Efficiency, reproducibility and agreement of five different hemodynamic measures for optimization of cardiac resynchronization therapy

Transcription

1 International Journal of Cardiology 129 (2008) Efficiency, reproducibility and agreement of five different hemodynamic measures for optimization of cardiac resynchronization therapy Zachary I. Whinnett, Justin E.R. Davies, Gemma Nott, Keith Willson, Charlotte H. Manisty, Nicholas S. Peters, Prapa Kanagaratnam, D. Wyn Davies, Alun D. Hughes, Jamil Mayet, Darrel P. Francis International Centre for Circulatory Health, St Mary's Hospital and Imperial College, North Wharf Road, W2 1LA, London, United Kingdom Received 2 December 2006; received in revised form 28 June 2007; accepted 3 August 2007 Available online 18 September 2007 Abstract Background: Several hemodynamic measures have been used for optimization of the AV delay of cardiac resynchronization therapy (CRT), including pulse (PP), systolic (SBP) and cardiac output (CO). We aimed to determine whether these measures identify the same optimum and whether they have the same efficiency and reproducibility at identifying this optimum. Methods and results: In 22 patients with cardiac resynchronization therapy, we adjusted the AV delay while atrially pacing at 110 bpm and simultaneously recording SBP, diastolic (DBP), PP, mean arterial (MAP) and CO. SBP, PP and CO all had essentially the same signal-to-noise ratios (15.4±5.4, 15.5±6.4, 15.3±7.4 respectively p=ns). In contrast, MAP and DBP had significantly worse signal-to-noise ratios than SBP (14.2±5.6, p=0.003 and 12.1±4.4, pb respectively). The optimal AV delay was very similar between SBP, PP, MAP and DBP. For example, the optima identified by SBP correlated strongly with those identified by PP (r=0.94), MAP (r=0.96) and DBP (r=0.90). In contrast, the optima detected by CO was poorly related to these (e.g. r=0.36 with SBP optima). Reproducibility was best for optima detected by SBP followed by MAP and PP. Conclusions: Essentially the same AV optimum is identified, regardless of whether the parameter chosen for maximization is SBP, PP, MAP or DBP. We conclude that optimizing the CRT AV delay using SBP gives the best combination of efficiency and reproducibility, with PP and MAP being reasonable alternatives Elsevier Ireland Ltd. All rights reserved. Keywords: Atrioventricular delay optimization; Cardiac resynchronization therapy; Finometer; Hemodynamic optimization 1. Introduction Cardiac resynchronization therapy improves hemodynamic status [1 8]. Additional hemodynamic improvements can be obtained by selecting the appropriate atrioventricular delay (AV) [9,7,10,11] and interventricular delay [5,12,13] for an individual patient. Conflict of interests' disclosures: Dr Whinnett (FS/05/068), Dr Davies (FS/05/006) and Dr Francis (FS/04/079) are supported by fellowships from the British Heart Foundation. Dr Manisty (077049/Z/05/Z) is funded by the Wellcome Trust. Our institution has filed a patent on some of the methods described in this manuscript. Corresponding author. Tel.: ; fax: address: zacharywhinnett@yahoo.com (Z.I. Whinnett). A number of different hemodynamic markers have been used to guide this optimization process. These include systolic (SBP) [11], pulse (PP) [9], mean atrial (MAP) and cardiac output (CO). However, it is unclear whether all of these measures give comparable results when they are used for AV delay optimization both in terms of signal quality, the delay identified as providing the maximal hemodynamic response and reproducibility. This information is potentially important when selecting which hemodynamic parameter to use in clinical practice and when designing studies assessing the effect of optimization on clinical outcome measures. It is possible to acquire these hemodynamic measures noninvasively using the Finometer device (Finapres Medical /$ - see front matter 2007 Elsevier Ireland Ltd. All rights reserved. doi: /j.ijcard

2 Z.I. Whinnett et al. / International Journal of Cardiology 129 (2008) Systems, Amsterdam, Holland). Measures of systolic, diastolic, cardiac output, mean arterial can all be acquired continuously on a beat-tobeat basis using this device. Signal-to-noise ratio is an important feature of any measure. It is a simple way to characterize the efficiency of optimization by that measure. One simple definition of the signal-to-noise ratio is the range of values obtained for different pacing settings, divided by the standard error of the measurements at each pacing setting. By prolonging the optimization session, signal-to-noise ratio can be improved, but for a given duration (i.e. a given number of data points) signal-to-noise ratio is a suitable way of comparing efficiency between measures. The second question that needs to be assessed is the degree of similarity of the optimal pacing configuration determined by the different hemodynamic measures. If all measures are in agreement then it lends support to the Fig. 1. Calculation of the relative change in (SBP rel ) at one tested AV delay, 40 ms in this example. Each tested AV delay is compared to a reference AV delay of 120 ms and the relative change in is calculated by determining the difference between the mean from the 10 beats before a transition and the 10 beats after a transition, the mean SBP rel is then calculated from a total of 6 transitions (3 forward and 3 back). This process, lasting 1 min, is then repeated with each of the other tested AV delays, to create the curve seen in Fig. 2. The same method was used for each of the different tested hemodynamic parameters.

