Mortality and morbidity in cardiac resynchronization patients: impact of lead position, paced left ventricular QRS morphology and other characteristics on long-term outcome

Abstract

Aims

To investigate the effect of implantation-related characteristics, especially lead position and left ventricular (LV)-paced QRS morphology, on long-term mortality and morbidity in cardiac resynchronization therapy (CRT) patients.

Methods and results

The study retrospectively analysed 362 consecutive patients who underwent CRT device implantation over a 6 year period. Pre-implantation, LV-only paced, and biventricularly paced 12-lead electrocardiograms were obtained. Left ventricular and right ventricular (RV) lead positions were determined using biplane fluoroscopy and roentgenograms. The Kaplan–Meier method was used to estimate the survival function for all-cause death/hospitalization and cardiovascular death/hospitalization. Univariate and multivariate Cox proportional hazards models were also applied. The mean follow-up time was 24.7 ± 16.9 months. There were 79 deaths (62 cardiovascular) and 99 unplanned hospitalizations (72 cardiovascular). One year and 2 year all-cause mortality rates were 8.5 and 18.0%, respectively. Electrocardiographic and fluoroscopic descriptors of the LV lead position were found to be predictors of mortality/morbidity (as were functional class, heart failure aetiology, hyponatremia, and chronic atrial fibrillation). In particular, the antero-apical pattern of LV-only paced QRS showed a hazard ratio (HR) of 1.8 in univariate and 1.7 in multivariate analysis for predicting all-cause death/hospitalization (P = 0.006). The apical/paraseptal LV lead position showed an HR of 2.1 in univariate and 1.9 in multivariate analysis for predicting cardiovascular death/hospitalization (P = 0.018).

Conclusion

To achieve better long-term outcomes in CRT patients the antero-apical pattern of LV QRS complexes and apical or paraseptal LV lead position should be avoided.

Introduction

Cardiac resynchronization therapy (CRT) is known to improve symptoms and decrease mortality in selected heart failure patients.^1–6 However, data concerning the effect of potentially important operator-dependent factors on mortality and morbidity are either limited [right ventricular (RV) lead position],^7–11 conflicting [left ventricular (LV) lead position],^12–22 or absent (LV-only paced QRS morphology).

Our goal was to investigate the effect of pre-implantation and implantation-related characteristics on long-term mortality and morbidity in an unselected cohort of CRT patients treated in a single centre. In particular, the study assessed the effect of potentially modifiable factors such as LV-paced QRS morphology, LV lead position, RV lead position, and RV–LV lead distance.

Methods

Inclusion/exclusion criteria

This study involved 377 consecutive patients who underwent CRT device implantation in our institution from January 2006 to January 2012. Patients with a failed LV lead implantation and patients whose status could not be determined as of January 2012 were excluded from further analyses.

The CRT system was implanted in a standard manner.²³ The RV lead implantation site was at the physician's discretion, with most physicians favouring non-apical positions (e.g. the RV outflow tract or the interventricular septum). The LV lead was prioritized to be on the LV-free wall, and where possible anterior or posterior paraseptal positions in the left anterior oblique (LAO) view were avoided. In addition, most physicians avoided placement at very distal parts of a cardiac vein, as that results in an anterior position on the lateral chest roentgenogram, and hence a small RV–LV lead separation.²³

