Measurement Precision in the Optimization of Cardiac Resynchronization TherapyClinical Perspective
Jump to
Abstract
Background— Cardiac resynchronization therapy improves morbidity and mortality in appropriately selected patients. Whether atrioventricular (AV) and interventricular (VV) pacing interval optimization confers further clinical improvement remains unclear. A variety of techniques are used to estimate optimum AV/VV intervals; however, the precision of their estimates and the ramifications of an imprecise estimate have not been characterized previously.
Methods and Results— An objective methodology for quantifying the precision of estimated optimum AV/VV intervals was developed, allowing physiologic effects to be distinguished from measurement variability. Optimization using multiple conventional techniques was conducted in individual sessions with 20 patients. Measures of stroke volume and dyssynchrony were obtained using impedance cardiography and echocardiographic methods, specifically, aortic velocitytime integral, mitral velocitytime integral, Awave truncation, and septalposterior wall motion delay. Echocardiographic methods yielded statistically insignificant data in the majority of patients (62%–82%). In contrast, impedance cardiography yielded statistically significant results in 84% and 75% of patients for AV and VV interval optimization, respectively. Individual cases demonstrated that accepting a plausible but statistically insignificant estimated optimum AV or VV interval can result in worse cardiac function than default values.
Conclusions— Consideration of statistical significance is critical for validating clinical optimization data in individual patients and for comparing competing optimization techniques. Accepting an estimated optimum without knowledge of its precision can result in worse cardiac function than default settings and a misinterpretation of observed changes over time. In this study, only impedance cardiography yielded statistically significant AV and VV interval optimization data in the majority of patients.
 cardiac resynchronization therapy
 pacing optimization
 AV delay
 interventricular interval
 echocardiography
 impedance cardiography
Clinical Perspective on p 404
Cardiac resynchronization therapy (CRT) decreases morbidity and mortality in populations with reduced cardiac function and conduction abnormalities^{1–3}; however, approximately one third of these patients do not experience benefit.^{4} Atrioventricular (AV) and interventricular (VV) pacing interval optimization improve hemodynamics acutely^{5–7} and may enhance the response to CRT relative to default interval settings among both the traditional responder and nonresponder groups. To date, however, results of prospective randomized trials have been mixed.^{8–12}
The clarification of fundamental issues in pacing interval optimization is still at an early stage. For example, the extent to which optimum AV/VV intervals change with body position and exertion remains unclear,^{13–16} and data assessing the stability of optima over time have been inconsistent.^{17,18} Even the definitions of pacing intervals vary among manufacturers, with identical programmed intervals corresponding to very different ventricular pace timing.^{19} Perhaps the most important fundamental weakness is the lack of statistical tools to quantitatively evaluate the significance of the measured data and to characterize the precision of the estimated optimum AV/VV interval.^{11,20–22}
In this study, we examined multiple commonly used AV/VV interval optimization techniques. Central to our approach is the recognition that the underlying dependence of cardiac function on AV/VV interval is obscured to some degree by measurement noise and that the optimum interval identified by a given technique is in fact an estimation of the true physiologic optimum. The degree to which measured optimization data demonstrate a significant dependence on pacing interval and the precision of estimated optima were rigorously evaluated using new statistical tools based on bootstrapping, a computational approach that makes knowledge of the underlying statistical properties of the data unnecessary.^{23}
Methods
Patient Selection
Data were obtained from consecutive patients referred for clinical pacing interval optimization with institutional review board approval. Patients with dualchamber or biventricular pacemakers or implantable cardioverter defibrillators were included without regard to underlying etiology (Table 1).
Optimization Techniques
AV/VV interval optimization was conducted using multiple techniques, as follows:
Noninvasive, beattobeat estimates of stroke volume (SV) were acquired continuously using impedance cardiography (ICG; BioZ, CardioDynamics, San Diego, Calif) (Figure 1A).^{24–26} With ICG, changes in thoracic impedance are measured using surface electrodes and then processed by a proprietary algorithm to estimate SV and other hemodynamic parameters. Each test interval was delivered for 60 seconds, with all beats recorded during the last 30 seconds included in the analysis.
