| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Departments of 1Psychiatry and 4Rehabilitation Medicine, College of Physicians and Surgeons, and 2Department of Biostatistics, Mailman School of Public Health, Columbia University, New York 10032; 3Division of Developmental Psychobiology, New York State Psychiatric Institute, New York, New York 10032; and 5Department of Psychiatry, University of California, San Diego 92093, and Veterans Affairs Healthcare System, San Diego, California 92169
Submitted 4 December 2002 ; accepted in final form 20 June 2003
 |
ABSTRACT |
|---|
 TOP ABSTRACT METHODSRESULTSDISCUSSIONDISCLOSURESACKNOWLEDGMENTSREFERENCES |
|---|
psychophysiology; methodology; cardiovascular reactivity
The reliability of using BP waveforms to calculate HPV has not been well studied, but there are potential sources of error. First, identifying precisely the SBP peak time compared with the corresponding R-wave peak is subject to greater error. Because the SBP pulse waveform represents the travel of fluid through a closed system, it rises and declines more slowly than the R waveform, which represents an electrical event.
A second source of potential error is variability in the interval between R waves and the succeeding SBP waves. The speed with which the BP pulse wave travels to the periphery, measured as pulse wave velocity (PWV) or the related measure of pulse transit time (PTT), is affected by arterial structural and functional state (10, 17, 18). Various stressors and stimuli have been shown to cause fluctuations in PWV and PTT within short time frames (2, 3, 16). In laboratory studies, BP pulse speed tends to be more sensitive to physical than to psychological stressors (8, 20). Although these studies demonstrate that stressors can affect PWV and PTT variability, the data from analytic methods used in these studies do not elucidate whether PWV varies enough on a beat-to-beat basis to affect variability in R wave-SBP intervals. Thus it is possible that effects of acute stressors on PWV would produce significant error when SBP-SBP intervals are used to calculate HPV.
Only two published studies have examined directly the validity of HPV indexes calculated from BP waveforms in several conditions: seated and supine rest, paced breathing, standing, aerobic exercise on a bicycle ergometer, and recovery from exercise (5, 9). In both studies, correlations between indexes of HR and HPV derived from R-R and SBP-SBP series were very high except for indexes of high-frequency power. During exercise, correlations dropped considerably compared with other study conditions. Despite high correlations, the SBP-SBP series produced significantly different values of HR and HPV than did the R-R series in many conditions.
We propose that several methodological features of these studies contributed to the discrepancies reported between the BP- and ECG-derived measures. The purpose of this paper is to examine more thoroughly the correspondence between measures of HR and HPV derived from simultaneous beat-to-beat measures of ECG and BP.
|
METHODS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONDISCLOSURESACKNOWLEDGMENTSREFERENCES |
|---|
The findings presented here are secondary analyses of data from a total of 234 participants in four laboratory studies of cardiovascular reactivity. The studies and participant populations were 1) an exercise intervention study of healthy young adults; 2) a study of cardiac transplant patients that included a comparison group of healthy adults, whose data are presented here; 3) a study of gender and hostility in healthy adults (19); and 4) a study of a respite intervention for healthy spouses or other relatives who were caregivers for a relative with Alzheimer's dementia. Table 1 provides demographics of the participants from each study.
|
Participants in all studies underwent a standard, laboratory-based psychophysiology protocol. Because some studies included repeated testing sessions over time to assess the effects of an intervention, all data presented here are from the first psychophysiology session (preintervention) completed by participants. The specific experimental conditions varied slightly across studies, but each protocol included several of the following. Conditions in a seated position were seated baseline, speech presentation task, and handgrip task. Conditions in a supine position were supine baseline, psychological stressors of mental arithmetic, and/or the Stroop color-word matching test. Finally, passive orthostatic tilt from supine to 70° upright position was used as a physical stressor. The varying sample sizes in the results presented here reflect these protocol differences and missing data.
Psychophysiology Data Processing
Heart period variability. COLLECTION OF ECG SIGNALS. During all experimental conditions, participants' ECGs were continuously monitored with a standard three-lead configuration. The ECG signal was amplified, and the analog waveform was digitized at 500 Hz by a National Instruments 16XE50 16-bit analog-to-digital card and collected by a microcomputer. Software developed for our laboratory was used to mark R waves and create files of R-R intervals for analysis as described below.
ARTIFACT CORRECTION AND REJECTION. All computer-generated marks of R waves were reviewed for accuracy, and errors in marking R waves were corrected interactively. Ectopic beats were replaced by the average of the surrounding intervals.
