(Hypertension. 1997;29:1225-1231.)
© 1997 American Heart Association, Inc.
Articles |
Placebo-Controlled Biofeedback Blood Pressure Effect in Hypertensive Humans
From the Cooperative Research Centre for Cardiac Technology and Cardiovascular Research Unit, Department of Cardiology (S.N.H., R.J.H., S.K.L.L., H.K., M.J., A.S.M.), and Department of Psychiatry Royal North Shore Hospital (R.W.B.), St Leonards; the National Acoustics Laboratory (N.L.C.); and Department of Health Science, University of Technology (A.C.), Sydney, Australia.
Correspondence to Prof Stephen N. Hunyor, CRC for Cardiac Technology, Block 4, Level 3, Royal North Shore Hospital, St Leonards (Sydney), NSW 2065, Australia. E-mail crcct{at}blackburn.med.su.oz.au
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsResultsDiscussionReferences |
|---|
Abstract The role of biofeedback in blood pressure control remains ill-defined because of nonspecific (placebo) effects, small study numbers, and the technical limitations of continuous pressure feedback. Clarification of its potential is awaited by those seeking a nonpharmacological approach to blood pressure control. This study examines the capability for systolic pressure lowering of 5 mm Hg or more using continuous pressure feedback in a statistical sample of untreated, well-characterized, mildly hypertensive individuals. Subjects were randomized in a double-blind study to active or placebo biofeedback. Placebo consisted of a modified contingency approach, using a partial disguise based on a digital high pass filter with 15 elements. Blood pressure–lowering capability was assessed during two laboratory sessions. Continuous visual feedback resulted in 11 of 28 subjects on active treatment and 12 of 28 on placebo treatment lowering their systolic pressure by 5 mm Hg or more (11±5.6 and 12±8.4 mm Hg, respectively; P=NS). Prestudy pressure was well-matched (153±9/97±4 and 154±8/98±4 mm Hg, respectively). An initial small difference in diurnal profile did not change. These findings indicate that among mildly hypertensive individuals, almost half can lower systolic pressure at will for short periods. This capability is independent of the real or placebo nature of the feedback signal. We conclude that there is no specific short-term biofeedback pressure-lowering capability in hypertensive individuals. Further exploration is needed to determine whether specific components of the placebo effect can be delineated, whether personality characteristics influence the response, and whether further biofeedback training can alter the outcome.
Key Words: blood pressure monitoring • placebo effect • biofeedback
|
Introduction |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Sixty million Americans are at increased health risk from elevated blood pressure (BP),1 the most common diagnosis (in 3.9% of visits) for individuals visiting physicians.2 Also, one in three adults uses "unorthodox therapies," resulting in 425 million visits compared with 338 million to general physicians. This led the National Institutes of Health to set up the Office of Alternative Medicine with a mandate to integrate effective alternate treatments into mainstream medical practice. However, testing of complementary medicine presents problems. The sifting out of possible placebo effects is regarded as particularly difficult.3
A meta-analysis of 14 randomized antihypertensive drug trials involving 37 000 people concluded that a mean reduction in diastolic BP of 5 to 6 mm Hg avoided at least one third of the risk of stroke and one fifth of the risk of coronary heart disease.4 However, these trials demonstrated problems with side effects5 and compliance.6 Accordingly, nonpharmacological alternatives such as dietary intervention, weight manipulation, and exercise have been sought. There is also a compelling theoretical basis for considering cognitive behavioral techniques as an alternative strategy. The evidence relates to four broad areas: (1) autonomic nervous system activation in some hypertensive situations,7 8 9 especially before the establishment of vascular and cardiac amplifier effects10 ; (2) centrally mediated increases in sympathetic outflow in some experimental models11 ; (3) inhibition of autonomic activity by many forms of effective antihypertensive therapy12 ; and (4) marked BP decline during sleep13 and elevations during periods of mental engagement or "stress."14
Early evidence for conscious influence on the autonomic nervous system was provided by Miller and DiCara15 in curarized animals able to differentially control ear skin resistance, renal blood flow, urine output, and blood flow in tail vessels. In humans, the frequency of transient vasoconstrictive events was also shown to be responsive to wholly operant methods.16
A large number of studies have subsequently examined the mechanism of action and applicability of cognitive behavioral techniques in hypertension. The American College of Physicians' report on "Biofeedback for Hypertension" summarizes the results of 18 studies using direct systolic biofeedback, diastolic biofeedback, or both, noting an average BP decline of 7.8/5.6 mm Hg.17 Several of these studies present useful insights into the possible mechanism of action of biofeedback, such as the demonstration by Messerli et al18 that in borderline hypertensive individuals, the effects are similar to those of ß-blockers and diuretics, where the fall in arterial pressure is initially mediated by reduced cardiac output, followed by long-term pressure reductions mediated by decreased peripheral resistance.
