Subjects

A distinct set of 19 patients with cardiac pacemakers and a clinical diagnosis of heart failure were recruited from the outpatient cardiac service of our hospital. The diagnosis of heart failure was based on a history of clinical symptoms, appropriate clinical signs at the time of diagnosis, and original echocardiographic evidence of abnormal systolic function. Exclusion criteria were atrial fibrillation, an intrinsic ventricular rate of greater then 80 beats per minute, and implantable cardiac defibrillators with antitachycardia therapy set at an unusually low rate (<120 beats per minute), because it would limit the ability to vary the HR during the experiment. The other exclusion criteria were significant respiratory disease (FEV₁ <50% predicted), any condition precluding lying comfortably on a bed for 90 minutes, a recent deterioration in condition ie, admission in previous 6 weeks, a brittle condition and renal failure requiring dialysis.

Patients were screened to confirm the absence of daytime periodic breathing during daytime assessment in clinic. On the day of the study, they were further monitored for 30 minutes while recumbent to exclude the presence of any form of apnea with a pause of respiration for 10 s or more, either with or without respiratory effort.³ Periodic breathing was defined as an oscillatory breathing pattern characterized by cyclic rises and falls in ventilation without true periods of apnea^4–6 and any patients showing such stereotypical oscillations, with a period of around the order of 60 s in end-tidal CO₂, end-tidal O₂, or ventilation, were not enrolled in the study. In total, 25 patients were assessed. Four patients were excluded because they demonstrated periodic breathing, one of whom also had atrial fibrillation. A further 2 patients were excluded because of atrial fibrillation.

All patients gave informed consent for the study which was approved by the local Research Ethics Committee. The investigation conforms to the principles outlined in the Declaration of Helsinki.

Measurements

Patients relaxed, recumbent on a couch, while breathing through a calibrated pneumotachograph attached to a Multicap (Datex Instumentarium, Helsinki, Finland) measuring ventilation, inspiratory, and expiratory respiratory gases. An ECG signal was recorded using a Hewlett-Packard 78351A, from which HR was derived. Beat-by-beat BP and cardiac output were measured noninvasively using a photoplethysmograph device (Finometer, Finapres Medical Systems, The 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 pressure. The device yields a continuous beat-to-beat arterial pressure waveform and incorporates the “Modelflow” algorithm which tracks changes in cardiac output by relating pressure changes to changes in a nonlinear 3-element aortic impedance model⁷ and has been extensively validated against invasive measurements for noninvasive measurement of changes in cardiac output.^8–13

Pacemaker Protocol

Pacemaker reprogramming was performed via a pacemaker telemetry head positioned on the subjects’ skin over their implanted device, to enable the HR to be changed according to protocol. Initially, the HR was varied by 30 beats from the resting value, and the effect on cardiac output was measured. This process was repeated for a range of baseline HRs, from the resting rate up to a baseline of 80 beats per minute. We then selected the baseline HR which allowed the greatest increment in cardiac output from the 30-bpm HR elevation.

In those patients in whom the cardiac output failed sufficiently to engender oscillations in ventilatory gases with any of the 30-bpm HR steps, we used alteration in AV delay between 30 and 120 ms instead of alternation of HR. If cardiac output still failed to change appreciably, we combined changing HR and AV delay simultaneously (Figure 1 depicts the protocol).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1. Protocol for deciding how to alter cardiac output. In each algorithm (HR alone, AV delay alone, and both), both square wave and sine wave changes were made in each subject. Changes in cardiac output were deemed “sufficient” if a clear change in end-tidal CO₂ and ventilation could be seen following the alternation in cardiac output.

To enable us to control the HR during the study, all subjects whose clinical pacing configuration and underlying disease gave them atrial or ventricular sensing at rest had their devices reprogrammed with a lower pacing rate 5 beats per minute above their intrinsic rate. This ensured that all subjects were paced throughout the study session.

In all patients, the cardiac output changes (whether it was by HR alone or AV delay or both) were made as a step alternation (or square wave) and as a gradual alternation (based on a sine wave).