3 218 Z.I. Whinnett et al. / International Journal of Cardiology 129 (2008) concept of hemodynamic optimization as whole. Alternatively if the measures produce dramatically different optima then the use of hemodynamics for optimization is cast into doubt. The third important characteristic of a measure put forward for use in optimization is reproducibility over time. If there is a large change in measured optimum then either the physiological optimum is changing a great deal (in which case more frequent optimization would be worthwhile) or the optimization method does not have sufficient reproducibility. In this study, we aimed to compare the signal quality for the tested hemodynamic measures, establish whether they identified similar optimal AV delays for individual patients, and compare the reproducibility of the different measures. 2. Methods 2.1. Subjects Twenty-two outpatients who previously had biventricular pacemakers or biventricular defibrillators implanted for clinical indications (NYHA III or IV heart failure, QRS N 120 ms, maximal medical therapy) were enrolled into this study, between 3 and 30 months after implantation. At the time of implant the mean NYHA class had been 3.1 and the mean reduction after implant was 0.9. At the time of study 4 patients were in NYHA class I, 8 were in NYHA II and 10 were in NYHA III. Fourteen patients were male and 8 female, with age range years (mean 68 years). Cause of heart failure was ischemic in 12, and idiopathic dilated in 9 and hypertensive in 1. Mean systolic by sphygmomanometer was 123 ± 19 mm Hg. Mean left ventricular ejection fraction of the patients at the time of the study was 32 ± 4%. Fourteen patients were taking angiotensin-converting enzyme inhibitors, 12 were taking angiotensin-ii receptor antagonists, 18 were taking beta-blockers, 16 were taking spironolactone, 17 were taking a diuretic (loop or thiazide) and 7 were taking digoxin. Patients gave informed consent for this study which was approved by the local ethical committee Measurements Data acquisition We recorded beat-by-beat systolic and diastolic using a Finometer (Finapres Medical Systems, Amsterdam, Holland). This technique, developed by Peňa z [14] and Wesseling [15] uses a cuff that is placed around the finger, a built-in photo-electric plethysmograph and a volume-clamp circuit that dynamically follows arterial. This technique is well validated for measuring instantaneous changes in [16 20]. We measured beat-by-beat cardiac output using the Modelflow algorithm built into the Finometer. The Modelflow method tracks changes in cardiac output by relating changes to changes in a non-linear three-element aortic impedance model. We entered the patients' age, sex, Fig. 2. Calculation of the signal-to-noise ratio. The signal is the range of SBP rel across different AV delays, while the noise is the size of the uncertainty in each SBP rel. By calculating a ratio, we obtain a dimensionless value that can be compared between the different hemodynamic measures tested.

