The paced electrocardiogram cannot be used to identify left and right ventricular pacing sites in cardiac resynchronization therapy: validation by cardiac computed tomography

AIMS
Paced electrocardiogram characteristics to confirm left ventricular (LV) and right ventricular (RV) pacing sites in cardiac resynchronization therapy (CRT) have not been validated with accurate knowledge of pacing lead positions. We aimed to evaluate the ability of the paced QRS morphology to differentiate between various LV and RV lead positions using cardiac computed tomography (CT) as the reference for LV and RV pacing site.


METHODS AND RESULTS
Ninety-seven CRT patients were included. The QRS morphology was evaluated during forced LV-only and RV-only pacing. Pacing lead positions were assessed in a standard LV 16-segment model and a simplistic RV 6-segment model using cardiac CT. Ten patients with LV lead displacement or a LV pacing site outside the non-apical free wall were excluded from the analysis of the LV paced QRS complex. Pacing within the LV free wall was associated with a superior and a right-axis deviation (P = 0.02 and 0.04, respectively). Pacing from basal LV segments mainly produced a late (V5 or later) precordial QRS transition as compared with mid-LV pacing (P = 0.001). No significant associations were found between RV pacing site and QRS axis or precordial transition. Different QRS morphologies were observed during single-chamber pacing from identical LV or RV myocardial segments.


CONCLUSION
Weak associations exist between LV and RV pacing sites and the paced QRS axis. None of the paced QRS characteristics can be used to reliably confirm specific LV and RV pacing sites in CRT patients.


Introduction
Cardiac resynchronization therapy (CRT) is an established treatment in selected heart failure patients exhibiting depressed left ventricular (LV) function, an electrocardiogram (ECG) with prolonged QRS duration, and who remain symptomatic despite optimal medical therapy. 1 However, using the current selection criteria, a significant proportion of patients do not benefit from CRT. 2,3 The LV pacing site is an important determinant of response to CRT, whereas the impact of the right ventricular (RV) lead position on clinical outcome remains controversial although a non-apical RV pacing site has been shown to increase the risk of ventricular tachyarrhythmias. 4 -10 Procedural biplane fluoroscopy or chest radiography is routinely used to evaluate pacing lead position in CRT. 11 However, the accuracy of these imaging modalities to determine the exact LV and RV lead positions is modest when compared with cardiac computed tomography (CT). 12,13 It has been suggested that the QRS morphology of the paced ECG may be helpful to confirm LV and RV lead positions both intra-and post-operatively. 14 -17 However, the paced ECG patterns for different LV and RV pacing sites have never been evaluated applying an imaging modality capable of displaying the exact pacing lead position in three dimensions (3D). The aim of the current study was to evaluate simple characteristics of the paced QRS complex to differentiate between various LV and RV lead positions in CRT using cardiac CT as the reference for LV and RV pacing sites.

Methods Patients
Between April 2011 and December 2012, we included 97 patients with chronic heart failure receiving a CRT pacemaker or a CRT-implantable cardioverter defibrillator at the Department of Cardiology, Aarhus University Hospital, Skejby. All patients were in New York Heart Association functional class II -IV, had LV ejection fraction (EF) ,35%, and a preimplant ECG with QRS duration .120 ms and left bundle branch block. Patients with a recent myocardial infarction (,3 months), those who were pregnant or lactating, and those with severe renal dysfunction (estimated glomerular filtration rate ,30 mL/min) were excluded.
Transvenous CRT implantation using commercially available leads and devices was performed in all patients. Only active fixation RV leads were used. The LV leads were bipolar or quadripolar programmed in a bipolar configuration. No lead revisions were performed between the implant procedure and cardiac CT.
All patients were enrolled in an on-going randomized study evaluating the clinical impact of imaging-guided LV lead placement in CRT. Cardiac CT was performed according to the specific study protocol. 18 Echocardiographic LV EF, end-diastolic, and end-systolic volumes were assessed prior to CRT implant and at 6-month follow-up using Simpson's biplane method (Vivid E9, GE Medical Systems).
All patients gave informed written consent before the implant procedure. The Central Denmark Regional Committee on health research ethics and the Danish Data Protection Agency approved the study.