To achieve an approximation of a sinusoidal pattern of pacing, we calculated a sequence of pacing configurations and times based on a sine wave pattern using custom written software written in Matlab (Natick, Mass). This software determined the closest programmable actual values of HR and AV delay, taking into account that these values cannot be varied continuously but rather only to a set of values specific to each model of pacemaker (Figure 2 shows an example of this and of the step change). The software allowed the magnitude of change in HR or AV delay and the possible values that could be programmed (eg, multiples of 10) to be tailored to the individual patient as well as the period of the cycle desired eg, 60 s. Examples of this for HR and AV delay are shown in Tables 1 and 2⇓, respectively.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2. Upper panel, an example of a square wave alternation of HR. Lower panel, an example of a sine wave alternation with “ideal” sine wave (red) and programmable HRs (blue) Alternations of AV delay and AV delay and HR were made in a similar way.

We monitored cardiorespiratory variables while alternating the pacing rate and or AV delay (via the pacemaker telemetry head) as a square wave for 5 cycles of 60 s and as a sine wave, also for five 60-s cycles. Each pattern used the same range of HRs (and/or AV delay), for example in an individual patient the square wave might consist of altering HR between 60 and 90 bpm and the sine wave consisted of a series of gradual changes between 60 and 90 bpm. The order in which this was done was allocated randomly for each patient with at least a 5-minute washout period between the square wave and sine wave protocols.

Data Acquisition

The data were sampled at 1000 Hz using a custom data acquisition system consisting of an analogue-to-digital card (DAQCard 6062E, National Instruments, Austin, Tex) and a workstation running software written in Labview instrument control language (v7.0, National Instruments). This system allowed data to be collected simultaneously from all the devices. The data were later analyzed off-line using software^1,14,15 written in Matlab (Natick, Mass). HR, BP, cardiac output, end-tidal gas concentrations, and ventilation were digitally interpolated and resampled to obtain signals at 1 Hz for subsequent analysis. Interpolation was done between breaths so that a value was available at each 1-s time-point to allow averaging across multiple cycles.

Data Analysis

The amplitude of the hemodynamic and respiratory oscillations in response to the cardiac output alternation were quantified using signal averaging. Data from each of the individual 60-s cycles were time-aligned, and then the mean and standard error at each point in time were calculated. A signal-averaged single cycle was produced for each patient, for each shape. The amplitude and timing of the oscillations were calculated using Fourier analysis at a frequency of 1/60 Hz, corresponding to the stimulus cycle time of 1 minute. An example of this is shown in Figure 3. The peak-to-trough excursion for each parameter was defined as the difference between the maximum and minimum value. The peak rate of change of BP was defined as the maximum rate of BP change over any 5 s of the cycle, as determined by linear regression.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3. An example of the variables recorded in one subject after 4 cycles of square wave HR manipulation (left panel) and the resultant signal-averaged cycle (right panel). Each individual cycle on the left is represented by a different color (blue, red, black, and green) and the resultant signal-averaged cycle on the right is depicted in red. Square wave changes in cardiac output can be seen to elicit, rapid and sizeable blood pressure oscillations (MAP) and sinusoidal fluctuations in end-tidal CO₂ and O₂.

Statistical Analysis

Continuous values are expressed as the mean±standard deviation. Paired t-tests were performed to compare square wave alternations with sinusoidal alternations within individuals. Patient subgroups were compared in a posthoc analysis by ANOVA. A value of P<0.05 was considered statistically significant. The sample size was designed for the principal question namely looking for a difference between square and sine wave. On the basis that the peak to trough BP would be of the order of 20 mm Hg in the square wave intervention, from previous experience, and the standard deviation approximately 5 mm Hg, to detect a change in peak to trough pressure of 5 mm Hg between the square wave and the new sine wave intervention, with 90% power at the 2-tailed 5% significance level, we would need 12 patients for the paired comparison.

Statement of Responsibility

The authors had full access to and take full responsibility for the integrity of the data. All authors have read and agree to the manuscript as written.

Subject Characteristics

Nineteen consecutive patients (7 women, 12 men) met the eligibility criteria and were enrolled. Table 3 shows the baseline characteristics. Sixteen patients had undergone biventricular pacemaker implantation for relief of heart failure symptoms in accordance with national guidelines. The remaining 3 patients had undergone conventional dual chamber pacemaker implantation, 2 for heart block, and 1 for sick sinus syndrome.