However, none of the clinical studies reported has adhered even
to the minimal qualitative requirements, the six "McMaster
criteria" specified as necessary for valid
conclusions.19 These criteria include (1) random
participant assignment, (2) reporting of all clinically relevant
outcomes, (3) characterization of study participants, (4) consideration
of clinical and statistical significance, (5) exclusion of
contamination or cointervention on the therapeutic maneuver, and (6)
accounting for all individuals entered at the conclusion of the study
and an acceptable dropout rate. Additionally, in the case of
biofeedback BP effects in hypertensive individuals, it is important to
add other basic criteria to lessen the likelihood of either ß or 
statistical experimental error. These are (1) use of a demonstrated
optimal BP feedback modality, (2) adequate treatment-free interval, and
(3) sufficient participant numbers.
The area of most frequent concern is the nonspecificity of response either because of combinations of treatment modalities, such as meditation and biofeedback, or use of either inappropriate or inadequate control (placebo) modalities. The ability to blind both participants and observers relates to the credibility of the placebo, an area that has been poorly served with respect to BP biofeedback. Other frequent confounders relate to the use of BP surrogates such as skin temperature, galvanic skin resistance, pulse transit time, and heart rate.
With respect to sample size, Jacob et al20 have calculated that 30 or more participants are required per intervention group to detect a clinically significant change in systolic BP of 5 mm Hg at the 5% significance level with a statistical power of 0.8.
The conclusions of the 1993 review by Eisenberg et al21 of cognitive behavioral techniques for hypertension treatment remain basically unchallenged. In an extensive and careful meta-analysis of 875 widely sourced articles, only 26 fulfilled minimal criteria for reliability. Furthermore, no more than 5 of the 26 had adequate sample size to detect a systolic BP decline of 5 mm Hg, and none of these studies specifically examined biofeedback. Potential confounding by noncognitive behavioral factors was reported to some extent in all 26 trials. Overall, the technical quality score of the 26 studies—33.8±5.1 out of a possible maximum of 104—indicates gross deficiencies across the board, leading the authors to conclude that interventions were poorly standardized, poorly categorized, and often involved combined or overlapping therapies.
We have examined the capacity of untreated mildly hypertensive individuals to voluntarily manipulate their systolic BP in response to continuous BP feedback provided by volume-clamp plethysmography of the finger.22 Sample size was based on preliminary studies,23 and a novel modified contingency placebo was applied in a randomized, double-blind design.
|
Methods |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
We used a simple, randomized, controlled trial design because the carryover effects of biofeedback are not determined and thus a number of alternative ideal experimental designs, such as a crossover trial, are not applicable. It is highly likely that once a subject has been trained in biofeedback, skills learned will confound attempts to use crossover designs. Subjects were randomized to either active or placebo continuous systolic BP feedback, and both subject and observer remained blinded.
The sample size was derived from a pilot study that examined the
biofeedback capabilities of 18 untreated hypertensive and 13
normotensive subjects in an open design.23 This had
demonstrated mean systolic BP lowering of 10.0±4.0 (±SD)
mm Hg in the normotensive subjects, whereas among the hypertensive
subjects, 10 of 18 significantly lowered systolic BP
(P<.05, one-tailed t test) by an average of
7.3±8.3 mm Hg. Using a power (1-ß) of 0.80, a type 1 error
() of 0.05% (two-tailed), and anticipating a biofeedback
systolic treatment effect at least 6 mm Hg greater than
placebo, with an SD of ±8, led to a requirement of 28 subjects for
each treatment arm.