PROCESSING OF R-R INTERVAL FILES. All measures were calculated on 240-s epochs of recordings from each experimental period. HR in beats per minute was computed from the R-R intervals. Time domain indexes of HPV were computed as the standard deviation (SD) and root mean squared successive differences (RMSSD) of R-R intervals. Standard estimates of low-frequency (0.04-0.15 Hz, LF) and high-frequency (0.15-0.40 Hz, HF) spectral power (21) were calculated by use of an interval method for computing Fourier transforms similar to that described by DeBoer et al. (6). Before computation of Fourier transforms, the mean of the R-R interval series was subtracted from each value in the series. The residual series was then filtered by using a Hanning window (11), and the power, i.e., variance (in ms2), over the LF and HF bands was summed. Estimates of spectral power were adjusted to account for attenuation produced by this filter (11). Because of the 240-s duration of each experimental condition in our protocol, estimates of very low-frequency spectral power (0.003-0.04 Hz) would not be accurate; thus very low-frequency data are not reported.
Blood pressure. During all experimental conditions, beat-to-beat BP was continuously measured by use of an Ohmeda Finapres model 2300 noninvasive monitor. This device uses a photoplethysmographic finger cuff to assess BP continuously by using the vascular unloading principle. The arterial pressure waveform was digitized at 500 Hz.
The BP data were processed with the same event-detection software used to process the ECG data. The peaks and troughs of the waveform were marked, from which beat-to-beat series of SBP and diastolic BP intervals were constructed. For the analyses presented here, only the SBP-SBP interval series were used to estimate HPV. In the same manner as with ECG recordings, the marked waveforms were reviewed, and errors in marking were corrected. Intervals between systolic peaks corresponding to ectopic beats from the ECG were replaced by using the same strategy as that used for replacement of nonnormal R-R intervals. The same spectral analysis methods used for R-R intervals were conducted on 240-s epochs of the SBP-SBP intervals to produce the same LF and HF power bands.
The ECG and BP data reported here include records from all subjects in the four studies that were analyzed with the same release version (December, 1999) of our laboratory's proprietary event-detection software. All subjects who had valid data for a resting baseline period were included in analyses, resulting in the omission of six subjects. In addition, a small number of data epochs (n = 21) from various experimental periods were excluded if the R-R and SBP-SBP series were marked differently because of intervals of noisy, unmarkable signals in one series but not the other. Specifically, if the number of valid events differed by 
3 between the two series for any 240-s epoch, data for that period were excluded from analyses.
Stress reactivity. Reactivity to psychological and physical stressors was computed by using both the R-R and SBP-SBP series. For the speech and handgrip tasks, reactivity was computed as the difference between the task and seated baseline condition for all measures of HR and HPV. Responses during the mental arithmetic and Stroop task were averaged for use in data analyses to increase response stability (13). Reactivity was then computed as the difference between this aggregated psychological stressor task measure and the supine baseline. Reactivity to tilt was the difference between the passive tilt and supine baseline periods.
Data Analysis
All variables except HR were log-transformed to correct for positive skew. HR data are reported in beats per minute, but data for all other variables are reported in natural log units. Data were analyzed by using intraclass correlation reliability coefficients. Reliability between the ECG- and BP-derived indexes of HR and HPV for each experimental condition and the reactivity change scores was calculated by using the fixed-effect form of the intraclass correlation coefficient because the two measurement devices were fixed across all participants. Unlike Pearson correlations, reliability can be interpreted as a proportion of variance (7). When used to compare two measurement devices or methods, as in the data presented here, the reliability coefficient indicates the proportion of variance due to between-subjects variability, as opposed to error variance between the two measurement devices.
Data from the four studies were analyzed separately to determine whether there were differences across participant populations. Although there were differences among the four study groups in demographic profiles (Table 1) and mean HR and HPV values (Table 2), the reliability coefficients between the BP and ECG data in calculating HR and HPV were very similar across all four samples. As a result, data for all four groups were combined and the reliabilities recomputed.
