Study protocol
All included patients underwent implantation of a CRT device and were tested for each of the 3 pacing configurations (MPP, SPP-FOI, MPP-FOI). The study protocol is summarized in Figure 1 . Written consent was obtained for each patient. The protocol was reviewed and approved by the local institutional research ethics committee (HCB/2018/0388).
a) CRT device implantation
All patients underwent CRT device implantation according to routine clinical practice. Each patient was implanted with a commercially available MPP-capable CRT system that included a quadripolar LV lead and a compatible CRT device. The indications for pacemakers or defibrillators were based on comorbidities and clinical factors limiting the quality of life or life expectancy. The atrial lead was positioned in the right atrial appendage, and the right ventricular (RV) lead was positioned in the RV apical septum. The LV quadripolar leads (four electrodes: D1, M2, M3, and P4 from distal to proximal) preferentially targeted a lateral or posterolateral branch of the coronary sinus. Positions were defined using radiological criteria and checked in 30-degree right and left anterior oblique projections immediately after implantation using fluoroscopy and the next day using X-ray.
b) Electrocardiographic optimization
Device optimization was performed immediately after implantation in the operating room. The QRS duration was measured during intrinsic rhythm and with the three specific configurations: MPP, SPP-FOI, and MPP-FOI. The 12 leads of a surface electrocardiogram (ECG) were simultaneously displayed in vertical alignment and digitally recorded for off-line measurement (WorkMate Claris; St. Jude Medical, Abbott). Time measurements were taken at a screen velocity of 200 mm/s by two experienced observers, unaware of the configurations, using a digital cursor with an accuracy of 1 ms; the mean across 3 consecutive cycles was calculated and recorded. The onset of QRS was considered to be the start of the fast deflection9 to avoid the average 45-ms transmural delay due to epicardial pacing of the LV lead. At discharge, the device was programmed with the configuration that provided the narrowest QRS duration.
Multipoint pacing programming: Multipoint pacing was programmed as previously described.3,4,14 The MPP configuration was chosen to first pace using the distal electrode (primarily D1, secondarily M2) as the cathode (LV1) and then the proximal electrode (primarily P4, secondarily M3) as the cathode (LV2). Two LV vectors with wide interelectrode cathode spacing were selected as the final LV1/LV2 vectors, utilizing the following order of preference: D1/P4 > D1/M3 or M2/P4 (Figure S1 ). When the LV pacing vector showed a capture threshold larger than 3.5 V and/or phrenic nerve stimulation (PNS), it was excluded. Programming included the following intervals: (1) fixed AV interval, 130 ms; (2) RV-LV2 (∆1) interval, 5 ms; and (3) LV1-LV2 (∆2) interval, 5 ms (∆1 and ∆2 with the minimal programmable delay).
Fusion-optimized intervals method: FOI optimization was performed as described elsewhere.7,10 Briefly, the fusion band was determined with atrial sensing and ventricular pacing with the anatomically furthest electrode in the LV lead. Pacing was initially tested with the longest AV interval that allowed LV capture; the AV interval was subsequently shortened in 20-ms decrements until LV-only capture was achieved. The AV interval that yielded the narrowest QRS was chosen and considered the AV FOI (Figure S2 ). This procedure was repeated during atrial pacing at 10 beats/min above spontaneous sinus rhythm. Afterwards, the interventricular (VV) interval was adjusted by programming the one that achieved the shortest QRS (simultaneous RV and LV pacing [VV 0 ms], 30 ms LV pre-excitation, and 30 ms RV pre-excitation). These VV intervals were chosen because previous studies indicated that most patients showed the best VVs in this range.7,10 The VV value that achieved the narrowest QRS was considered the VV FOI.
Combined MPP-FOI strategy: The AV and ∆1 (RV-LV2) intervals were optimized as in the FOI method. Using the pacing vectors LV1 and LV2 obtained for the MPP strategy, ∆2 was ultimately programmed to 5 ms (minimum programmable delay).
c) Electrocardiographic imaging
The biventricular activation time was evaluated using ECGI recordings (CORIFY Care SL) obtained while running four programming configurations: (1) intrinsic rhythm, (2) MPP, (3) SPP-FOI, and (4) MPP-FOI.
The torso surface potential was obtained using body surface potential mapping, wherein 64 electrodes attached to the skin were recorded at a sampling rate of 1000 Hz for off-line processing. Electrode location and torso and heart volumes were determined based on (1) cardiac images acquired via computed tomography or magnetic resonance imaging obtained before device implantation and (2) photogrammetry techniques that enable 3D reconstruction of the torso from a standard video recording.
An average PQRST interval was calculated for each lead using template matching averaging after removing pacing artifacts. An inverse computed electrogram was reconstructed for the heart’s surface by solving the inverse problem using zero-order Tikhonov and L-curve methods.15
The biventricular activation time was calculated by (1) a set of sinusoidal wavelets for all time instants with a negative slope within the QRS complex and with an amplitude proportional to the slope at that time and (2) the instantaneous phase of Hilbert’s transform of a composite signal of all sinusoidal wavelets. The activation times corresponded to the points where phase inversions occurred in that instantaneous phase. The total biventricular activation time was defined as the difference between the last and first activation times.
d) Echocardiographic measurements
An echocardiography specialist blinded to device programming obtained standard Doppler echocardiographic images using a commercially available system (Vivid 7; General Electric, Milwaukee, WI, USA).
Each patient was evaluated at baseline (device in OFF) and with each pacing configuration (MPP, SPP-FOI, and MPP-FOI). Cardiac asynchrony was evaluated at several levels: (1) left intraventricular asynchrony using septal flash16 and 2D strain, (2) AV asynchrony using the diastolic filling ratio (diastolic filling time/RR interval duration),17 and (3) VV asynchrony quantified as the difference between the LV and RV pre-ejection times with pulsed-wave Doppler (Figure S3 ).18 In the 2D strain analysis, the time difference between the peak systolic strain at opposing LV walls (mid-septal and mid-lateral) was measured. Finally, the LV outflow tract velocity time integral, a measure of cardiac systolic function, was quantified.
e) Estimated battery longevity
At 45 days post-implantation, data for estimating battery longevity with the final programmed configuration (MPP, SPP-FOI, or MPP-FOI) were collected, and estimated battery life was compared between patients who had MPP and SPP activated. Right atrial and RV pacing outputs were programmed at 2.5 V/0.5 ms, and LV pacing output was programmed at 1.5 times the pacing threshold.