Participants ranged in age from 18 to 69 years. Any suspicion of secondary hypertension on history or physical examination led to investigation with abdominal ultrasound, 24-hour urinary catecholamines (measured twice), intravenous pyelography, renal angiography, or specialized endocrine investigations. Evidence of target-organ damage led to exclusion. When the electrocardiogram showed only voltage evidence of left ventricular hypertrophy, quantitative echocardiography was performed. The study was approved by the Institutional Ethics Committee and carried out according to its guidelines. Of 137 subjects screened, 58 were enrolled in the study. Forty-six were found to have BP too low for entry despite at least 3 months off antihypertensive treatment. Five had BP too high for entry (ie, systolic BP>200 or diastolic BP>115 mm Hg), and 28 others were excluded on other grounds, including left ventricular hypertrophy (n=3), retinal hemorrhages (n=1), and the inability to make the necessary time commitment (n=8). Two failed to finish the protocol, and 1 was retrospectively excluded because of echocardiographic left ventricular hypertrophy.
Study Protocol
Fig 1
shows the study protocol. Eight
laboratory biofeedback sessions each used three 12-minute trials that
included an instruction to ignore BP (3 minutes) before attempts to
lower systolic BP (5 minutes) and to raise it (1 minute). In
each session, the result was expressed as the average of the three
trials. The first four sessions constituted training, and biofeedback
capability was assessed on the average of sessions 5 and 6. Sessions 7
and 8 served to assess other cardiorespiratory parameters
that are not considered in this analysis. The study was
conducted in a temperature-controlled laboratory, with the operator
using a remote slave monitor to follow events.
|
Manipulation of BP was prompted by screen prompts, followed by BP "zones" that encouraged subjects to raise, ignore, or lower their BP. Inability to achieve the BP change indicated led to the task being made easier to avoid frustration. During the ignore phases, the BP signal was removed from the screen and subjects were given bland literature. BP raising was included to test both limbs of the control loop and to confirm situations likely to lead to BP elevations. All subjects viewed a prestudy audiovisual slide presentation with facts about high BP and its control. Although the emphasis was on ideation and imagery, somatic mechanisms such as altering muscle tone or breathing pattern were not restricted. If the systolic-lowering capability was 5 mm Hg or less in the first two sessions, suggestions were made for strategies found successful by others. Only systolic BP was displayed, but diastolic pressure and heart rate were also recorded. Studies were performed at the same time of day with subjects in a 2-hour postprandial state. Coffee and tea were not allowed for 2 hours and alcohol intake for 12 hours before the study. Weight and physical activity were kept constant.
BP Measurement
Standardized arm cuff BP measurement used the Hawksley
random-zero sphygmomanometer to exclude observer bias, with phase V
Korotkoff sounds taken as the diastolic end point. Readings
were taken in triplicate after subjects had sat for 5 minutes on two
occasions over at least 2 weeks after a minimum 4-week treatment-free
interval. The average of all readings was used as the entry criterion.
BP was also measured in duplicate before and at the end of each
laboratory session.
Continuous finger BP was monitored during laboratory studies by volume-clamp plethysmography,24 25 with systolic BP displayed on a computer screen. This technique maintains a stable photoplethysmographic signal (set at 940 nm for optimal sensitivity to hemoglobin) in a finger segment by means of a rapidly responding computer-controlled feedback loop (90% response time of approximately 0.5 second). The implementation varies the compression pressure in a finger-encircling cuff to defeat any BP-induced vascular volume change. Such monitoring is well-tolerated in excess of an hour and tracks BP change reliably during contrived interventions.22
Choice of BP Modality for Biofeedback
To optimize the BP biofeedback signal,26 we studied
36 normotensive individuals (age, 42±15 years) who undertook
laboratory training before receiving visual, auditory, or combined
visual and auditory feedback of beat-to-beat systolic BP in
random order in a sound-proofed room (35dbA-NR30). We also examined BP
modality presentation in relation to density of information
(1:3, 1:10, or each beat) and specificity of target (relative or
absolute).