Clinical and implantation-related data

Pre-implantation and implantation-related data for all patients were collected retrospectively from hospital files, and these included medical history, heart failure aetiology, cardiac medications, biochemical data, echocardiographic study data, electrocardiograms (ECGs), chest-roentgenograms and the stage of heart failure according to the New York Heart Association (NYHA) functional classification system. The estimated glomerular filtration rate (eGFR) was calculated according to the abbreviated MDRD Study equation.²⁴ We routinely registered continuous ECGs throughout the whole CRT device implantation procedure, and stored them in an electrophysiology system (LabSystem Pro, BARD, Lowell, MA, USA) database. This enabled precise computer-based assessment of pre-implantation QRS duration, QRS duration during biventricular pacing (BiV), delta QRS (pre-implantation-QRS duration minus BiV-QRS duration), and LV-paced QRS morphology (the zone of transition in the pre-cordial leads). These measurements were made using a simultaneous 12-lead recording with a sweep speed of 100 mm/s, an amplitude augmentation of eight-fold, and digital calipers. Categorization of LV-only paced QRS complexes was based on observations by our group and other data from the literature,^25–27 that LV-paced QRS patterns related to apical, anterior, or posterior-paraseptal LV lead positions result in QRS patterns that are distinct from QRS patterns resulting from pacing at other LV-free wall sites (Figure 1). Antero-apical QRS morphology was considered to be present when LV-paced QRS complexes were negative at V1 and/or V2. As shown by us elsewhere, this pattern has very good sensitivity and specificity for predicting unfavourable LV lead position.²⁷ Antero-apical QRS morphologies included three patterns (i) early abrupt polarity change (from positive QRS in V1 to negative in V2–V6), (ii) negative QRS concordance, and (iii) left bundle branch block (LBBB)-like QRS morphology (negative V1 and positive V6). Pre-implantation QRS morphology was categorized as LBBB or non-LBBB, where the latter included RV-paced QRS, right bundle branch block, non-specific interventricular conduction delay and narrow QRS.

Fluoroscopic LV lead position classification was based on the analysis of two post-implantation views, in the left and right oblique projections (30–40°). The short-axis circumferential position was classified as lateral or septal/paraseptal. The lateral position comprised a broad spectrum of lateral positions including leads deployed from the 1 o'clock to 5 o'clock positions on LAO. The paraseptal position included anterior positions (from 12 o'clock to 1 o'clock) and posterior positions (from 6 o'clock to 5 o'clock). Using a right anterior oblique (RAO) view, the heart silhouette was divided into three equal parts from the apex to the base. The base of the heart was approximated by the course of the lead in the coronary sinus. The RAO view was used to classify the lead position as apical or non-apical. The RAO and LAO data were combined in order to categorize the LV lead position as apicoparaseptal [which was considered non-optimal (i.e. apical on RAO or anterior on LAO or posterior on LAO)], or non-apicoparaseptal (i.e. all other positions) (Figure 2).

Lateral chest roentgenograms were used to assess the horizontal LV–RV lead tip distance in the sagittal axis. Postero-anterior roentgenograms were used to categorize RV lead positions as apical or non-apical (RV outflow tract or interventricular septum).

All lead positions were analysed by a cardiologist with extensive experience in CRT, and who was ‘blinded’ from the clinical and outcome data.

Outcome data

Data concerning life status, mode of death, and unplanned hospitalizations were obtained by telephone contact, office visits, and hospital chart review. Deaths and hospitalizations were classified as cardiac or non-cardiac. The primary study endpoint was a composite of all-cause mortality and all-cause hospitalizations (as in the COMPANION study), and the secondary endpoint was a composite of cardiovascular mortality and cardiovascular hospitalization.

Statistical analysis

Continuous variables are presented as mean and standard deviations or as median and quartiles, as appropriate. Categorical variables are presented as number and percentage. The effect of individual variables on survival was investigated using univariate Cox proportional hazards models. Variables that showed a statistically significant effect on survival in univariate analyses or were believed to be clinically important (e.g. age, gender, LV ejection fraction, LV end-diastolic dimension and eGFR) were entered into multivariate Cox proportional hazards models. Cox model findings are presented as hazard ratios (HRs), tests of significance and 95% confidence intervals. There were no significant violations of the proportionality assumption that underlies the Cox proportional hazards method. The Kaplan–Meier method was used to estimate the survival function for the primary endpoint. Statistical analysis was performed using STATISTICA 10.0 and R (a language and environment for statistical computing). P values of <0.05 were considered to indicate significance.

Results

The device implantation procedure was successful in 364 of the 377 patients (96% success rate). Two implanted patients were permanently lost to follow-up and were excluded from the study. Therefore, 362 patients were included for further analysis. Patient clinical characteristics are presented in Tables 1 and 2. The mean follow-up time was 24.7 ± 16.9 months. Of the 79 deaths, 62 were classified as cardiovascular. There were 99 unplanned hospitalizations, and 72 of those were classified as cardiovascular. The 1 year and 2 year all-cause mortality rates were 8.5 and 18.0%, respectively.

