The Acute Effects of Changes to AV Delay on Blood Pressure and Stroke Volume: Potential Implications for Design of Pacemaker Optimization Protocols

Abstract

Background

Atrioventricular (AV) delay optimization of biventricular pacemakers (cardiac resynchronization therapy, CRT) may maximise hemodynamic benefit, but consumes specialist time to conduct echocardiographically. Non-invasive blood pressure monitoring is a potentially-automatable alternative, but it is unknown whether it gives the same information, and similar precision (signal-to-noise ratio). Moreover, the immediate blood pressure increment on optimization has been reported to decay away: it is unclear whether this is the result of an (undesirable) fall in stroke volume or a (desirable) compensatory relief of peripheral vasoconstriction.

Methods and Results

To discriminate between these alternative mechanisms, we measured simultaneous beat-to-beat stroke volume (flow) using Doppler echocardiography, and blood pressure (BP) using finger photoplethysmography, during and after atrioventricular delay changes from 40 to 120 ms in 19 subjects with cardiac pacemakers. BP and stroke volume both increased immediately (p<0.001, within one heartbeat). BP showed a clear decline a few seconds later (average rate −0.65mmHg/beat, r=0.95 [95% CI 0.86 to 0.98]); in contrast, stroke volume did not decline (p=0.87). The immediate BP increment correlated strongly with the stroke volume increment (r=0.74, p<0.001). Signal-to-noise ratio was threefold better for BP than stroke volume (6.8±3.5 versus 2.3±1.4, p<0.001).

Conclusions

Improving atrioventricular delay immediately increases blood pressure, but the effect begins to decay within a few seconds. Reassuringly this is due to compensatory vasodilatation rather than reduction in cardiac function. Pacemaker optimization will never be reliable unless there is adequate signal-to-noise ratio. Using BP rather than Doppler minimises noise. The early phase – before vascular compensation – has the richest signal lode.

Introduction

The increasing prevalence of devices specifically designed to improve timings within the cardiac cycle (i.e. cardiac resynchronisation therapy) has created a clinical need to accurately monitor the effects on cardiac performance of changes to pacing configurations.

Optimization of pacemaker atrioventricular (and interventricular) delay settings for individual patients maximises the hemodynamic benefit of pacing ^1-4, and might be approached by monitoring stroke volume or blood pressure (BP), while changes are made to the settings. Currently the most widely-used quantitative approach for pacemaker optimization uses echocardiography to measure cardiac output using Doppler, however this is time-consuming, relies on experienced operators and has limited reproducibility^{3, 5}.

Blood pressure, whether measured invasively or non-invasively, has been shown to track trends in stroke volume well during pacemaker optimization³ and recent studies have therefore proposed using continuous beat-to-beat non-invasive BP as a marker during pacemaker optimization ^6-10. Although potentially simple, highly sensitive and reproducible^{6, 7}, it is notable that the initial increment in blood pressure that follows an AV delay change from 40ms to 120 ms may not be sustained: over the course of the subsequent minute this increment declines. A reverse phenomenon can be observed when going from the optimal to the worst configuration, Figure 1.

Figure 1.

Representative non-invasive blood pressure data from one subject during acute alternation of atrioventricular delay settings. As the programmed atrioventricular delay is changed from a physiological AV delay (120ms) to an unphysiogically short AV delay (40ms) there is a clear measureable reduction in blood pressure. This decrement however reduces in the initial 10-20 seconds back towards the baseline, and when the atrioventricular delay is changed in the converse direction (from 40-120ms), the initial increment in blood pressure also reduces over a similar time period to a proportional degree.

It is currently not known whether the decline in blood pressure increment results from decay in the stroke volume increment, or is the result of reflex compensatory reduction in peripheral resistance following the stroke volume increase. This distinction is clinically important because a decline in stroke volume means that the cardiac efficiency improvement achieved by optimization may be short-lived. Conversely, if it is caused by a decay in peripheral resistance, then the hemodynamic benefit from optimization causes not only an increase in cardiac output, but also a reflex reduction in the chronic systemic vasoconstriction of heart failure.

