Biofeedback and Self-Regulation, Vol. 7, No. 1, 1982
Skin Temperature Biofeedback for Raynaud's Disease: A Double-Blind Study 1 R. Sergio Guglielmi University of Minnesota Alan H. Roberts 2 Scripps Clinicand Research Foundation Robert Patterson University of Minnesota
The lack o f control procedures inherent in most o f the experiments conducted to assess the effectiveness o f skin temperature biofeedback in the treatment o f Raynaud's disease renders the results inconclusive. In this study, control groups and a double-blind approach are adopted. Thirty-six patients, carefully screened f o r a diagnosis o f primary Raynaud's disease, were assigned to a skin temperature increase group ( N = 12), to an E M G relaxation control group (N = 12), or to a notreatment control group (N = 12). A II patients kept records o f their symptoms f o r the duration o f the study. Each subject in the two training groups received 20 sessions, the last 2 conducted under coM stress. Data analysis according to original group assignment, as well as following regrouping o f subjects according to several learning criteria, showed that while all patients reported a marked decrease in the number o f vasospastic attacks, no significant differences were f o u n d among the three groups on the clinical measures used to assess symptomatic relief. The general improvement reported must therefore be attributed to nonspecific factors. ~This study was supported in part by Rehabilitation Services Administration Grant No. 16-P-56810/5-17 to the University of Minnesota Medical Rehabilitation Research and Training Center. We are grateful to Gaff Gaebe, Carla Grossman, Steve Janousek, Linda Rubbelke, and Scott Williamson, who served as "blind" assistants, and to Steve Sheffield for his technical support. 2Address all correspondence to Alan H. Roberts, Scripps Clinic and Research Foundation, 10666 N. Torrey Pines Road, La Jolla, California 92037. 99 0363-3586/82/0300-0099503.00/0
© 1982 Plenum Publishing Corporation
100
Guglielmi, Roberts, and Patterson
The therapeutic applicability of peripheral skin temperature biofeedback training to such medical conditions as migraine headaches and Raynaud's disease has lately become the object of considerable interest. Raynaud's disease, in particular, would appear to be ideally suited for this form of treatment, not only because the very nature of the vascular derangement involved makes the potential benefit of learning to control peripheral vasomotor responses readily evident, but also because no other satisfactory therapeutic approach is currently available. Most commonly, the management of Raynaud's disease on the part of the physician is limited to reassurance and exhortations to keep the extremities warm and to avoid injuries, tobacco, and emotional upsets. The basic issue in the treatment of Raynaud's disease is prevention; the frequency and duration of the vasospastic episodes must be reduced to avoid the occurrence of irreversible organic changes. The hypothesis that biofeedback therapy might well serve this prophylactic function would appear to be in keeping with the generally encouraging results presented in the experimental reports currently available (Jacobson, Hackett, Surman, & Silverberg, 1973; May & Weber, 1976; Sedlacek, 1976; Shapiro & Schwartz, 1972; Stephenson, 1976; Surwit, 1973; Peper, Note 1). Exhaustive reviews of the literature by Blanchard and Epstein (1977), Blanchard and Miller (1977), and Shapiro and Surwit (1976, 1979) have already critically evaluated these reports. In addition, a Task Force Report of the Biofeedback Society of America (Taub & Stroebel, 1978) summarized research involving approximately 130 patients with vasoconstrictive syndromes and made several suggestions for future research. More recently, Sundermann, and Deik (1978), Adair and Theobald (1978), Freedman, Lynn, and Ianni (1978), and Taub (1977) have all claimed successful treatment of so-called Raynaud's disease with skin temperature biofeedback. These and the previous studies, however, suffer from one or more of the following methodological problems. 1. Anecdotal single cases are reported. 2. No control groups are used to assess the nonspeciflc contribution of general relaxation and, especially, placebo effects, which are likely to be significant in a condition like Raynaud's disease, known to be remarkably influenced by psychogenic factors. 3. Spontaneous remissions induced by warmer weather or by natural fluctuations in symptomatology ("regression. to the mean") are construed as reflecting potent treatment effects. 4. Baseline periods are too short or not used, so that the normal adaptational tendency of hand temperature to increase, sometimes dramatically, after the patient is allowed to sit quietly in a room kept at a comfortable ambient temperature is erroneously interpreted as evidence for voluntary vasodilatation.
Biofeedback for Raynaud's Disease
101
5. No attempt is usually made to preserve the diagnostic homogeneity of the sample studied, and several other peripheral vascular disorders are confused with idiopathic Raynaud's disease (Guglielmi, in press). 6. Many other treatments such as psychotherapy, hypnosis, EMG relaxation, or autogenic training are administered to the same subject in addition to thermal biofeedback. While this multiple treatment approach might be in the patient's best interest, it obviously makes it impossible to determine which of the procedures introduced is actually responsible for the therapeutic benefit claimed. Surwit, Pilon, and Fenton (1978) should be credited with the first serious attempt to remedy some of the deficiencies evident in the previous studies. Their effort needs to be reviewed in detail because of its relevance to the present investigation. The experiment was designed to assess the relative value of a general relaxation technique like autogenic training and of skin temperature biofeedback in the treatment of Raynaud's disease. The effectiveness of laboratory training, as opposed to self-training of the patients at home, was also evaluated. After being diagnostically screened, 30 patients with primary Raynaud's disease were treated either with autogenic training or with a combination of autogenic training and thermal biofeedback. Training, which took place either in the laboratory or at home, was given in two phases: While one group was treated during the month of February, the second group was treated during March, thus serving as a notreatment control, albeit not an independent one, for the month of February. Clinical improvement was assessed by observing the patient's hand temperature response following a cold stress test administered before and after treatment, as well as by examining the frequency and severity of the attacks recorded by the patients on data sheets. The most important finding in this study was that all subjects showed clinical benefit; the introduction of thermal biofeedback did not add to the therapeutic effectiveness of a simple relaxation technique. Moreover, self-training at home was just as effective as laboratory training. While the results of this study seriously challenge the specific therapeutic value of skin temperature biofeedback, limitations in the experimental design do not permit assessment of the contribution to the therapeutic outcome of other nonspecific variables, like placebo and seasonal changes. In fact, the authors' conclusions that clinical improvement was obtained and that this improvement should be attributed to the general relaxation induced with autogenic training is itself very questionable. Comparisons of skin temperature changes during the cold stress test before and after treatment, showing higher temperatures following training, were interpreted by the authors as the most convincing evidence of therapeutic effectiveness.
