Pre- and Intra-procedural Predictors of Reverse Remodeling After CRT
Pre- and Intra-procedural Predictors of Reverse Remodeling After CRT
Studying 60 patients undergoing CRT according to standard guidelines, this study demonstrates that (i) MRI-defined scar extent and intra-LV mechanical dyssynchrony are independent predictors of RR 6 months after CRT; (ii) electrical dyssynchrony as assessed by measuring QRS duration on surface ECG is less predictive than mechanical dyssynchrony, mainly because of scar-related moderate QRS enlargements that do not translate into significant mechanical dyssynchrony in the viable myocardium; (iii) the presence of scar at pacing site, even when nontransmural, is associated to a lesser RR after CRT, whereas the mechanical delay at this site does not seem to influence the response.
In this study, significant RR at 6 months was found in 70% of the patients, which is superior to the response rate reported in prior studies. This is likely due to the inclusion of patients with less severe symptoms of heart failure than in reference studies (mean NYHA class 2.77 ± 0.53 vs over 3.2 in most studies). However, this clinical status is in the range of values reported in standard practice. The high response rate might also be due to the assessment of RR at 6 months after CRT, versus 3 months in most studies. Our results show that intra-LV mechanical dyssynchrony and myocardial scar extent are independent predictors of RR after CRT. The use of mechanical dyssynchrony to improve the selection of CRT candidates has already been proposed, mostly with the use of echocardiographic indexes. Results of the PROSPECT study have outlined the limitations of the echocardiographic method in terms of acquisition and quantification reproducibility. In this study, intra-LV mechanical dyssynchrony was quantified from cine MRI images acquired at high TR, and defined as the maximal delay between first peaks of radial wall motion over 20 segments. The use of time-to-peak radial motion has already been reported for dyssynchrony quantification on cine MRI data. In a past study, the standard deviation of time-to-peak radial displacement over 16 segments was shown predictive of echocardiographic response to CRT. In this work we used a higher TR (17 milliseconds), and quantified dyssynchrony from time-to-first peak instead of time-to-maximal peak of radial wall motion, in order to be able to depict early septal mechanical activations. A similar strategy has already been applied on circumferential strain data in a tagged MRI study. In this study, this enabled the quantification of the mechanical delay at pacing site as a percentage of LV total activation, using the maximal delay between first displacement peaks as a surrogate for LV total mechanical activation. In addition, we used a 20-segment model instead of the standard 16-segment one because we believed a higher segmentation was mandatory to characterize the LV pacing site, which was part of our study objectives. Our results show that intra-LV mechanical dyssynchrony is independently associated to RR after CRT, which is consistent with prior studies using echocardiography, SPECT, or MRI. Mechanical dyssynchrony was more accurate for the identification of nonresponders (Sp = 0.75), than for the identification of responders (Se = 0.67). This result is also in line with previous reports. However, a significant number of patients remain nonresponders despite the presence of both mechanical and electrical dyssynchrony. This can be explained either by the spontaneous progression of the underlying disease, or by the presence of additional factors altering the response. One of these factors, myocardial scar, is accessible to MRI with the use of delayed-enhanced imaging. Our results show that RR after CRT is inversely correlated to myocardial scar extent, and that myocardial scar extent predicts CRT response independently from other characteristics, including mechanical dyssynchrony. The lower response rate in ischemic patients is well known, and an association between scar extent and CRT response has also been reported in prior MRI studies. However, it is to our knowledge the first report of an independent association between RR, mechanical dyssynchrony, and myocardial scar extent. It is interesting to note that unlike mechanical dyssynchrony, myocardial scar extent shows a higher accuracy for the detection of responders (Se = 0.79) than nonresponders (Sp = 0.64). These results might have an impact on the use of preprocedural imaging to improve the selection of CRT candidates. The assessment of mechanical dyssynchrony might be well suited, because specific, to lower the nonresponse rate in the population referred to CRT according to current guidelines. The assessment of myocardial scar extent might be well suited, because sensitive, to the identification of potential responders beyond the current indication criteria, such as patients with moderate heart failure.
All 60 patients included in this study showed significant electrical dyssynchrony, as defined by measuring QRS duration on the surface ECG. However, 17/60 patients did not show significant intra-LV mechanical dyssynchrony. A discrepancy between electrical and mechanical dyssynchrony has already been reported. Our results show that patients with electrical but not mechanical dyssynchrony are less likely to respond to CRT. In these patients, myocardial scar extent was found higher, which might suggest that the discrepancy between electrical and mechanical dyssynchrony is related to alterations of the electro-mechanical coupling in scarred areas. This discrepancy might be one of the causes responsible for the lesser response to CRT in ischemic patients. Indeed, the viable myocardium is the only accessible to resynchronization, and some ischemic patients might be misclassified as dyssynchronous because of scar-related QRS enlargements that do not translate into mechanical dyssynchrony, because the electrically delayed myocardium is nonviable. However, our results also show that patients with electrical but not mechanical dyssynchrony have lower QRS duration. This indicates that this cause of nonresponse is mostly encountered in patients with intermediate QRS enlargements. Intermediate QRS duration of LBBB morphology is a class IIa indication for CRT, according to recent guidelines. Our results indicate that this population is a primary target for the use of MRI to select CRT candidates.