In this study we measured blood pressure and stroke volume simultaneously to discriminate between these two potential hemodynamic mechanisms, in order to improve the protocols used in studies of the effects of pacing parameter settings optimization.

Methods

Subjects

22 consecutive outpatients with implanted pacemakers were recruited from clinics. Patients with both normal and impaired systolic function were included into the study, to ensure that any observations seen were applicable to any paced patient, rather than being limited to heart failure. Three were excluded after screening because their echocardiographic windows were too poor for reliable prolonged Doppler measurements needed for this study. Of the remaining patients, ten had normal systolic function and nine had heart failure (mean LV ejection fraction 61±8% versus 30±12%, p<0.001). Patients gave informed consent for this study which was approved by the local ethical committee.

Data acquisition

Patients lay awake, recumbent in the left lateral position on a couch having rested supine for at least 10 minutes. Systolic and diastolic blood pressures (SBP and DBP) were measured non-invasively using a photoplethysmograph device (Finometer, Finapres Medical Systems, Netherlands). This uses a cuff that is placed around the finger, a built-in photo-electric plethysmograph and a volume-clamp circuit that dynamically follows arterial blood pressure. The device yields a continuous beat-to-beat arterial blood pressure waveform and has previously been extensively validated against invasive measurements for changes in BP and cardiac output^11-13. Data were acquired using Labview (National Instruments, Austin, TX) and analysed with custom software based on the Matlab platform.

Echocardiographic measurement of stroke volume

Stroke volume was calculated from Doppler flow velocity measurements sampled in the left ventricular outflow tract (LVOT) according to published protocols¹⁴ using the aortic annulus leading edge method and measurements of the time-velocity integral (VTI) of pulsed wave Doppler blood flow ^{15, 16}. Thus cardiac output (Q) was calculated as:

$$\mathrm{Q}={\mathrm{VTI}}^{\ast }\pi {(\mathrm{D}∕2)}^{2\ast }\text{Heart Rate}$$

Measurements were made using a Philips ATL 5500 machine with a 2.5 MHz transducer. All measurements were made by a single experienced operator. Doppler data were transferred to HDI Lab in digital format and then further analysed off line with custom Matlab software.

Design of protocol of changes in pacemaker setting

It was already known that changing a pacemaker setting causes changes in both stroke volume and blood pressure. It was also suspected (as shown in Figure 1) that these immediate changes may not be sustained, but it was not known whether this progressive decrement in the initial change affected both stroke volume and blood pressure, and if it did, whether it affected them to the same extent. The study was therefore designed exclusively to examine the temporal characteristics of secondary progressive changes in stroke volume and blood pressure after the initial instant effect of change in AV delay. We knew that these secondary changes were likely to be smaller than the initial instant effects. To be able to confidently describe the pattern of these secondary changes it was essential that the signal-to-noise ratio would be large enough for these response patterns to be distinguished from the spontaneous natural fluctuations that always occur in both variables. To achieve this, we required the signal to be large and the noise to be small.

To make the signal large, we designed the protocol to change the AV delay setting from one which would be far from optimal in every patient studied, to another setting which would be much better in every patient studied. We settled on 40ms and 120ms respectively because our previous experience is that 40ms is always far from optimal⁶, while 120ms is a common default setting that has frequently been used as a reference setting in our research ^{6, 7}.

To make the noise (standard error) small for each patient, we performed multiple replicates of each change in setting and averaged the replicates. Maximizing the number of replicate measurements per patient allowed the physiological response to be discriminated more precisely. We have previously found, in common with other investigators^{5, 17, 18}, that the effect of changing AV delay is much larger than the effect of changing VV delay⁶. We therefore chose to focus on the AV delay in this study, programming a fixed VV delay with the nominal setting of 0ms as our standard throughout the patient group.

The paced heart rate was maintained at 120bpm as previous data has shown that alterations in AV delay have a more pronounced effect on BP and cardiac output at higher heart rates⁷. The optimal AV delay at 120bpm correlates well with the optimal at rest¹⁹, but the greater difference in hemodynamic parameters at the higher heart rate is due to the greater importance of LV filling characteristics when diastole is shorter. This elevated heart rate also means that a uniform protocol can be used with all subjects being both A and V paced throughout the study.