Modulating Ventilation Via Cardiac Output

In each patient, the purpose of the initial evaluation was to select a mechanism by which cardiac output would be altered in that individual in the main study: HR alone, AV delay alone or both (Figure 1). In all patients, the initial evaluation successfully identified a means of producing oscillations in end-tidal CO₂ and ventilation. In 14 of the 19 patients, the initial evaluation showed that alternation of HR by 30 beats per minute produced distinct changes in cardiac output on the Finometer. In a further 2 patients a greater change in cardiac output was elicited by instead altering AV delay from 30 to 120 ms: in them the AV delay alternation was used. In the remaining 3 patients, cardiac output failed to change appreciably using either HR or AV delay alone and therefore both HR and AV delay were altered together.

In a posthoc analysis, we found that the average size of oscillations in cardiac output and end-tidal CO₂ achieved by the individually-selected pacing intervention, were comparable among patients in whom HR, AV delay or both were selected (cardiac output: 2.7±3.0 versus 1.3±0.4 versus 2.3±1.2 L/min P=NS by ANOVA; end-tidal CO₂: 0.40±0.30 versus 0.36±0.30 versus 0.30±0.17 kPa, P=NS by ANOVA). Thus, if HR alone fails to elicit cardiac output changes in a particular individual, it may still be possible to achieve an appreciable effect using an alternative mechanism.

Shape of the Cardiac Output Waveform Elicited

Application of square wave changes in pacing configuration achieved fluctuations in cardiac output as previously described¹ (Figure 3). Application of the sine wave pattern in pacing configuration, by sequential manual programming under guidance from real-time computer software, also yielded fluctuations in cardiac output (Figure 4). The shape of the fluctuation in cardiac output produced by sine wave intervention fitted well to a sine wave (r²=0.73, P<0.01, Figure 4). The peak-to-trough difference in cardiac output was similar in both square wave alternations and sine wave alternations (2.8±1.6 versus2.2±1.2 L/min, P=0.2).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4. An example of the variables recorded in one subject after 4 cycles of sine wave HR manipulations (left panel) and the resultant signal-averaged cycle (right panel). Again, each individual cycle on the left is represented by a different color (blue, red, black, and green) and the resultant signal-averaged cycle on the right is depicted in red. The HR manipulation resembles a sine wave and the resultant cardiac output change elicits less rapid BP (MAP) changes but nevertheless achieves comparable oscillations in ventilatory gases.

Effect of Alternation of Cardiac Output on Sequence of Respiratory Parameters

Sine wave fluctuations in cardiac output produced sinusoidal oscillations in end-tidal CO₂ (fit to sine wave r²=0.80 P<0.01). Interestingly, even with square wave intervention the shape of fluctuation in CO₂ produced still fitted fairly well to a sine wave (r²=0.77, P<0.01) and indeed better than it did to a square wave shape (r²=0.22, P=0.04; Figure 3). This smoothing effect suggests that the temporal response of the end-tidal CO₂ is a gradual process of moving from one steady state value to another, which takes in the order of 30 s, so that the waveform produced is gently curved rather than a sharp square wave pattern.

The sequence of events was first an increase in the cardiac output followed by a rise in end-tidal CO₂ and then, after a delay, an increase in ventilation as shown in Figures 5 and 6⇓. The delay between the waveform of cardiac output and the waveform of end-tidal CO₂ was the same regardless of whether pacing intervention shape was square wave or as a sine wave (12.2±11.7 versus 13.0±12.1 s, P=0.84). The chemoreflex delay (measured as the time difference from peak end-tidal CO₂ to peak ventilation) was also the same regardless of the shape of modulation of cardiac output, with a mean of 23.4±7.7 s for square wave and 24.9±9.6 s for sine wave (P=0.59).

• Download figure
• Open in new tab
• Download powerpoint

Figure 5. A single alternation of HR (preceded by 30 s of free breathing without any pacemaker manipulation) demonstrates a rise in HR (top panel), is first followed by a rise in end-tidal CO₂ (middle panel) and then a rise in ventilation (bottom panel).

• Download figure
• Open in new tab
• Download powerpoint

Figure 6. Recording from one subject, in whom alternations in cardiac output were established, demonstrating sinusoidal manipulations of cardiac output (upper panel) were followed by end-tidal carbon dioxide oscillations (middle panel) and then fluctuations in ventilation (bottom panel).

Differences in Peak-to-Trough BP, End-Tidal CO₂, and Ventilation Between Square Wave and Sine Wave

The BP oscillations induced were more dramatic with square wave interventions than sinusoidal interventions: peak-to-trough BP swing was almost twice as large with square wave than with sinusoidal (22.4±11.7 versus 13.6±4.45 mm Hg, P<0.01; Figure 7).