The remaining techniques were derived from echocardiography using a Philips iE33 System (Philips International B.V., Amsterdam, The Netherlands). The aortic velocitytime integral (AVTI), which is directly proportional to SV, was obtained by numerically integrating the ejection velocity envelop obtained by continuouswave Doppler directed in line with aortic flow in the apical 5chamber view (Figure 1B).^{20,21} For this and the other echocardiographic techniques, a 10 to 20second equilibration period followed each programming change. Data then were recorded over 1 to 2 respiratory cycles, with premature and postpremature beats excluded.
Mitral inflow velocitytime integral (MVTI), which is directly proportional to inflow volume, was obtained using pulsedwave Doppler with the sample volume placed just apical to the mitral leaflets in the apical 4chamber view (Figure 1C).^{20,21}
In contrast to the techniques described above, which attempt to characterize cardiac function by estimates of forward flow, septalposterior wallmotion delay (SPWMD) provides an assessment of left ventricular (LV) mechanical synchrony.^{20} As shown in Figure 1D, the transducer signal was directed across the septum and the posterior LV wall in the parasternal longaxis view. Color Mmode Doppler was used to highlight the relative motion of the 2 walls, with minimum lag taken to represent optimal ventricular synchrony.
Awave truncation identifies the optimum AV delay as the shortest pacing interval that avoids truncation of the Awave^{21} and is based on the same Doppler waveform as MVTI. Figure 1E shows an Awave truncated at 150 milliseconds and untruncated at 180 milliseconds.
Optimization techniques were compared in terms of their ability to detect underlying physiologic changes with pacing interval. Specifically, the statistical significance of the data was quantitatively estimated, as described below. Because Awave truncation yields a binary assessment at each test interval (A wave is or is not truncated), it is not amenable to the analytic paradigm used for the other techniques. Therefore, for Awave truncation, we report the number of times each of the 3 readers, blinded to other results, was able to estimate an optimum pacing interval.
Summary statistics were obtained based on all patients referred for clinical optimization and, in addition, with hypertrophic cardiomyopathy patients excluded. Because ICG allows a greater number of data points to be obtained compared with echocardiographic methods, the AV interval optimization analysis was repeated for ICG data using the same number of measurements that were obtained in the corresponding AVTI data set at each test interval.
Statistical Methods
A thirddegree polynomial was fit to the data, and the location of the maximum was taken as the estimated optimum interval. For SPWMD, the location of the minimum absolute value of the polynomial was taken as the optimum. Use of a continuous function, such as a polynomial, allows interpolation between test intervals and averaging both at and across test intervals provided that the number of free parameters of the function is smaller than the number of unique test intervals.
The test for statistical significance was based on the formulation of the alternative hypothesis that the measured data do not depend on pacing interval. The probability of obtaining the observed data under this null hypothesis was estimated using bootstrapping.^{22,23} A test statistic s was defined to be equal to the area bounded above by the bestfitting polynomial and below by the minimum value of the polynomial, as illustrated in Figure 2. The test statistic is not unique; other measures also would be acceptable if they possess the property that their value varies depending on how “physiological” the optimization data are. Specifically, if an underlying physiologic optimum exists, and if the range of test intervals was appropriately selected to span the optimum, then the test statistic should be larger than what would be obtained under the null hypothesis in which there is no dependence on pacing interval. The greater the difference in cardiac function at the optimum pacing interval and the extreme of the test range, the greater the value of s. For each optimization data set, 1000 surrogate bootstrapped data sets were generated by randomly selecting data points from the original data set with replacement and without regard to test interval. This process yields surrogate data sets with the same number of data points at each test interval as the original but replaces the original mapping between measured data and test interval with a random pairing. Each surrogate bootstrapped data set thus represents a single realization of data that would be observed under the null hypothesis while preserving the amplitude statistics of the original data. The test statistic s′ was calculated for each surrogate data set, and the fraction that was greater than or equal to that of the original data estimates the probability that a test statistic at least as large as that associated with the original data would be observed under the null hypothesis. For P≤0.05, the null hypothesis was rejected, and the data were interpreted as demonstrating a statistically significant dependence on pacing interval.