We investigated for predictors of mortality and morbidity. A number of previously known predictors were confirmed, namely functional class, heart failure aetiology, hyponatremia, and chronic atrial fibrillation. In addition, we found that electrocardiographic and fluoroscopic descriptors of the LV lead position were strong predictors (Table 3). In particular, the apical/paraseptal LV lead position showed a HR of 2.1 in univariate analysis and an HR of 1.9 in multivariate analysis for predicting cardiovascular death or cardiovascular hospitalization (P = 0.018). Similarly, LV-only paced QRS morphology typical for activation of the LV from the periapical/anterior regions had an HR of 1.8 in univariate and 1.7 in multivariate analysis for predicting all-cause death and all-cause hospitalization (P = 0.006). A pre-implantation ECG showing LBBB morphology was associated with a better outcome.

Details of the univariate and multivariate analysis for the primary and secondary endpoints are presented in Tables 3 and 4, respectively. The Kaplan–Meier curves for the primary endpoint illustrating the impact of pre-implantation QRS morphology, LV-lead position, LV-paced QRS morphology and QRS narrowing post-implantation are presented in Figure 3.

Discussion

The major finding of this study of CRT patients was that the placement of an LV lead into the apical and/or paraseptal regions was associated with an increased risk of death or hospitalization in patients with advanced heart failure. Moreover, this increased risk was predicted not only from fluoroscopic data but also from LV-paced QRS morphology findings.

Fluoroscopic left ventricular lead position

Several studies have shown that apical or paraseptal LV lead positions are related to a poor patient outcome.^{12,14,16,17,19–22} Therefore, in this study, we combined those two suboptimal LV lead positions in a group termed as the ‘apicoparaseptal’ LV lead position. This LV lead position is avoided by some implanters on pathophysiological grounds (LV activation during LBBB), and the present long-term outcome data support such a policy. When the LV lead position is assessed only in the longitudinal axis, the results might be confounded by an uneven distribution of paraseptal positions. In addition, the converse may also be true in which circumferential position assessment may be confounded by an uneven distribution of apical positions. Therefore, we believe that potentially detrimental lead positions according to longitudinal and circumferential assessment should be grouped together.

Although two previous large studies (MADIT-CRT and REVERSE) showed that an apical LV lead position was linked to a poor outcome,^19,20 both of those studies involved patients with mild heart failure (NYHA class I–II). Importantly, this study showed that this association was present in patients with advanced heart failure. Paraseptal/septal LV lead positions result from lead implantation into the posterior or anterior interventricular vein. Pacing from such sites results in LV activation not dissimilar to that achieved by pacing the interventricular septum from the side of the right ventricle—as indicated by similar LBBB morphology of the paced QRS complexes.²⁵ Several studies have shown that paraseptal lead positions are related to worse clinical, echocardiographic and/or hemodynamic outcomes.^{12,14,16,21,22} This study corroborate those findings with mortality and morbidity data.

In contrast to the present findings and the works cited above, three important studies have concluded that LV lead position has no effect on clinical outcome.^13,15,18 We believe the discrepancies between those and the current findings may reflect differences in study methodologies. In the study based on an analysis of the COMPANION trial dataset,¹⁸ and also for the majority of patients in the study by Kronborg et al.,¹⁵ the apical LV lead position was mostly determined based on the postero-anterior view. However, as shown by Rickard et al.,²⁸ lead positions that appear to be apical using such a view are more often non-apical. This occurs due to the oblique position of the heart within the chest cavity, which results in a posterior mid-ventricular LV lead position often projecting itself as apical.

The study by Foley et al.¹³ excluded patients with leads implanted in the anterior or posterior interventricular vein. Hence, that study essentially compared various lateral sites among themselves. That study design might also explain the lack of a difference in outcomes.

It may be that apical and paraseptal positions are harmful only when they are truly apical and septal/paraseptal. It is unlikely that there exists a definitive line demarcating ‘good’ and ‘bad’ positions; the transition between those positions is likely to be a continuum. We believe that the anterior and posterior positions should not be extended too far to avoid classifying ‘good’ positions as paraseptal. However, such extensions have been a feature of a number of studies. In the study based on an analysis of the COMPANION trial data, potentially ‘bad’ anterior and posterior positions spanned 50% of the circumference in LAO, unavoidably including many ‘good’ lateral leads.¹⁸