• Download figure
• Open in new tab
• Download powerpoint

Figure 7. Comparison of peak-to-trough BP, end-tidal CO₂ and ventilation elicited via (i) sine wave alternations and (ii) square wave alternations in all patients.

The rapidity of BP rise, measured as the fastest 5-s slope, during the cycle, was 250% steeper with square wave oscillation than sinusoidal (19.8±10.0 versus 7.9±3.2 mm Hg over 5 s, P<0.01; Figure 8).

• Download figure
• Open in new tab
• Download powerpoint

Figure 8. An example of five 60-s averaged ventilatory cycles from one patient of square wave changes of HR (upper left) and resultant change in BP with marked upstroke and down-stroke (middle left) and change in end-tidal CO₂. Right-hand panel demonstrates sinusoidal HR change (upper right) in the same patient and resultant more gradual change in BP (middle right) and comparable oscillation of CO₂.

The change in the peak-to-trough end-tidal carbon dioxide produced, was also higher with square wave alternations than and sinusoidal, but only by 45% (0.45±0.18 versus 0.31±0.13 kPa, P=0.01, Figure 7). Peak-to-trough ventilation showed only a nonsignificant trend to being larger with square wave alternations than sine wave alternations (square: 0.05±0.03 L/s, sine: 0.04±0.02 L/s, P=0.24) as shown in Figure 7.

Ventilatory Disorders in Heart Failure

A substantial proportion of patients with heart failure suffer from respiratory disorders such as periodic breathing.^6,16 Periodic breathing is associated with increased mortality in heart failure.^17,18 During sleep, there are mechanical interventions which can alleviate ventilatory instability, reduce daytime fatigue and somnolence and markers of neuroendocrine activation.¹⁹ However, these mechanical interventions, including CPAP, BiPAP, and AutosetCS, generally require a firmly-held mask so that they can deliver the altered pressure to the lung. As a result, some patients find the treatment uncomfortable.¹⁹ Whereas patients with obstructive sleep apnea experience often marked symptomatic relief from CPAP and therefore find it highly acceptable,²⁰ the symptomatic relief is not always as marked in central sleep apnea in heart failure and therefore compliance ranges from good to moderate.¹⁹ There may still, therefore, be a role for exploring opportunities to develop treatments for periodic breathing that might have a higher acceptability to patients with central apneas.

The concept of using pacing to attenuate sleep apnea is not new. Atrial overdrive pacing has previously been mooted as a potential tool for the attenuation of sleep disordered breathing²¹ but studies failed to consistently demonstrate its efficacy^22,23. Cardiac resynchronization therapy using biventricular cardiac pacemakers is increasingly being used in heart failure and has been demonstrated to have a beneficial effect on mortality.²⁴ Moreover, biventricular pacemakers have been demonstrated to reduce the apnea-hypopnea index and can improve sleep quality in central sleep apnea,^25–27 making them a useful adjuvant to other therapies.²⁸ Nevertheless, implantation of a biventricular pacemaker does not uniformly extinguish periodic breathing.^25–28

Static carbon dioxide administration has previously been used and demonstrated to be effective at attenuating the severity of periodic breathing. However, this supplementary administration has been associated with adverse consequences such as an overall increase in mean ventilation²⁹ and adrenergic activation.³⁰ An optimal oscillatory cardiac pacemaker algorithm that delivers a dynamic counteraction to an intrinsic tendency to ventilatory oscillations allows the “redistribution” of carbon dioxide within the body at the correct times with no additional carbon dioxide being added to the system. Moreover, this intervention could be achieved without a closely-fitting mask and therefore might conceivably be used in situations other than sleep. Because an increasing proportion of patients with heart failure are receiving pacemaker devices³¹ it may be possible to have such an algorithm programmed within a device being implanted anyway, with the intention of additionally attenuating daytime periodic breathing beyond the benefits already recognized from biventricular pacemaker implantation itself.^24–28

Mechanism: Pacemaker Manipulation of CO₂ fluxes

This study has demonstrated that manipulations of HR, or AV delay or both can elicit changes in end-tidal CO₂ followed by ventilation. This is the first time that this effect has been demonstrated using AV delay rather than HR. This finding suggests the effect is secondary to a change in cardiac output, rather than being some direct result of HR variation.