Previous reports have shown that the response is highly dependent on LV pacing location, both in ischemic and nonischemic patients. Additionally, the optimal site is known to be highly variable between patients. Therefore, the characterization of segmental predictors at imaging could be used to personalize LV pacing site and optimize response. Our results show that presence of scar at pacing site is detrimental to CRT response. This finding is consistent with a prior study reporting a lower response when LV pacing is performed in an area of transmural scar. However, in this study the presence of scar at pacing site was defined as an average scar transmurality >25% on the segment, and in our population no pacing site was located in an area of transmural scar. This result supports the integration of MRI-derived scar maps to guide the implantation of the LV lead. Indeed, if segments exhibiting transmural scar can be identified on lead thresholds, our results suggest that subendocardial scar also alters the response, and these areas might have no impact on pacing thresholds.
The second pacing site characteristic evaluated in this study was the mechanical delay of activation. Prior studies have shown that patterns of electrical activation are highly variable between patients. When considering mechanical activation, this variability is even higher because of electromechanical coupling heterogeneities, particularly when an ischemic substrate is present. Thus, mechanical activation is a patient-specific characteristic that might be used to personalize the optimal site for LV pacing. To this day most studies have used the most delayed site as a target, with contradictory results. In this study, the predictive value of mechanical delay at pacing site was assessed from MRI data. To study its correlation with RR, this delay was expressed quantitatively as a% of LV total activation. The mean delay at LV pacing site was 90.6 ± 12% of LV total activation, and 65% of the patients were paced within the last 10% of LV activation. These results are consistent with prior reports, and indicate that in the absence of pacing site optimization, the usual strategy aiming at the most lateral stable site most often result in pacing a mechanically delayed area. In this study, no relationship was found between mechanical delay at pacing site and RR after CRT. This finding is consistent with results from several previous animal and clinical studies, and suggests that the most delayed site of LV activation might not be the appropriate characteristic to define the optimal pacing site. This does not imply that the patient-specific sequence of activation does not influence response to CRT, but the choice of the optimal site might have to integrate a number of potentially confounding factors: location of conduction block, heterogeneities of electro-mechanical coupling and contractility, biventricular interactions. Therefore, an appropriate use of activation data for treatment personalization will likely benefit from integrative strategies and multiparametric modeling approaches.
The first limitation of this study is related to the registration between pacing site and MRI characteristics. In order to improve the accuracy of pacing site characterization, we used a 20-segment model. This higher segmentation might be more susceptible to mis-registration. In order to optimize the registration, the pacing site was identified on MDCT data reformatted in short axis slices in every patients in whom MDCT had been performed during the 6 months following CRT (N = 38). In other patients (N = 22), the pacing site was identified on a 2 incidence chest x-rays, and we acknowledge that registration errors might have occurred in this population. However, our results showed good agreement between pacing segment identification on chest x-rays and MDCT in 38 patients (data available in Appendix). Another limitation is related to the use of displacement rather than strain data to quantify mechanical dyssynchrony and mechanical delays. However, displacement data derived from high TR retrospectively triggered cine SSFP imaging enabled the depiction of early septal mechanical activation, and in this study the maximal delay between first peaks of radial displacement was used to estimate total LV activation and quantify segmental activation times as a percentage of this total activation. Tagged MRI is less accurate for the assessment of mechanical events occurring immediately before the R wave peak (because of tag fading), and immediately after the R wave peak (when tagging is being applied). Thus, the tagging method is in our experience less suitable for the quantification of early septal mechanical activation, particularly in patients with wide QRS complexes. Another limitation is related to the definition of response on echocardiographic parameters only. The intraobserver variability of ESV measurements might have impacted the assessment of RR. Additionally, because a discrepancy between echocardiographic and clinical responses has been reported, the predictive value of MRI metrics remains to be confirmed on long-term clinical outcome. Finally, the main limitation of this study is related to the methodology used to define the optimal pacing site characteristics. In this study, only 1 site per patient was tested. The impact of pacing site characteristics on response is therefore highly dependent on patient baseline characteristics, and our results show that these baseline characteristics are major response predictors. A more robust strategy for the characterization of optimal pacing site characteristics would be to compare different sites, but then only the acute response would be studied. An accurate definition of optimal pacing site characteristics would ultimately require a randomized controlled study comparing different implantation strategies with respect to long-term outcome.
