Mechanism studies in people

A mechanism study asks what happens in the body when a needle goes in, rather than whether patients improve. This page reports the studies done in people, which fall into four groups: experiments that gave naloxone, a drug that blocks opioid receptors, to see whether it cancelled acupuncture’s effect on pain; brain scans taken during needling; a PET study of opioid receptors in patients with fibromyalgia; and measurements of a chemical messenger, adenosine, in the tissue around the needle. Animal work is named as animal work where it is mentioned. Whether patients benefit, and by how much over sham, is on Sham acupuncture.

Does naloxone block acupuncture analgesia?

The human studies disagree. If acupuncture eased pain by releasing the body’s own opioids, the endorphins, then naloxone should cancel the effect, and in 1977 Mayer, Price and Rafii reported in Brain Research antagonism of acupuncture analgesia in man by naloxone.

Chapman and colleagues at the University of Washington Pain Center tried to replicate it in 1983. Mayer’s group had measured pain thresholds, the intensity at which a stimulus first becomes painful. Chapman’s group first found that 0.4 mg of naloxone did not reverse analgesia from acupuncture for experimental tooth pain, measured through the brain’s electrical response. They then repeated Mayer’s threshold method: volunteers who showed analgesia during electrical stimulation of LI04 (Hegu) on the hand received 1.2 mg of naloxone or saline, double-blind, and the raised thresholds did not fall with naloxone. The authors concluded that the human and animal evidence gave little convincing support for endorphins playing a significant role, and noted that endorphins are also released by stress, which could explain why they sometimes rise with acupuncture.

The studies that followed split the same way.

The studies differ in nearly every respect that could matter: experimental pain in volunteers or clinical pain in patients, manual or electrical stimulation, the dose of naloxone and when it was given. Kenyon’s group read their null result as evidence against an endorphin mechanism for acupuncture therapy; Ernst and Lee read theirs as showing the opioid system is partly involved. A further difficulty sits in the drug itself: as Richard Harris and colleagues note in the study below, the same opioid system also carries placebo analgesia, so a naloxone effect alone cannot say whether it was the needle or the expectation that set the system going. Placebo and expectation covers that side of the ritual.

Opioid receptors imaged in fibromyalgia

Harris and colleagues at the University of Michigan imaged opioid receptors directly in 2009, in 20 women with fibromyalgia randomised to nine sessions of traditional Chinese acupuncture or of a sham whose needles did not pierce the skin, ten in each group, over four weeks. A PET scan with carfentanil, a radioactive tracer that binds to μ-opioid receptors, measured how many receptors were available to bind, once during the first session and again a month later after the eighth.

Real acupuncture increased receptor binding during the first session, in the cingulate, insula, caudate, thalamus and amygdala, and the increase was still present in several of the same regions a month later. In the sham group binding did not change or fell slightly, which the authors described as closer to earlier PET studies of placebo. Among the real-acupuncture patients, larger long-term increases went with larger falls in pain. Both groups’ pain fell by a clinically meaningful amount and by a similar amount, and patients in both groups guessed their assignment alike. The authors concluded that different opioid-receptor processes may underlie the relief seen with acupuncture and with sham. AcuiQ’s fibromyalgia page lists the protocols individual studies prescribed for the condition.

What brain scans show during needling

Functional MRI (fMRI) measures changes in blood oxygen that follow nerve activity, and it can be recorded while a needle is in place. K. K. Hui and colleagues at Massachusetts General Hospital scanned 13 healthy volunteers in 2000 while a practitioner turned a needle at LI04. In the 11 who felt the aching, spreading sensation called de qi, the signal fell in a group of deep structures involved in emotion and reward, including the nucleus accumbens, amygdala, hippocampus and hypothalamus. In the two who felt pain instead, several of those regions showed increases. Light touch at the same spot raised the signal in the touch-mapping cortex and lowered nothing deeper. The authors presented this as a preliminary result and a hypothesis. De qi itself is described on Needling and de qi.

Younbyoung Chae and colleagues at Kyung Hee University in Seoul pooled 28 fMRI studies in 2013, covering 51 needling experiments and 10 experiments of plain touch. Needling went with activation in regions that process sensation and pain, among them the insula, thalamus, anterior cingulate and the somatosensory cortices, and with deactivation in a set of regions including the medial prefrontal cortex, caudate, amygdala and posterior cingulate. Touch produced weaker responses in similar areas. The authors read the pattern as the brain registering the sensory, cognitive and emotional sides of a painful stimulus.