4 Z.I. Whinnett et al. / International Journal of Cardiology 129 (2008) height and weight and these were used to estimate the three elements of the aortic impedance model (based on population averages). This method has previously been validated with thermodilution measurements of cardiac output. We used only non-invasive measurements in this study and therefore did not calibrate cardiac output with invasive thermodilution measures. We also derived beat-bybeat pulse and mean arterial from the trace acquired from the Finometer. An ECG signal was also recorded using the Hewlett- Packard 78351A. Analog output feeds of these signals were taken via a National instruments DAQ-Card AI-16E-4 (National Instruments, Austin, TX) and acquired using Labview (National Instruments, Austin, TX). They were analysed off line with custom software based on the Matlab platform (MathWorks, Natick, MA) [21]. The cardiac output signal is released by the Finometer with a one second time delay, which we removed by post-processing prior to analysis Measurement of relative change in hemodynamic measures across different AV delays Hemodynamic measures were recorded simultaneously during adjustment of the AV delay of the biventricular pacemaker, while pacing atrially at a rate of 110 bpm. Testing was performed at this elevated heart rate because previous work from this unit has demonstrated that the hemodynamic peak for AV delay is clearer when heart rate is elevated [11]. Since the aim of this study was to compare different hemodynamic measures, we performed testing at a heart rate known to generate unambiguous hemodynamic differences which also resembles the heart rate at which heart failure symptoms become limiting in day-to-day life in such patients. However, in current clinical practice, those centres that do routine optimization perform it at resting heart rate. We therefore performed a substudy to examine whether the results obtained at higher heart rates were also applicable to optimization carried out at resting heart rates. In order to minimize the effects of background variation in the hemodynamic trace, we applied our algorithm [11] comparing each tested AV delay with a fixed reference AV delay (120 ms). The VV delay was left at 0 ms (or as close as was possible to 0 ms) while the AV delay was adjusted. We calculated the relative change in the measured hemodynamic parameter by comparing the mean of the 10 beats immediately after a transition with the 10 beats immediately before. We repeated the transitions, reversing the signs of the tested hemodynamic measure for reverse transitions, to obtain at least 6 replicate measurements for each tested AV delay (Fig. 1). We combined these to obtain, for each tested delay, a mean relative change in hemodynamic measure (SBP rel, DBP rel, MAP rel,pp rel,co rel, SVR rel ). The relative change in hemodynamic parameter was measured in the manner described above for a series of different paced and sensed AV delays (40, 80, 120, 140, 160, 240, 280 and 320 ms with the protocol stopped when intrinsic conduction occurred). The optimal AV delay was identified as that corresponding with the maximal value of the hemodynamic measure tested. The maximal value was identified by applying a quadratic equation to the hemodynamic results from the range of AV delays tested, with the maximal value taken as that corresponding to the peak of the parabola. We have previously demonstrated that the curve of response to changes in a range of AV delays fits extremely closely to a parabola [22]. We calculated the signal-to-noise ratio for each tested hemodynamic parameter (Fig. 2) Statistics The SBP rel value was determined for each tested AV delay in relation to a reference AV delay (120 ms) and VV delay (0 ms) by taking the mean of observed changes from at least 6 individual transitions. Paired comparisons were made using Student's paired t test. A p value of b 0.05 was taken as statistically significant. The statistical package Statview 5.0 (SAS Institute Inc., Cary, Table 1 Comparison of signal-to-noise ratio for the tested hemodynamic measures, at a heart rate of 110 bpm Patient number Signal-to-noise ratio for detection of AV timing optimum systolic diastolic cardiac output mean arterial pulse Mean Standard deviation p versus systolic b

5 220 Z.I. Whinnett et al. / International Journal of Cardiology 129 (2008) NC) was used for analysis. We used linear regression and the Bland Altman method to compare the optimal AV delays obtained using the different tested hemodynamic measures. 3. Results 3.1. Comparison of signal-to-noise ratio of hemodynamic measures The signal-to-noise ratios of the hemodynamic measures tested were compared in order to determine whether they were equivalent in terms of signal quality (Table 1). Systolic, pulse and cardiac output all had essentially the same signal-to-noise ratio (15.4, 15.5, 15.3 respectively p=ns for all three pairwise comparisons). In contrast mean arterial and diastolic were significantly worse in terms of signal-to-noise ratio in comparison with SBP (14.2, p = and 12.1, p b respectively). The individual signal-to-noise ratio for each patient and each hemodynamic parameter are displayed in Table 1. Values of signal-to-noise ratio are displayed for each tested hemodynamic measure. The signal-to-noise ratio was equivalent for systolic, pulse and cardiac output, whereas mean arterial and diastolic had poorer signal-to-noise ratios Comparison of the optimal AV delay identified using the different hemodynamic measures tested It was possible to identify an individual AV delay which corresponded to the maximal acute hemodynamic measure being tested in all patients. There were close similarities between the hemodynamically optimal AV delay identified using SBP, PP, MAP and DBP, as shown in Fig. 3. For example the optima identified by SBP correlated strongly with those identified by PP Fig. 3. Correlation of the optimal AV delay determined using the tested hemodynamic parameters, at a heart rate of 110 bpm. The selected optimal delay is correlated between individual hemodynamic measures. Systolic (SBP), pulse (PP), mean arterial (MAP) and diastolic (DBP) all correlate well with one another. Cardiac output (CO) does not correlate well with the other hemodynamic parameters. The line of identity is shown (dotted line), as well as the mean difference (Mean Diff) and standard deviation of the difference (SDD).

