Unpacking the effects of acupuncture


  • Liana Fraenkel

    Corresponding author
    1. Yale University School of Medicine, New Haven, and VA Connecticut Healthcare System, West Haven
    • Yale University School of Medicine, Section of Rheumatology, 300 Cedar Street, TAC #525, PO Box 208031, New Haven, CT 06520-8031
    Search for more papers by this author

Controlled studies have shown that acupuncture decreases pain. What is less clear is how. Randomized controlled trials that include a sham acupuncture arm, as did the well-designed study in this issue of Arthritis Care & Research, suggest that placebo-associated effects are responsible for the beneficial effects of this treatment (1–3). The study by Suarez-Almazor et al (3) adds to a growing body of literature focused on elucidating the mechanisms underlying placebo-associated effects in knee osteoarthritis and other disorders.

Biologic mechanisms

Placebo-associated responses for several disorders, including pain, depression, and Parkinson's disease, are associated with measurable central nervous system effects. Studies using positron emission tomography imaging (4) and quantitative electroencephalography (5) reveal regional changes in brain metabolism in responders to placebo analgesics and antidepressants that differ from the patterns of nonresponders and from subjects who respond to an active drug. Activation of dopaminergic pathways appears to explain the strong placebo response seen in patients receiving sham fetal neural cell transplantation in Parkinson's disease. Using functional magnetic resonance imaging, Cho et al (6) demonstrated that both sham and traditional acupuncture reduce brain activity in areas involved with perception of pain. In experimental pain models, the placebo response to analgesics can be blocked with naloxone, suggesting that some placebo-associated effects are mediated by opioid receptors (7). Moreover, inhibiting cholecystokinin (which has anti-opioid actions) potentiates the placebo effect (8).

Treatment characteristics

The influence of treatment characteristics on patient outcomes has been well established. The color of placebo pills can predictably cause stimulation or sedating effects (9). The strength of placebo-associated effects is positively correlated with the size and number of pills or capsules, as well as the invasiveness of the treatment (10). The latter explains the impressive benefits seen with sham acupuncture and sham arthroscopic lavage and/or debridement in subjects with knee osteoarthritis (11). Placebo-associated benefits are not limited to subjective reports of pain. Deactivated pacemakers reduce syncopal episodes in patients with hypertrophic cardiomyopathy as effectively as do functional pacemakers (12), sham internal mammary ligation reduces angina as or more effectively than actual ligation (13), and sham transplantation of fetal neural cells is as effective at reducing patient global disease severity ratings as actual surgery in patients with Parkinson's disease (14).

Patient-provider relationship

The strong influence of the provider-patient relationship on patient-reported outcomes is well known to clinicians and has been documented in extensive literature (15). Consultants who are warm and supportive are more effective than those viewed as more formal and/or neutral. A fairly recent systematic review found that interventions that encourage providers to offer positive information and reassurance significantly improve health outcomes (16).

Randomized controlled trial evidence supporting the effect of the provider-patient relationship on outcomes in acupuncture date back to a small study published in 1977, in which Berk et al (17) randomized subjects with soft tissue disorders (bursitis or tendinitis) to one of 4 groups: acupuncture-positive environment, acupuncture-negative environment, sham acupuncture–positive environment, and sham acupuncture–negative environment. The study found that although both real and sham acupuncture improved pain, subjects in the positive environment reported greater reductions in pain than those randomized to the negative environment.

As mentioned by Suarez-Almazor et al (3), Kaptchuk et al (18) performed a study in which they randomized patients with irritable bowel syndrome to a waiting list, sham acupuncture with limited provider-patient interactions, and augmented sham acupuncture in which patient expectations were enhanced by an encouraging and supportive provider. The effect size was 0.51 for the limited versus waiting list group, 0.46 for the augmented versus limited group, and 0.99 for the augmented versus waiting list group, thus providing strong evidence for the contribution of patient-provider interactions on placebo-associated effects.

Patient expectations

Patient expectations, whether positive or negative, predict outcomes. Women participating in the Framingham Heart Study who believed they were at risk for heart disease were 4 times more likely to die compared with women with similar risk profiles who did not share this belief (19). Subjects' ratings of their expected responses to treatment explain up to 80% of the variance found in placebo-controlled pain trials (20).