From our study, we cannot be certain whether the mechanism by which cardiac output changes lead to the changes in carbon dioxide and ventilation is the result of a primary effect on carbon dioxide (leading to a change in ventilation) or ventilation (leading to a change in carbon dioxide) because ventilation and carbon dioxide affect each other mutually. Significantly more invasive experimentation, especially if carried out in animals, might be able to establish whether the change in cardiac output induced by pacemaker manipulation affects pulmonary vagal irritant receptors sufficiently to affect ventilation.³² Within the protocol for this study, we do have a small amount of information to guide us (Figure 5) that suggests that after cardiac output increases, end-tidal CO₂ seems to rise first, after which ventilation rises.

By increasing the cardiac output, either suddenly as a square wave or more gradually as a sine wave, one possible mechanism for this effect is that a greater quantity of CO₂-rich blood is delivered via the systemic veins to the lungs. The greater quantity of CO₂ delivered to the pulmonary capillaries shifts the dynamic equilibrium so that the alveolar CO₂ levels gradually rise: this is detectable as a rise in end-tidal CO₂. Blood draining from the pulmonary capillaries, which has been in equilibrium with the alveolar gases, correspondingly rises in CO₂ concentration. This higher arterial CO₂ concentration (and the parallel lower arterial O₂ content) is soon sensed by chemoreceptors, which elicit a ventilatory response: an increase in ventilation. The chemoreflex gain is the degree of cardiorespiratory response of the system to dynamic changes in respiratory gases. Hence, dependent on the individual’s chemoreflex gain, there will be a corresponding change in ventilation.

An alternative mechanism may be that an increase in cardiac output may produce a fall in pulmonary capillary wedge pressure which produces a rise in ventilation. Invasive studies would be required to discriminate with total confidence between these 2 possibilities.

Sine Wave or Square Wave Oscillations of Cardiac Output

If this was to be developed into a means of manipulating ventilation, it would be preferable to achieve this while minimizing any induced sudden changes in BP. Square wave changes in cardiac output are easy to program manually but are associated with large sudden changes in BP. We found that by manipulating cardiac output in a gradual, approximately sinusoidal, configuration, changes in end-tidal CO₂ and ventilation can still be elicited with comparable efficiency as if a step change were used. Importantly, the very sudden surges in BP are greatly reduced.

Potential Clinical Impact

A system for treating unstable ventilatory control in heart failure patients that used cardiac pacemakers might be completely implantable within the patient and need not impose any potentially uncomfortable equipment in the face or nose area. Devices such as biventricular pacemakers that are already in situ or that are going to be implanted on current clinical grounds could be opportunistically used.

In most patients altering HR alone is likely to be sufficient to bring about changes in ventilation, although AV delay is a potential tool in patients who are paced using a dual chamber pacemaker.

Study Limitations

This study deliberately only included patients with heart failure, as this is the population likely to gain maximum benefit from this form of therapeutic manipulation of ventilation, and we aimed to assess the size of the effects attainable in them. Therefore, the effects of different algorithms for changing cardiac output, on BP and end-tidal CO₂ in a nonheart-failure population remains untested. The comparison between those patients who had HR manipulated alone, those who had HR and AV delay manipulated and those who had AV delay manipulated alone was not a planned analysis and therefore is likely to have been underpowered to detect any true differences.

In this study, we have focused on measuring the effects of a fixed sequence of cardiac output alternations of a defined frequency and amplitude with fixed patterns (square or sine wave). In reality, for practical clinical benefit it would be necessary to have a real-time algorithm which monitors either blood composition or ventilation or both, to appropriately adjust the pacemaker configuration multiple times per minute. It will need the ability to adjust the timing of therapy to the timing of the underlying oscillations, and may also need the ability to vary the intensity of therapy (duration and/or amplitude) to suit the severity of the underlying oscillations. Although this technology is not yet available, this study suggests that it may be worthwhile to explore.

The sine wave modulations of cardiac output were approximations that were limited by both individual pacemaker programmers and the speed at which changes could practically and safely be made. Despite this, changes in HR or AV delay produced changes in cardiac output that were a good approximation to a sine wave. The study protocol allowed a standardized gradual change in cardiac output to be compared to a standardized sudden change in cardiac output. An eventual therapeutic system would have to be designed to have much finer changes in HR or AV delay, so that sudden BP changes would be even further attenuated.