6 Z.I. Whinnett et al. / International Journal of Cardiology 129 (2008) Fig. 4. Reproducibility of the AV delay identified as optimal, by each of the methods. Visit 2 occurred after a mean interval of 6 months after the first. The peak AV delay identified using SBP showed the best correlation between the two visits, r=0.80. The mean difference (Mean Diff) and standard deviation of the difference (SDD) are also displayed, along with the line of identity (dotted). (r=0.94), MAP (r=0.96) and DBP (r=0.90). In contrast, there was a poor correlation of the optima between SBP and cardiac output (r = 0.36, Fig. 3). We also performed comparisons using the Bland Altman approach, whose numerical results are given within Fig Reproducibility Eleven patients took part in the reproducibility substudy, returning for repeat testing after a mean of 6 months. Only patients who were clinically stable were included into the reproducibility study. The AV and VV delays were programmed back to their original settings after the first visit. The AV delays determined as optimal for the tested parameters at the two visits were compared. Systolic showed the best correlation between the first and second visits (r = 0.80) followed by mean arterial (r=0.80) and pulse (r=0.50) as shown in Fig. 4. The mean absolute difference (±standard deviation of the difference) between the optimal AV delay identified between the two visits was 2.5 ms (±12.4 ms) for SBP, 2.5 ms (±18.8 ms) for DBP, 5.1 ms (±14.0 ms) for MAP, 4.2 ms (±18.3 ms) for PP and 9.4 ms (±31.6 ms) for CO Optimization at resting heart rate We examined the effect of heart rate on the efficiency and ability to detect a peak AV delay, in a substudy of 18 patients. Values of signal-to-noise ratio are displayed in Table 2 for each tested hemodynamic measure. At resting heart rate, the signal-to-noise ratio was lower than that measured during optimization at higher heart rate, for each hemodynamic parameter. At resting heart rate, the signal-to-noise ratio was equivalent for all the tested parameters. Identification of an optimal hemodynamic AV delay was more difficult at resting heart rate because the hemodynamic peak was often less curved than with elevated heart rate. Again there were close similarities between the hemodynamically optimal AV delay identified using SBP, MAP, PP and DBP, as shown in Fig. 5. For example the optima identified by SBP correlated well with those identified by PP (r=0.81), MAP (r=0.75) and DBP (r=0.67). The correlation between SBP and cardiac output was fairly high (r = 0.65), but the correlation of cardiac output with the other measures was less good (Fig. 5). Overall the correlation between the optima identified by different hemodynamic markers was not as strong at lower higher heart rates compared with higher rate. We also performed comparison using the Bland Altman approach as displayed in Fig Discussion In this study we have found that it is possible to identify a hemodynamically optimal AV delay for cardiac resynchronization therapy with any one of a number of beat-by-beat measures: systolic, mean arterial,

7 222 Z.I. Whinnett et al. / International Journal of Cardiology 129 (2008) Table 2 Comparison of signal-to-noise ratio for the tested hemodynamic measures with optimization performed at resting heart rate Patient number Signal-to-noise ratio for detection of AV timing optimum systolic diastolic cardiac output mean arterial pulse Mean Standard deviation p versus systolic pulse, diastolic and non-invasively estimated cardiac output. Signal quality is essentially the same for SBP, PP and CO, whereas for MAP and DBP it is less good. The AV delay identified as optimal is very similar between SBP, PP, MAP and DBP. In contrast, the AV delay identified using CO is poorly related to those optima. Reproducibility is best for optima detected by SBP and MAP Mechanism of change in cardiac function with change in AV delay With changes in AV delay, the timing of passive and active filling are likely to be major factors determining the changes in cardiac function and resultant change in stroke volume. If AV delay is programmed to too short a value, the contribution to ventricular filling is curtailed by the onset of ventricular systole which results in a fall in stroke volume. At the other extreme, if too long an AV delay is programmed then the latter part of diastole is wasted. Attempts to use echocardiography to determine the optimum relationship between these two filling patterns are difficult, particularly at higher heart rates because the two components of filling merge and analysis is inevitably subjective. The use of an outcome measure with an easily defined endpoint, such as SBP, allows the conditions under which the best