In many scenarios, patients' expectations are prerequisites for placebo-associated effects. Covert interventions have little or no placebo-associated response compared with overt ones. Injection of saline has the same effect of morphine as long as the study subjects are able to see the injection (21). Anticipation of an analgesic response results in changes in brain activity in regions that stimulate endogenous opioid production (22).

Patients' evaluations of how a treatment works have been described as a specific form of expectation (23). Arthroscopy is associated with strong placebo-associated effects because it makes sense that a torn meniscus causes pain and that removing or repairing it should alleviate the pain (23). In contrast, treatments with more obscure mechanisms of action may appear less rational to patients and result in lower expectations and lower placebo response rates. Acupuncture is associated with variable appraisals depending on one's cultural background and health beliefs. However, it is likely that most subjects agreeing to participate in a trial view acupuncture as a rational method of treating pain.

Acupuncture in knee osteoarthritis

The benefits of sham acupuncture in knee osteoarthritis have been well documented. Possible mechanisms underlying observed improvements (outside of a possible biologic effect of nonspecific needling) include subjects' anticipated analgesic response to the intervention, the positive meaning patients attach to the therapy (i.e., endorsement of the restoration of the normal flow of Qi as a means to treat pain), and enhancement of subjects' expectations by encouraging and supportive providers.

In this issue of Arthritis Care & Research, in addition to comparing the effectiveness of traditional and sham acupuncture, Suarez-Almazor et al (3) sought to examine the specific effects of provider-patient interactions on acupuncture outcomes in knee osteoarthritis by manipulating provider communication styles. The investigators first randomized subjects to one of 3 treatment groups (waiting list, high expectations, or neutral expectations) and subsequently randomized subjects in the latter 2 groups to sham or traditional acupuncture. Consistent with the study by Scharf et al (2), the investigators found no difference between the sham and traditional acupuncture arms. The high expectation therapists reassured patients that they would likely benefit. In contrast, the neutral expectation therapists conveyed messages of uncertainty. As predicted, the authors found an incremental benefit among the subjects randomized to the high expectations group. Because patients are generally “ambiguity averse” and tend to view neutral providers as being nonsupportive and/or cold, it is possible that the uncertainty conveyed by the providers in this study actually had a negative, or nocebo, effect (24). Regardless, the manipulations used in this study reflect attitudes commonly seen in clinical practice.

Implications for research and clinical practice

Given the frequent strong placebo-associated response in clinical trials for knee osteoarthritis and other conditions, several approaches have been proposed to minimize placebo response rates. Some investigators advocate restricting trials to those with more severe disease because placebo response rates tend to be lower in this group (25). However, there are several pitfalls associated with this approach, including negative effects on enrollment, generalizability, and lower expected responses to the actual biologic effects of treatment. Biologic markers may be less prone to placebo-associated effects than subjective outcomes. However, such markers are frequently not available and usually do not reflect the outcomes that are most valued by patients. Run-in periods have also been proposed as a possible way to minimize placebo response rates in clinical trials (25). In these studies, subjects are given a placebo for one to two weeks prerandomization, and those that improve by at least a prespecified amount (after accounting for the regression to the mean) are excluded. Criticisms of this approach include the effects of this design on generalizability as well as ethical concerns related to unjustifiable exclusion of subjects who may derive benefit from the proposed treatment. Other innovative designs have been proposed, but to date there is no evidence demonstrating that these modifications actually decrease the placebo response rate or improve the clinical impact of randomized trials.


Improved benefits in many treatment trials for knee osteoarthritis are due to placebo-associated effects. The well-blinded, high-quality trial by Suarez-Almazor et al (3) in this issue of Arthritis Care & Research provides further strong evidence that invasive procedures with positive meaning, delivered in a supportive environment, result in significant benefits. However, given the strength of the data demonstrating that the benefits related to acupuncture are likely fully explained by placebo-associated effects, recommending acupuncture for patients with knee osteoarthritis now raises real ethical concerns. What is greatly needed is a better understanding of the pathophysiology of osteoarthritis and targeted treatment options that result in significant improvements over and above placebo-associated responses. Meanwhile, tomorrow in the clinic, I will be as warm, positive, and caring as possible as I inject corticosteroids into my patients' knees.