The 95% CI of the estimated optimum pacing interval was obtained by generating an additional 1000 bootstrapped data sets in which the original data were randomly selected with replacement while preserving the mapping to test interval. In this case, with the mapping between test interval and measured data retained, the collection of bestfitting polynomials of the bootstrapped data sets reflects the variability in the bestfitting polynomial of the original data that is attributable to the statistical variability in the measurements. The locations of the maxima of the surrogate data sets were ordered from smallest to largest, and the cutoff point of the smallest and largest 2.5% of the surrogate optima were taken as the 95% CI of optimum estimated from the original data.
Assessment of AVTI Variability
The variability of the measured AVTI data was quantified by calculating the average and SD over all measurements made at an AV delay of 120 ms, which by protocol were acquired over at least 1 respiratory cycle at 4 different times during the recording session. To allow comparison with previously published values, a transformation was derived that converts the coefficient of variation (defined as SD divided by mean) to the average difference of successive measurements. Specifically, μ_{y}=2CV
Results
Patient demographics are presented in Table 1. Of the 20 sequential, unique patients referred for clinical pacing optimization, 18 had biventricular pacemakers or implantable cardioverter defibrillators and 2 had dualchamber devices. The LV lead in 1 patient failed to capture, and another patient with chronic atrial fibrillation had a biventricular device without an atrial lead. Patients who underwent AV delay optimization were in sinus rhythm with the AV delay initiated by an atrial sensedevent in all cases.
AV delay optimization data from a single patient are presented in Figure 3. As with the ICG data shown here (left panel), when statistically significant data were obtained they typically exhibited an inverted U appearance, indicating that the estimated optimum interval was in the interior of the range of test intervals. Statistically insignificant data typically had a bestfitting polynomial that was flat relative to the intrinsic variability of the data.
The average and SD of the measured AVTI data were calculated over all beats (10 to 31, median 20) recorded from each patient during AV optimization at the 120ms test interval. The ranges of the perpatient average and SD of the AVTI data were 6.7 to 56 cm and 0.73 to 5.1 cm, respectively. The median sample coefficient of variation was 0.075, which corresponds to an average normalized difference in successive measurements of μ_{y}=0.12, consistent with previous reports of μ_{y}=0.1±0.1.^{27}
The VV optimization data shown in Figure 4 are from the same patient whose AV optimization results are presented in Figure 3. As with this individual, for most patients, VV optimization yielded insignificant results more frequently than did AV optimization.
The number of statistically significant data sets is summarized in Table 3. For both AV and VV optimization, AVTI, MVTI, and SPWMD yielded data that were indistinguishable from the null hypothesis (ie, failed to demonstrate a statistically significant dependence on pacing interval) the majority of the time. As shown in Table 3, this finding remained true whether patients with hypertrophic cardiomyopathy were included or excluded from the analysis. ICG was significantly different from the null hypothesis 84% of the time for AV delay optimization and 75% of the time for VV interval optimization. Excluding patients with hypertrophic cardiomyopathy from the analysis, the null hypothesis could be rejected 81% and 75% of the time, respectively.