102
Guglielmi, Roberts, and Patterson
Inspection of those curves, however, reveals that posttreatment temperatures were higher because they started at a higher point before the cold stress, possibly due to the climatic transition from winter to spring. The point-by-point difference between pre- and posttreatment curves actually appears quite similar at the beginning and at the end of the session, indicating a comparable temperature drop in response to the cold stress. This conclusion is clearly supported by the results of the analysis of variance. A significant main effect was found for time (pre- vs. posttraining), which simply confirms the finding of a higher average hand temperature following treatment, and for blocks (7-minute intervals), which implies that for all subjects, decreasing ambient temperature produced a sharp reduction in skin temperature. Surwit and his associates, however, did not find a significant interaction between time and blocks. Only this interaction, if significant, could have demonstrated that, in spite of the initial differences in hand temperature, subjects were able to maintain higher skin temperatures during the cold stress following the treatment. Such finding would have convincingly argued for the clinical effectiveness of training, which must instead be seriously questioned in view of the results reported. Furthermore, when subjects who received the treatment were compared with the waiting-for-treatment controls, no statistically significnt differences were found in the frequency or severity of vasospastic attacks, although there was a trend for the treated patients to show greater improvement. All subjects reported a dramatic reduction in symptom activity with the passage of time, a finding again suggesting the important contribution of seasonal changes. Finally, it seems difficult to reconcile the therapeutic benefit, which the authors assume to be treatment-specific, with the patients' demonstrated inability to learn voluntary vasodilatation either with autogenic training or with skin temperature biofeedback. The investigators admitted that, after a small increase in hand temperature, which followed the 10-minute stabilization period, and which, in the present authors' opinion, could be reasonably explained as a baseline effect since the adaptation period was too short, "subjects actually showed decreases in temperature during the rest of the training" (Surwit et al., 1978, p. 333). The absence of a physiological correlate of the positive clinical findings forcefully argues for the presence of strong placebo or seasonal effects. The study reported here evaluates the relative effectiveness of skin temperature biofeedback and EMG relaxation training in the treatment of Raynaud's disease, while introducing rigorous controls for the contribution of extraneous or nonspecific variables. Subjects were carefully screened to eliminate those with irreversible organic changes in their blood vessels. The ubiquitous influence of placebo was controlled for by adopting a doubleblind experimental design. A separate notreatment control group contin-
Biofeedback for Raynaud's Disease
103
uously provided reference clinical data that allowed assessment of the therapeutic outcome independent of seasonal changes and of spontaneous remissions. Finally, relatively long baselines were used in order to remove the contaminating influence of adaptational effects from the physiological data recorded. METHOD
Subjects A total of 472 volunteers, mostly females, responded to radio announcements, newspaper press releases, and contacts with physicians encouraging people whose "fingers go dead" as a result of cold weather or emotional upset to participate in a research project investigating the possibility of treating the problem. Males, those not living in the city, and those who admitted by telephone or in writing that they had complicating organic disorders (most commonly rheumatoid arthritis or scleroderma) were immediately excluded. The remaining 133 volunteers were sent a cover letter, an informed consent form, and a 16-page, 80-item questionnaire developed to ascertain the severity, frequency, and precise nature of their vasospastic symptoms. The responses from 109 patients were reviewed to detect evidence of organic disease, and 49 patients were excluded at this stage for failing to meet one or more of the following criteria: (1) episodes of bilateral discoloration of the fingers precipitated by cold or emotional stimuli; (2) absence of nutritional changes in the fingertips and nails; (3) no evidence of scalenus anticus, cervical rib, or other neurovascular compression syndromes; (4) absence of excessive swelling in the extremities; (5) absence of skin changes typical of scleroderma or lupus erythematosus; (6) absence of pain in the joints or deformity of the fingers; (7) absence of high blood pressure; (8) no evidence of occupationally induced symptoms; (9) not taking medications known to cause vasospastic symptoms; (10) not taking medications for Raynaud's disease ; (11) no history of sympathectomy; and (12) being naive about the clinical applications of biofeedback. In the interest of diagnostic homogeneity, the criteria listed above were chosen to be considerably more stringent than those suggested by Allen and Brown (1932), which have been typically adopted in the literature. While a step in the right direction, Allen and Brown's criteria still allow a large number of organic disorders to be classified as primary Raynaud's disease (Guglielmi, in press). Sixty patients who met the above criteria were examined to check further for these inclusion variables and in addition were administered a
104
Guglielmi, Roberts, and Patterson
reactive hyperemia test (Abramson, 1974; Roos, 1969; Sanchez, 1966; Strandness, 1969; Strandness & Bell, 1965; Sumner, 1977, 1978). The procedure was conducted by attaching two mercury strain gauge plethysmographs to the right and left middle fingers of the reclining subject. Pulse volume curves were recorded at rest for 5 minutes; then, two blood pressure cuffs, which had been applied around the subject's upper arms, were inflated above systolic pressure for the duration of 5 minutes. They were then quickly deflated and the pulse volume curves were recorded for 5 additional minutes. The timing and pattern of the hyperemic reaction that followed deflation of the cuffs were carefully studied. The amplitude of the pulse volume curves as well as their qualitative characteristics were observed in order to assess the ability of the peripheral blood vessels to dilate normally. Thirty-nine patients passed this evaluation and also agreed to participate in the study. Twelve o f these were assigned to an experimental group and were trained to increase digital skin temperature, 12 were assigned to an E M G control group and were trained to relax their forehead muscles, and the remaining 15 patients were assigned to a no-treatment control group that did not receive laboratory training but instead kept records of their symptoms at home. Three patients in the no-treatment control group dropped out or did not keep adequate records. Patients were randomly assigned to the three groups except that they were matched as closely as possible for age, and frequency and duration of attacks. They ranged in age from 17 to 54. Data for subjects in all three groups are summarized in Table I. Subjects gave informed consent to participate in this study.