Discussion
Studying 60 patients undergoing CRT according to standard guidelines, this study demonstrates that (i) MRI-defined scar extent and intra-LV mechanical dyssynchrony are independent predictors of RR 6 months after CRT; (ii) electrical dyssynchrony as assessed by measuring QRS duration on surface ECG is less predictive than mechanical dyssynchrony, mainly because of scar-related moderate QRS enlargements that do not translate into significant mechanical dyssynchrony in the viable myocardium; (iii) the presence of scar at pacing site, even when nontransmural, is associated to a lesser RR after CRT, whereas the mechanical delay at this site does not seem to influence the response.
Preprocedural Predictors of Reverse Remodeling
In this study, significant RR at 6 months was found in 70% of the patients, which is superior to the response rate reported in prior studies. This is likely due to the inclusion of patients with less severe symptoms of heart failure than in reference studies (mean NYHA class 2.77 ± 0.53 vs over 3.2 in most studies). However, this clinical status is in the range of values reported in standard practice. The high response rate might also be due to the assessment of RR at 6 months after CRT, versus 3 months in most studies. Our results show that intra-LV mechanical dyssynchrony and myocardial scar extent are independent predictors of RR after CRT. The use of mechanical dyssynchrony to improve the selection of CRT candidates has already been proposed, mostly with the use of echocardiographic indexes. Results of the PROSPECT study have outlined the limitations of the echocardiographic method in terms of acquisition and quantification reproducibility. In this study, intra-LV mechanical dyssynchrony was quantified from cine MRI images acquired at high TR, and defined as the maximal delay between first peaks of radial wall motion over 20 segments. The use of time-to-peak radial motion has already been reported for dyssynchrony quantification on cine MRI data. In a past study, the standard deviation of time-to-peak radial displacement over 16 segments was shown predictive of echocardiographic response to CRT. In this work we used a higher TR (17 milliseconds), and quantified dyssynchrony from time-to-first peak instead of time-to-maximal peak of radial wall motion, in order to be able to depict early septal mechanical activations. A similar strategy has already been applied on circumferential strain data in a tagged MRI study. In this study, this enabled the quantification of the mechanical delay at pacing site as a percentage of LV total activation, using the maximal delay between first displacement peaks as a surrogate for LV total mechanical activation. In addition, we used a 20-segment model instead of the standard 16-segment one because we believed a higher segmentation was mandatory to characterize the LV pacing site, which was part of our study objectives. Our results show that intra-LV mechanical dyssynchrony is independently associated to RR after CRT, which is consistent with prior studies using echocardiography, SPECT, or MRI. Mechanical dyssynchrony was more accurate for the identification of nonresponders (Sp = 0.75), than for the identification of responders (Se = 0.67). This result is also in line with previous reports. However, a significant number of patients remain nonresponders despite the presence of both mechanical and electrical dyssynchrony. This can be explained either by the spontaneous progression of the underlying disease, or by the presence of additional factors altering the response. One of these factors, myocardial scar, is accessible to MRI with the use of delayed-enhanced imaging. Our results show that RR after CRT is inversely correlated to myocardial scar extent, and that myocardial scar extent predicts CRT response independently from other characteristics, including mechanical dyssynchrony. The lower response rate in ischemic patients is well known, and an association between scar extent and CRT response has also been reported in prior MRI studies. However, it is to our knowledge the first report of an independent association between RR, mechanical dyssynchrony, and myocardial scar extent. It is interesting to note that unlike mechanical dyssynchrony, myocardial scar extent shows a higher accuracy for the detection of responders (Se = 0.79) than nonresponders (Sp = 0.64). These results might have an impact on the use of preprocedural imaging to improve the selection of CRT candidates. The assessment of mechanical dyssynchrony might be well suited, because specific, to lower the nonresponse rate in the population referred to CRT according to current guidelines. The assessment of myocardial scar extent might be well suited, because sensitive, to the identification of potential responders beyond the current indication criteria, such as patients with moderate heart failure.
Discrepancy Between Electrical and Mechanical Dyssynchony
All 60 patients included in this study showed significant electrical dyssynchrony, as defined by measuring QRS duration on the surface ECG. However, 17/60 patients did not show significant intra-LV mechanical dyssynchrony. A discrepancy between electrical and mechanical dyssynchrony has already been reported. Our results show that patients with electrical but not mechanical dyssynchrony are less likely to respond to CRT. In these patients, myocardial scar extent was found higher, which might suggest that the discrepancy between electrical and mechanical dyssynchrony is related to alterations of the electro-mechanical coupling in scarred areas. This discrepancy might be one of the causes responsible for the lesser response to CRT in ischemic patients. Indeed, the viable myocardium is the only accessible to resynchronization, and some ischemic patients might be misclassified as dyssynchronous because of scar-related QRS enlargements that do not translate into mechanical dyssynchrony, because the electrically delayed myocardium is nonviable. However, our results also show that patients with electrical but not mechanical dyssynchrony have lower QRS duration. This indicates that this cause of nonresponse is mostly encountered in patients with intermediate QRS enlargements. Intermediate QRS duration of LBBB morphology is a class IIa indication for CRT, according to recent guidelines. Our results indicate that this population is a primary target for the use of MRI to select CRT candidates.