Experimental Protocol

Continuous beat-to-beat BP and cardiac output were recorded using the Finometer, with simultaneous measurement of aortic Doppler flow velocity from which stroke volume and cardiac output were calculated, during adjustment of the AV delay of the subjects’ pacemaker. We had continuous ECG recording and confirmed in all subjects that the QRS morphology was consistent at both AV delay settings, thereby ensuring that there was full capture of the ventricles irrespective of the programmed AV delay.

Pacemaker reprogramming was performed via a pacemaker telemetry head positioned on the subjects’ skin over their implanted device, to enable the atrioventricular delay to be changed according to protocol. Changes to AV delay were programmed without concomitant alterations to the programmed heart rate. Finometer and Doppler data were collected for 10 beats before and 20 beats after the change in AV delay. Mean arterial pressure (MAP) was approximated using SBP and DBP in the usual way, MAP≈DBP+⅓(SBP-DBP).

Each AV delay change was performed in quintuplicate, and the data averaged in order to minimise the effect of noise, and of variation in stroke volume caused by, for example, respiration.

In a small subset of 5 patients (3 male and 2 female), we also performed prolonged data acquisition for 1 minute (120 beats) after the AV delay change. These patients had similar characteristics to the overall patient group (3 with depressed and 2 with normal cardiac function), but were the final 5 subjects recruited because the extended recording was thought worthwhile to add as an adjunct substudy.

Calculation of relative change in BP and cardiac output between 2 atrioventricular delays

The baseline values for SBP, DBP, MAP and cardiac output for each AV delay change were defined as the mean of the values for the 10 beats before the change in AV delay. We then calculated the relative change (Δ) in each of the hemodynamic parameters by comparing the mean of the data from the 10 beats after a change in AV delay with the baseline value. The results of the 5 changes were then averaged.

Beat-to-beat data analysis

In order to accurately assess whether the initial increments in blood pressure and stroke volume following the change in AV delay from 40 to 120 ms declined following the atrioventricular delay change, we analysed beat-to-beat data. For each beat before and after the change in atrioventricular delay we measured blood pressures and Doppler-derived stroke volume. We also calculated a baseline consisting of the mean of the final 10 beats of 40 ms before the AV delay change. All beat-to-beat data were then expressed relative to this baseline as ΔSBP, ΔDBP, ΔMAP and ΔQ.

We then quantified the decline by applying linear regression, and tested whether the slope was significantly different from zero.

Calculation of the relevant signal-to-noise ratio

The relevant characteristic of a variable used to detect a change is not the percentage change in response to an intervention, but rather its signal-to-noise ratio. In detecting and quantifying a change, the comparison should be not with the magnitude of the baseline value, but rather the degree of beat-to-beat variability of the baseline value.

The signal-to-noise ratio was calculated by dividing the ‘signal’, defined as the change in a variable (ΔSBP, Δ DBP, ΔMAP and ΔQ) by the ‘noise’. Signal was defined as the mean of the first 10 beats after the change in atrioventricular delay (the acute peak) minus the mean of the immediately preceding 10 beats. Noise was defined as the standard error of the mean of those preceding 10 beats.

A high signal-to-noise ratio for a variable means that the magnitude of the effect induced by the AV delay change is large in comparison to the uncertainty (arising from random beat-to-beat variation) in that measured magnitude of effect.

Statistics

Distributions are described by the mean and standard deviation, as all data met criteria for normality. Precision of the estimate of the mean is described by the standard error of the mean. Comparisons between groups of numerical values are made using Student’s t test, with 2-tailed hypothesis testing. Correlations were measured using the Pearson correlation coefficient (with calculation of 95% confidence intervals of r based on the Fisher r-to-z transformation) , and the slope of the reduction in the blood pressure and stroke volume data was measured using least-squares linear regression. In order to calculate the response across all subjects, we calculated the beat-to-beat relative change in each parameter compared to baseline, and then averaged the data across the entire subject group. A p value of <0.05 was considered significant. Statistical analysis was performed using StatView software (version 5.0, SAS Institute Inc, Cary, NC).