Visual systolic BP was provided as a beat-to-beat varying horizontal blue trend line. Auditory feedback through stereo headphones (Koss stereophones, Pro/4XTC) was delivered as a sinusoidal tone proportional to systolic BP. The headphones were calibrated by an "artificial ear" (type 4153, Bruel and Kjaer Ltd).
Placebo Design
The aim of a placebo algorithm for use with continuous BP
biofeedback is to eliminate the display of BP variations that may be
produced by successful attempts at BP control. However, it should
retain physiological fluctuations related to both
faster respiratory effects and slower diurnal variations. In a
double-blind experiment, it is necessary for both the subject and
operator not to recognize the application of the placebo by an obvious
absence of response to BP manipulation.
Our placebo feedback signal filters out the intermediate frequency band
from 0.1 to 0.02 Hz but retains the normal BP variations. Thus, it
greatly attenuates the observable results of conscious control. It must
be expected that the two signals are not completely orthogonal so that
one cannot be completely separated from the other. In our placebo
application, low-frequency filtering subtracts a running average of 15
beats from the BP signal. To reconstitute the slowest variations, the
signal is reset every 50 pulses to the actual BP. The results of
implementing the algorithm are shown in Fig 2.
|
24-Hour Ambulatory BP
Twenty-four-hour ambulatory BP (ABP) was examined on three
occasions (Fig 1
) with a Takeda (model TM-2420) device27
programmed to measure BP every 15 minutes. The ambulatory studies
assessed any transfer of training outside the laboratory and possible
persistence of effect after the sessions.
Statistical Analysis
Customized software averaged beat-to-beat BP for the baseline,
raise, and lower tasks. BP and heart rate changes were calculated by
subtracting the value during the task from the preceding baseline. All
results are reported as mean±SD. Statistical analysis was
performed with SPIDA Version 6.0 (Statistical Package for Interactive
Data Analysis, Macquarie University). To determine the net
treatment effect of BP biofeedback, unpaired Student's t
test was performed between the placebo and active BP responses assessed
during sessions 5 and 6. Statistical significance was set at a value of
P<.05, and clinical significance was defined as a
systolic BP decline of 5 mm Hg or more. The presence of
any training effect (across sessions 1 through 4) was examined with a
generalized estimating equation using repeated
multivariate analysis. Covariates and
confounders such as body mass index and age were correlated against BP
lowering with Spearman's rank correlation. Smoking and alcohol
consumption were regressed on BP lowering using a linear
regression.
Twenty-four-hour ABP was analyzed in three ways: (1) The shape
of the 24-hour profile was compared in the active and placebo groups;
(2) any alteration in the similarity (or dissimilarity) between
profiles was examined in relation to the type of feedback or time; and
(3) the BP response in sessions 5 and 6 (mean performance of
sessions; active versus placebo; "clinical," ie, 
5 mm Hg,
systolic BP lowerers versus non-lowerers) was related to the
24-hour ABP level in two ways (24-hour average and daily variability,
ie, SD). For the first two methods of analysis, we used a
generalized linear model approach in which standard errors for
significance tests were estimated with a linearization algorithm
implemented in SUDAAN28 statistical software with
significance set at a value of P<.01. For the third method,
Spearman's rho correlation coefficients were obtained.
|
Results |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
BP Levels and Response to Biofeedback
The 56 subjects completing the study had mean entry BP levels of 153±9/97±4 (active) and 154±8/98±4 (placebo) mm Hg. Twenty-four of the subjects had not previously been treated. Known duration of hypertension was 9.5±9.2 (range, 0 to 45) years. The modality study established that systolic BP lowering was greater with visual and visual plus auditory feedback than with auditory feedback alone (P=.03). Fifteen subjects achieved a systolic decrease of 5 mm Hg or more with visual feedback and 18 with visual plus auditory feedback. There was no difference in response to density of information or target type (Fig 3
).
|
In the whole group, systolic BP was lowered 5±7.2
(active) and 6±7.6 (placebo) mm Hg (P=.56, unpaired
t test). Eleven subjects receiving active BP feedback and 12
receiving placebo lowered their systolic BP by 5 mm Hg or
more (active, 11±5.6; placebo, 12±8.4; P=.83) (Fig 4). Diastolic BP and heart rate changed
significantly only during attempted systolic BP raising and
were not different in active and placebo groups (Fig 5).
|
|
There was a small training effect confined to the placebo group from session 1 (2±5.0 mm Hg) to session 4 (5±4.9 mm Hg) (P<.002). The placebo group had a higher body mass index (26.5±4.2 versus 24.5±2.4 kg/m2, P=.03) and consumed more alcohol (125±110.4 versus 94±59 g/d, P=.01), but neither these factors nor smoking and age were correlated to BP-lowering capability.