Clinical Data
All subjects received a packet of identical diary forms and instructions, except that subjects in the two biofeedback groups were instructed to use the laboratory training at home for at least 5 minutes before using any other method to relieve their attacks. Diaries given to these two laboratory training groups included items to determine whether they tried to use the laboratory training to relieve the attacks, for how many minutes they tried it, and how much relief they obtained?
3Detailed information on the questionnaire and diaries used in the study, as well as on the extensive physiologicaldata collected,may be found in Guglielmi(1979).
Biofeedback for Raynaud's Disease
105
Table I. Group Means and Standard Deviations for Five Preexperimental Variables Skin temperature Group ( N = 12)
Age Number of years since onset Average number of attacks per month during previous winter Average duration of attacks in minutes Actual number of attacks from October 1 to December 1, 1976
EMG Group ( N = 12)
No-treatment group ( N = 12)
Mean
SD
Mean
SD
Mean
SD
33.25
9.62
33.83
8.72
34.33
11.13
9.50
7.42
11.00
9.61
10.08
8.36
22.67
8.55
26.67
10.71
24.50
10.89
21.92
15.11
20.67
9.04
27.92
15.99
29.67
23.62
29.58
30.13
21.42
21.85
Each diary contained 13 items to determine precipitating incidents, symptoms, and distribution and severity. A diary was completed for each attack that was defined as not just "episodes of cold hands, but rather specific changes that take place in your fingers; in particular, color changes (white and/or blue) and numbness." Subjects in the training groups brought in completed diaries when they came in for weekly training sessions. The untrained group was provided with prestamped envelopes to mail in diaries weekly. These subjects were frequently contacted by phone by an assistant in order to maintain their interest in the experiment and to remind them of the importance that they fill out the dairies scrupulously and send them in weekly.
Equipment Skin temperature was measured by Yellow Springs thermilinear components with a Type 727 surface temperature probe. Thermistors were taped to the center of the distal pad of each middle finger. Subjects in the skin temperature group received feedback for temperature changes from the middle finger of their dominant hand. The time constant of the probe in a liquid Was .5 sec. Electronic circuitry was developed to provide four channels of direct temperature readout and one channel at a higher gain with zero suppression for temperature change information. The full-scale
106
Guglielmi, Roberts, and Patterson
output of the feedback channel was _+ 1V for a +_ 1 °C temperature change, which also represented the full-scale change for the audio and visual channels. The feedback was reset to midrange each time the subject's finger temperature exceeded i °C from the midpoint of the scale. The E M G was recorded using a remote preamplifier built with an Analog Device model 521 amplifier. Beckman Ag-AgC1 electrodes were attached to the subjects' forehead following Lippold's (1967) suggestions about lead placement. The preamplifier had a gain of 100. The output of the E M G preamplifier was connected to a Gould model 11-4307-02 E E G coupler. The band width of the amplifying system was from 10 to 1,000 Hz. The output of the E E G coupler was connected to a full wave rectifier and averager, and was recorded on one channel of the Gould recorder. The averaging system had a time constant of 33 sec for increases in E M G signals and a time constant of 15 sec for E M G decreases. These relatively long time constants made the auditory and visual feedback signal controlled by the averaged E M G output much more sluggish and comparable to the feedback derived from skin temperature changes. A fast-acting limit detector circuit, which had a time constant of 100 msec, was developed in order to prevent large and rapid increases in the E M G feedback that may result from brief or long-acting motor responses. When the raw E M G signal exceeded the limit of the fast-acting detector, the averaged E M G signal that controlled the feedback was held, thus eliminating the danger that subjects might deduce the muscular origin of the feedback when phasic increases in muscle tension were produced. Finger pulse volume changes were recorded using two Park Electronics model 270 mercury strain gauges, which were applied around the two index fingers with the copper wire junction resting on the base o f the nails. In order to eliminate movement artifacts from the feedback signal, a double-sided limit detector circuit completely cut off the feedback signal when the output from either plethysmograph exceeded the high or low level limits o f the detector. The limit detector circuits would hold or cut o f f the feedback signal whenever undesired movements from the face or the hands were produced and were activated whether the feedback originated from skin temperature or from E M G changes.
Double-Blind Design In order to control placebo effects and to minimize experimenter bias, a double-blind design was adopted. The subjects selected for the experiment were told that they would be assigned to several treatment groups so that the therapeutic effects of different training conditions could be compared. They were also told that if one of the treatments proved to be more benefi-
Biofeedback for Raynaud's Disease
107
cial, the subjects assigned to the other conditions would be given, at the conclusion of the experiment, the opportunity to receive the more beneficial treatment. Four assistants served as primary therapists and carried out all necessary interactions with the subjects. Both the patients and the assistants were asked to refrain from reading about biofeedback and from discussing the experiment with each other as long as the study was in progress. Neither the subjects nor the assistants were informed about how many treatment conditions were being studied, the particular physiological functions recorded, the overall research design, or the working hypothesis. The assistants scheduled the appointments, "hooked up" the subjects with the devices used to measure various physiological functions, and sat with the patients in the isolated experimental room, offering support and encouragement for their performance. Throughout the duration of the study the experimenter, who operated the recording and feedback equipment in an adjacent room, had no interaction with the subjects except to greet them when they came in for their weekly sessions, or when they left, and to signal the beginning and end of each trial. All subjects who received laboratory training were "prepared" exactly in the same manner: Thermistors, mercury strain gauges, and EMG electrodes were applied to the skin of each subject. They all received both auditory and visual feedback, but they were not told about the nature of direction of the physiological change upon which feedback was contingent. They were simply instructed to "drive" the feedback meter and tone in one direction. All subjects who received laboratory training were informed that their task would be to move the meter and the tone, which were centered at the beginning of each trial, as far to the left as possible. For both the skin temperature and EMG groups, therefore, good performance was reflected by a deviation to the left of the feedback tone and meter proportional to the magnitude of the physiological change. When the equipment was tested on pilot subjects in preparation for the experimenter, it was observed that even small movements of the facial musculature, and especially twitching the eyebrows or tightening the jaws, would produce immediate and large deviations of the meter and tone, thus making it possible for subjects in the EMG group to infer the muscular origin of the feedback they received. This would, of course, undermine the efforts to preserve the double-blind approach. It was necessary to make the feedback given to the EMG group more sluggish, with the large phasic changes in muscle activity cut out in order to better reflect the tonic levels of muscle tension maintained over a period of time. In short, it was necessary to make the EMG feedback more comparable to the feedback produced by skin temperature changes. This was accomplished by the electronic circuitry just described.