Electrocardiogram acquisition and analysis
Before the CRT implant procedure, a standard supine 12-lead ECG (50 mm/s, 10 mm/mV) was recorded during intrinsic rhythm. After final lead positioning, a forced LV-only and RV-only pacing ECG was acquired (both V00-mode, 90 b.p.m., and an output ,3.5 V at 0.5 ms). A dedicated LV lead bipolar pacing configuration was used without crosschamber pacing to avoid RV anodal capture. All ECGs were stored for offline analysis. Two reviewers without knowledge of the patients' clinical characteristics and pacing lead positions evaluated the paced QRS morphology. In case of disagreement, the reviewers determined the final QRS morphology by consensus. Commercially available software (CardioLab IT, GE Healthcare) was used for ECG analysis. The following parameters were evaluated during LV-only and RV-only pacing: (1) QRS duration measured simultaneously in all 12 leads as global intervals from the earliest onset to latest offset of the QRS complex. (2) QRS-axis deviation classified into right-or left-axis deviation according to net QRS area in lead I and into a superior-or inferior-axis deviation according to net QRS area in lead aVF. (3) Notching of the RV paced QRS complex in inferior ECG leads (II, III, or aVF). (4) Precordial QRS transition (the lead closest to V1 with the same net QRS amplitude before changing to predominantly a negative or positive QRS complex).
These paced QRS characteristics have previously been applied to confirm regional LV and RV lead positions. 14 -16,19 Subgroup analyses of the different QRS characteristics and pacing lead positions were performed for patients with non-ischaemic and ischaemic heart failure (medical records documenting a history of revascularization or myocardial infarction). 20

Cardiac computed tomography protocol and analysis
Cardiac CT was performed 6 months after CRT implantation using a second-generation dual-source CT scanner (Siemens Somatom Definition Flash, Siemens Healthcare). Computed tomography scanner settings and protocol were described previously. 18 During breath hold, a contrast-enhanced [70 mL (Optirayw 350 mg/mL, Covidien)] helical retrospective ECG-gated scan timed according to contrast filling of both the RV and LV cavity was performed. The median (25th; 75th percentile) estimated radiation dose was 4.9 (3.8; 6.8) mSv. Pacing sites were determined using axial, multiplanar, and 3D volume-rendered reconstructions. Images were analysed using commercially available software (Syngo.via, Siemens Healthcare). Excellent intra-and interobserver agreements for the assessment of LV and RV lead positions by cardiac CT have previously been reported. 12

Assessment of left ventricular pacing site
The LV pacing site was determined using the standardized LV segmentation dividing the LV myocardium into 16 segments. 21 The LV long axis was divided into equal thirds: basal, mid-LV, and apical. The LV short axis was divided into opposing segments: anterior, anterolateral, inferolateral, inferior, septal, and anteroseptal ( Figure 1).
We excluded 10 patients from the analysis of the LV paced QRS complex. Five patients exhibited macroscopically visible LV lead displacement between the procedural fluoroscopy and follow-up cardiac CT. As previously described, this was determined by comparing the procedural fluoroscopy and a volume-rendered 3D cardiac CT reconstruction in the same right anterior oblique projection. 12 Four of these patients showed LV lead movement within the coronary sinus tributary and small LV lead threshold changes (0.4 V at 1 ms; 0.1 V at 0.5 ms; 0.3 V at 0.4 ms; and 0.1 V at 0.4 ms, respectively). One patient demonstrated LV lead displacement to the CS without LV lead capture at follow-up. They were excluded to compare procedural acquired paced ECGs with CT images only in patients without visible LV lead movement between the procedure and the follow-up cardiac CT. Furthermore, a small minority of five patients had a LV pacing site outside the non-apical free wall: two patients had an apical LV pacing site producing a right inferior and a left inferior QRS axis, two patients had an anterior LV pacing site creating a What's new? † The left ventricular (LV) pacing site is an important determinant of response to cardiac resynchronization therapy (CRT), whereas the impact of the right ventricular (RV) lead position on clinical outcome remains controversial. † Procedural biplane fluoroscopy or chest radiography is routinely used to determine LV and RV lead positions, but the accuracy of these imaging modalities is modest. † We evaluated the ability of the paced QRS morphology to differentiate between cardiac CT verified LV and RV pacing sites. † Different QRS-axis deviations and precordial transition patterns were observed during single-chamber pacing from identical cardiac CT verified LV or RV myocardial segments. † No paced QRS characteristic can reliably confirm specific LV and RV pacing sites in CRT.
LV paced right inferior and a left inferior QRS axis, and one patient had an inferior LV pacing site producing a right superior-axis deviation, respectively. They were excluded from further analysis to evaluate only the paced ECG in patients with a pacing site within the non-apical anterolateral or inferolateral LV free wall. Achieving a LV pacing site in selected myocardial segments within this region is often considered optimal for CRT benefit. 4,5,7 Assessment of right ventricular lead position Evaluation of the RV lead tip position was performed using a simplistic segmentation dividing the RV long-axis cavity into equal thirds (basal, mid, and apical region). Subsequently, the RV lead position was evaluated in the short axis according to a six-segment model [two basal segments (basal septum and free wall), two mid-RV segments (mid septum and free wall), and two apical segments (apical septum including RV apex and apical free wall)] (Figure 1).

Statistics
Normally distributed continuous variables are presented as mean + SD and were compared using Student's t-test. Continuous variables not normally distributed were compared using a Wilcoxon's rank-sum test.
Proportions were compared by a two-sample test of proportions. A Pearson's x 2 or Fisher's exact test wherever appropriate was used to compare categorical variables. Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated to assess the ability of the paced QRS axis to identify lead positions between two categories of binary variables. Values of P , 0.05 were considered statistically significant. Commercially available software (Stata version 12, StataCorp) was used for statistical analysis.

Study population
Clinical characteristics of the 97 patients are presented in Table 1. All patients received optimal medical therapy at maximum tolerated dosages. A significant increase in LV EF and reduction in LV volumes were observed at 6-month follow-up (Table 1).

Left ventricular paced QRS duration and axis
Similar LV paced QRS duration was found for patients with an anterolateral or inferolateral pacing site (197 + 23 and 198 + 21 ms; P ¼ 0.80) and with a basal or mid-LV lead position (196 + 21 and 201 + 22 ms; P ¼ 0.29), respectively. The distribution of the paced QRS-axis deviations according to LV pacing site is illustrated in Table 2. Pacing from a LV non-apical free wall position produced a superior axis in 51 (59%) patients and an inferior axis in 36 (41%) patients (P ¼ 0.02) while right-and left-axis deviations were observed in 50 (57%) and 37 (43%) patients (P ¼ 0.04), respectively ( Figure 2). An inferolateral LV pacing site was associated with a superior and a right-axis deviation (both P ¼ 0.01), respectively. No significant associations were found between a basal  or mid-LV pacing site and superior-to-inferior or right-to left-axis deviations.
The ability of different QRS-axis deviations to distinguish between regional LV lead positions is illustrated in Table 4.

Right ventricular paced QRS duration and axis
The RV paced QRS duration was similar for patients with a septal or free wall (184 + 22 vs. 191 + 19 ms; P ¼ 0.16) and with a mid-RV or apical RV lead position (183 + 19 vs. 190 + 22 ms; P ¼ 0.11), respectively.
The distribution of RV lead positions according to paced QRS-axis deviation is shown in Table 3. No significant associations existed between RV pacing site and QRS-axis deviation (Figure 3). The diagnostic performance of RV paced QRS-axis deviations to distinguish between regional RV lead positions is presented in Table 4.
No other associations between LV and RV pacing sites and paced QRS axis or precordial transition pattern were found in patients with non-ischaemic or ischaemic heart failure. Irrespective of the heart failure aetiology, no associations between RV pacing site and inferior QRS notching were observed.

Discussion
In the present study, we evaluated simple paced QRS characteristics applicable in the routine clinical setting for confirming LV and RV pacing sites in CRT using cardiac CT as a reference for exact lead position. We found weak associations between the paced QRS axis and pacing sites and none of the paced QRS characteristics could be used to reliably confirm specific LV and RV pacing sites.
Along with device interrogation, the paced ECG has been applied to ensure adequate device function and provide an understanding of lead positions in CRT patients. 16 The LV pacing site is an important determinant of response to CRT. 4 -7 Several studies have demonstrated an improved clinical outcome with LV pacing according to a pre-implant selected optimal LV myocardial segment; frequently within the non-apical anterolateral and inferolateral free wall. 4,5,7 The optimal RV pacing site remains controversial. 8 -10 The paced QRS is readily available during the implantation and could be valuable if reflecting specific pacing sites. Several ECG markers of the paced QRS complex to confirm LV and RV pacing sites in CRT have been proposed using fluoroscopy, electroanatomical mapping, or chest X-ray as reference for LV and RV lead positions.
Commonly, a LV paced QRS axis pointing towards the right inferior or right superior quadrant and a net positive QRS in V1-V3 is reported to confirm pacing from the LV anterolateral and inferolateral free wall. 14,16,19 Furthermore, a basal pacing site has been associated with a later precordial QRS transition than   apical locations. 17 In agreement with these criteria, we found the majority of patients with a free wall LV pacing site having a superior and a right QRS-axis deviation and a net positive QRS in the precordial leads V1-V3, with basal pacing sites demonstrating the latest precordial QRS transition. However, a large percentage of anterolateral LV pacing sites produced a left-axis deviation, mainly in patients with ischaemic heart failure. A leftward axis during free wall LV pacing has previously been reported but the reasons for this unusual axis deviation remain unclear. 19 Varying the epicardial LV pacing output has been shown to influence the LV electrical activation sequence possibly by RV anodal capture or by capturing a larger myocardial area. 22 Using dedicated LV bipolar pacing, we avoided RV anodal stimulation. However, extending LV myocardial capture beyond or away from local conduction blocks and thereby altering QRS morphology may possibly explain a left-axis deviation during free wall LV pacing. The proposed RV paced QRS criteria to confirm an apical RV lead location is the presence of a superior axis and mainly a net negative QRS in V1 while a septal RV pacing site produces a right inferior axis and a negative QRS in V1. 15,16,19 Furthermore, RV free wall pacing has been reported to produce a left-axis deviation, QRS notching in the inferior leads, and a longer QRS duration than septal RV pacing in patients without heart failure. 15,16 Concordantly, we recorded a net negative QRS in V1 in the majority of patients during RV pacing from all analysed RV regions. In contrast to previous reports, we found no significant associations between regional RV lead position and paced QRS-axis deviation, precordial QRS transition, inferior QRS notching, or QRS duration in the total study population. This lack of agreement may likely be explained by different imaging methods applied for determining lead positions. We used cardiac CT. Previous studies applied fluoroscopy or electroanatomical mapping as reference for the RV pacing site. 15,16,19 Fluoroscopy has been demonstrated to be inaccurate and poorly reproducible for localizing RV pacing site as compared with cardiac CT. 12 In addition, electroanatomical mapping can only approximate cardiac anatomy and not illustrate detailed cardiac morphology as displayed by cardiac CT.
Despite several associations between pacing site and paced QRS characteristics, the present study demonstrates different QRS-axis deviations during single-chamber pacing from identical LV or RV myocardial segments. Accordingly, the diagnostic performance of the paced ECG to confirm specific LV and RV lead positions is low. This may be explained by a varying degree of cardiac dilatation, rotation, and positioning in the thoracic cavity in heart failure patients altering the direction of electrical activation. Also, in patients with ischaemic cardiomyopathy, scar areas may change electrical activation patterns. Hence similar cardiac anatomical pacing sites may produce different paced QRS morphologies.

Limitations
We acknowledge the inherent limitations of a single-centre study with a moderate sample size. Nevertheless, this is the first study evaluating the diagnostic value of the paced QRS complex characteristics to distinguish between regional LV and RV pacing sites using cardiac CT as the reference for exact pacing lead position.
We applied a 6-month follow-up cardiac CT as a reference for pacing lead positions and LV remodelling was observed during follow-up. However, the anatomical relationship between the LV myocardium and the coronary sinus tributary containing the LV lead was presumably unaltered despite cardiac remodelling. We excluded patients with visible LV lead displacement from the analysis and only active fixation RV leads were used.
Our analysis of the LV paced QRS complex was limited to patients with a non-apical free wall LV lead position and the results are not applicable for other LV lead positions.
Introduction of cardiac CT into the diagnostic algorithm in CRT patients will increase their cumulative radiation exposure and should not be used routinely outside protocolled studies. In clinical practice, cardiac CT may be useful to confirm exact LV lead position and assess the presence and location of additional cardiac veins before considering lead revision. Several approaches to minimize radiation dose are applied in this study, including the use of iterative reconstruction algorithms, individual settings of tube voltage and  current, and tube current modulation with a narrow full current window, respectively. 23

Conclusion
Different QRS morphologies exist during single-chamber LV or RV pacing from identical cardiac CT verified LV or RV myocardial segments. The paced QRS characteristics are not valid for confirming specific LV and RV pacing sites in CRT patients.