filling relationship occurs to be more reliably defined Comparison of signal-to-noise ratio The signal quality of systolic, cardiac output and pulse were all similar, whereas diastolic and mean arterial had a significantly lower signal-to-noise ratio. Systolic, cardiac output and pulse therefore appear to be more sensitive to the changes in cardiac function associated with AV delay optimization. Systolic shows a proportionally greater increase compared with diastolic for a given increase in stroke volume, this results in increased sensitivity of this measure to changes in AV delay. In effect this allows a magnification in the response to a given change in AV delay. This difference in response to changes in AV delay settings for SBP and DBP can be explained by the differing effects MAP and PP have on these parameters. An acute increase in stroke volume effects systemic hemodynamics in two ways: it increases mean arterial and also increases pulse. Each increase raises SBP. In effect, SBP has two reasons to increase. In contrast, DBP has one reason to increase (the rise in MAP) and one to decrease (the rise in PP). Hence we would expect SBP to be more sensitive to changes in cardiac function compared with DBP. Pulse also shows a proportionally larger increase in for a given change in stroke volume compared with diastolic. This is likely to explain the greater sensitivity to changes in AV delay for pulse when compared with diastolic. Cardiac output is a direct measure of stroke volume when the heart rate is fixed (as was the case in this study) therefore it would be expected to provide a sensitive measure of changes in stroke volume and this is reflected by the high signal-to-noise ratios Comparison of identified optimal AV delay Reassuringly almost all the hemodynamic parameters tested showed good agreement between the AV delays they selected as optimal. The exception was cardiac output, an unexpected finding. There are a number of potential explanations for the discordant behaviour of cardiac output. One explanation might be that there is a genuine difference physiologically in the effect of changing AV delay, between cardiac output and the other tested hemodynamic parameters. But, before this finding is taken at face value, we must consider the reproducibility of optimizing by cardiac output, because any measure of a physiological variable with poor reproducibility cannot be expected to correlate with a different variable any better than it correlates with itself [23]. We believe that the poor reproducibility of

8 Z.I. Whinnett et al. / International Journal of Cardiology 129 (2008) Fig. 5. Correlation of the optimal AV delay determined using the tested hemodynamic parameters at resting heart rate. The selected optimal delay is correlated between individual hemodynamic measures. Systolic (SBP), pulse (PP), mean arterial (MAP) and diastolic (DBP) all correlate reasonably well with one another, although not as strongly as at a higher heart rate. Cardiac output (CO) does not correlate as well with the other hemodynamic parameters. The line of identity is shown (dotted line), as well as the mean difference (Mean Diff) and standard deviation of the difference (SDD). the CO optimum is the most likely explanation for the lack of correlation of cardiac output with the other hemodynamic variables. Secondly, there are a number of potential sources of error in the cardiac output measure we have used. The Modelflow method is a non-linear model which uses measures derived from population averages for sex, age, height and weight in order to predict aortic compliance at a given. It may be that these population averages are not accurate for our patient population who have impaired LV function and may also have aortic abnormalities. This may reduce the accuracy of this technique for assessing cardiac output. The Modelflow method has been reported to work best when it is calibrated to an invasive measure of cardiac output [24]. However, it is not practical to calibrate invasively during routine outpatient optimization visits. Therefore our study applied the clinically achievable strategy of using uncalibrated cardiac output measures. This may account for the lack of sensitivity when using this method for identifying optimal AV delays, as there may be relatively small hemodynamic changes around the peak AV delay Reproducibility Systolic is best in terms of reproducibility followed by MAP and PP. The error within the cardiac output measure used appears to be too high to be able to recommend it as a reproducible measure for optimization of the AV delay of cardiac resynchronization therapy. The other measures are less good than SBP but similar to each other in terms of reproducibility. The Bland Altman data (Fig. 4) show that the variability between sessions (SDD) is not very different between pulse (18.3 ms), mean arterial (14.0) and diastolic (18.8 ms). It is the systolic which has the least variability between episodes (12.4 ms). The time between measures was relatively long (mean 6 months) by comparison with previous assessments of reproducibility (mean 3 months) [11]. However, patients were enrolled into the study some time after implantation (median 17 months, range 3 30 months) it would therefore be anticipated that much of the remodelling observed acutely after implantation will have occurred prior to patients being