Intraprocedural Predictors of Reverse Remodeling
Previous reports have shown that the response is highly dependent on LV pacing location, both in ischemic and nonischemic patients. Additionally, the optimal site is known to be highly variable between patients. Therefore, the characterization of segmental predictors at imaging could be used to personalize LV pacing site and optimize response. Our results show that presence of scar at pacing site is detrimental to CRT response. This finding is consistent with a prior study reporting a lower response when LV pacing is performed in an area of transmural scar. However, in this study the presence of scar at pacing site was defined as an average scar transmurality >25% on the segment, and in our population no pacing site was located in an area of transmural scar. This result supports the integration of MRI-derived scar maps to guide the implantation of the LV lead. Indeed, if segments exhibiting transmural scar can be identified on lead thresholds, our results suggest that subendocardial scar also alters the response, and these areas might have no impact on pacing thresholds.
The second pacing site characteristic evaluated in this study was the mechanical delay of activation. Prior studies have shown that patterns of electrical activation are highly variable between patients. When considering mechanical activation, this variability is even higher because of electromechanical coupling heterogeneities, particularly when an ischemic substrate is present. Thus, mechanical activation is a patient-specific characteristic that might be used to personalize the optimal site for LV pacing. To this day most studies have used the most delayed site as a target, with contradictory results. In this study, the predictive value of mechanical delay at pacing site was assessed from MRI data. To study its correlation with RR, this delay was expressed quantitatively as a% of LV total activation. The mean delay at LV pacing site was 90.6 ± 12% of LV total activation, and 65% of the patients were paced within the last 10% of LV activation. These results are consistent with prior reports, and indicate that in the absence of pacing site optimization, the usual strategy aiming at the most lateral stable site most often result in pacing a mechanically delayed area. In this study, no relationship was found between mechanical delay at pacing site and RR after CRT. This finding is consistent with results from several previous animal and clinical studies, and suggests that the most delayed site of LV activation might not be the appropriate characteristic to define the optimal pacing site. This does not imply that the patient-specific sequence of activation does not influence response to CRT, but the choice of the optimal site might have to integrate a number of potentially confounding factors: location of conduction block, heterogeneities of electro-mechanical coupling and contractility, biventricular interactions. Therefore, an appropriate use of activation data for treatment personalization will likely benefit from integrative strategies and multiparametric modeling approaches.
Study Limitations
The first limitation of this study is related to the registration between pacing site and MRI characteristics. In order to improve the accuracy of pacing site characterization, we used a 20-segment model. This higher segmentation might be more susceptible to mis-registration. In order to optimize the registration, the pacing site was identified on MDCT data reformatted in short axis slices in every patients in whom MDCT had been performed during the 6 months following CRT (N = 38). In other patients (N = 22), the pacing site was identified on a 2 incidence chest x-rays, and we acknowledge that registration errors might have occurred in this population. However, our results showed good agreement between pacing segment identification on chest x-rays and MDCT in 38 patients (data available in Appendix). Another limitation is related to the use of displacement rather than strain data to quantify mechanical dyssynchrony and mechanical delays. However, displacement data derived from high TR retrospectively triggered cine SSFP imaging enabled the depiction of early septal mechanical activation, and in this study the maximal delay between first peaks of radial displacement was used to estimate total LV activation and quantify segmental activation times as a percentage of this total activation. Tagged MRI is less accurate for the assessment of mechanical events occurring immediately before the R wave peak (because of tag fading), and immediately after the R wave peak (when tagging is being applied). Thus, the tagging method is in our experience less suitable for the quantification of early septal mechanical activation, particularly in patients with wide QRS complexes. Another limitation is related to the definition of response on echocardiographic parameters only. The intraobserver variability of ESV measurements might have impacted the assessment of RR. Additionally, because a discrepancy between echocardiographic and clinical responses has been reported, the predictive value of MRI metrics remains to be confirmed on long-term clinical outcome. Finally, the main limitation of this study is related to the methodology used to define the optimal pacing site characteristics. In this study, only 1 site per patient was tested. The impact of pacing site characteristics on response is therefore highly dependent on patient baseline characteristics, and our results show that these baseline characteristics are major response predictors. A more robust strategy for the characterization of optimal pacing site characteristics would be to compare different sites, but then only the acute response would be studied. An accurate definition of optimal pacing site characteristics would ultimately require a randomized controlled study comparing different implantation strategies with respect to long-term outcome.