Friday, November 9, 2012
Fun With Thought Field Therapy: Why Some Randomized Controlled Trials Are More Controlled Than Others
Elsevier Publishers kindly send me links to articles in their CAM journal Explore:The Journal of Science and Healing. Today I actually read a couple of the articles, and although one is about a technique that is usually used with adults rather than children, I thought they might serve as good object lessons in why research design is every bit as important as statistical analysis done with packaged programs, and why basic statistical concepts are needed for interpretation of data.
The first of these papers is by Irgens, Dammen, Nysatter, and Hoffart (“Thought Field Therapy (TFT) as a treatment for anxiety symptoms: A randomized controlled trial”, 2012, Vol.8, pp. 331-338).
Let me start by describing TFT for those who are not familiar with this CAM treatment. TFT assumes that mood and emotional disturbances are caused by disruptions in energy flow within the mind-body totality, and that these can be corrected by tapping (physically) on points associated with meridians or lines along the body that are thought to correspond with a spiritual energy similar to the Chinese qi concept.
A second aspect of this study that I should mention is the idea of a randomized controlled design. The point of any experimental design is to discriminate between the effects of the factor being investigated-- such as a psychotherapy-- and the effects of the many other factors that can influence an outcome-- such as patients’ and therapists’ expectations or communications. Controlled studies employ various ways to make sure that those other factors are not mistaken for the important factors in the study-- the effects of the treatment, and how they compare with the outcome when that treatment is not present. Randomizing, or choosing by a random method which people will receive the treatment and which will not, is a way to control factors like choices made by patients or by therapists. For instance, if the patients choose their own treatments, their expectations and beliefs about the treatments can affect the outcomes. Similarly, if therapists choose which patients will receive which treatment (this includes no treatment), they may inadvertently put people in each group according to their expectations or their wishes about what will “work”. (This is not an attack on their integrity, just acknowledgement of human errors.) But randomization is no panacea in itself, and if other aspects of the design are not appropriate, a randomized trial may not control for all factors that can confuse our interpretation of the results.
Here’s what the Irgens group did in their randomized controlled trial of TFT. They randomly assigned people who had anxiety symptoms to one of two groups, those who would receive TFT right away and those who would be placed on a waiting list and told they would receive TFT 2 ½ months later. This meant that one group received treatment that included social interaction and physical contact with the therapist, and the other did not. In other words, the study did not control for the effects of simply talking to and being touched by the therapist, and the effects of those factors were mixed with the possible effects of TFT itself, with its methods of tapping on meridian lines. The treatment group experienced all three of these things, and the non-treatment group experienced none of them-- but the researchers nevertheless concluded that TFT was responsible for an improvement in anxiety symptoms following treatment. They had randomized, yes, but they had not controlled, and therefore their conclusion was an invalid one.
The Irgens group did not show that anxious people feel better because of TFT. What they showed was congruent with the common-sense expectation that anxious people feel better when they have more experiences with caring social contact, and possibly that touch is an important part of those experiences. In order to show that TFT itself caused improvement in symptoms, they would have needed to control for these confounding factors by creating a sham treatment group in which patients would have received social contact, touch, and tapping as well-- but the tapping would not have used the meridians that are the focus of TFT. Without an approach like that one, we have to say that the researchers’ randomized design did not successfully control for powerful factors that could bring about their reported results.
In addition, I should point out that the longer-term improvement reported can easily be explained by considering regression to the mean-- that is, if you feel pretty bad today (and are not suffering from a progressing physical illness), chances are that you will feel better tomorrow. In addition, chances are that you seek treatment for emotional distress when you feel your worst, and therefore it is likely that you will feel better later on, with or without treatment. Without wishing to suggest that serious mental illness be left untreated, I would point out that both unpleasant moods and pleasant ones will generally alter over time without any changes in circumstance.
There’s one more paper in this issue of Explore that I’d like to mention. By Jensen and Parker, it’s called “Entangled in the womb? A pilot study on the possible physiological connectedness between identical twins with different embryonic backgrounds” (2012, Vol. 8, pp. 339-347). The different embryonic backgrounds part had to do with whether they had shared a placenta and an amniotic sac, although it was not clear how that was known. “Connectedness” was the term used in describing posited events in which one twin was subjected to a mild shock or surprise, and the other responded by changes in blood pressure and other measures. Jensen and Parker concluded that there was no significant association between the reactions of the co-twins. However, in the time-honored manner of parapsychologists, they selected events in one of the four pairs of twins tested and reported that there were significant numbers of “hits” because on three of the 10 events the non-stimulated twin showed some reaction. This form of cherry-picking completely ignores the fact that such physiological changes can occur by chance or because of some other event in the environment. (Actually, what surprises me is that these researchers did not conclude that a failure to show “hits” was due to a motivation to develop an autonomous identity in the face of constant experiences of “connectedness”.)
I should point out, by the way, that Jensen and Parker, who obviously could not randomize people into twin and non-twin pairs, did not attempt to provide a comparison group of non-twins. If this kind of idea is to be pursued at all, surely it would be of interest to see whether a similar number of “hits” occurred in one non-twin pair. But, I guess, for Jensen and Parker--- not so much!