9 224 Z.I. Whinnett et al. / International Journal of Cardiology 129 (2008) entered into the study. Additionally only patients who had remained clinically stable were enrolled into the reproducibility substudy. Moreover, the optima detected by systolic, diastolic and mean arterial remained consistent between visits which makes it unlikely that there had been unnoticed underlying changes in clinical state. Cardiac output had a high signal-to-noise ratio but poor reproducibility. We speculate that this is because the Modelflow method of CO calculation is a complex analysis of the waveform, which may be affected by features that vary over time (over several months) that distort the 6-month data, in a way that does not happen with the simpler measures. It therefore appears that cardiac output and pulse may well have intrinsically poorer suitability for optimization Optimization at resting heart rate At resting heart rate, we found that the numerical value of signal-to-noise ratio is lower for all the tested hemodynamic measures compared to optimization performed at higher heart rates. This is a result of the smaller range observed in the hemodynamic measure at lower heart rates in comparison with higher heart rates [11]. The lack of a significant difference between signal-tonoise ratios of different measures at lower heart rates is likely to be due to the overall smaller numerical value of the signalto-noise ratio. Indeed both diastolic and mean arterial, which had significantly worse signal-tonoise ratios at higher heart rates, showed non-significant trends towards worse signal-to-noise ratio at resting heart rate. In selecting the optimal AV delay, SBP, DBP, PP and MAP all appear to correlate reasonably well at resting heart rate. However, the correlation is not as strong at rest measurements as it is at higher heart rates. Why is the hemodynamic peak more pronounced at higher heart rates? We believe this can be physiologically explained by the reduced time within the cardiac cycle at higher heart rates and therefore less opportunity for the heart to compensate for non-optimal AV delay settings. In principle resynchronization has two important effects on myocardial activity. First, it makes contraction more synchronous, and therefore systole more efficient. Second, more time becomes available for left ventricular filling. For many stable ambulatory outpatients, at rest, this improvement in filling is less significant than at higher heart rates, because filling time may already be sufficient. However, at higher heart rates, filling time may become a more limiting factor, and therefore selecting the ideal AV delay may have a greater beneficial impact on cardiac function and thus hemodynamics. While current standard clinical practice is for optimization to be carried out with resting heart rate, patients typically become symptomatic at higher rates associated with exercise. Therefore a knowledge of the properties of these hemodynamic variables at higher heart rate could prove clinically useful in developing optimization strategies. It may be preferable to perform optimization at a series of different heart rates. This would require pacemaker companies to modify their devices to allow the AV delay to be programmed for individual heart rates Which hemodynamic parameter should be used in clinical practice? It was possible to identify a hemodynamic optimal AV delay by performing optimization using any of the hemodynamic measures tested in this study. However, systolic appears to give the best combination of properties. It has a high signal-to-noise ratio, generates optima that agree well with most of the other hemodynamic measures, and is highly reproducible. Mean arterial is probably the next best of the hemodynamic measures tested. It correlates well with the other tested hemodynamic measures, is highly reproducible, and has a respectable signal-to-noise ratio. Pulse has a good signal-to-noise ratio, and selects very similar values for optimal AV delay to the other hemodynamic markers, and has fairly good reproducibility. Diastolic has a lower signal-to-noise