No difference in arm cuff BP was observed across biofeedback sessions
(Table).
|
24-Hour ABP
The 24-hour ABP averages of the active and placebo groups were
practically superimposable at all three time points of the study,
except that nighttime systolic BP appeared lower in the placebo
group after completion of the biofeedback sessions. Less than 2% of
readings were rejected because of technical problems. It was found that
a linear, quadratic, and sinusoidal model best fitted the 24-hour ABP
profile. However, the diurnal profiles between active and placebo
groups were different at all three time points for both
systolic and diastolic BPs, respectively (prestudy,
P=.017 and P=.004; poststudy, P=.026
and P=.002; 4 weeks poststudy, P=.014 and
P=.012) (Fig 6). These differences did not
change over time and were not affected by biofeedback training.
|
|
Discussion |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
This study examines specific BP biofeedback effects using a credible placebo in a tightly controlled experimental design applied to well-characterized hypertensive subjects. We used an electronic BP display to enhance precision and verification29 and based group size on preliminary experiments to avoid type II error. Such design considerations have been lacking in studies of cognitive behavioral treatments for hypertension.21 Our findings show a high proportion (23 of 56) of hypertensive subjects capable of clinically relevant reductions of systolic BP (average, 12 mm Hg) during laboratory biofeedback sessions. Yet this ability is independent of the true (active) or placebo nature of the feedback signal. In previous studies, variations in the type of signal feedback with respect to information density, modality emphasis, and physiological function displayed have led to unstandardized applications.30 In a separate study, we determined the optimal BP presentation to be a beat-to-beat, visual analog display.
Two outstanding issues remain after common shortcomings of previous BP biofeedback studies,21 such as BP profiling and experimental design, are addressed, namely, the method of BP feedback and the nature of the placebo. Currently available technology24 25 for continuous BP display has previously been used only in feasibility studies of BP biofeedback.23 31 Design, validation, and application of a suitable BP placebo has remained elusive.30 32 33
The control of nonspecific effects is seen as critical to the establishment of a scientific foundation for biofeedback32 33 and in the past has addressed potential confounders such as adaptation, habituation, and suggestibility by using a varied noncontingent feedback. This has included recorded data from another subject, a constant tone, no treatment, waiting list, regular monitoring, and various "placebo" sham interventions.21 The latter consisted of yoking controls,34 general physiotherapy in place of BP exposure,35 and attentional placebo groups.32 Such fully noncontingent placebos are likely to degrade the performance of control groups and tend to bias results in favor of true biofeedback.36 The lack of evidence for differing control procedures eliciting the same nonspecific effects raises the possibility that they are not equivalent inert treatments. The removal of contingency between biofeedback and behavior was seen as logically similar to removing the active chemical ingredient from a medicine. But when tried with electromyography pseudofeedback, which provides little or no veridical information, there was a suggestion of response bias and breaching of the double-blind method.32
On the other hand, our BP placebo relies on a modified contingency approach that controls for boredom, frustration, failure experience, and related factors. It maintains the character of an active placebo with a reduced level of information feedback in which the true information is degraded in a random fashion. This placebo algorithm has a demonstrable distorting effect (both qualitative and quantitative) on the true BP. We believe this approach served to neutralize differences in expectancies for improvement and other nonspecific influences on BP outcome. Awareness of the contingency of the feedback was no different in the active and placebo groups.
Traditional assumptions about the null response "placebo" effect are being questioned by work such as in depression,37 in which a 40% placebo response compares with the effect of antidepressants. Also, the paradox in attributing "no effect" to placebo yet having to take it into account when "real" drugs are being tested seems illogical. When placebo works as well as the investigational treatment, it is taken as evidence that the "real" drug does not work. An alternative interpretation is that it works no better than a placebo.