Results

Patient characteristics

Twelve patients were male, 7 female, age range 46-88 years (mean 72 years), Table 1. Indications for pacemaker implantation in the patients with normal systolic function comprised sinoatrial node disease (2), complete heart block (3), atrioventricular block (4) and syncope (1). Eight of the patients with heart failure had biventricular pacemakers implanted for conventional resynchronization indications, and one had a dual chamber pacemaker implanted for complete heart block. Mean baseline sensed and paced AV delay was 120ms (range 80-160ms).

Table 1.

Patient characteristics.

Patient characteristic	Normal systolic function (n=10)	Heart Failure (n=9)
Mean age /years	75 (10)	69 (12)
Mean ejection fraction /%	61 (9)	30 (12)
Male (%)	6 (60)	6 (67)
CRT device (%)	0 (0)	8 (89)
Paced QRS duration (ms)	162 (75)	98 (52)
Hypertension (%)	7 (70)	2 (22)
Diabetes mellitus (%)	3 (30)	4 (44)
Valvular heart disease (%)	1 (10)	2 (22)
Ischaemic heart disease (%)	1 (10)	5 (56)
ACE inhibitors/ AIIRB (%)	2 (20)	9 (100)
β blockers (%)	1 (10)	8 (89)
Spironolactone (%)	0	5 (56)
Diuretics (%)	1 (10)	8 (89)

Continuous variables are reported as mean (standard deviation) and categorical variables are reported as count (%).

In the heart failure group most patients were NYHA class II/III at the time of study (NYHA I-1, NYHA II-3, NYHA III-5).

All patients tolerated the protocol well without symptoms or ECG changes.

Immediate effect of change in atrioventricular delay on blood pressure and stroke volume

When atrioventricular delay was changed from 40ms (unphysiologically-short) to 120ms (a more physiological value), there was an immediate increment in all hemodynamic parameters (Table 2) across all subjects, as expected. Systolic blood pressure (SBP) increased by an average of 19.7 ±9.5mmHg (mean± SD) or, in relative terms, 16.5±7.9% of its prior value. Cardiac output measured beat-to-beat by Doppler echo VTI in the LV outflow tract rose by an average of 9.7±4.8 ml/beat (20.7±10.3%) across all patients. Individual patient data is shown in Table 3. There was a strong correlation between the initial increments in MAP and cardiac output (r=0.74, p<0.001).

Table 2.

Average response to a change in programmed atrioventricular delay from 40 to 120ms across all subjects.

Hemodynamic Parameter	Absolute Increase (mean ± SD)	Relative Increase (mean ± SD)	p value
Systolic Blood Pressure	19.7 ± 9.5 mmHg	16.5 ± 7.9 %	<0.0001
Diastolic Blood Pressure	6.8 ± 3.9 mmHg	10.6 ± 6.1 %	<0.0001
Mean Arterial Pressure	11.0 ± 5.3 mmHg	13.3 ± 6.4 %	<0.0001
Pulse Pressure	12.4 ± 7.7 mmHg	19.3 ± 3.6 %	<0.0001
Cardiac output	9.7 ± 4.83 ml/beat	20.7 ± 10.3 %	<0.0001

Data presented as mean ± standard deviation.

Table 3.

Individual patients’ responses to a change in programmed atrioventricular delay from 40 to 120ms. Data shown are relative increases in each hemodynamic parameter.

Subject No	Group	SBP change (%)	DBP change (%)	MAP change (%)	Stroke volume change (%)
1	HF	24.9	19.6	21.8	26.3
2	NHF	25.4	18.4	22.0	16.2
3	HF	12.1	9.7	10.3	16.1
4	HF	11.4	8.0	9.6	28.3
5	NHF	12.2	3.1	6.0	23.3
6	NHF	20.0	16.2	18.2	19.4
7	NHF	21.0	12.9	16.6	25.7
8	HF	14.3	8.0	11.1	14.0
9	HF	25.3	16.1	20.5	22.8
10	NHF	19.7	12.7	15.7	21.8
11	NHF	28.1	10.6	20.3	13.1
12	NHF	−0.5	−0.9	−0.7	3.5
13	NHF	14.6	8.4	11.1	8.8
14	HF	10.6	6.9	8.6	13.9
15	HF	16.6	7.8	12.7	13.4
16	HF	1.5	−1.3	0.1	8.7
17	HF	15.4	12.0	13.4	28.4
18	NHF	9.6	9.1	6.9	41.7
19	NHF	31.7	28.8	30.0	41.3