Left ventricular-paced QRS morphology

Endocardial pacing close to the apex or a ventricular tachycardia with an anterior/apical exit site produces a negative concordant pattern (i.e. QRS complexes negative in leads V1–V6). In contrast, a ventricular tachycardia with a postero-basal exit produces a positive concordant pattern (i.e. QRS complexes positive in leads V1–V6), while a lateral, mid-segmental exit results in right bundle branch block (RBBB) morphology with a QRS transition from positive to negative somewhere between leads V2–V6. In our experience, similar patterns characterize different epicardial LV-pacing sites (see Figure 1).²⁷ In particular, periapical sites result in either a very early transition from a positive to a negative QRS complex, or in a negative concordance; such patterns are almost never seen when basal to mid-ventricular postero-lateral segments are paced. Only occasionally are negative QRS complexes in V1–V2 seen with a non-apical lead, albeit mainly in anterior paraseptal positions. Therefore, we have named this pattern ‘antero-apical’. Very good correlation between LV-only paced QRS morphology and LV lead position was shown by us elsewhere.²⁷ This study shows for the first time that the antero-apical QRS pattern seen during LV pacing has a similar prognostic value as the LV lead position assessed using biplane fluoroscopy. However, in contrast to fluoroscopy, ECG is easily available for assessment during follow-up of CRT patients, and therefore it can be very useful for predicting poor outcome/unfavourable LV lead position. Moreover, it can be useful for double checking of the LV lead position already during implantation.

Other electrocardiogram parameters

We also found that QRS duration shortening was a significant ECG parameter. This concurs with previous studies showing greater QRS narrowing in CRT responders.^29–31

This study also examined the effect of two other pre-implantation QRS-based parameters that were found by others to impact on long-term prognosis or response to CRT, namely, the presence of LBBB morphology and the presence of a QRS duration >150 ms.^32,33 Our univariate analysis, but not the multivariate analysis, showed a better prognosis for LBBB patients (HR = 0.62). In multivariate analysis, a pre-implantation QRS duration >150 ms and pre-implantation LBBB morphology approached but did not reach significance (HR = 0.69, P = 0.09, and HR = 0.69 and P = 0.07, respectively) in terms of association with a lower rate of cardiovascular deaths/cardiovascular hospitalizations.

Right ventricular lead position

Previous major CRT studies did not provide data regarding the RV lead position. However, the presumption is that the position was apical in most if not all cases. Some smaller studies have compared the effect of apical vs. non-apical RV lead position on outcomes in CRT patients (n = 79–112)^7–11, as did one larger study (n = 345)²⁰. However, those studies involved only a moderate follow-up time (6–12 months), and mostly used ‘soft’ outcome endpoints (mainly echocardiographic data).

Of studies comparing apical and non-apical RV lead positions in CRT patients, this study has the largest number of subjects and has the longest follow-up. Moreover, apart from the REVERSE trial, this study is the only one to use mortality and morbidity as endpoints. Like the REVERSE study, we found that RV lead position had no effect on those outcomes, indicating that a priori selected RV lead position in CRT does not impact on outcomes.

General clinical predictors of poor prognosis in heart failure

The rate of all-cause mortality/morbidity in this study (18%) was very similar to that observed in the CARE-HF study (16%) and in the single-centre registry based study by van Bommel et al. (17.1%).^3,21 Several risk factors identified in this study were also identified by others including the presence of chronic atrial fibrillation, hyponatremia, ischaemic aetiology, lower functional class, diabetes mellitus (borderline significance) and lower LV ejection fraction (borderline significance). Of those factors, atrial fibrillation deserves special attention as it was once considered a contraindication for CRT, and many major CRT trials did not include atrial fibrillation patients. In our study, the risk related to atrial fibrillation (HR = 1.5) was similar to that reported by van Bommel et al. (HR = 1.78). However, a recent meta-analysis showed that atrial fibrillation patients do benefit from CRT despite being at a higher risk.³⁴ Perhaps the higher risk of poor outcomes in atrial fibrillation patients should not be interpreted as a lower relative benefit from CRT but rather as the presence of more advanced cardiac disease in those patients.

Conclusion

This study suggests that in order to achieve better long-term outcomes in CRT patients apical or paraseptal LV lead position and the antero-apical pattern of LV-paced QRS complexes should be avoided.

Conflict of interest: none declared.

References