Our finding of equivalence in systolic BP–lowering capacity from active and placebo BP biofeedback has several possible explanations, including (1) true inability to exert a significant short-term influence on systolic BP by cognitive means, (2) use of incorrect BP-lowering strategies because of incomplete understanding by subjects or inadequate training by the experimenters, (3) confounding of cognitive by somatic effects, (4) dilution of cognitive response caused by processing needs of extensive feedback data, (5) inadequate training period,38 (6) shortcomings of our modified contingent placebo, or (7) advanced structural cardio-vascular changes. However, the extent of BP lowering achieved by the responsive subjects in both active and placebo groups (11 and 12 mm Hg) would have a significant effect on outcomes if translated to a chronic response.4 The latter could possibly result from a change in the BP set point with repeated acute BP manipulations, according to Hebbian principles of central synaptic functioning.39
Our findings fit broadly with those of a recent meta-analytic review of cognitive behavioral effects on BP that found such interventions to have an effect superior to no treatment at all but no different from placebo. However, these conclusions were based on only 90 heterogeneous subjects spread over a number of seriously flawed studies.
We believe that our technique of BP presentation, including the modified contingency placebo, also takes into account central nervous system mechanisms related to attention and awareness. For example, the more distracters there are in BP data presentation, the longer searches take through the usual serial search mechanism.40 Other presentations may lead to the dramatic example of preattentive processing known as "pop out."41
The effects of attention have been observed at the neural level in sensitized monkeys, in which neuronal firing was influenced by the attentional status.42 Also, involvement of the thalamus in visual attention has been demonstrated in human positron-emission tomographic scans.43 Other studies of thalamic stimulation in humans show that differing somatosensory signals can influence behavior even without producing awareness. The relevance of attention and awareness to BP biofeedback studies is evident when it is considered that an attended event is reacted to more rapidly, at a lower threshold, and more accurately.
The outcome resulting from BP feedback is also highly dependent on subject motivation, perhaps the most salient variable of all in determining the effectiveness of biofeedback. Although electronic instrumentation is not an essential component of biofeedback efficacy, as evidenced by Ekman's use of a mirror with the Facial Action Coding System,29 we believe more elaborate displays aid in subject motivation. Our approach to BP biofeedback using software-manipulated continuous BP information allows presentation of an optimized signal that can also be modified to provide the best approximation to a true placebo. This is likely to enhance attention and subject motivation and opens up the possibility of dissecting elements of the BP response to biofeedback by excluding nonspecific effects.
This study indicates that a cognitive approach to lowering systolic BP is effective in nearly half of untreated mildly hypertensive individuals during laboratory periods when continuous BP is monitored and displayed. However, the BP response is independent of the feedback display representing either the subject's real-time BP or a modified contingency placebo alternative. There is no difference in the demographic and baseline BP characteristics of subjects related to their ability to achieve such BP alterations.
The paradox posed by our results is not so much a lack of biofeedback BP effects, since 11 of 28 subjects receiving active biofeedback were able to lower systolic BP by an average of 11 mm Hg. Rather, it is that a similar response was evident when placebo feedback was provided. Insights provided by our work should stimulate further clinical studies addressing the interfacial issues between neuroscience, cognitive behavioral effects, and cardiovascular physiological responses.
|
Acknowledgments |
|---|
This work was supported by grants from the National Health and Medical Research Council, National Heart Foundation (Australia), The Government Health Employees Research Fund (NSW), and The North Shore Heart Research Foundation.
Received July 29, 1996; first decision August 29, 1996; accepted November 15, 1996.
|
References |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
1. Stokes J III, Kannel WB, Wolf PA, D'Agostino RB, Cupples LA. Blood pressure as a risk factor for cardiovascular disease: The Framingham Study—30 years of follow-up. Hypertension. 1989;13(suppl I):I-13-I-18.
2. Schappert SM. National Ambulatory Medical Care Survey: 1991 Summary. US Department of Health and Human Services, Centers for Disease Control and Prevention, Hyattsville, Md. Vital Health Stat.. 1994;13:1-110.