108
Guglielmi, Roberts, and Patterson
Training Sessions All training sessions were conducted between January 17 and June 19, 1977. Each of the 24 subjects selected for laboratory training, equally divided into two treatment groups, received 20 weekly training sessions. The subjects reclined in a semirecumbent position on a hospital bed in a dimly lit, electrically shielded, temperature-controlled, and soundproof room, with ambient temperature set at 22.5°C _+ .5°C. Feedback was provided by means of stereo headphones and a visual meter located above the bed, visible also to the assistant, who sat on a chair a few feet away from the subject. The experimenter operating the equipment in an adjacent room could observe the subject through a window. Subjects were asked to abstain from drugs and from cigarettes at least 1 hour before their weekly appointment. Except for the difference in the origin of the feedback given to the two groups, all sessions were identical for all subjects. Each session was approximately 1 hour in length. An initial 20-minute baseline period was followed by three 8-minute learning trials. A 5-minute rest period followed each trial. The experimenter signaled the beginning and end of each trial via intercom. During baseline and rest periods the feedback was turned off. The feedback meter and tone were centered by the experimenter to the midpoint of their full-scale range at the beginning of each trial. In order to facilitate transfer of learning from the laboratory to real life and to obtain an objective assessment of the practical applicability of voluntary vasomotor control, the last two sessions for each subject were conducted under vasoconstricting conditions simulated in the laboratory by lowering the room temperature to 18.5°C. All subjects in the two treated groups were also instructed to practice every day at home for 10 minutes the training they were receiving in the laboratory. Each of the four assistants, two males and two females, was assigned six subjects, three from each of the two treatment groups. The assistants, of course, had no knowledge about the number of treatment conditions studied and the particular representation of those different conditions among the subjects assigned to them. The assistants were instructed to offer copious encouragement and reinforcement in order to maintain a positive expectancy in all subjects. Each patient worked with the same assistant for the duration of the study.
Double-Blind Questionnaire After all experimental sessions had been completed, both the assistants and the subjects were asked to fill our a short questionnaire developed to assess how successfully the double-blind approach had been preserved.
Biofeedback for Raynaud's Disease
109
The questionnaire included items requiring subjects and assistants to identify the specific bodily functions recorded by each of the three sets of devices used, to specify from which of the three sites or combination of them the visual and auditory feedback originated, to cite the specific bodily changes causing the feedback meter and tone to move in each particular case, and to state whether the subject had been assigned to the treatment condition expected to be clinically most effective (i.e., the skin temperature group) or to one expected to be less effective (i.e., the EMG relaxation group). This latter question appeared most useful in establishing the adequacy of double-blind procedures as a way of controlling for the effects of patient expectancy and experimenter bias. RESULTS
Double-Blind Questionnaire Data Table II shows the percentage of subjects and assistants correctly answering critical items on the questionnaire. It is evident from these data that patients and assistants remained essentially unaware of the basic research design and experimental hypothesis. Their answers reveal that correct identifications were made often considerably below, and certainly not significantly above, chance level. It might be of interest to note that most of the patients thought that the forehead electrodes measured "brain waves," and only 13°70 recognized the muscular nature of the response recorded. It is also noteworthy that the majority of the subjects (78%) believed that the feedback originated from physiological changes recorded by the forehead electrodes. This finding explains the discrepancy between the two groups in response to question 2.
Dependent Variables and Data A nalyses Nine clinical measures, extracted from the diary forms filled out by the patients during the course of the study, were analyzed in order to detect possible differential treatment effects among the three groups. These dependent variables included frequency of attacks, duration in minutes, rated severity of attacks, extent of hand involvement, number of symptoms experienced, rating of pain, amount of impairment, length of time spent using laboratory training to relieve the attack, and amount of relief obtained from that effort. For each of the nine outcome variables, group means were obtained separately from each of 5 months and cumulatively for the total period of
110
Guglielmi, Roberts, and Patterson
I I I = =
0
I I I 6 0
< r~
r~ r~ 0
=
c
= -~
0.~
~.~ ~ ~ 0 0 ~ ~
~/
~'~
CD
~
~ ~ ~'~
Biofeedback for Raynaud's Disease
11 |
the experiment, which was 22 weeks. Separate analyses were performed on those six time blocks. In order to eliminate small differences among subjects with respect to the time they began and stopped filling out the diaries, the first and last weeks of the experiment were not used in forming the monthly data periods, and each month was composed of 4 of the remaining 20 weeks. Analyses of variance and covariance were performed on each outcome. For the covariance analyses, three preexperimental measures collected before the subjects were assigned to the three groups were used as covariates. These included the subject's estimate of the average number of attacks per month experienced during the previous winter, the average duration of those attacks in minutes, and the actual number of attacks experienced from October 1 to December 1, 1976. Although an attempt had been made to match subjects on these variables when they were assigned to the three groups, it seemed advisable to use covariate adjustment to control for any residual preexperimental differences among the three groups (Kirk, 1968; Winer, 1971). In spite of the reduction in experimental error and added precision expected from this type of statistical control, the results obtained were basically the same as those produced by simple analyses of variance, and only the latter will be reported in this paper. Statistical analyses were first performed in order to detect differential treatment effects among subjects as originally assigned to the three groups. However, it became difficult to interpret the results obtained because it was evident that not all subjects who received laboratory training had learned to produce reliable physiological changes in the prescribed direction, and some had actually "learned" to produce the opposite response. The clinical data, therefore, were subsequently reanalyzed after subjects had been assigned to new groups formed according to several learning criteria to be described later. The 5% significance level was adopted in all cases.
Data Analysis According to Original Group Assignment The results of the analyses of variance on each dependent variable were tabulated for each of the six time blocks. In order to properly interpret the significance of the differences among the group means, it is important to note that the number of subjects in each group contributing to those means and the number of attacks they experienced became conspicuously smaller as the warmer months are considered, particularly months 4 and 5 of the study. In those months, a very deviant score of one subject could produce large differences in group means, which, although statistically significant, do not necessarily reflect differential treatment effects.
112
Guglielmi, Roberts, and Patterson
Table III. Total Number of Attacks Experienced by Each Group and Number of Subjects
Who Experienced at Least One Attack During Each Month and for the Total Period of the Experiment Number of attacks
Month 1 Month 2 Month 3 Month 4 Month 5 Total period
Skin temperature group 92 69 39 5 6 225
EMG group 91 58 26 14 1 204
Notreatment group 88 53 42 21 11 254
Number of subjects Skin temperature group 9 10 9 4 3 12
EMG group 12 10 5 5 1 12
Notreatment group 11 8 7 4 2 12
Table III shows the total number of attacks experienced by each group for each of the 5 months and for the total period. The number of subjects in each group who experienced at least one attack during each time period and who therefore contributed to the computation of the group means is also shown in Table III. It can be seen, for example, that the average values obtained for the E M G group on each outcome variable for m o n t h 5 are based on one attack experienced by one subject. The figures shown in Table III remain the same for all outcome variables except for frequency of attacks, in which case daily averages over the six time periods were obtained for each group. No significant differences were found a m o n g the three groups for the five monthly time periods analyzed separately or for the total period of the experiment for the following variables: frequency of attacks, duration of attacks, severity of attacks, extent of hand involvement, rating of pain, and, for the two laboratory groups, time spent using laboratory training and a m o u n t o f relief following use of laboratory training. The number of symptoms checked by subjects f r o m a s y m p t o m checklist showed a significant difference only for month 4, F(2, 10) = 17.99, p < .001. The meaning of this finding, however, is questionable in view of the small number of attacks reported during that period (see Table III). When ratings of the a m o u n t of impairment experienced in association with each attack were analyzed, a statistically significant difference was found only for m o n t h 4, F(2, 10) = 9.62, p < .005.
Data Analyses Aecording to Learning Criteria Average hand temperature in degrees Celsius, and frontalis E M G in microvolts were plotted against the 57 minutes constituting each session for
Biofeedback for Raynaud's Disease
l 13
each subject separately. Inspection of these data revealed that one subject assigned to the E M G relaxation group showed the largest and most consistent increases in hand temperature across sessions. This observation, and the possibility that other subjects included in the E M G group might have actually learned to raise skin temperature and vice versa, made it advisable to perform additional analyses on each clinical variable after subjects had been reassigned to new groups formed by some criteria for learning, first considering the number of trials during Which changes in the prescribed direction were produced, and then paying attention to the magnitude of those changes. These criteria for choosing learners and nonlearners, regardless of original assignment, were applied independently to both skin temperature and E M G data. Learners and nonlearners were then compared on the various clinical measures with the group that did not receive laboratory training. Learning Criterion 1. Each training session included three 8-minute trials. For both skin temperature and E M G data the average of the first 4 minutes of each trial was subtracted from the average of the last 4 minutes. Only the sign of the difference was considered, not its magnitude. Skin temperature learners ( N = 11) were defined as those subjects whose difference had a positive sign in 50% or more of the trials, while skin temperature nonlearners ( N -- 13) were those patients who showed a positive difference in 49% or less of the trials. Similarly, E M G learners ( N = 10) were identified as those subjects who had a negative difference in 50% or more of the trials, while E M G nonlearners ( N = 14) were those subjects who showed a negative difference in 49% or less o f the trials. Analyses of variance were again performed on each outcome variable for the regrouped subjects compared with those subjects who did not receive laboratory training. The only significant differences among groups reassigned according to this skin temperature learning criterion were found in the duration of the attacks for the 4th m o n t h of the experiment, F(1, 10) = 4.18, p < .05, with the no-treatment group having the lowest mean, and in the amount o f time spent using laboratory training for the total period, F(1, 20) = 4.72, p < .05, which was greater for the temperature increase group but which was not paralleled by a significantly greater amount of relief, F(1, 20) = 1.58, n.s. Analyses based on E M G learning criterion 1 yielded statistically significant difference only in the amount of impairment for month 3, F(2, 18) = 4.27, p < .05, and for the total period, F(2, 33) = 4.62, p < .02, which, in both cases, was greater for the E M G nonlearners and lowerst for the no-treatment control group. Learning Criterion 2. The 50% correct trials cutoff adopted for learning criterion 1 allowed, on the one hand, utilization of all subjects, thus making comparisons more meaningful, but it also led to the inclusion of borderline cases that might have mitigated possible differences among the three groups. Therefore, the following more stringent criteria were
114
Guglielmi, Roberts, and Patterson
used to form new skin temperature and E M G learning groups. Skin temperature learners ( N = 5) were those subjects who showed a positive difference between the second half and the first half of each trial in 75°70 or more of the trials. Skin temperature nonlearners ( N = 5), on the other hand, were defined as those having a positive difference in only 27°70 or less of the trials. E M G learners ( N = 6) were defined as those subjects who showed a negative difference in 55°70 or more of the trials. E M G nonlearners ( N = 6) showed a negative difference in only 26070 or less of the trials. These new groups were again compared with the 12 subjects assigned to the no-treatment control group. While the skin temperature learners, according to this criterion, spent a significantly larger amount of time trying to abort their attacks with the use of laboratory training for the total period, F(1, 6) = 9.44, p < .025, they again did not receive significantly greater relief than the nonlearners, F(1, 6) = .05, n.s. The only other statistically significant differences found among the three groups were in the amount of impairment associated with the vasospastic attacks. Those differences, however, were not in the expected direction; for both skin temperature, F(2, 19) = 5.08,p < .02, and EMG, F(2, 21) = l l . 1 2 , p < .001, learning criteria, the no-treatment group reported the least amount of impairment. Learning Criterion 3. Failure to detect clear treatment effects among the subjects as originally assigned to the three groups, as well as reassigned according to the two learning criteria just described, suggested the desirability of testing the effects of another method of regrouping subjects based on the magnitude of the physiological changes produced during each trial rather than just on the direction of those changes. Different ways of quantifying magnitude of change were tested. Here only the learning criterion that yielded the largest group differences is described. The average of the first 4 minutes of each trial was again subtracted from the average of the last 4 minutes. Only the differences reflecting changes in the prescribed direction were considered and summed over the 54 trials. Skin temperature learners ( N = 15) were those subjects whose sum was equal to or greater than 29.02°C, while nonlearners ( N = 5) had a sum equal to or lower than 2.69°C. E M G learners ( N = 5) were those subjects who showed a sum equal to or smaller than - 12.89 V; nonlearners ( N -- 5) had a sum equal to or greater than - 3 . 3 7 V. These new learning groups were again compared with the notreatment group. Groups formed according to skin temperature learning criterion 3 showed significant differences in the frequency of attacks during the 1st, F(2, 19) = 5.73, p < .02, and the 2nd, F(2, 19) = 4.78, p < .03, months, in their ratings of pain during the 2nd month, F(2, 14) = 4.79, p <
Biofeedback for Raynaud's Disease
115
.03, and in the amount of impairment experienced during he 1st, F(2, 17) = 6.68, p < .01, and 4th, F(2, 7) = 5.91, p < .04, months. Mean differences were not in a consistent direction, however. No significant differences emerged for groups formed according to EMG learning criterion 3. CoM Stress As noted earlier, the last two training sessions were conducted under cold stress conditions. It was hypothesized that the ability to increase skin temperature, or at least to prevent marked decreases, in the presence of environmental conditions normally producing vasoconstriction would provide the most reliable indication of learning and would therefore be associated with the greatest degree of clinical improvement. Five new skin temperature learning criteria, similar to the ones described earlier and reflecting both the direction and the magnitude of the temperature changes, were derived from the six trials given during the two cold stress sessions, and were used to again reassign subjects to new groups that were compared on all clinical variables with the no-treatment control group. Analyses of variance revealed no statistically significant differences beyond chance level except for the amount of impairment experienced during the attacks. For all five learning criteria the skin temperature learners tended to score highest and the notreatment group lowest on this variable.
DISCUSSION The results of this experiment do not support the hypothesis that skin temperature biofeedback or EMG relaxation training are effective in the treatment of Raynaud's disease. While patients in all three groups showed a marked decrease in the number of attacks experienced as the study progressed, no significant differences were found, with the exception of the cases indicated above, for any of the clinical measures used to assess symptomatic remission among the group trained to increase skin temperature, the EMG relaxation group, and the no-treatment controls. The therapeutic benefit reported, therefore, must be ascribed not to the treatments administered but rather to nonspecific factors, which might include placebo, regression to the mean, and seasonal changes. The improvement reported by subjects who did not receive laboratory training would at first seem to rule out the contribution of placebo effects. It should be noted, however, that these patients were in regular contact with an assistant who offered constant reinforcement for good record-keeping
116
Guglielmi, Roberts, and Patterson
and reminded them of the importance of their participation in the study. Moreover, these patients' expectancy and motivation were maintained high by the promise that they could receive any effective laboratory training following the conclusion of the experiment. Finally, it has been noted (Keefe, Kopel, & Gordon, 1978) that the task of keeping clinical records can itself be associated with a tendency to report feeling better. Certainly the largest portion of the improvement observed in the three groups must be attributed to the climatic changes that occurred during the course of the study, which started January 17 and terminated June 19. The consistent decrease in symptomatic activity, as measured by attack frequency, recorded by all subjects for each month of the experiment closely parallels the warming trend associated with the seasonal changes. Since cold is known to be the primary stimulus for the vasospastic attacks, this finding should be fully expected, and it only reaffirms the importance of including a no-treatment control group in experiments designed to evaluate the effectiveness of biofeedback procedures in the treatment of Raynaud's disease. It could be argued that this study has failed to conclusively rule out the specific therapeutic efficacy of thermal biofeedback because the reduction and almost disappearance of attacks observed during the warm weather may have concealed or "washed out" possible treatment effects. It should be noted, however, that subjects in this experiment received more training sessions (20) than have been usually reported in the literature. In the study by Surwit et al. (1978), for example, only six sessions were given to the patients. It appears sensible to assume, therefore, that if treatment effects were present, they should have emerged but did not after 2 or 3 months of training, when symptom activity was still substantial (see Table III). Rather than refute the findings reported here as being inconclusive because environmental factors may have allegedly obscured the presence of treatment effects, it seems more reasonable to dismiss the "positive" results claimed by many of the previous studies on the grounds that they were contaminated by seasonal and other nonspecific variables that were not controlled. The absence of specific treatment effects in this experiment seems also to be demonstrated by the observation that while all subjects experienced fewer and fewer attacks as the weather became warmer, no corresponding improvement or group differences were obtained on the other outcome variables. This would suggest that once an attack occurred, its severity, as determined by the clinical measures used in its study, was not mitigated either by the experiments introduced or by the climatic changes. It could also be argued that the failure to uncover specific treatment effects might be attributed to the finding that few subjects actually learned to produce large and reliable increases in hand temperature. While it is possible that a training program much more effective in teaching voluntary
Biofeedback for Raynaud's Disease
117
vasodilatation might improve the therapeutic outcome, it should be remembered that the difficulty in learning to raise peripheral skin temperature is not at all peculiar to this study. The same problem has been reported in normal subjects (Alberstein, 1977; Lynch, Hama, Kohn, & Miller, 1976; Packer & Selekman, 1977; Shapiro & Surwit, 1979; Surwit, Shapiro, & Feld, 1976; Turin, 1977) as well as in patients with Raynaud's disease (Surwit et al., 1978), and it recently led Surwit (1978) to conclude that "a review of the experimental literature produces only equivocal evidence for the ability of humans to learn to voluntarily vasodilate with the use of temperature feedback alone" (p. 10). Furthermore, when the subjects of this study were reassigned to new groups formed according to several skin temperature learning criteria derived from the first 18 training sessions, so that those subjects who did learn some degree of vasodilation could be compared with those who learned vasoconstriction and with the no-treatment control group, again no differences were found on the clinical outcome variables. Not even regrouping of subjects based on the most powerful criterion of learning, as determined by the patients' ability to increase hand temperature, or at least to prevent vasoconstriction, during the cold stress sessions, could demonstrate the therapeutic advantage of learning such control. Five separate temperature learning criteria were used to isolate learners and nonlearners, and while several subjects were able to vasodilate under cold stress conditions, they did not differ from the nonlearners on the clinical outcome variables. In fact, the only consistently significant statistical differences obtained were in the amount of impairment suffered during the vasospastic attacks, which was highest for the skin temperature learners and lowest for the subjects who did not receive laboratory training. A third argument that could be used to challenge the results of this study might be that the subjects' ignorance of the feedback-relevant response inherent in the double-blind studies necessarily led to inadequate training and therefore to negative findings. The success in training animals, obviously unaware of the target response, to control a variety of autonomic functions would seem to invalidate this contention. Furthermore, in recent years a body of literature has accumulated indicating that what is true for rats also applies to humans. Several experiments designed to compare the performance of subjects who were aware and those who were not aware of the response contingencies were recently reviewed in the report of a task force of the Biofeedback Society of America (Carlson, Note 2). The majority of those studies indicates that knowledge of the feedback-relevant response is unnecessary, if not detrimental, in training humans to control physiological functions like heart rate, EEG patterns, galvanic skin response, and finger pulse volume. It has been shown that subjects can even overcome deceiving information about the target response (Headrick,
118
Guglielmi, Roberts, and Patterso n
Feather, & Wells, 1971). There is no experimental evidence indicating that response awareness is essential or even helpful in learning to control peripheral skin temperature, whereas at least two studies have demonstrated that such awareness is unnecessary (Gardner & Keefe, 1976; Sdorow & Masterson, Note 3). The task force concluded their review of this issue by noting that "there appears to be no basis for the claim by many clinicians that awareness of the feedback-relevant response is necessary in order to achieve self-control over the response . . . . In fact, the weight of the evidence to date indicates that nonawareness produces results equal to or better than awareness" (p. 7). This study essentially replicates the negative findings of another investigation, which failed to uncover specific therapeutic effects of skin temperature biofeedback in the treatment of migraine headaches (Kewman & Roberts, 1980). It is noteworthy thatthe only other controlled study on a large sample of Raynaud's disease patients (Surwit et al., 1978) determined that thermal biofeedback added nothing to the benefit derived from general relaxation. It was pointed out earlier in this paper that methodological limitations inherent in that investigation, particularly the absence of an independent no-treatment control group, made it difficult to reveal the contribution of other nonspecific variables uncovered by this study. In fact, when closely examined, the outcome of that experiment is certainly consistent with the hypothesis that the symptomatic improvement reported could simply be a function of seasonal warming. The results of the present investigation clearly indicate that the best treatment for Raynaud's disease is warm weather. As obvious as this finding may be, it assumes great significance in interpreting the existing literature where the contaminating influence of this insidious environmental variable has been surprisingly unrecognized. As indicated earlier, however, these results also invalidate the contention that specific treatment effects might have been present but were overshadowed by the seasonal changes. The evidence available to date compels the conclusion that therapeutic applications of skin temperature biofeedback are at least premature. Before those clinical possibilities can be entertained, the difficulties reported in learning significant degrees of peripheral vasoditatation as well as the parameters and conditions necessary to optimize that training must be carefully examined. It is very disconcerting that "despite the clear lack of experimental evidence to support assertions of the effectiveness of biofeedback as therapy.., increasing numbers of clinicians are incorporating biofeedback into their practice" (Katkin, Fitzgerald, & Shapiro, 1978, p. 286). If additional investigations of the therapeutic value of skin temperature biofeedback are conducted, the adoption of adequate controls, possibly in a double-blind situation, seems essential. These results fully agree with the exhortation made by Katkin and co-workers, who concluded their
Biofeedback for Raynaud's Disease
119
review of the clinical applications of biofeedback by observing that "there is powerful desire among health and health-related professionals to be able to provide treatment. It is incumbent upon the serious scientist to temper that noble desire with an equally noble appreciation for the value of hard evidence, and the need for caution and patience" (Katkin et al., 1978, p. 286).
REFERENCE NOTES 1. Peper, E. Case report. Paper presented at the annual meeting of the Biofeedback Research Society, Boston, 1972. 2. Carlson, J. G. (Chair). Biofeedback as a research tool (Biofeedback Society of America Task Force Study Section Report). Denver: Biofeedback Society of America, 1978. 3. Sdorow, L. M., & Masterson, J. Effects o f knowledge o f target response, feedback, and direction o f change on self-control o f fingertip temperature. Unpublished manuscript, Department of Behavioral Sciences, St. Francis College, Loretto, Pennsylvania, 1977.
REFERENCES Abramson, D. I. Vascular disorders o f the extremities (2nd ed.). Hagerstown, Maryland: Harper & Row, 1974. Adair, J. R., & Theobald, D. E. Raynaud's phenomenon: Treatment of a severe case with biofeedback. Journal o f the Indiana State Medical Association, 1978, 71, 990-993. Alberstein, B. Biofeedback and skin temperature control. Psychophysiology, 1977, 14, 115. (Abstract). Allen, E. V., & Brown, G. E. Raynaud's disease: A critical review of minimal requisites for diagnosis. American Journal o f the Medical Sciences, 1932, 183, 187-200. Blanchard, E. B., & Epstein, L. H. The clinical utility of biofeedback. In M. Hersen, R. M. Eisler, & P. M. Miller (Eds.), Progress in behavior modification (Vol. 3). New York: Academic Press, 1977. Blanchard, E. B., & Miller, S. T. Psychological treatment of cardioavascular disease. Archives o f GeneraI Psychiatry, 1977, 34, 1402-1413. Freedman, R., Lynn, S. J., & Ianni, P. Biofeedback treatment of Raynaud's phenomenon. Biofeedback and Self-Regulation, 1978, 3, 230. (Abstract) Gardner, E. T., & Keefe, F. J. The effects of knowledge of response on temperature biofeedback training. Biofeedback and Self-Regulation, 1976, 1, 314. (Abstract) Guglielmi, R. S. A double-blind study o f the effectiveness o f skin temperature biofeedback as a treatment f o r Raynaud's disease. Unpublished doctoral dissertation, University of Minnesota, 1979. Guglielmi, R. S. The neurogenic basis o f Raynaud's disease. Montreal: Eden Press, (in press). Headrick, M. W., Feather, B. W., & Wells, B. T. Undirectional and large magnitude heart rate changes with augmented sensory feedback. Psychophysiology, 1971, 8, 132-142. Jacobson, A. M., Hackett, T. P., Surman, O. S., & Silverberg, E. L. Raynaud's phenomenon: Treatment with hypnotic and operant technique. Journal o f the American Medical Association, 1973, 225, 739-740. Katkin, E. S., Fitzgerald, C. R., & Shapiro, D. Clinical applications of biofeedback: Current status and future prospects. In H. L. Pick, H. W. Leibowitz, J. E. Singer, A. Steinschneider, & H. W. Stevenson (Eds.), Psychology: From research to practice. New York: Plenum Press, 1978. Keefe, F. J., Kopel, S. C. Gordon, S. B. A practical guide to behavioral assesment. New York: Springer, 1981.
120
Guglielmi, Roberts, and Patterson
Kewman, D. G., & Roberts, A. H, Skin temperature biofeedback and migraine headaches: A double-blind study. Biofeedback and Self-Regulation, 1980, 5, 327-345. Kirk, R. E. Experimental design: Procedures for the behavioral sciences. Belmont, California: Brooks/Cole, 1968. Lippold, O. C. J. Electromyography. In P. H. Venables & I. Martin (Eds.), A manual ofpsychophysiological methods. Amsterdam: North-Holland, 1967. Lynch, W. C., Hama, H., Kohn, S., & Miller, N. E. Instrumental control of peripheral vasomotor response in children. Psychophysiology, 1976, 13, 219-221. May, D. S., & Weber, C. A. Temperature feedback training for symptom reduction in primary and secondary Raynaud's disease. Biofeedback and Self-Regulation, 1976, 1, 317. Packer, L. E., & Selekman, W. L. Within-subjects control in EMG and thermal biofeedback training: The baseline effect. Biofeedback and Self-Regulation, 1977, 2, 293-294. Roos, D. B. Plethysmography: A simple method of studying and following peripheral vascular disorders. Surgical Clinics of North America, 1969, 49, 1333-1342. Sanchez, S. A. Strain gauge plethysmography: A diagnostic aid in obstructive arterial disease. Postgraduate Medicine, 1966, 39, 451-457. Sedlacek, K. EMG and thermal feedback as a treatment for Raynaud's disease. Biofeedback and Self-Regulation, 1976, 1, 318. (Abstract) Shapiro, D., & Schwartz, G. E. Biofeedback and visceral learning: Clinical applications. Seminars in Psychiatry, 1972, 4, 171-184. Shapiro, D., & Surwit, R. S. Learned control of physiological function and disease. In H. Leitenberg (Ed.), Handbook of behavior modification and behavior therapy. Englewood Cliffs, New Jersey: Prentice-Hall, 1976. Shapiro, D., & Surwit, R. S. Biofeedback. In O. F. Pomerleau & J. P. Brady (Eds.), Behavioralmedicine: Theory and practice. Baltimore: Williams and Wilkins, 1979. Stephenson, N. E. Two cases of successful treatment of Raynaud's disease with relaxation and biofeedback training and supportive psychotherapy. Biofeedback and Self-Regulation, 1976, 1, 318-319. (Abstract) Strandness, D. E., Jr. Peripheral arterial disease: A physiologic approach. Boston: Little, Brown, 1969. Strandness, D. E., Jr., & Bell, J. W. Peripheral vascular disease: Diagnosis and objective evaluation using a mercury strain gauge. Annals of Surgery, 1965, 161(Suppl.), 1-35. Sumner, D. S. Digital plethysmography. In R. B. Rutherford, V. M. Bernhard, F. E. Maddison, W. S. Moore, M. O. Perry, & D. S. Sumner (Eds.), Vascular surgery. Philadelphla: Saunders, 1977. Sumner, D. S. Mercury strain-gauge plethysmograph. In E. F. Bernstein (Ed.), Non-invasive diagnostic techniques in vascular disease. St. Louis: Mosby, 1978. Sundermann, R. H., & Delk, J. L. Treatment of Raynaud's disease with temperature biofeedback. Southern Medical Journal, 1978, 71, 340-342. Surwit, R. S. Biofeedback: A possible treatment for Raynaud's disease. Seminars in Psychiatry, 1973, 5, 483-490. Surwit, R. S. Peripheral vasomotor: Raynaud's disease. Behavioral Medicine Newsletter, 1978, l(3), 9-I1. Surwit, R. S., Pilon, R. N., & Fenton, C. H. Behavioral treatment of Raynaud's disease. Journal of Behavioral Medicine, 1978, 1, 323-335. Surwit, R. S., Shapiro, D., & Feld, J. L. Digital temperature autoregulation and associated cardiovascular changes. Psychophysiology, 1976, 13, 242-248. Taub, E. Self-regulation of human tissue temperature. In G. E. Schwartz & J. Beatty (Eds.), Biofeedback: Theory and research. New York: Academic Press, 1977. Taub, E., & Stroebel, C. F. Biofeedback in the treatment of vasoconstrictive syndromes. Biofeedback and Self-Regulation, 1978, 3, 363-373. Turin, A. C. Biofeedback and suggestion in finger temperature training: An effect for the controls but not the "treatments." Biofeedback and Self-Regulation, 1977, 2, 296. (Abstract) Winer, B. J. Statisticalprinciples in experimental design. New York: McGraw-Hill, 1971. (Revision received October 18, 1981)