In a previous, more adventurous stage of life, I hitchhiked around Australia. I got a lift out of Adelaide from a man with one leg, the other having been amputated after a car accident when he was six. There were two remarkable things about this man. The first was that he drove a manual - three pedals, one leg. The second was his remarkable method of gaining relief from an excruciating pain in his missing foot: he would put his prosthetic leg in the exact location he felt his own leg to be, and then drive a screwdriver into the painful spot. As long as he could see it, driving the screwdriver into the exact site of his pain turned it off just like a switch-he called it his 'magic button'. Either he was lying, or something about the screwdriver, and seeing the phantom limb, relieved his pain. I don't reckon he was lying.

Such phantoms of missing body parts are generated by the same brain mechanisms that generate the experience of an existing limb. Nearly everyone who has a limb amputated or the nerve supply removed reports experiencing some kind of phantom limb, but only some report persistent phantom limb pain. Those who do, however, report that the pain is very nasty - it is usually untouched by drugs or implanted pain relief devices. It would be ideal to find an inexpensive self-treatment that has almost no side effects, nor risks. Some think that such a treatment may be found in mirrors. Although 19th century neurologists, such as Pierre Janet, had experimented  with the mirror treatment, modern versions of the treatment were developed by Vilayanur S. Ramachandran at the University of California, San Diego.

Fixing Phantom Pain
A recent letter by Jack Tsao, et. al  in the New England Journal of Medicine describes one of the better attempts to elucidate the true value of mirror therapy for phantom limb pain. The researchers randomly assigned 22 lower-limb amputees with phantom limb pain to one of three groups:

1) Mirror movements: patients watched the reflected image of their intact foot in a mirror while they moved both feet simultaneously. (Obviously, amputees can't move their missing foot, but they can move their phantom foot.)
2) Covered mirror movements: patients performed the same movements but the mirror was covered so they did not see a moving limb.
3) Imagined movements: subjects mentally pictured moving the phantom foot, with eyes closed.

All patients performed 15 minutes per day of their assigned therapy and recorded the number, duration and intensity of pain episodes. After four weeks, there were two key findings. First, pain decreased significantly in all six patients doing the mirror movements, with the median decrease, as reported on a 100 point scale, being from 30/100 to about 5/100. Second, three out of six patients in the covered mirror movements group, and four out of six patients in the imagined movements group, got worse, not better.

Let's get the research questions out of the way first, and then get back to the potential importance of the study. I think it is a pity that this paper was a letter, because a letter's need for brevity meant that key information was left out. As a researcher, I want to know:

1) What happened to the three who didn't finish the study? Which group were they in? If they were all in the mirror group, then the results would be far less convincing.
2) To what question did the pain measure relate? If it related to pain at the end of the 15 minutes, then the study tells us about instantaneous effects. If it related to pain over the last few days or week, then the study gets more exciting.
3) How did they remove bias? Reporting bias is a big player in pain studies and could easily impart reductions in pain of 30/100.

Without the answers to those questions, and a few slightly more pedantic ones, it is very difficult to know what the study actually showed. However, let's give the authors the benefit of the doubt and presume that they actually did show that mirror movements decrease phantom limb pain. The most obvious next question is "Why?" The authors suggest that the effect might be due to activation of mirror neurons in the brain, which fire when we observe another person doing a movement, and presumably also when we observe ourselves doing a movement. I don't think this explains the result because these mirror neurons also fire when we move, which means they should have been firing during the other two conditions. The authors also suggest that "seeing" the missing limb move might deactivate systems that perceive certain types of pain, which, roughly interpreted, means that seeing the missing limb move might decrease phantom limb pain. Not an explanation so much as a restatement of the result.

What is Pain?
To consider how mirror movements might reduce pain, it might be helpful to first consider what pain is. I argue that pain emerges from the brain in accordance with the brain's unconscious perception of danger to the body part in question. Sensory input of danger (called nociception) is important, but not sufficient (nor necessary actually), for pain. Perhaps mirror movements simply convince the brain that all is exactly as it should be, which removes the brain's need to evoke pain. Alternatively, perhaps mirror therapy is a great distracter: distraction remains our most effective analgesic. Such explanations are very reasonable, but not particularly exciting.

Here is a more exciting theory, one the authors' introduction suggests they had in mind. It is based on the idea that phantom limb pain results from an internal conflict in the brain. Although sensory feedback, from the nerves that used to supply the missing limb, tells the brain that the limb is still present, visual feedback tells the brain it is not. According to that theory, seeing the phantom would remove the conflict and attempting to move the phantom (i.e. the other two conditions) might exacerbate it (although removing visual feedback didn't help, which doesn't fit neatly into the theory). That anaesthetizing the stump can eliminate phantom limb pain seems consistent with this theory. Perhaps anaesthetizing the stump brings sensory feedback into line with visual feedback, whereas mirror movements bring visual feedback into line with sensory feedback.

Of course, perhaps the key to mirror movements is seeing the movements, not necessarily seeing them in a mirror. A previous study in 14 lower limb amputees with phantom limb pain used a control condition in which patients watched the intact limb but couldn't see its reflection. There was no added benefit of seeing the phantom limb. However, that study looked at single sessions of ten movements each performed ten times, whereas the Chan et al. paper concerned four weeks of 15 minutes a day. Their data suggest that there is a cumulative effect, such as that observed with virtual walking, an approach that utilizes visual feedback to give paraplegic patients the perception that they are watching themselves walk. In my experience, patients don't get long-term relief from phantom limb pain by performing mirror movements. On that basis, I devised an alternative program that involved first doing implicit and then explicit motor imagery. Motor imagery refers to activation of the brain's movement-related processes without actually executing the movement. The same brain areas and processes underpin motor imagery and motor execution. The advantage of the former, however, is that it is not disrupted by an inability of the body to execute the command. The difference between implicit and explicit motor imagery is that you don't know you are doing the former, whereas you intentionally do the latter. The implicit imagery we used was a task in which we asked participants to judge whether a pictured hand was a left or a right hand. To determine the side, a person mentally maneuvers his or her own hand to match the posture of the one shown in the picture--but the person isn't aware of how the brain does it. The program progressed from implicit motor imagery to explicit motor imagery and then to mirror movements. That randomized controlled trial included patients with complex regional pain syndrome and showed long-term (six months) improvements that were similar in magnitude to those reported by Chan et al.

If there is a cumulative effect, then resolution of conflict between sensory and visual feedback is unlikely to be the key mechanism. Perhaps, then, mirror movements actually serve to correct faulty maps of the body part, which are held throughout the brain, most famously in the primary sensory cortex. We know the brain's map of the limb is disrupted in phantom limb pain; we know the map returns to normal when phantom limb pain resolves; and we know visual input of a body part can affect the body map. The disrupted body map theory seems worthy of investigation, but at the moment, we don't really know.

In summary, the article by Chan et al re-energizes the debate about mirrors and phantom limb pain. I don't think, in its published form, it proves an effect, but it does remind us that we need tight, well controlled, and well reported clinical trials. A cheap, safe and side-effect free treatment for what is a nasty and debilitating problem would be absolutely terrific.