3. Hunyor SN, Henderson RJ. The role of stress management in blood pressure control: why the promissory note has failed to deliver. J Hypertens. 1996;14:413-418.[Medline] [Order article via Infotrieve]
4. Collins R, Peto R, MacMahon S, Herbert P, Fiebach NH, Ebertein KA. Blood pressure, stroke and coronary heart disease, Part 2: short-term reductions in blood pressure: overview of randomised drug trials in their epidemiological context. Lancet. 1990;335:827-838.[Medline] [Order article via Infotrieve]
5. Multiple Risk Factor Intervention Trial Research Group. Baseline rest electrocardiographic abnormalities, antihypertensive treatment, and mortality in the Multiple Risk Factor Intervention Trial. Am J Cardiol. 1985;55:1-15.[Medline] [Order article via Infotrieve]
6. Klein LE. Compliance and blood pressure control. Hypertension. 1988;11:1161-1164.
7. Panfilov VV, Reid JL. Brain and autonomic mechanisms in hypertension. J Hypertens. 1994;12:337-343.[Medline] [Order article via Infotrieve]
8. Esler M, Ferrier C, Lambert G, Eiseindorfer G, Cox H, Jennings G. Biochemical evidence of sympathetic hyperactivity in human hypertension. Hypertension. 1991;17(suppl III):III-29-III-35.
9. Julius S. Autonomic nervous system dysregulation in human hypertension. Am J Cardiol. 1991;67:3B-7B.[Medline] [Order article via Infotrieve]
10. Folkow B, Hallback M, Lundgren Y, Sivertsson R, Weiss H. Importance of adaptive changes in vascular design for establishment of primary hypertension, studied in man and in spontaneously hypertensive rats. Circ Res. 1973;32(suppl I):I-2-I-16.
11. Chalmers JP, Pilowsky PM. Brainstem and bulbospinal neurotransmitter systems in the control of blood pressure. J Hypertens. 1991;9:675-694.[Medline] [Order article via Infotrieve]
12.
Wallin BG, Sundlof G, Stromgren E, Aberg H.
Sympathetic outflow to muscles during treatment of hypertension with
metoprolol. Hypertension. 1984;6:557-562.
13. Littler WA, Honour AJ, Carter RD, Sleight P. Sleep and blood pressure. Br Med J. 1975;3:346-348.
14.
Henry JP, Cassel JC. Psychosocial factors in
essential hypertension: epidemiologic and animal experimental
evidence. Am J Epidemiol. 1969;90:171-180.
15. Miller NE, DiCara LV. Instrumental learning of peripheral vasomotor responses by the curarised rat. Am J Physiol. 1968;215:677-683.
16. Snyder C, Noble M. Operant conditioning of vasoconstriction. J Exp Psychol. 1968;77:263-268.[Medline] [Order article via Infotrieve]
17. Health and Public Policy Committee. American College of Physicians. Biofeedback for hypertension. Ann Intern Med. 1985;102:709-715.
18. Messerli FH, De Carvalho JGR, Christie B, Frohlich ED. Systemic haemodynamic effects of biofeedback in borderline hypertension. Clin Sci. 1979;57(suppl):437s-439s.
19. Sackett DL, Haynes RB, Tugwell P. Clinical Epidemiology. A Basic Science for Clinical Medicine. Boston, Mass: Little, Brown & Co; 1985.
20.
Jacob RG, Fortmann SP, Kraemer HL, Farquhar JW, Agras
WW. Combining behavioral treatments to reduce blood pressure: a
controlled outcome study. Behav Modif. 1985;9:32-53.
21.
Eisenberg DM, Delbanco TL, Berkey CS, Kaptchuk TJ,
Kupelnick B, Kuhl J, Chalmers TC. Cognitive behavioral
techniques for hypertension: are they effective? Ann
Intern Med. 1993;118:964-972.
22.
Lal SKL, Henderson RJ, Cejnar M, Hart MG, Hunyor
SN. Physiological influences on continuous
finger and simultaneous intra-arterial blood
pressure. Hypertension. 1995;26:307-314.