Among the statistically significant data sets, the optimum AV delay estimated by ICG differed from the default value (120 ms) by an average magnitude of 57 ms, and 63% of the estimates differed from default by at least 50 ms. Among the statistically significant, ICGderived VV interval data sets, estimated optima differed from the default value (0 ms) by an average of 52 ms, and 75% of the estimates differed by at least 30 ms. Repeating the ICG AV interval optimization analysis using the same number of measurements at each test interval as the AVTI recordings continued to yield statistically significant results in the majority of patients, with 81% of the reduced ICG data sets having P values <0.05.
An optimum AV delay could be estimated by 3 independent readers using Awave truncation 69%, 75%, and 94% of the time. The estimates of each reader are compared with ICG results in Table 4. The relationship between optima predicted by ICG and Awave truncation was variable, with the Pearson correlation coefficient ranging from 0.02 to 0.67 for the 3 readers.
Although the analysis presented above is based on unique patient visits, 1 patient underwent both ICG and echocardiographic optimization on 2 separate occasions separated by 3.5 months. As shown in Figure 5, in both cases ICG yielded similarappearing, statistically significant results. The estimates of the optimum pacing intervals were precise, with narrow 95% CIs, and had good concordance, with 91 ms at the first optimization and 67 ms 3.5 months later. In contrast, in neither optimization session did AVTI yield statistically significant results. The wide 95% CI of the initial AVTI optimum predicted that subsequent optimization attempts would be unlikely to identify a similar optimum interval.
Discussion
In this study, multiple clinically accepted AV and VV interval optimization techniques were used in single sessions in a heterogeneous group of 20 patients. For 62% to 86% of patients, AVTI, MVTI, and SPWMD yielded data that were statistically indistinguishable from the null hypothesis; that is, the majority of the time these measures yielded data that did not show a significant dependence on AV or VV interval. With Awave truncation, it was possible for 3 independent readers to estimate an optimum AV delay 69%, 75%, and 94% of the time. ICG performed better than the echocardiography methods, yielding statistically significant data 84% of the time for AV delay optimization and 75% of the time for VV interval optimization. Notably, the echocardiography techniques tested here represent the most commonly used approaches to AV/VV interval optimization.^{28}
In previous studies, SPWMD failed to predict a response to CRT,^{29,30} and interval optimization using AVTI neither improved clinical outcomes^{12} nor yielded acute data that were distinguishable from a negative control.^{31} The poor precision of these techniques demonstrated in this study may account for the lack of clinical utility seen in the earlier work.
AV/VV interval optimization traditionally has been conducted without consideration of the intrinsic variability of the measured data. Test intervals often are delivered in a nonrandom order, sweeping systematically from one end of the test range to the other, and the AV or VV interval associated with the best average measure of cardiac function typically is taken at face value to represent the underlying physiologic optimum.^{11,20} Although efficient, the traditional approach has important drawbacks. A nonrandom order of test intervals allows systematic error to be introduced into the measured data in a way that cannot be subsequently corrected by averaging (eg, with subtle drift of the Doppler angle over the course of data collection). Furthermore, without considering the intrinsic variability of the measured data, it is not possible to characterize the precision of the estimated optimum AV/VV interval.
In the absence of knowledge about its precision, an estimated optimum AV/VV interval may in fact be spurious and associated with worse cardiac function than the populationderived, default setting (eg, a sensed AV delay of 120 milliseconds) (Figures 2, 3, and 5). In addition, although multiple studies have suggested that optimum pacing intervals change over time,^{12,17,18} the lack of characterization of the precision of the estimated optima makes it impossible to determine whether the observed change reflects changes in underlying physiology or is simply due to statistical variability in measured data^{31} (Figure 5).
The fact that any optimization technique carries some degree of imprecision requires caution when designing randomized prospective trials. Two electrogrambased optimization techniques sponsored by competing manufacturers are currently undergoing clinical trials. In 1 design,^{32} patients are randomized to the experimental optimization technique or a control arm that may or may not include AV and VV interval optimization. If optimization is performed, it is conducted at the physician's discretion without a requirement to examine the precision of the measured data, raising the possibility of a programmed AV or VV interval that yields worse cardiac function than populationderived default settings. By design, the trial will not address the question of whether the experimental technique is superior to default settings. In contrast, another clinical trial^{33,34} uses a 3arm design in which patients are randomized to the experimental technique, a control arm in which populationderived default settings are used, and a conventional optimization arm in which a specific optimization technique is uniformly used. This study design allows direct comparisons between the experimental technique and both default settings and the specific conventional optimization method, and for comparison of the conventional technique to default settings.
Statistical significance testing may provide a useful way to evaluate the quality of competing optimization techniques. It avoids prespecifying a gold standard, and although demonstration of improved clinical outcomes in welldesigned trials ultimately is required, the approach presented here offers a way to narrow the very wide field of plausible optimization techniques without the resource requirements of a prospective clinical trial.
Although a rigorous, quantitative analysis is desirable, statistical significance can be evaluated informally by plotting the measured data against pacing interval. With test intervals delivered in a random order, if the data exhibit the expected inverted U shape and are tightly clustered about the overall curve, then one can be confident that the estimation of optimum AV/VV interval is precise; repeating the process likely would yield a similar result. On the other hand, if the plot is relatively flat compared with the intrinsic variability of the measured data, then the estimate of the underlying optimum is imprecise and heavily influenced by measurement variability. In this case, repeating the optimization process likely would yield a very different estimated optimum AV/VV interval. The statistical tools used here along with plotting capability have been made available on the Internet.^{35}
In the majority of patients examined in the present study, ICG generated precise estimates and Awave truncation yielded data from which an optimum could be inferred, although often with significant interreader variability and marginal correlation with the ICGpredicted optima. Notably, neither ICG nor Awave truncation has been clinically validated in prospective interval optimization studies. In addition, unanswered fundamental questions include whether the physiologic optimum interval evolves over time and whether it changes between rest and exertion or between supine and upright posture. A study that compares estimated optimum intervals obtained at both supine rest and upright exertion in the same patient, perhaps using motiontolerant ICG, would add important insight into these basic questions. In the absence of such data and given the theoretical potential benefit of pacing optimization, our approach is to accept an estimated optimum interval if quantitative and qualitative analyses suggest that the estimate has good precision. Particular attention should be paid to the effect of outliers and the overall shape of the curve compared with the measurement variability.
That ICG continued to yield statistically significant data in the majority of patients, even when using an identical number of data points as the AVTI analysis, suggests that it has a superior intrinsic signaltonoise ratio and that the acquisition time can be substantially shortened from the 60 s per test interval that was used in this study. The superior noise properties of ICG may be partly due to the automatic and objective nature of data acquisition and analysis in contrast to AVTI, which requires the sonographer to physically hold the probe in a fixed position and the reader to manually demarcate the envelope of the velocity waveform.
A wide variety of optimization techniques have been advocated, including multiple approaches to the assessment of systolic function, diastolic function, and electrical and mechanical synchrony.^{9,19–21,24–26,30,36–40} Although each has a rationale that is mechanistically plausible, consideration of the neurohormonal derangements of heart failure and the therapeutic interventions that have been successful lead us to view SV and its surrogates as parameters that when optimized are most likely to translate into clinical benefit. Specifically, it is now well established that ameliorating the effects of sympathetic tone in these patients leads to improved clinical outcomes.^{41} For a given cardiac output, maximizing SV would minimize sympathetic tone. Indeed, the effects on the neurohormonal system of increased mechanical efficiency may contribute to the salutary effects of CRT, which remains the only contractilityenhancing intervention demonstrated to prolong life.^{42} Theoretical arguments may not account for important effects, however. For example, an increase in SV at the expense of greater oxygen consumption may not benefit the patient with ischemic heart failure. Ultimately, any proposed optimization technique must be validated by randomized prospective trials with hard clinical end points, a goal which to date has not been frequently achieved.^{8–12}
Limitations
The study was based on a small and heterogeneous population comprising successive patients referred for clinical pacing interval optimization. Multiple conventional optimization techniques were examined for their ability to yield statistically significant results. This property is necessary but not sufficient for improved clinical outcomes, which were not examined in this study.
Conclusions
Optimization of AV and interventricular intervals in CRT requires assessment of the variability of the measured data. Accepting an estimated optimum without considering its precision may result in worse cardiac function than default settings in the individual patient and confound results in clinical trials. In the small, heterogeneous pacemaker population examined here, echocardiographic techniques yielded statistically insignificant data in the majority of patients. In contrast, ICG yielded precise estimates of the optimum AV and VV interval in most patients. Further research is necessary to confirm these results, to validate the accuracy of the impedancepredicted optima, and to demonstrate clinical improvement with pacing interval optimization compared to populationderived default settings.
Footnotes

Sources of Funding
Dr Turcott was supported by a Heart Failure Society of America Research Fellowship and by the National Library of Medicine (Biomedical Informatics Training Grant LM 07033). Dr Ashley was supported by the National Institutes of Health (grant K08 HL083914).
Disclosures
Dr Witteles has received honoraria from Medtronic. Dr Wang has received honoraria and research support from and has served as a consultant or advisor to Boston Scientific, Medtronic, and St Jude Medical. Dr Fowler has received honoraria from Medtronic and Boston Scientific.
 Received August 6, 2009.
 Accepted January 20, 2010.
 © 2010 American Heart Association, Inc.
References
 1.↵
 2.↵
 3.↵
 4.↵
 5.↵
 6.↵
 Auricchio A,
 Stellbrink C,
 Block M,
 Sack S,
 Vogt J,
 Bakker P,
 Klein H,
 Kramer A,
 Ding J,
 Salo R,
 Tockman B,
 Pochet T,
 Spinelli J
 7.↵
 Kass DA,
 Chen CH,
 Curry C,
 Talbot M,
 Berger R,
 Fetics B,
 Nevo E
 8.↵
 Boriani G,
 Muller CP,
 Seidl KH,
 Grove R,
 Vogt J,
 Danschel W,
 Schuchert A,
 Djiane P,
 Biffi M,
 Becker T,
 Bailleul C,
 Trappe HJ
 9.↵
 10.↵
 Rao RK,
 Kumar UN,
 Schafer J,
 Viloria E,
 De Lurgio D,
 Foster E
 11.↵
 12.↵
 Boriani G,
 Biffi M,
 Muller CP,
 Seidl KH,
 Grove R,
 Vogt J,
 Danschel W,
 Schuchert A,
 Deharo JC,
 Becker T,
 Boulogne E,
 Trappe HJ
 13.↵
 Ritter P,
 Daubert C,
 Mabo P,
 Descaves C,
 Gouffault J
 14.↵
 Melzer C,
 Korber T,
 Theres H,
 Nienaber CA,
 Baumann G,
 Ismer B
 15.↵
 16.↵
 Mehta D,
 Gilmour S,
 Ward DE,
 Camm AJ
 17.↵
 18.↵
 Valzania C,
 Biffi M,
 Martignani C,
 Diemberger I,
 Bertini M,
 Ziacchi M,
 Bacchi L,
 Rocchi G,
 Rapezzi C,
 Branzi A,
 Boriani G
 19.↵
 20.↵
 21.↵
 Jansen AH,
 Bracke FA,
 van Dantzig JM,
 Meijer A,
 van der Voort PH,
 Aarnoudse W,
 van Gelder BM,
 Peels KH
 22.↵
 Turcott RG,
 Ashley EA
 23.↵
 Efron B,
 Tibshirani R
 24.↵
 Heinroth KM,
 Elster M,
 Nuding S,
 Schlegel F,
 Christoph A,
 Carter J,
 Buerke M,
 Werdan K
 25.↵
 26.↵
 27.↵
 28.↵
 Gras D,
 Gupta MS,
 Boulogne E,
 Guzzo L,
 Abraham WT
 29.↵
 30.↵
 Chung ES,
 Leon AR,
 Tavazzi L,
 Sun JP,
 Nihoyannopoulos P,
 Merlino J,
 Abraham WT,
 Ghio S,
 Leclercq C,
 Bax JJ,
 Yu CM,
 Gorcsan J III.,
 St John Sutton M,
 De Sutter J,
 Murillo J
 31.↵
 32.↵
FREEDOM: A Frequent Optimization Study Using the QuickOpt Method. http://www.clinicaltrials.gov. Identifier: NCT00418314. Accessed January 2, 2010.