HF=Heart Failure, NHF=Non-Heart Failure

Divergent behaviour of blood pressure and stroke volume after the initial synchronous increment

Following the initial rise, the beat-to-beat blood pressure-derived measurements showed a gradual but distinct decline after 5-10 beats (Figure 2).

Figure 2.

Beat-to-beat data for each of the hemodynamic measures during a change in atrioventricular delay from 40ms to 120ms – averaged across all subjects. Almost immediately following the change in atrioventricular delay (at beat 0) there is a sharp significant increase in all measures of blood pressure and cardiac output. However after a few beats, BP gradually falls, whilst stroke volume remains elevated. Data represents mean ± SEM. SBP – systolic blood pressure, DBP – diastolic blood pressure, MAP – mean arterial pressure.

All measures of increments in BP declined progessively from 5 beats after the change in atrioventricular delay. Across all subjects SBP decreased by an average of 1.28 mmHg per second, r=0.95 (95% CI 0.86, 0.98). Similarly, DBP decreased by 0.75 mmHg/s, r=0.97 (95% CI 0.92, 0.99).

In contrast the beat-to-beat stroke volume measurements showed that the initial increment in cardiac stroke volume at the AV delay change was sustained throughout the recording period, with no significant decline (r=0.46 (95% CI −0.05, 0.78), Figure 3.

Figure 3.

Rate of decline of hemodynamic measures between beats 5 and 20 after the change in atrioventricular delay – averaged across all subjects. There was a clear decline in all blood pressure-derived measures from 5 beats after the change in atrioventricular delay In contrast, stroke volume as measured by Doppler echocardiography, retained almost all of the increment caused by the change, throughout the recording period. Data represents mean ± SEM. SBP – systolic blood pressure, DBP – diastolic blood pressure, MAP – mean arterial pressure.

Extended recording substudy

In a subset of 5 patients we used repeated experimental sequences to acquire stroke volume and blood pressure data over a longer composite time frame than could be acquired in a single recording, in order to determine the pattern and timeframe of change.

The first part of these extended recordings was consistent with the findings in the overall patient group. The initial BP increment then declined by about one-third from the peak (SBP 28%, MAP 32%, DBP 39%), to form a plateau after approximately 30 beats. Stroke volume, however, did not decline from its initial increment (Figure 4).

Figure 4.

Beat-to-beat data for an extended measurement period of 120 beats post change in atrioventricular delay. Data averaged from the 5 subjects who underwent this extended measurement. Immediately following the change in atrioventricular delay there are initial increments in both blood pressure and stroke volume Doppler data. In the subsequent 20 beats, the increment in blood pressures decline by approximately one-third from their peak values and then plateau. In contrast the stroke volume Doppler data retains the initial elevation for the whole 120-beat recording. SBP – systolic blood pressure, DBP – diastolic blood pressure, MAP – mean arterial pressure.

Comparison of signal-to-noise ratio for mean arterial BP and cardiac output

The signal-to-noise ratios of the arterial BP increment measured using the Finometer and cardiac output increment measured using Doppler echocardiography were compared in order to determine whether they were equivalent in terms of information quality (Figure 5). We quantified the “signal-to-noise ratio”, namely the change in a hemodynamic parameter elicited by pacemaker reprogramming divided by the within-patient standard deviation in that measure in the preceding beats.

Figure 5.

Signal-to-noise ratio in each of the hemodynamic measures. There was a significantly better signal-to-noise ratio in all of the parameters measured using the Finometer than in the Doppler-derived stroke volume data. Data represents mean ± standard deviation. SBP – systolic blood pressure, DBP – diastolic blood pressure, MAP – mean arterial pressure.

