Author + information
- Received November 11, 2016
- Revision received February 10, 2017
- Accepted February 16, 2017
- Published online August 21, 2017.
- Niraj Varma, MA, MD, PhDa,∗ (, )
- Jason Lappe, MDa,
- Jiayan He, ScDb,
- Mark Niebauer, MD, PhDa,
- Mahesh Manne, MDa and
- Patrick Tchou, MDa
- aHeart and Vascular Institute, Cleveland Clinic, Cleveland, Ohio
- bDepartment of Quantitative Health Sciences, Cleveland Clinic, Cleveland, Ohio
- ↵∗Address for correspondence:
Dr. Niraj Varma, Cardiac Electrophysiology, J2-2, Cleveland Clinic, 9500 Euclid Avenue, Cleveland, Ohio 44195.
Objectives In this study, the authors sought to assess the impact of body and heart size on sex-specific cardiac resynchronization therapy (CRT) response rate, according to QRS duration (QRSd) as a continuum.
Background Effects of CRT differ between sexes for any given QRSd.
Methods New York Heart Association functional class III/IV patients with nonischemic cardiomyopathy and “true” left bundle branch block (LBBB) were evaluated. Left ventricular mass (LVM) and end-diastolic volume were measured echocardiographically. Positive response was defined by left ventricular ejection fraction (LVEF) improvement post-CRT.
Results Among 130 patients (LVEF 19 ± 7.1%; QRSd 165 ± 20 ms; 55% female), CRT improved LVEF to 32 ± 14% (p < 0.001) during a median 2 years follow-up. Positive responses occurred in 103 of 130 (79%) (78% when QRSd <150 ms vs. 80% when QRSd ≥150 ms; p = 0.8). Body surface area (BSA), QRSd, and LVM were lower in women, but QRSd/LVM ratio greater (p < 0.0001). Sexes did not differ for pharmacotherapy and comorbidities, but female CRT response was greater: 90% (65 of 72) versus 66% (38 of 58) in males (p < 0.001). With QRSd as a continuum, the overall CRT–response relationship showed a progressive increase to plateau between 150 and 170 ms, then a decrease. Sex-specific differences were conspicuous: among females, a peak effect was observed between 135 and 150 ms, thereafter a decline, with the male response rate lower, but with a gradual increase as QRSd lengthened. Sex-specific differences were unaltered by BSA, but resolved with integration of LVM or end-diastolic volume.
Conclusions Sex differences in the QRSd–response relationship among CRT patients with LBBB were unexplained by application of strict LBBB criteria or by BSA, but resolved by QRSd normalization for heart size using LV mass or volume.
Cardiac resynchronization therapy (CRT) is an important heart failure therapy imparting survival benefit (1). However, effects range from individuals normalizing left ventricular (LV) function to others who do not respond (or worsen) (2). Intriguingly, there may be sex differences (3,4). Reasons may include frequency of underlying nonischemic pathology and/or incidence of “true” left bundle branch block (LBBB) (i.e., QRS notching/slurring in leads I, AVL, V5, or V6 ). QRS duration (QRSd) itself may have different implications in women versus men. For example, women with LBBB and QRSd <150 ms tend to benefit, whereas males in this category do not (6–8). When QRSd is treated as a continuous function, sex-specific differences are even more pronounced. This has prompted calls for different guideline recommendations for males and females. The reasons why any single QRSd in LBBB predicates different probabilities of response between men and women are unclear. Sex differences in prevalence of ischemic pathology and “false LBBB,” and effects of body and/or heart size have been proposed as underlying factors, but not examined.
The premise of CRT is that pre-exciting late-activated LV regions, due to LV depolarization delays caused by LBBB, restores synchronized LV contraction. QRSd indexes extent of conduction delay: wider QRSd specifies an electrical substrate more likely to be responsive to CRT (9,10). However, this concept does not account for modulation of the QRSd by LV size, itself affected by body size and/or remodeling in heart failure (typically producing LV dilatation) (11–13). Thus, in the presence of LBBB, prolongation of QRSd may occur from reduced myocardial conduction velocity and/or extension of the travel distance (conduction “path length”) of the wavefront across the LV. Reversal of delayed LV activation due to LBBB is the aim of CRT, but on the other hand, LV dilation limits CRT response (14). This raises the interesting notion that, because women generally have smaller hearts (15), any single QRSd may represent greater retardation of myocardial conduction compared with men, and account for better CRT response.