 33.↵
 Stein KM,
 Ellenbogen KA,
 Gold MR,
 Lemke B,
 Lozano IF,
 Mittal S,
 Spinale FG,
 van Eyk JE,
 Waggoner AD,
 Meyer TE
 34.↵
Comparison of AV Optimization Methods Used in Cardiac Resynchronization Therapy (CRT) (SMARTAV). http://www.clinicaltrials.gov. Identifier: NCT00677014. Accessed January 2, 2010.
 35.↵
Stanford University. Optimize CRT (Development Version) Available at http://optimizecrt.stanford.edu. Accessed January 2, 2010.
 36.↵
 Bertini M,
 Ziacchi M,
 Biffi M,
 Martignani C,
 Saporito D,
 Valzania C,
 Diemberger I,
 Cervi E,
 Frisoni J,
 Sangiorgi D,
 Branzi A,
 Boriani G
 37.↵
 38.↵
 39.↵
 Vidal B,
 Tamborero D,
 Mont L,
 Sitges M,
 Delgado V,
 Berruezo A,
 DiazInfante E,
 Tolosana JM,
 Pare C,
 Brugada J
 40.↵
 Turcott RG,
 Pavek TJ
 41.↵
 Brophy JM,
 Joseph L,
 Rouleau JL
 42.↵
Clinical Perspective
Cardiac resynchronization therapy improves morbidity and mortality in appropriately selected patients. Whether further clinical benefit is possible with atrioventricular and interventricular pacing interval optimization remains unclear. Tools to assess the statistical significance of the measured optimization data have not been available previously. In the study reported here, an objective methodology for quantifying the statistical precision of estimated optimum pacing intervals was developed and applied to a number of commonly used optimization techniques. Many of the techniques did not yield statistically significant data in a majority of patients referred for atrioventricular and interventricular interval optimization, a finding that raises questions about the ability of pacing interval optimization to enhance clinical outcomes. The data demonstrate that accepting an estimated optimum interval without consideration of its statistical significance can result in worse cardiac function than default settings and can lead to the erroneous conclusion that the physiological optimum has changed over time. These results highlight the importance of evaluating the precision of measured data when conducting pacing interval optimization for the individual patient and when interpreting the results of clinical trials.
This Issue
Jump to
Article Tools
 Measurement Precision in the Optimization of Cardiac Resynchronization TherapyClinical PerspectiveRobert G. Turcott, Ronald M. Witteles, Paul J. Wang, Randall H. Vagelos, Michael B. Fowler and Euan A. AshleyCirculation: Heart Failure. 2010;3:395404, originally published May 18, 2010https://doi.org/10.1161/CIRCHEARTFAILURE.109.900076
Citation Manager Formats
Share this Article
 Measurement Precision in the Optimization of Cardiac Resynchronization TherapyClinical PerspectiveRobert G. Turcott, Ronald M. Witteles, Paul J. Wang, Randall H. Vagelos, Michael B. Fowler and Euan A. AshleyCirculation: Heart Failure. 2010;3:395404, originally published May 18, 2010https://doi.org/10.1161/CIRCHEARTFAILURE.109.900076Permalink: