"Direction of Temperature Control in the Thermal Biofeedback Treatment of Vascular Headache", by Edward B. Blanchard et al., (Applied Psychophysiology and Biofeedback, 22:4, December 1997 [Received in May, 1998], p. 227-246
Just when we were getting close to celebrating the complete Internet publication of Shellenberger & Green's "The Ghost in the Box" along comes another "experimental" Headache study that proves there's no ghost in the book, either. That is, it is perfectly possible to READ the book without learning anything from it.
In "Direction of Temperature Control in the Thermal Biofeedback Treatment of Vascular Headache", by Edward B. Blanchard et al., (Applied Psychophysiology and Biofeedback, 22:4, December 1997 [Received in May, 1998!], p. 227-246 the pursuit of the demons "specific effects" and "confounding variables" proceeds with a vengeance. In an attempt to cover criticism, the authors make a bold attempt to apply "criteria" to assert that at least some "learning" took place, in spite of doing everything scientifically possible to prevent it.
In their discussion, the authors admit that "our TBF-Warm condition with no home practice does not mirror conventional clinical practice (p. 243)." Indeed, hardly any aspect of the biofeedback "therapy" they tested bears any resemblance to "conventional clinical practice", except perhaps for the presence of the magic box with its implied "ghost" inside.
For example, the 70 patients in four treatment conditions were admonished NOT to practice on their own at home, without biofeedback, as this "might interfere with learning the appropriate skill (p. 234)". Absent also was the biofeedback practitioner as "coach"; for most of the sessions (10 out of 12) the coach was not even in the room. Absent also was honest feedback from the experimenter, who essentially lied to most subjects most of the time. Instead of acknowledging a "bad day", the experimenters gave all subjects "mild" encouragement regardless of whether it was warranted or not.
Even more aggrievious, the patients were apparently never advised to generalize their training to the real world. There is no mention of any relevance of training strategies to the onset of HA symptoms. [In contrast, I require my HA patients to constantly monitor finger temperature and engage in counter-measures as soon as trouble starts.]
A GHOST EVALUATION
If we look at the "Methodological Errors" described in Chapter 2 of The Ghost, we find that in spite of having had more than a decade to learn to avoid them, Blanchard et al. still commit almost every error in the book. Well, a goodly number, anyway.
Blanchard's patients were allowed 12 biofeedback sessions spread over a period of six weeks. Since the TBF-Warm patients did show a trend towards higher scores, it would have been interesting to see what would happen if they were allowed to continue for additional weeks. Often changes become progressively larger as sessions continue.
Patient sessions were limited to 16 minutes of biofeedback, twice a week for six weeks. That equals 204 minutes of training, or about 3.4 hours of total training. In my own practice, patients generally get 30-40 minutes of biofeedback per weekly session, or about 210 minutes in six weeks, which is often (but not always) sufficient, so we won't fault them here. (But see #7.)
In spite of being asked NOT to practice at home, we learn that 10 of the 70 subjects did some home practice. They were reportedly "equally distributed" among the four conditions, although at, for example, 2,2,3,3 in the respective categories it is apparent that subjects in two groups got 50% more home practice than the others; I wonder which ones? We aren't told.
The absence of home training exercises is very troublesome. Blanchard et al. acknowledge that such practice appears to account for 20-35% of HA reduction in other studies, but still they did not include it. From a clinical perspective, we normally tell patients that home training accounts for almost ALL of their symptom improvement, and without it they just won't get much better.
A current migraine patient in my office reported tremendous improvement in symptoms after doing twenty 35 minute practice sessions the first week. During the second week she slacked off to 14 sessions of 20-30 minutes and even she noted the reduced level of improvement. There is a big difference in results when 700 minutes of home practice is added to the 32 to 35 minutes per week of office practice. And her 350 minutes of practice during the second week is still 350 minutes more than Blanchard et al., but cutting the home practice in half did produce less results for my patient. The miracle is that Blanchard got any results at all. Or did he? (See below!)
Blanchard acknowledges that in others' research, "Those who were led to believe they were doing well showed more HA relief than those who were led to believe they were only modestly successful (p. 243);" yet only "mild encouragement" was given to these 70 patients, regardless of whether it was warranted or not. We can only speculate on the confusing effect of providing each patient with the same coaching feedback, at every session, even when the patient was having a bad day and probably knew it.
Blanchard concludes that "the therapeutic effect derives from expectations and success experiences rather than peripheral physiological change (p. 243)," and that is clearly the case in THIS study, since peripheral change was so minimal. Blanchard did monitor expectations; while the TFB-warm group remained confident from week one to week three, all the other conditions suffered a 10% loss. [Since there are no "norms" for his three-item scale, we have no idea if a 10% drop is really significant or not.]
Apparently these subjects were given NO FEEDBACK from the experimenters (i.e., coaching) about their success, since they were only told to utilize the feedback screen, and it is not known how clearly they understood its interpretation. [I.e., did they really KNOW that most of them were doing very mediocre?] Since there was no coach present, many opportunities for ensuring that they understood the moment-to-moment effect of their manipulations were missed.
For instance, I always cheer the patient on, noting modest improvements verbally (Encouragement ranges from "That's it, good, your hands are getting warmer, watch the red line go up" to "Wow, what are you doing now? See how rapidly your temperature is going up!) My patients are given credit for what happens, maximizing their sense of internal locus of control. One of my favorites was my own mother, who initially disavowed any influence over the feedback screen. "I'm not doing anything", she protested. "Mother, you're the one who's connected to the machine, not me!", I replied!
Aside from an initial two-sessions of autogenic phrases in the TFB-Warm conditions, (and NO progressive relaxation), subjects were essentially "on their own" to make sense out of the feedback. Only mild explanations for the therapy were given, and apparently only initially. The EEG group was even told NOT to relax, but to sit quietly and watch the screen.
Since it is well known that patients "hear" only a tiny fraction of the explanations and information given by "doctors", it is clinically essential to repeat these "cognitive supports" and to verify, in conversations, that the patient fully understands them. This was NOT done in the present study. Although "treatment credibility" was scored at the start and middle, it was NOT scored at the end of treatment, and it did decline 5% for the TFB-Warm group, and 10% for the rest. I wonder how much more it would have dropped by week six, when they knew they weren't much better?
This Ghost "error" seldom appears in research anymore, thank goodness.
Here we have a major misunderstanding of the concept of training criteria on the part of Blanchard et al. In the "Mastery Model" of Shellenberger and Green (drawing on the work of many clinical researchers), training to criteria means continuing the biofeedback training UNTIL the patient is able to demonstrate a specific skill, in this case, until the patient is able to consistently and rapidly increase finger temperature to 95 or 96 degrees F. The "criteria" is known to the patient as a goal, and work continues until that goal is met, regardless of how many sessions it takes. [In the words of the Internet, EPID (Every Person Is Different) and YMMV (Your Mileage May Vary)!]
Typical of "grant research", which is often based more on academic year schedules than appropriate scientific considerations, a fixed number of sessions was offered. Such designs invariably manage to conduct all treatment during the course of a single semester, so graduate student vacations are not interrupted! If the patients acquire the skill in the designated timeframe, they are lucky. If not, the researcher still gets three graduate credits, regardless of what happens to the so-called patients.
Although "criteria" are discussed in this article, their "criteria" concept is totally different from that described in The Ghost. Here there are three "criteria" discussed, with different levels of stringency. It is suggested that at least one of the levels can be used to classify "learners", but this suggestion is entirely post-hoc and without justification. All three "levels" are drastically inferior to the "criteria" mentioned above. It appears that NONE of the subjects would be classified as "learners" under The Ghost's criteria, so it should not be a surprise that all four groups had somewhat equal, and equally insignificant, reductions in headache reports and medication reductions. Indeed, that's what they found.
Blanchard offers, completely post hoc, three levels of "criteria" from "weak" to "strong". In the weakest definition, TBF-Warm patients are judged to have "learned" the skill if they EVER reached 95 degrees and EVER showed a two degree rise, regardless of how long they were able to maintain or how rapidly they approached that target; 79% are judged to have "learned" the skill. In the medium condition, a paltry one degree rise in three sessions is defined as "learning" (63%), and in the strong condition 1.5 degrees rise in each of any six sessions would qualify (47% did). It is important to note that NONE of these subjects would qualify under the standard clinical definition of 95 degrees in 20 minutes, achieved consistently, on command. In other words, by clinical standards, NONE of these 70 subjects would be judged to have learned anything, so there would be no basis for comparing the effect of their "learning" on headache outcome. We would not expect any significant change in headaches.
The reduction in headaches in all groups was "statistically significant", but was it clinically significant? The over-all daily headache index went from 3.7/20 to 3.0/20, an average 17% decline. A 17% decrease in the headache index does not seem very significant from a clinical perspective; I would not expect many word-of-mouth referrals if my patients only decreased their headaches by 17%. Similarly, medication consumption was cut an average of only 28% (Range = 8-51%), hardly a big deal. This represents a decrease from 22 pill-units a week to 16 a week (or more than three a day down to more than two a day). They are still using too many medications.
We are reminded of several incontinence studies which boast about a 65-70% symptom reduction rate that is "statistically significant". But is it "clinically significant"? A woman who has been using 21 pads a week (not unusual) would still be using 7 pads a week. She now only changes pads once a day, but she is still wearing pads every day of her life!
Fortunately, that doesn't apply to this study, which used three implausible treatments to compare with the clinical standard.
Blanchard et al. considers mental/emotional variables to be "confounding" variables producing "non-specific effects", and tried to eliminate them from consideration by monitoring patient expectations and credibility of the treatment. Granted that there were no between-groups differences, it is still hard to attach meaning to an average score of 21 to 23 (on a scale of zero to 27). We only know that the average subject rated each of these items a 7 out of a possible 9, but we have no way of knowing if that is good or bad. 7 out of 9 is only 78%, or these patients only had a "C+ to B-" expectation anyway; that hardly seems like much when 16 weeks of daily record keeping, medication monitoring, and 6 weeks of bi-weekly clinic visit.
I could not determine if these patients paid for their therapy, or were paid to participate in the study, although that would certainly seem significant in evaluating their activities.
Blanchard was testing four hypothetical measures -- finger warmth, finger cold, finger temperature stability, and EEG alpha suppression, all of which have been postulated to have an effect on vascular headaches. Only the first is widely recognized in clinical practice as being a reliable measure of a relevant self-regulation skill.
All of Blanchard's patients would seem to fall victim to Rich Sherman's recent study, in which his subjects showed more random variation in finger temperature than these subjects showed in "learning". Clearly Sherman would not be happy with the mere four-minute adaptation period used here. The Ghost seems to recommend at least 20 minutes adaptation period, while Sherman seems to suggest even more!
Sherman, unfortunately, does not give the absolute temperature values for his subjects; only the change scores. Temperature is not, however, normally distributed in the population, and there is a practical limitation of 96 or 97 degrees at the upper end of the scale.
Blanchard, however, gives "initial baseline" temperatures, and they are distressingly high; the average starting temperature of the three biofeedback groups is 87.9 degrees, so his patients had an practical upper limit of a only 8 or 9 degree improvement. The Ghost, on the other hand, mentions studies where initial baseline was 80.5, 78.7, and 84 degrees. I don't have clinical data complied, but in my experience most migraine patients start out in the 78-80 degree range, and I did have one start at 65 degrees (yes, they were blue!).
Give that his patients had already obtained 50% of the thermal improvement that my patients seek, one has to ask if his were really "that sick", at least in a vascular sense, at the start of therapy.
The only evidence we have is the "Headache Index" and "Medication Index", provided in Tables III and IV. It is notable that the migrainers had a 2.6 (out of a possible 20) on the HI, compared with 4.75 for the "mixed HA" groups. Likewise, the migrainer's used fewer meds before treatment (21.3 vs. 26). [I haven't investigated the validation of the Medication Index, but I have initial trouble rating an Aspirin at a "1" and a Fiorinal as only a "2". Maybe it's because I take an Aspirin every day, but I've never even seen a Fiorinal.]
According to The Ghost, "generalization to the real world" and developing a "sense of control" are essential elements in the biofeedback treatment model; yet these are emphatically missing from the Blanchard model. Not only were they admonished NOT to practice relaxation in everyday life, but their sense of control was intentionally confounded by giving all patients "mild encouragement" regardless of whether it was appropriate or not.
I once had a migraine patient who refused to believe that thermal biofeedback would help. Very quickly, I refused to treat her because she didn't do her homework. She then spent $5,000 at a famous headache clinic in Boston, but they could find nothing wrong and advised her to return to my office. Again, she was soon ejected for failure to follow the homework requirements, so she spent $7,000 for a workup at a famous headache clinic in New York City. They, too, could find nothing, but this time she didn't return. Some months later she was in the local ER in complete collapse of self-regulation. The hospital, upon learning her history, eventually got her chemistry back to normal and discharged her to the care of myself and a local shrink -- an unheard of move in 1978.
She later scored an incredible 99+ on the Zuckerman Sensation-Seeking Scale, and with a little psychotherapy, began to follow my advice -- and avoid both migraines and collapse. Frankly, I can't imagine any value to biofeedback without Training people to the Mastery Model.
In summary, Blanchard (who obviously must have read The Ghost in the Box when it came out, since he gives lip service to concepts like "criteria"), appears to have been fundamentally unaffected by his exposure to the book. Not only is there is no Ghost in the Box, there is no Ghost in the Book, either.