In this study, we hypothesized that integrating surrogate measures of increased path length (i.e., body surface area (BSA), left ventricular mass (LVM) or dilation) into the QRSd value in heart failure patients with true LBBB would explain sex-specific differences in the CRT response. We tested this by measuring the probability of occurrence of structural remodeling by QRSd normalized for BSA or LVM and size, assessed as a continuous function.
This retrospective institutional review board–approved study comprised patients with only nonischemic cardiomyopathy (NICM) and true LBBB (confirmed by 2 electrophysiologists, N.V., P.T.) undergoing CRT following guideline-directed medical therapy (5,6) (Figure 1). Clinical characteristics and laboratory values were obtained from the electronic medical record.
Patients included had undergone echocardiographic studies within 3 months before the procedure and at least 3 months post-procedure, with technical quality permitting measurement of LVM and volumes. Parameters were measured by echocardiographers blinded to baseline patient characteristics and CRT outcomes. LVM was calculated by the traditional linear method using measurements of the septum, ventricular diameter, and posterior wall obtained from a parasternal long-axis image (0.8 [1.05 (LVDD + IVS + PW)3 − (LVDD)3] + 0.8 g, where LVDD = left ventricular diastolic dimension; IVS = interventricular septal thickness; and PW = posterior wall thickness). These calculations assume that the ventricle is a prolate ellipse of revolution. However, diseased hearts are known to not have this shape. Therefore, we also determined LVM using biplane imaging with measurements of the endocardial and epicardial borders, application of the modified Simpson’s rule, and correction for myocardial density (16,17). We observed a strong correlation of results between the linear and biplane imaging methods (R = 0.90). Biplane measurements were used in our analysis because in a separate test sample (n = 19) of NICM patients with LBBB undergoing magnetic resonance imaging, absolute LVM values correlated more closely with the biplane method (R = 0.84), aligning with prior reports (18).
Response was defined as an increase in left ventricular ejection fraction (LVEF) occurring from baseline to follow-up echocardiogram. Firstly, we assessed the incidence of positive CRT effect overall, and contrasted baseline characteristics (including BSA, LVM, and QRSd normalized for each of these) and CRT response rates of responders to non-responders and between sexes. The relationship of QRSd to the probability of chronic response was then examined across the continuum of the QRSd range. The influence of BSA, LVM, and left ventricular end-diastolic volume (LVEDV) were tested separately by normalizing each to QRSd, and then contrasted between sexes. We tested the discriminatory power of QRS dichotomization by 150 ms and of cut points of QRSd/LVM ratio for predicting CRT effect.
Continuous variables are presented as mean ± SD, and compared using the Wilcoxon rank sum test. Categorical variables are presented as numbers with percentages, and compared using chi-square test or Fisher exact test, as appropriate. Overall and sex-specific changes in echocardiographic parameters from baseline to post-CRT were tested using a paired Student t test.