Each of the Finometer-measured indices had significantly higher signal-noise ratio than the gold standard non-invasive measure of cardiac output using Doppler (signal-to-noise ratio for SBP versus Doppler stroke volume (mean ± SD) 6.8±3.5 versus 2.3±1.4,p<0.001), Figure 5.

The higher signal-to-noise ratio for the Finometer-derived measurements compared to the Doppler-derived measurements is due to the significant beat-to-beat variability of the Doppler trace. This was seen in all of the individual patients’ data (Figure 6 and all individual patient data published in Online Supplemental Data).

Figure 6.

Individual example beat-to-beat data for SBP and stroke volume Data shown is the mean (and standard error) beat-to-beat blood pressure and stroke volume data for 5 replicate transitions in AV delay from 40-120ms in one subject. Systolic blood pressure had significantly less beat-to-beat variability than the Doppler-derived stroke volume data. This resulted in a much larger ‘noise’ denominator of the signal-to-noise ratio calculation for the Doppler stroke volume than for blood pressure. All other individual patient data can be found online in the Online Supplementary Data.

Comparison of responses between subjects with heart failure and those with normal systolic function

The magnitude of the responses in all of the variables to the change in AV delay from 40-120ms was not significantly different between heart failure patients and those with normal systolic function (SBP % change 14.7±7.3 versus 18.2±9.5% (mean±SD), p=0.39; DBP % change 9.7±5.9 versus 11.9±8.2%, p=0.50; MAP 12.0±6.5 versus 14.6±9.0%, p=0.48; stroke volume 19.1±7.4 versus 21.5±12.5%, p=0.63).

We compared the rate of decline of the blood pressure and stroke volume increments in the patients with impaired systolic function with those with normal systolic function and found that the pattern of response was very similar in the two groups. The rate of decline of SBP across the heart failure subjects was 1.29mmHg/s (r=0.88,p<0.001), compared to 1.28mmHg/s (r=0.96,p<0.001) in the subjects with normal systolic function, with no significant difference between the groups. In neither subject group was there a significant downward trend in stroke volume with time after the initial rise at the change in AV delay.

Discussion

We find that, on changing atrioventricular delay, both stroke volume measured using Doppler, and BP measured using the Finometer device, show discernable reproducible increments and therefore have potential for use as physiological markers to guide pacemaker optimization. The effect of atrioventricular delay modification on stroke volume is immediate and persistent. However BP, despite an initial parallel effect, subsequently partly decays back towards its prior value. We can conclude that this decay in blood pressure is therefore due to compensatory vasodilatation (which may be beneficial to the patient) rather than a loss of the increment in stroke volume.

While the persistent nature of the stroke volume elevation permits a longer window of opportunity to measure it, this has to be balanced against the much greater operator skill needed to obtain and quantify Doppler data, and its significantly poorer signal-to-noise ratio. If BP is to be used as a hemodynamic marker to guide pacemaker optimization to avoid the acquisition and interpretation difficulties inherent in a Doppler approach, the BP recording should be made in the initial phase following a change in pacemaker parameter when the signal is richest in information.

Optimization of pacemaker atrioventricular (and interventricular) delay settings

Many methods for device optimization have been proposed including invasive hemodynamics ^{3, 7, 20}, impedance cardiography ²¹ and implantable hemodynamic monitors ²². The methods most commonly employed are based on Doppler – either using the velocity time integral of aortic outflow ^{23, 24} or the Ritter method²⁵ which determines the longest filling time associated with complete atrial systole uninterrupted by ventricular systole. There are however substantial technical difficulties with using Doppler echocardiography for pacemaker optimization in that it is time-consuming, needs skilled operators, and is difficult to perform during exercise and in non-echogenic patients.

Non-invasive beat-to-beat measurement of BP has been recently developed as an alternative approach to pacemaker optimization ^{6, 7, 9, 10, 26}. This is significantly faster to perform, does not require skilled operators and can in principle be performed on any patient, including under physiological conditions other than rest, if desired. It could also potentially be automated (further speeding up the optimization process) and the data has a significantly higher signal to noise ratio.