ratio than for the other tested hemodynamic measures. But never less correlates well with the other hemodynamic measures when selecting the optimal AV delay, and has fairly good reproducibility. The measure of cardiac output used in this study (Modelflow), calculated from non-invasive finger, does not seem to be a good measure for AV delay optimization. While its signal-to-noise ratio is high, it has poor reproducibility and does not correlate well with any of the other tested hemodynamic measures used in this study in identification of an optimal AV delay. We would like to make it clear that in this study we only measured cardiac output using the Modelflow method. We can therefore only conclude that this particular measure and not cardiac output in general is imperfect when used for optimization. In this study we investigated only the effect of changing AV delay. In principle the relationship observed should be the same for VV delay. We did not test this specifically as similar albeit smaller acute effects on hemodynamics are known to occur Study limitations It is potentially a limitation of this study that it used only hemodynamics measured non-invasively from the finger: there was no comparison with invasive hemodynamics. However, while the absolute values for individual hemodynamic measures are likely to vary between the sites at which they are recorded, the algorithm we use for AV delay

10 Z.I. Whinnett et al. / International Journal of Cardiology 129 (2008) optimization works by calculating only the relative change in the tested hemodynamic parameter and is therefore in principle independent of the absolute recorded value of the hemodynamic measure. All the tested hemodynamic measures are recorded from the same site at the same time which should make the comparison fair. Non-invasive hemodynamics measured using the Finometer device have previously been validated as being good at tracking acute changes in hemodynamics when compared with invasive measures [18,19]. We did not make a direct comparison with echocardiographic measures. This was because we have previously investigated the reproducibility of echocardiographic measures during AV delay optimization and have found them to be so poor as to preclude useful comparison with hemodynamics [11]. Since echocardiographic optimization does not agree well with itself, a failure to agree with any other measures would give no useful information. This finding does not conflict with the importance of echocardiography as a means for identifying patients who may benefit from cardiac resynchronization [25,26]. In addition echocardiographic methods are routinely carried out at rest with resting heart rates. This study is not designed to demonstrate whether one choice of hemodynamic parameters for optimization gives better clinical symptomatic outcome for the patients than another. Rather, it aims to compare the intrinsic suitability of each of these parameters for use in studies where efficiency, agreement between methods, and reproducibility are important considerations. The main part of the study was performed while pacing at higher heart rates. We did this because the higher signalto-noise ratio at these rates allowed comparison between the different measures to be made more readily. The optima identified at higher heart rates were different from those identified during atrial sensing at resting heart rate. It is not known which setting would provide a better outcome for the patients (optimization at resting heart rates or higher heart rates) and this information cannot be obtained from this type of study. The aim of this study was to compare the different hemodynamic markers in terms of efficiency and reproducibility. 5. Conclusions It is possible to perform optimization of the AV delay of cardiac resynchronization therapy using a number of different non-invasively measured hemodynamic measures. The same AV optimum is identified through hemodynamic optimization, regardless of whether the parameter chosen for maximization is SBP, PP or MAP. There is equally high efficiency of detection, for SBP, PP and CO. Reproducibility is best for SBP, followed by MAP and PP. We conclude that SBP appears to be the most suitable measure for detecting the hemodynamic optimal AV delay, with PP and MAP being reasonable alternatives. References [1] Yu Y, Kramer A, Spinelli J, Ding J, Hoersch W, Auricchio A. Biventricular mechanical asynchrony predicts hemodynamic effect of uni- and biventricular pacing. Am J Physiol Heart Circ Physiol 2003;285: H [2] Butter CD, Auricchio A, Stellbrink C, et al. Effect of resynchronization therapy stimulation site on the systolic function of heart failure patients. Circulation 2001;104: [3] Breithardt O, Stellbrink