Multivariable logistic regression was performed to identify factors other than QRSd that were associated with CRT response. The factors considered in the analysis included age, sex, New York Heart Association (NYHA) functional class (I to IV), body surface area, comorbidities (chronic obstructive pulmonary disease, hypertension, history of smoking, atrial fibrillation), diabetes mellitus, creatinine, glomerular filtration rate, hemoglobin, hyperlipidemia), medication (beta-blocker, angiotensin-converting enzyme inhibitor or angiotensin receptor blocker, diuretic agents, hydralazine, aldosterone antagonist), echocardiographic measures (LVEF, mitral regurgitation), and date of operation (years since January 1, 2000) and interval between implant and follow-up echocardiogram. Variable selection was performed using the bagging method (19–21). In brief, 1,000 bootstrap datasets with the same size as the original were generated by sampling with replacement. Multivariable logistic regression analysis was performed on each of these datasets using forward stepwise variable selection with a retention criterion of ≤0.05. Variables appearing as significant in 50% or more of the 1,000 models were incorporated into a multivariable model. To examine the effect of QRSd on CRT response, QRSd and indexed QRSd (QRSd/BSA, QRSd/LVM, and QRSd/LVEDV) were included in the multivariable logistic model adjusting for significant variables. To account for missing values of those variables in multivariable modeling, we performed 5-fold multiple imputation. Of 22 variables considered in the multivariable modeling, only 3 variables had missing data. Using the multivariable logistic models, nomograms were plotted to show the effect of QRSd and sex on the predicted probability of response. Natural log transformation of scale for QRSd was evaluated with respect to response in order to obtain the best model fit and examine the shape of the relationship of QRSd with the response outcome. Statistical analyses were performed using SAS statistical software (SAS version 9.4, SAS Institute, Cary, North Carolina) and R version 3.2.3 (R Foundation for Statistical Computing, Vienna, Austria).
Patient demographics (N = 130, female 55%) of the overall cohort, and separated by response to CRT, are shown in Table 1. Baseline characteristics were typical for CRT recipients in the studied epoch: QRSd 165 ± 20 ms; LVEF 19 ± 7.1%; 83% NYHA functional class III, 12% class II, and 3.6% class IV. Distribution was normal for QRSd and LVM (Figure 2). The Spearman correlation coefficient of QRSd with BSA was 0.34, with LVM 0.44, and with LVEDV 0.41. LVM correlated positively with LVEDV and negatively with LVEF (Spearman correlation coefficients 0.90 and −0.24, respectively) (Figure 3). Sexes were well balanced in proportions (female 55%), age, and comorbidities (Table 2). Although QRSd was shorter in women, the ratios of QRSd/BSA and QRSd/LVM ratio were 10% and 54% greater (because BSA was 14% and LVM 38% lower compared with men).
Post-CRT LVEF was 32 ± 14%, representing an increase of 13 ± 12% during a follow-up interval of 2.2 ± 1.3 years. CRT pacing was consistently >98% and did not differ between sexes (6). LV volumes decreased (LVEDV by 40 ± 63 ml, from 208 ± 100 ml to 161 ± 111 ml; p < 0.0001; left ventricular end-systolic volume [LVESV] by 45 ± 64 ml, from 172 ± 95 ml to 121 ± 108 ml; p < 0.0001). Positive response was observed in 103 of 130 (79%) individuals (LVEF increased ≥5% in 102 of 130 and by 1% to 4% in 1 patient). Among CRT responders, LVEF improved by 17 ± 11% and LV volumes diminished (LVESV decreased by 63 ± 55 ml and LVEDV by 55 ± 58 ml), contrasting with a decline in LVEF of 8.3 ± 12% and increased LV volumes (LVEDV by 17 ± 49 ml and LVESV by 24 ± 47 ml) in non-responders (p < 0.001). Baseline electrocardiogram (ECG) characteristics and LV function did not differentiate responders from non-responders (QRSd 163 ± 18 ms vs. 170 ± 25 ms; p = 0.09; LVEF 20 ± 7.3% vs. 18 ± 6.8%; p = 0.20, respectively). However, LV volumes and LVM were greater in non-responders. Responders had a higher proportion of females versus non-responders (90% vs. 9.7%; p = 0.005), lower incidence of atrial fibrillation (23% vs. 56%; p = 0.001), and 32% greater QRSd/LVM (but QRSd/BSA was similar). When QRSd was dichotomized by 150 ms, 78% of patients with <150 ms and 80% of those ≥150 ms responded to CRT (p = 0.80) in the whole cohort (Figure 4).
Females (Table 2), demonstrated greater incidence of response (90% [65 of 72] vs. males 66% [38 of 58]; p = 0.0005) and larger remodeling effects. LVEF in women increased from 21 ± 7.3% to 37 ± 13% and in men from 17 ± 6.6% to 25 ± 14% (p = 0.0001). LVEDV decreased by 51 ± 52 ml versus 25 ± 74 ml, and LVESV by 57 ± 52 ml versus 30 ± 75 ml, in females and males, respectively. This difference was seen despite shorter baseline QRSd in females (160 ± 16 vs. 170 ± 23; p = 0.002). Among patients with QRSd <150 ms (n = 27), 50% of men and 94% of women (p = 0.02), and with ≥150 ms (n = 103), 69% of men and 89% of women, sustained positive CRT effect CRT (p = 0.01).
Following multivariable analysis, the important factors besides QRSd associated with CRT response included female sex and lack of atrial fibrillation. Response was unaffected by time from implant to follow-up echocardiogram. The multivariable model contained all considered variables listed in the preceding text except QRSd. We then included QRSd and its different transformations in the model to test whether the effect of QRSd on response differed by sex. Results are depicted in Figure 5. Adjustment for atrial fibrillation did not affect curve profiles.
The relationship between response and QRSd exhibited a continuous, but nonlinear, relationship (Figure 5A). Overall probability of chronic CRT response progressively increased with increasing baseline QRSd to peak and plateau at QRSd 150 to 170 ms, followed by a decline. When plotted by sex, curve profiles were significantly different (p = 0.05) (Figure 5B). Women demonstrated high probability of response at QRSd = 135 to 150 ms (equivalent to males with 150 to 175 ms), but a pronounced decline at >170 ms. Men demonstrated lower probability at shorter QRSd with a lower peak and plateau. The widest sex difference was observed when QRSd was <150 ms.
Sex differences in curve profiles remained when QRSd was normalized for BSA (Figures 5C and 5D). However, use of the QRSd/LVM ratio resulted in significant changes. Overall, the predicted probability of response ascended progressively to peak and plateau with progressive increases in QRSd/LVM (Figure 5E). Significantly, this curve showed no terminal decline as QRSd/LVM became greater, in contrast to when QRSd alone was plotted as the dependent variable (Figure 5A). When plotted by sex, male and female trajectories coincided, that is, sex-specific differences were resolved (test for sex interaction p = 0.90) (Figure 5F). With a QRS/LVM cut point of 0.8, those patients below (n = 72) had 66.7% chance of response, contrasting with 94.8% in those above (n = 58; p < 0.0001). Curve profiles for QRSd normalized to LV volume were similar to QRSd/LVM (Figures 5G and 5H). Notably, males were concentrated in the ascending limb of these probability curves because male sex was associated with smaller QRSd/LVM (or QRSd/LVEDV) ratios (Figure 2).
Sex-specific differences in probability of CRT response for any given QRS duration in LBBB have been recently recognized (3,6–8). Differences in QRS configuration or heart size, related to body size and/or effects of structural remodeling from heart failure, have been advanced as possible reasons. We tested these and found that sex-specific differences were resolved only when normalizing QRSd to LV mass (or volume), on an individual basis (Figure 6).
Markers of electrical LV dyssynchrony on surface electrocardiography (i.e., QRS complex with LBBB configuration and duration >150 ms) in heart failure patients define Class 1 recommendations for CRT. This is a blunt selection tool, because some patients with <150 ms respond positively (especially among women) and others with >150 ms do not. Refinement of the QRS criteria to include only “true LBBB” (to exclude those with “false-positive LBBB,” i.e., QRS abnormalities due to effects of hypertrophy/increased LVM rather than conduction delay) has been proposed as a means to improve candidate selection (5). Our study population comprised only such patients, yet demonstrated a 21% nonresponse rate, with no difference between those with QRSd <150 ms or ≥150 ms, indicating that these enhanced criteria were insufficient (Figure 4). Although it is important to differentiate mass effect from LBBB—important because the former is not a recognized substrate for CRT—the surface ECG is known to be insensitive for this purpose (22). Here, we show a wide range of increased LVM among HF patients, all with strictly defined LBBB, that is, the ECG of “true LBBB” did not exclude patients with increased LVM effectively (Figure 3A).