Recently, contrary to previous individual studies ^{4, 23}, SMART-AV, a carefully-designed, prospectively-recruited, externally-monitored, randomised controlled trial has shown that neither the qualitative iterative method (whereby an echocardiogram operator qualitatively selects the visually most desirable transmitral Doppler profile) nor a proprietary electrocardiogram-based optimization (designed to match qualitative echocardiographic optimization) confer clinical benefit chronically to heart failure patients ²⁷ Trials of quantitative optimization are now underway.

Evaluation of the clinical impact of optimization algorithms may be prolonged, expensive and disappointing unless a rational process of selection of algorithms is applied early. First, test-retest reproducibility of the optimum delay should be evaluated in dispassionate, blinded hands. Poor test-retest reproducibility disqualifies an optimization method early (and cheaply), although good reproducibility is no guarantee of suitability (and good between- and within-observer remeasurability on identical pre-acquired data sets gives no reassurance). Second, within the set of algorithms with good test-retest reproducibility, head-to-head comparison of optima in the same patients may show clustering – some algorithms frequently agreeing with each other on the optimum, while other algorithms rarely agree with the others. The clusters of test-retest reproducible algorithms with good mutual agreement are the most plausible as valid optimization methods. Since they frequently agree, any one of them might be chosen for the third stage, of assessment of clinical endpoint impact in a prospectively recruited, randomised controlled trial.

Mechanisms of hemodynamic changes during changes in atrioventricular delay

Until this study it had been unclear whether the reduction in the BP increment following a change in AV delay was purely due to compensatory vasodilatation, or whether there was a significant contribution from decay in the stroke volume increment, which would clearly be deleterious in the heart failure population where BP is such a significant prognostic marker²⁸. We found that over the 20-30 beats after the initial peak increment in BP, about 1/3 of that increment was lost; after that time, there was no further loss up to 120 beats. In contrast, stroke volume maintained its increment, with no sign of decay. This indicates that the important benefit of increased cardiac function (namely increased stroke volume), is preserved, and that the decline in blood pressure is due to compensatory vasodilatation and to the increase in the Windkessel charging. The Windkessel effect is the mechanical capacitor effect of the great vessels in response to cardiac ejection, that ‘cushions’ the peripheral vasculature from changes in stroke volume²⁹.

We speculate that there are 4 phases to the acute hemodynamic response to a change in atrioventricular delay. First, the improved left ventricular filling increases stroke volume and this partly goes to progressively charge the Windkessel and partly to raise the BP. Soon, the Windkessel capacitor is fully charged, and BP is maximal. The third stage occurs when reflexes stimulate peripheral vasodilatation meaning that the overall BP increment declines: in our data this seems to start about 10 beats (~5 seconds) after the rise in stroke volume, consistent with the response time of the baroreflex³⁰,. Phase 4 occurs when a steady-state is reached and BP joins the stroke volume in a plateau phase

In summary, while the first few beats after a change in atrioventricular delays reflect the changes in cardiac output, the subsequent beats reflect reflex changes in systemic vascular resistance. We also found that there was no difference between the subjects with heart failure and those with preserved systolic function in the patterns of response following a change in AV delay, and specifically that the rate of decline in the blood pressure increment was very similar in the two groups. The implication is that despite the likely differences in physiological responses between these groups, the phenomenon of secondary decline in pressure is conserved.

Implications for hemodynamic optimization

These results have importance for measurement during the clinical optimization of atrioventricular and interventricular delay of biventricular pacemakers. We have found that despite a fall in the non-invasive BP measured, there is no concomitant fall in stroke volume, meaning that if measurements are taken in the initial phase after a change in AV delay, non-invasive BP is a suitable surrogate marker for stroke volume and hence cardiac output.

Although stroke volume measured using Doppler echocardiography is a direct measure of cardiac function following a change in atrioventricular delay, the signal-to-noise ratio is threefold poorer than that of non-invasive BP measurements. The physiological sensitivity of Doppler stroke volume measurements is greater than that of the blood pressure measurements because of the influence of respiration, however these fluctuations reduce the detection of ‘genuine’ changes in stroke volume that result from a change in pacing parameters. There is also more measurement error contributing to the Doppler data, because of the practical difficulties with maintaining a constant Doppler angle and position for repeated, prolonged measurements during optimization. Because of the three-fold poorer signal-to-noise ratio with the Doppler data, it would take 9 (=3²) times as many beats to obtain the same discriminatory power using stroke volume rather than blood pressure.