|
|
RESULTS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONDISCLOSURESACKNOWLEDGMENTSREFERENCES |
|---|
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONDISCLOSURESACKNOWLEDGMENTSREFERENCES |
|---|
Specifically, Carrasco and colleagues (5) compared BP intervals measured with a Finapres 2300 monitor to ECG-derived R-R intervals. They report very high correlations between HR and HPV indexes from the two interval series during supine rest, paced breathing, standing, and recovery from exercise on a bicycle ergometer. During exercise, however, correlations dropped considerably for RMSSD, LF HPV (0.04-0.15 Hz), HF HPV (0.15-1.0 Hz), and total spectral power. In all five experimental conditions, the SBP-SBP series significantly underestimated HR by an average of 1.8-3.9 beats/min. During standing and exercise only, the SBP-SBP series significantly underestimated the percentage of LF HPV by 6.9% (standing) and 33.8% (exercise) and overestimated both RMSSD, by 0.2 (standing) and 2.3 (exercise) beats/min, and the percentage of HF HPV by 7.0% (standing) and 33.8% (exercise). Giardino and colleagues (9) measured ECG and BP with a finger photoplethysmograph to calculate LF HPV (0.07-0.15 Hz) and HF HPV (0.15-0.40 Hz) in two small studies. They reported very high correlations in both LF and HF HPV from the two interval series during both a resting condition and "active baseline," which included a nonchallenging task. They reported a somewhat lower correlation for HF HPV during a Stroop test challenge after the active baseline. Similar to the Carrasco et al. results, the BP-derived series significantly overestimated mean HF HPV compared with the R-R series; however, in contrast to Carrasco's results in similar nondynamic conditions, the BP-derived data underestimated LF HPV.
The most striking distinction in our results is in the high reliability of the BP-derived data. In contrast to these prior studies, our HR calculations from BP and ECG waveforms are extremely reliable in all conditions. Although reliability drops somewhat for measures of HF HPV, all coefficients are well within acceptable limits to suggest that the SBP-SBP series could be used as a proxy for R-R to calculate frequency- and time-domain indexes of HPV across all experimental conditions tested (7). During the upright tilt condition, we did not observe the mean differences in ECG- and BP-derived RMSSD and HPV that Carrasco et al. (5) found in a standing condition. The consistency in our data cannot be attributed solely to our specific laboratory setting or technicians, because the caregivers' data set was from a collaborative study using our data-collection methodology in a different setting.
We propose that methodological differences between our studies and the other two published reports contributed primarily to discrepancies in the results. We report findings from a large sample drawn from several healthy adult populations, in contrast to the small samples by Carrasco et al. (5) (n = 10) and Giardino et al. (9) (n = 10 and n = 16). In all our protocols, BP data were collected at the finger as an analog signal by using the Finapres and were digitized at the same 500-Hz sampling rate as the ECG waveform. Both types of digitized waveforms were subjected to identical peak-detection methods. In contrast, Carrasco et al. collected BP data through the Finapres's RS-232 serial port, which digitizes data at 50 Hz, but their ECG data were digitized at 500 Hz. Giardino et al. also collected BP data from the finger with a photoplethysmograph and used a uniform, sufficiently high sampling rate of 1,000 Hz for both ECG and BP signals. Evidence that sampling rate contributed to the cross-study discrepancies was offered by Giardino et al. They tested the effect of sampling rate on their data by subsampling their 1,000-Hz interval series to simulate 200-, 20-, and 10-Hz sampling rates. Correlations between the ECG- and BP-derived interval series dropped as sampling rates decreased.
Another source of error could be the type and setup of the BP measurement devices used. The underestimation of HR by BP data in the Carrasco study (5) could be due to the Finapres's stated pulse rate accuracy from the serial port of ±5 beats/min or ±5% of the reading, whichever is greater (4); nevertheless, this systematic error between BP- and ECG-derived HR in their data is somewhat puzzling. Giardino et al. (9) measured BP with a different device but at the same peripheral site. Their results are in closer agreement with those of Carrasco et al. than with ours, thus it is unclear whether use of the plethysmograph contributed to the different findings across studies. We utilized standardized frequency spectra to assess HPV (21). Both other groups used different spectral bands to define LF and HF HPV, which would cause slight between-study differences in absolute values of spectral power. It is doubtful, however, that different spectral bands would explain why we found high agreement between ECG- and BP-derived HPV values, whereas the other groups found substantial differences. Taken together, these methodological differences likely augmented the degree of peak-detection error in the SBP-SBP series in both prior studies. Further reliability studies are warranted to determine specifically which aspects of methodology affected the discrepant results among the studies and the degree of their influence.
Despite methodological differences, our results agree most consistently with the prior studies in one respect: increased error when calculating HF HPV from BP data compared with ECG data. The reasons for this finding are not so clear, however. In all these studies, BP was measured at the finger. Here, the arterioles are much less compliant, i.e., stiffer than the aorta, producing a faster PWV than at the aorta. Generally, the PWV at a given site is considered fairly stable within individuals unless there is a clinical change such as hypertension or aging (1). Nevertheless, Carrasco et al. (5) and other investigators have demonstrated that the acute effect of certain dynamic stressors or combinations of stressors on PWV or PTT might cause substantial R wave-SBP variability. Szabo (20) reported that PTT was longer when moving from the standing to the seated position. Another similar study (8) demonstrated a main effect of stationary bicycle exercise in shortening PTT. Psychological stressors, however, seem to affect PWV or PTT only in combination with the other stressors. A mental arithmetic task shortened PTT only during standing (20). Similarly, the impact of an arithmetic challenge was observed only as an interaction effect during certain phases of exercise and exercise recovery (8).