Our study showed significant sex-specific differences in probability of CRT response by QRS duration as a continuous function, among candidates all with “true” LBBB (Figure 5B). Men accounted for a larger proportion of non-responders, and >90% women with QRSd <150 ms responded. Similar findings were reported in patient-level meta-analysis and from the National Cardiovascular Data Registry (7,8). Mechanisms were attributed to a possibly higher prevalence of NICM and true LBBB in female CRT recipients but could not be evaluated in those reports. Our study population controlled for these factors, yet show the same sex-specific differences. One cited explanation for a sex-specific effect is that women naturally have a narrower QRSd (5 to 10 ms), so any single QRSd value implies relatively greater degree of electrical dyssynchrony compared with men (5,23). This baseline QRSd difference was observed in our heart failure population. However, the dramatically different QRSd–response curves between sexes are not superimposable with a simple frameshift of 10 ms. Another postulate, that QRSd should be normalized for body size (normally less in women) and, by implication, heart size, in the same fashion as structural heart measures, did not affect sex-specific differences in our population (Figure 5D). However, increased heart size in heart failure due to remodeling may not be simply related to body size. Our data show that QRSd normalized for LVM (or LV volume) resolved sex-specific differences (Figures 5F, 5H, and 6).
The interaction of QRSd with heart size in CRT patients is important. Increased LVM increases the path length to prolong QRSd, independently of Purkinje fiber lesions (11–13). Hence any single QRSd in LBBB represents a balance of conduction impairment and path length, which may differ among individuals at any unitary value of QRSd. The distinction is important because CRT is designed for the former and be likely ineffective (or worse) in the latter as exemplified among patients with greater LV dilation and mechanical dyssynchrony but lower QRSd prolongation (14,24). In our CRT population with relatively balanced sex proportions, both QRSd and LVM exhibited normal distributions. Although, women had lower QRSd and lower LVM, QRSd/LVM ratios was greater, supporting the concept of a substrate with greater electrical dyssynchrony that may be rectified by CRT (Figure 2). Overall, patients with a high index score (relatively large QRS duration and small LV size) have better responses following CRT compared with those with a low index score (smaller QRS and large LV size) (Figures 5E and 5G). Interestingly, the direction of effect of LV size on CRT response we report is opposite to that presumed to facilitate response in pivotal early CRT trials (i.e., beneficial in patients with larger LV size). However, those did not account for the interplay between QRS duration and LV size. Our results may explain effects of CRT at extremes of the QRSd range. Thus, the puzzling dip noted in overall and female response rates when QRSd >175 ms (Figures 5A and 5B) (also noted by others [7,25]) resolved with QRSd/LVM (Figure 5F) or QRSd/LVEDV (Figure 5H), suggesting mass or volume effect dominates when QRSd is excessively prolonged in heart failure patients. At the short end of the QRSd spectrum, mortality increased in men selected for CRT with QRS <130 ms and LV dilation, but not in women with a similar range of QRSd and smaller ventricles (24).
Study strengths and limitations
Our study population consisted of patients with only NICM with strictly defined LBBB and normally distributed QRS duration, with well-balanced sex proportions and few comorbidities, explaining the relatively high overall rate of reverse remodeling. This idealized set of largely NYHA functional class III CRT recipients facilitated the aims of our study. We evaluated QRSd (and its derivatives) as a continuous function, which better describes its relationship to CRT effect than arbitrary dichotomization by 150 ms (6–8). QRSd curve profiles (overall and by sex) from our single center report match those from a patient-level meta-analysis of randomized trials, validating our sample population (7). Separating mechanisms of QRS prolongation into either increased path length (size and mass) or conduction slowing (from LBBB) is an approximate method. Heart failure remodeling is a complex process affecting cardiomyocytes and interstitium. Impulse propagation may be affected by increased LVM, but is also affected at the tissue, cellular and subcellular levels. Thus fibrosis, scar, and reduced intercellular coupling will affect conduction velocity as well as path length. Echocardiographic estimation of LVM demands adequate definition of endocardial and epicardial borders, and assumes a lack of geometrical distortions. The correlation of QRSd with LVM in our study was very similar to that reported using CMR (0.4) in patients with LBBB and LV dysfunction, validating our method (13). NICM, studied here, may represent a form of remodeling less vulnerable to distortions and errors associated with scar, and in which there is colinearity with LV dilation. Responders were defined by positive LVEF change, accounting for variable intervals between examinations. (The alternative definition of change in end-systolic volume ≥15% did not alter male–female response profiles (Online Appendix)). LVEF is a complex summary measure incorporating ventricular size, contractile function and afterload, and the extent of changes following CRT or pharmacotherapy in heart failure patients has been shown to correlate with gain in survival (26,27). We did not assess individual LV lead positions, but anatomic locations (e.g., lateral, or apical vs. non-apical) do not identify sites of best hemodynamic improvement nor predict future CRT-mediated echocardiographic remodeling (28,29).
Although our study explains sex-specific differences, normalization of QRSd for LVM or volume is still an incomplete overall selection criterion because those with QRSd/LVM <0.80 (predominantly men) still had >50% probability of response. Our conclusions cannot be extended to patients with ischemic heart disease or non-LBBB QRS morphologies.
Sex differences in QRSd–CRT response relationship were not explained by application of strict LBBB criteria nor by BSA, but resolved by QRSd normalization for heart size using LV mass or volume. Thus, QRS prolongation in heart failure patients with LBBB may represent the sum effects of mass and conduction block, the balance of which affects ultimate CRT outcome. Sex, as a modulator of disease and therapy, should be considered during patient selection for CRT (30).
COMPETENCY IN PATIENT CARE: Response to CRT according to QRS duration differs significantly between sexes. Women with QRS duration <150 ms have a high probability of CRT response. Sex differences are unexplained by application of strict LBBB criteria, nor by body surface area, but are accounted for by differences in heart size (women have smaller hearts).
TRANSLATIONAL OUTLOOK: Extent of left ventricular remodeling as well as QRS duration and morphology should be considered during patient selection for CRT. Understanding the influence of left ventricular size (mass and volume) on QRS duration and as a modulator of CRT effect, irrespective of sex, are important objectives for future investigations.
The authors have reported that they have no relationships relevant to the contents of this paper to disclose.
All authors attest they are in compliance with human studies committees and animal welfare regulations of the authors’ institutions and Food and Drug Administration guidelines, including patient consent where appropriate. For more information, visit the JACC: Clinical Electrophysiology author instructions page.
- Abbreviations and Acronyms
- body surface area
- cardiac resynchronization therapy
- left bundle branch block
- left ventricle/ventricular
- left ventricular end-diastolic volume
- left ventricular ejection fraction
- left ventricular end-systolic volume
- left ventricular mass
- nonischemic cardiomyopathy
- New York Heart Association
- QRS duration
- Received November 11, 2016.
- Revision received February 10, 2017.
- Accepted February 16, 2017.
- 2017 American College of Cardiology Foundation
- Costanzo M.R.
- Varma N.,
- Mittal S.,
- Prillinger J.B.,
- Snell J.,
- Dalal N.,
- Piccini J.P.
- Zusterzeel R.,
- Curtis J.P.,
- Canos D.A.,
- et al.
- Stewart R.A.,
- Young A.A.,
- Anderson C.,
- Teo K.K.,
- Jennings G.,
- Cowan B.R.
- Lin F.Y.,
- Devereux R.B.,
- Roman M.J.,
- et al.
- Rubin D.B.
- Bacharova L.,
- Ugander M.
- ↵Macfarlane P, Oosterom A, Pahlm O, Kligfield P, Janse M, Camm J, editors. Appendix 1: adult normal limits. In: Comprehensive Electrocardiology. London, UK: Springer-Verlag London 2011:2057–125.
- Steffel J.,
- Varma N.,
- Robertson M.,
- et al.
- Sundaram V.,
- Bilchick K.,
- Waldo A.L.,
- Wilson B.,
- Reddy Y.N.,
- Sahadevan J.
- Rickard J.,
- Cheng A.,
- Spragg D.,
- et al.
- Goldenberg I.,
- Moss A.J.,
- Hall W.J.,
- et al.
- Derval N.,
- Steendijk P.,
- Gula L.J.,
- et al.
- Legato M.J.,
- Johnson P.A.,
- Manson J.E.