The fact that the BP increment partially declines following optimization has created a quandary for investigators designing a protocol based on blood pressure for hemodynamic optimization, because it is unclear when the blood pressure increment should be measured. Some investigators have elected to omit the first few beats after an AV delay change^{9, 20}, however they are then measuring not only the change in stroke volume resulting from a change in atrioventricular delay, but also incorporating a function of the subject’s vasomotor tone and reactivity – a variable that changes even within an individual subject with time. Our data indicate that it would be preferable to focus on the initial BP change and that it may be counterproductive to ‘skip’ beats after the AV delay change because those skipped beats would have been the richest source of information about cardiac output changes.

An easy, time-efficient, non-invasive, rapid and potentially reproducible method for optimization of biventricular pacemakers may therefore be to measure beat-by-beat blood pressure averaged over the first 10 beats after a change to the tested setting, in comparison to the beats before the transition (in a common reference state used for all the tested settings). The relative status of each atrioventricular and/or interventricular delay can then be compared with the common reference, and along a common scale. It would be important to perform each test in enough replicates within the patient and average them to ensure that the inevitable chance variation is not mistaken for optimization^{5, 31}.

Study Limitations

In this study we aimed to assess the cardiovascular mechanisms behind the acute hemodynamic responses to a change in atrioventricular delay, which could later be exploited for scientific development of pacemaker optimization. This was not itself a study of optimization and we were not assessing long term clinical outcomes. We did not attempt to either determine the optimal delay for each subject, to assess which method is the most sensitive for determining the optimal AV delay, nor to draw conclusions concerning the chronic effects of optimization. We carried out elaborate beat-to-beat velocity-time-integral measurements in an experimental protocol that focussed on a single, consistent, type of AV delay change. We designed the study to use the change in atrioventricular delay from 40 to 120 ms because our experience⁷ has shown that this change is likely to cause an increase in blood pressure in all subjects, and therefore would be suitable to use to explore the hemodynamic mechanisms.

We did not assess the acute beat-to-beat effects of each potential AV delay change because of the very much large number of measurements that would be needed to do this adequately. For example, suppose we had chosen to test in addition to 40 ms, another setting twice as close to 120 ms (i.e. 80ms). Because of the parabolic shape of haemodynamic responses, we would expect its haemodynamics to be 4 times closer to those of 120 ms, i.e. the signal would be ¼ the size of the signal for 40-to-120 ms. In order to preserve the signal-to-noise ratio so that the time course can be studied, because shrinkage of noise is proportional to the square root of the number of replicates, we would need 16 times more replicate measurements per setting tested, or 16 times more patients. In general studying the region N times closer to the optimum requires ~N⁴ more data to be acquired per setting.

Our study was limited to non-invasive measurements – conceivably invasive measurements might show different findings because of the ability to make more central measures of beat-by-beat pressure directly.

This study included both normal subjects as well as subjects with heart failure. Although these are two groups with different clinical characteristics, we found that the pattern of hemodynamic responses to changes in atrioventricular delay was similar, irrespective of clinical group. This showed that the responses were not limited to heart failure subjects, but were a phenomenon of any paced patient.

Conclusions

The early rise in BP is a valid marker of cardiac performance for optimization of pacemaker settings. Although it declines slightly after a few beats, this is not accompanied by any decline in stroke volume. The partial decline in BP is due to compensatory vasodilatation – which is clinically desirable – rather than any reduction of cardiac performance.

If economy of time is a concern, the 3-fold higher signal-to-noise ratio for BP should translate into a 9-fold shorter time to deliver an optimum of the same precision versus quantitative Doppler optimization. Moreover, rather than prolonged BP recording after the signal attenuates, replicating the transition in pacemaker setting and remeasuring the signal at full strength may be wiser.