C, Franke A, et al. Acute effects of cardiac resynchronization therapy on left ventricular Doppler indices in patients with congestive heart failure. Am Heart J 2002;143: [4] Auricchio A, Ding J, Spinelli JC, et al. Cardiac resynchronization therapy restores optimal atrioventricular mechanical timing in heart failure patients with ventricular conduction delay. J Am Coll Cardiol 2002;108: [5] Van Gelder BM, Bracke FA, Meijer A, Lakerveld LJ, Pijls NH. Effect of optimizing the VV interval on left ventricular contractility in cardiac resynchronisation therapy. Am J Cardiol 2004;93: [6] Nelson GS, Berger RD, Fetics BJ, et al. Left ventricular or biventricular pacing improves cardiac function at diminished energy cost in patients with dilated cardiomyopathy and left bundle-branch block. Circulation 2000;102(25): [7] Kass DA, Chen CH, Curry C, et al. Improved left ventricular mechanics from acute VDD pacing in patients with dilated cardiomyopathy and ventricular conduction delay. Circulation 1999;99: [8] Blanc JJ, Etienne Y, Gilard M, et al. Evaluation of different ventricular pacing sites in patients with severe heart failure: results of an acute hemodynamic study. Circulation 1997;96: [9] Auricchio A, Stellbrink C, Block M, et al. Effect of pacing chamber and atrioventricular delay on acute systolic function of paced patients with congestive heart failure. Circulation 1999;99: [10] Sawhney NS, Waggoner AD, Garhwal S, Chawla MK, Osborn J, Faddis MN. Randomized prospective trial of atrioventricular delay programming for cardiac resynchronization therapy. Heart Rhythm 2004;5: [11] Whinnett ZI, Davies JER, Willson K, et al. Determination of optimal atrioventricular delay for cardiac resynchronization therapy using acute non-invasive. Europace 2006;8: [12] Perego GB, Chianca R, Facchini M, et al. Simultaneous vs sequential biventricular pacing in dilated cardiomyopathy: an acute hemodynamic study. Eur J Heart Fail 2003: [13] Porciani MC, Dondina C, Macioce R, et al. Echocardiographic examination of atrioventricular and interventricular delay optimization in cardiac resynchronization therapy. Am J Cardiol 2005;95: [14] Peňa z J. Photoelectric measurement of, volume and flow in the finger. Digest of the international Conference on Medicine and Biological Engineering, Dresden; p [abstract]. [15] Wessling KH, De Wit B, Van der Hoeven GM, Van Goudover J, Settels J. Physiocal calibrating finger vascular physiology for Finapres. Homeostasis 1995;36: [16] Smith NT, Wesseling KH, dewit B. Evaluation of two prototype devices producing non-invasive, pulsatile, calibrated measurement from a finger. J Clin Monit 1985;1: [17] Van Egmond J, Hasenbos M, Crul JF. Invasive versus non-invasive measurement of arterial. Comparison of two automatic methods and simultaneously measures direct intraarterial. Br J Anaesth 1985;57: [18] Petersen ME, Williams TR, Sutton R. A comparison of non-invasive continuous finger measurement (Finapres) with intraarterial during prolonged head-up tilt. Eur Heart J 1995;16: [19] Jellema WT, Imholz PM, van Goudoever J, Wesseling KH, van Lieshout J. Finger arterial versus intrabrachial and continuous cardiac output during head-up tilt testing in healthy subjects. Clin Sci 1996;91:

11 226 Z.I. Whinnett et al. / International Journal of Cardiology 129 (2008) [20] Imholz BPM, Weiling W, Montfrans GA, Wesseling KH. Fifteen years experience with finger arterial monitoring: assessment of the technology. Cardiovasc Res 1998;38: [21] Davies LC, Francis DP, Jurak P, Kara T, Piepoli M, Coats AJS. Reproducibility of methods for assessing baroreflex sensitivity in normal controls and in patients with chronic heart failure. Clin Sci (Lond) 1999;97: [22] Whinnett ZI, Davies JER, Willson K, et al. Haemodynamic effects of changes in AV and VV delay in cardiac resynchronisation therapy show a consistent pattern: analysis of shape, magnitude and relative importance of AV and VV delay. Heart 2006;92: [23] Francis DP, Coats AJ, Gibson DG. How high can a correlation coefficient be? Effects of limited reproducibility on common cardiological measures. Int J Cardiol 1999;69: [24] Remmen JJ, Wim RM, Aengevaeren RM, et al. Finapres arterial pulse wave analysis with Modelflow is not a reliable non-invasive method for assessment of cardiac output. Clin Sci 2002;103: [25] Whinnett ZI, Davies JER, Lane R, Francis DP, Mayet J. Echocardiographic methods for selecting patients suitable for biventricular pacing therapy. Minerva Cardioangiol 2005;53(3): [26] Yu CM, Chau E, Sanderson JE, et al. Tissue Doppler echocardiographic evidence of reverse remodelling and improved synchronicity by simultaneously delaying regional contraction after biventricular pacing therapy in heart failure. Circulation 2002;105: