an increase in ejection fraction ≥25%. To our knowledge, no study has previously used these criteria to define response. Third, the authors concluded that “TUS yields greater quantitative results and CRT benefits than other asynchrony measurements”. TUS was only compared to one other dyssynchrony parameter (TsSDstrain). Thus, generalizing their conclusions to “other asynchrony measurements” is not supported by the data presented and requires future studies.
Methods to quantify dyssynchrony based on newer imaging modalities have been developed recently. Speckle tracking is an image processing technique for B-mode echocardiography which overcomes the angle dependence of traditional Doppler-based techniques and has therefore been used to assess both radial and longitudinal strain in the LV29. Several studies have used speckle tracking to demonstrate mechanical dyssynchrony and predict response to CRT30, 31. However, most speckle tracking dyssynchrony studies once again have utilized time-to-peak analysis instead of more sophisticated quantitative techniques like the TUS analysis used by Bertola et al23. In addition, speckle tracking dyssynchrony parameters are not as reproducibile as DTI dyssynchrony parameters19. Real-time three-dimensional echocardiography has also been used to predict response to CRT32, 33. Once again, these studies utilized time-to-peak analyses, and the study by Takeuchi et al reported high inter-observer variability from 26 – 56%32. Similarly, velocity-encoded MRI has been used to quantify dyssynchrony by utilizing the same time-to-peak parameters that were developed for DTI34.
In our opinion, simply applying time-to-peak methodology to other imaging modalities such as speckle tracking and three-dimensional echocardiography will not solve the problems that lead to the negative results of PROSPECT and RETHINQ. New techniques to quantify dyssynchrony need to be developed and these should:
While the above ideal description of a dyssynchrony parameter seems obvious, no parameter has yet been proven to have all these attributes. However, in addition to XCA24 and TUS23, several new methods – which utilize more than just peak values – show promise, as discussed below.
Initial publications using tissue Doppler to quantify LV mechanical dyssynchrony reported sensitivities and specificities as high as 100% for predicting response to CRT35. Thus, the presence of LV mechanical dyssynchrony was equated with “response to CRT” in the literature. Publications in the field considered only the timing of contraction and ignored the magnitude of the contraction. Resynchronizing a segment of myocardium that is completely infarcted (but still shows dyssynchrony) will unlikely improve heart failure. In his seminal paper from 1967, Herman et al quantitatively described patterns of myocardial contraction that included both akinesis (not contracting), asynchrony (contracting out of phase) and dyskinesis (paradoxical outward motion with no true contraction)36. Resynchronization therapy is targeted at asynchrony and will unlikely benefit akinetic or dyskinetic segments that do not contain an adequate substrate of viable myocardium. Thus, it is important to quantify the magnitude of contraction in segments targeted for resynchronization in addition to their timing, as evidenced by recent papers suggesting that the presence and location of scar tissue significantly affects response to CRT37–39.
Several recent papers have attempted to address both the magnitude and timing of contraction to predict response to CRT. Carasso et al recently predicted response to CRT based on a qualitative categorization of “mechanical strain patterns of LV function”40. The authors used speckle-derived longitudinal strain curves to categorize myocardial regions as one of six patterns: normal, mildly reduced shortening, severely reduced shortening, severe holosystolic stretching, dyssynchrony, or pseudo-dyssynchrony. The authors reported that the absence of passive segments with the holosystolic stretch pattern combined with the presence of lateral wall dyssynchrony was predictive of response to CRT with a sensitivity of 100% and specificity of 94%. Traditional echocardiographic parameters of mechanical dyssynchrony including SLD and TsSD failed to predict response to CRT. While these results are encouraging and, importantly, address both magnitude and timing of contraction, the classification system is qualitative and will likely be difficult to implement clinically.
Kirn et al hypothesized that mechanical discoordination (opposite strain within the LV wall) predicts reverse LV remodeling after CRT better than mechanical dyssynchrony41. The authors developed a novel metric, called internal stretch fraction, to quantify the relative lengthening versus shortening that was occurring throughout the cardiac cycle. In a preliminary study of 21 patients, the authors concluded that internal stretch fraction, but not mechanical dyssynchrony, could differentiate responders from non-responders.
As Kirn et al mention in their discussion41, internal stretch fraction is conceptually related to internal flow fraction, which has been measured by conductance catheter and shown to improve dramatically following CRT42. The concept of internal flow was first described in 1969: “With alternating zones of contraction and expansion, blood will be sequestered in the ventricle rather than ejected through the aortic valve”43. Internal flow fraction measures the overall inefficiency of the heart by comparing the amount of blood that is effectively ejected versus the amount that is shifted between dyssynchronous walls (thus wasting energy) in the LV. The ability to quantify internal flow fraction non-invasively with MRI has recently been reported44, and its ability to predict response to CRT remains to be studied.
Mechanical dyssynchrony, while likely a necessary requirement to respond to CRT, is just one of several factors which need to be included in a comprehensive assessment prior to CRT implantation. Thus, it is not surprising that the Bertola and Olsen papers in this issue of JASE were only able to predict response to CRT with an AUC of 0.65 – 0.6923, 24. We believe that the maximum AUC for predicting response to CRT with mechanical dyssynchrony assessment alone is likely around 0.60 – 0.70. The remaining predictive ability required for ideal accuracy (AUC of 1.0) is likely accounted for by the overall scar burden38, 39, the location of the scar in relation to the LV lead position37, the region of latest activation in relation to the LV lead position45 and co-morbidities such as diabetes, renal failure and atrial fibrillation46. Imaging of the coronary veins will likely be useful to determine potential LV lead positions prior to CRT to ensure placement in viable, dyssynchronous myocardium47, 48. Future studies should work towards addressing all of these factors in a comprehensive manner to predict response to CRT.
Time-to-peak analysis to quantify left ventricular mechanical dyssynchrony is easy and quick. However, we believe that the negative results of the PROSPECT and RETHINQ studies, along with several new studies on reproducibility and accuracy in normal controls, show that these parameters should no longer be used to diagnose LV dyssynchrony. Simply applying the same time-to-peak methodology to different imaging modalities such as speckle tracking and three-dimensional echocardiography will not solve the problems inherent in using time-to-peak methodology.
Left ventricular dyssynchrony assessment needs to be refined so it can be a part of evaluation for CRT – response to CRT cannot be entirely accounted for by the presence of dyssynchrony. The challenge for future research in the field of ventricular dyssynchrony is to develop methods that utilize the immense amount of quantitative data available to help guide clinical decisions, and several methods, including the papers by Olsen and Bertola in this issue of JASE23, 24, show promise in this regard.
1All authors have read and approved the manuscript.
2 The authors have one conflict of interest to disclose:Brandon Fornwalt and John Oshinski have applied for a patent on using cross-correlation methodology to quantify dyssynchrony. This technique is discussed in the manuscript.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.