23. Cejnar M, Hunyor SN, Liggins GW, Bartrop R. Voluntary blood pressure control using continuous systolic blood pressure biofeedback. Clin Exp Pharmacol Physiol. 1988;15:265-269.[Medline] [Order article via Infotrieve]
24. Penaz J. Photoelectric measurement of blood pressure volume and flow in the finger. In: Digest of the 10th International Conference on Medical and Biological Engineering. Dresden, FRG: International Federation for Medical and Biological Engineering; 1973:104.
25. Kobler H, Cejnar M, Hunyor SN. A continuous non-invasive blood pressure monitor. J Electric Electron Eng Aust. 1991;11:102-108.
26. Sherrick CE. Basic and applied research on tactile aids for deaf people: progress and prospects. J Acoust Soc Am. 1984;75:1325-1342.
27. Kelly JJ, Hunyor SN, Ho KY. Evaluation of the reproducibility and accuracy of ambulatory blood pressure monitoring using the Takeda TM-2420 automated blood pressure monitor. Clin Exp Hypertens. 1993;15:271-287.
28. Shah BV, Barnwell BG, Bieler GS. SUDAAN Users Manual: Software for Analysis of Correlated Data. Research Triangle Park, NC: Research Triangle Institute; 1995.
29. Ekman P. Methods for measuring facial action. In: Scherer KR, Ekman P, eds. Handbook of Methods in Nonverbal Behavior Research. New York, NY: Cambridge University Press; 1982:25-90.
30. Furedy JJ, Shulhan D. Specific versus placebo effects in biofeedback: some brief back-to-basic considerations. Biofeedback Self Regul. 1987;12:211-215.[Medline] [Order article via Infotrieve]
31. Sawada Y, Yamakoshi K-I. A preliminary study of beat-to-beat blood pressure biofeedback using the vascular unloading method. Jpn Psychol Res. 1991;33:42-47.
32. Hatch JP, Klatt K, Fitzgerald M, Jasheway LS, Fisher JG. Cognitive and physiologic responses to EMG biofeedback and three types of pseudofeedback during a muscular relaxation task. Biofeedback Self Regul. 1983;8:409-425.[Medline] [Order article via Infotrieve]
33. Goebel M, Viol GW, Orebaugh C. An incremental model to isolate specific effects of behavioral treatments in essential hypertension. Biofeedback Self Regul. 1993;18:255-280.[Medline] [Order article via Infotrieve]
34. Kotses H, Glaus KD, Crawford PL, Edwards JE, Sherr MS. Operant reduction of frontalis EMG activity in the treatment of asthma in children. J Psychosom Res. 1976;20:453-459.[Medline] [Order article via Infotrieve]
35. Patel C, Marmot MG, Terry DJ. Controlled trial of biofeedback-aided behavioural methods in reducing mild hypertension. Br Med J. 1981;282:2005-2008.
36. Carlson JG, Feld JL. Expectancies of reinforcement control in biofeedback and cognitive performance. Biofeedback Self Regul. 1981;6:79-91.[Medline] [Order article via Infotrieve]
37. Brown WA. Placebo as a treatment for depression. Neuropsycho-pharmacology. 1994;10:265-269.[Medline] [Order article via Infotrieve]
38. On practise, practice, and praxis. Lancet. 1993;342:1002-1003. Editorial.[Medline] [Order article via Infotrieve]
39. Hebb DO. Organization of Behavior. New York, NY: John Wiley & Sons Inc; 1964.
40. Treisman A, Gormican S. Feature analysis in early vision: evidence from search asymmetrics. Psychol Rev. 1988;95:15-48.[Medline] [Order article via Infotrieve]
41. Julesz B. Early vision is bottom-up, except for focal attention. Cold Spring Harbor Symposium on Quantitative Biology. The Brain. 1990;55:973-978.
42.
Spitzer H, Desimone R, Moran J. Increased
attention enhances both behavioral and neuronal
performance. Science. 1988;240:338-340.
43. Robinson DL, Petersen SE. The pulvinar and visual salience. Trends Neurosci. 1992;15:127-132.[Medline] [Order article via Infotrieve]
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||