On the basis of these reports, if PWV variability affected HF HPV in our data, it should have appeared primarily during the upright tilt condition as reduced reliability values or differences in mean HPV values from the two data series. Indeed, reliabilities were slightly lower, and the two data series differed more during tilt than in other experimental conditions, but differences were not as pronounced as Carrasco's differences during standing. Our findings suggest, in concert with studies of combined mental and dynamic stressors, that responses to psychological or orthostatic challenge alone do not produce enough change in PWV to result in substantial R wave-SBP interval variability.
Giardino et al. (9) reported secondary analyses from one of their studies, further suggesting that changes in PTT may help explain the greater error in BP-derived HF HPV. They reported a negative correlation of PTT with differences between corresponding R-R and SBPSBP intervals. When a given PTT was shorter than average, the corresponding R-R intervals tended to be longer than the SBP-SBP intervals. They then spectrally analyzed the series of differences between each R-R and SBP-SBP interval pair and found that the peak power occurred at about 0.20 Hz. The authors conclude that the periodic mechanical effects of respiration on BP may cumulatively augment HF variability in the SBP-SBP series. This is an interesting possibility but one that needs further investigation because of the very small sample on which they reported.
Our findings suggest that SBP waves can substitute for R waves to calculate HPV indexes; however, the high reliability we observed may be restricted to the specific experimental conditions and data-collection and analysis methods used. Our studies did not include dynamic stressors to compare with the anomalous findings during exercise reported by Carrasco et al. (5). Exercise is a much more complicated condition for interpreting HPV indexes, even when derived from R-R waveforms. During exercise, mechanical effects of respiration on HPV are greater than when breathing is more stable or controlled (15). Additional factors during exercise, e.g., metabolic or chemoreflex, exert differential effects on HPV and BPV (14).
These findings cannot yet be generalized to populations with cardiovascular disease or other clinical conditions. Because vasoactive medications and vascular changes associated with conditions such as hypertension cause changes in PWV (3, 18), the impact on SBP-SBP variability may be greater than was observed in the healthy adults included in our analyses. The reliability of using BP waveforms to calculate HPV and this method's generalizability to other BP measurement devices deserve further study, especially in various clinical populations, using different BP measurement devices, and in dynamic conditions such as exercise.
Practical applications of these findings include any research or clinical setting in which only beat-to-beat BP measurement is feasible or desirable. Another potential application is in situations in which precise identification of R waves is impossible because of electrical noise in the ECG signal, a phenomenon that occurs periodically in electronic data collection. Short interruptions in the middle of an epoch of data can be corrected with interpolation methods. Longer interruptions at the beginning or end of an epoch sometimes can be omitted while retaining a sufficiently long epoch for valid spectral analysis of HPV. When longer signal aberrations occur within an epoch, however, linear interpolation is invalid because it reduces variability in the signal. Under these conditions, spectral analysis cannot be conducted, and the epoch is lost as missing data. In such situations, the R-R time series could be reconstructed by using corresponding SBP-SBP intervals from the same time interval.
In conclusion, BP waveforms measured in healthy adults, digitized at a sufficiently high rate, and subjected to the same analysis procedures can produce reliable estimates of HPV. The correspondence between ECG- and BP-derived estimates of HPV may have practical value in settings in which only one type of data can be collected feasibly. Series of BP waveforms also may provide a valid way to restore relatively short segments of missing ECG waveform series in longer epochs of data.
|
DISCLOSURES |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONDISCLOSURESACKNOWLEDGMENTSREFERENCES |
|---|
|
ACKNOWLEDGMENTS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONDISCLOSURESACKNOWLEDGMENTSREFERENCES |
|---|
|
FOOTNOTES |
|---|
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
|
REFERENCES |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONDISCLOSURESACKNOWLEDGMENTSREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  A. Jones, A. Beda, A. M.V. Ward, C. Osmond, D. I.W. Phillips, V. M. Moore, and D. M. Simpson Size at Birth and Autonomic Function During Psychological Stress Hypertension, March 1, 2007; 49(3): 548 - 555. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Visit Other APS Journals Online |
, hostility study; 
, transplant study. Solid diagonal line, least squares regression line; dashed diagonal line, line of identity.
, caregivers study. Solid diagonal line, least squares regression line; dashed diagonal line, line of identity.