One PET study separated the needle from the expectation. Jérémie Pariente and colleagues in London scanned 14 patients with osteoarthritis in 2005 during real acupuncture, placebo acupuncture and a skin prick that patients were not led to expect anything from. None of the three changed the patients’ pain. The insula on the side of the needling was more active with real acupuncture than with placebo, and real and placebo acupuncture, which patients expected equally to help, both activated parts of the prefrontal cortex, anterior cingulate and midbrain more than the skin prick did. The authors linked the second finding to expectation and the brain’s reward system.

Do brain scans tell real acupuncture from sham?

Only in part, and not consistently. Wenjing Huang, Claudia Witt and colleagues at the Charité in Berlin reviewed the fMRI literature up to September 2009 in English, Chinese, Korean and Japanese: of 779 papers, 149 met their criteria and 34 could be pooled. Pooled real acupuncture activated the somatosensory, limbic, basal ganglia, brainstem and cerebellar regions earlier studies had described, while a difference between real and sham showed up in the middle cingulate; for other regions the answer depended on how the pooling was done. The four studies that compared real needles with the Streitberger retracting needle, all from the US, gave heterogeneous results, and several studies of other shams found no difference at all. The authors called for higher-quality studies with more transparent methods.

The visual-cortex finding that was retracted

In 1998 Zang-Hee Cho and colleagues reported in PNAS that stimulating a point on the outer side of the foot, one traditionally used for eye disorders, activated the visual cortex in 12 volunteers much as a flashing light did, while points 2 to 5 cm away did not.

Others could not find it. Isabel Gareus and colleagues in Freiburg needled GB37 (Guangming), a point documented for vision, in 2002, with soft and stronger stimulation, and detected no visual-cortex response to the acupuncture; the needle twisting did change the signal in other cortical areas. In 2006 Cho and four co-authors retracted the paper, writing that their later studies had led them to conclude there is no point specificity, at least for pain and analgesic effects; three of the original authors did not approve the retraction.

Florian Beissner and Christian Henke reviewed why such findings appeared in 2011, when more than 60 acupuncture fMRI studies had been published. They listed methods that some of these studies had not shared with the rest of the fMRI field: the choice of resting baseline, how a fall in signal was interpreted, whether the participant’s attention was controlled, and which group statistics were used. They judged that the early reports of visual and auditory cortex activation from points traditionally linked to sight and hearing were probably not a direct result of the needle but of one or more of these problems.

Adenosine at the needle

In 2010 Nanna Goldman, Maiken Nedergaard and colleagues at the University of Rochester reported in mice that needling released adenosine, a chemical messenger that dampens pain signals, and that the pain relief disappeared in mice lacking the adenosine A1 receptor; this was animal work.

The same group measured adenosine in people in 2012, by microdialysis: a fine probe under the skin of the leg collects fluid from the tissue around it. Fifteen healthy men aged 24 to 30 received 30 minutes of acupuncture at ST36 (Zusanli), below the knee, with the needle turned for five minutes in every ten; two were excluded because leg movement had raised their readings. In the remaining 13, adenosine near the needle rose on average from about 318 to 530 nanomolar during the session and stayed above baseline for the next 30 minutes. In three smaller comparisons of five men each, adenosine did not rise in the other, unneedled leg, did not rise when the needle was inserted but not turned, and showed little change when the needle went in 2 cm to the side of the point. The study measured the chemical, not pain: the volunteers were healthy and no pain outcome was recorded, and the authors named the small sample of healthy people as a limit.

Jin Cao, Jian Kong and colleagues at Massachusetts General Hospital, starting from caffeine’s action in blocking adenosine receptors, tested whether coffee drinkers get less from acupuncture in a preliminary crossover study published in 2021: of 27 healthy volunteers, 24 completed real acupuncture and a Streitberger sham. Real acupuncture raised heat-pain thresholds more than sham, and neither daily caffeine intake, which ranged from none to 500 mg, nor years of caffeine use was associated with the size of that rise.

What the mechanism literature cannot settle

None of these studies links a measured mechanism to clinical benefit in a way that excludes the alternatives. The naloxone trials disagree, and the opioid system they test also carries placebo analgesia. Harris’s receptor differences between real and sham came with the same fall in pain in both groups, in 20 patients. The adenosine rise was measured in 13 healthy men with no pain outcome, and the one human test of the caffeine prediction found no effect. The imaging studies show that needling changes brain activity, but the reviews found the real-against-sham differences inconsistent, and the point-specific visual-cortex finding was not found again by another group and was withdrawn by most of its own authors. Each line rests on few, small studies, mostly in healthy volunteers or with experimental pain. Reading an acupuncture trial sets out how to judge the clinical trials these mechanisms are offered to explain.