Article Text


Summaries and commentaries by Adrian White on a selection of recent acupuncture studies

Statistics from

Clinical trials of effectiveness

Acupuncture and expectation for knee pain

This study (n=560) was designed to compare the effectiveness of real and sham acupuncture on osteoarthritis (OA) knee pain while simultaneously controlling expectation.


Acupuncturists were trained to interact in one of two communication styles: high expectations (such as ‘I think this will work for you’) or neutral expectations (such as ‘it may or may not work for you’). A brochure was given to each patient, written with expectations appropriate to their group. Patients were randomised to one of three communication style groups—waiting list, high or neutral—and nested within style, traditional Chinese acupuncture (TCA) or sham acupuncture twice a week over 6 weeks.

TCA consisted of points Xi Yan (the two ‘eyes’ of the knee), He Ding (centre of superior border of patella), GB34, SP6, ear point ‘knee’ and one or two tender Ah shi points proximal to the knee. Electroacupuncture using a transcutaneous electrical nerve stimulation (TENS) machine (alternating 15/50 Hz to maximum tolerance) was applied across two pairs of points: the eyes of the knee, and SP6 and SP9. Sham acupuncture was performed in predefined, non-meridian points: four on the lateral side of the leg—two above and two below the knee—and two in the back of the wrist. Shallow needles and minimal stimulation were used. The TENS machine was started up at the wrist points and the two points below the knee but turned off as soon as the stimulus was felt.

Primary outcome measures were Joint-Specific Multidimensional Assessment of Pain (J-MAP), Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) pain scale, and satisfaction with knee procedure scores at baseline, 4 and 6 weeks and 3 months.


A total of 455 patients received treatment (TCA or sham) and 72 were waiting list controls. In the TCA and sham groups, 52% and 43%, respectively, thought they had received TCA (κ=0.05), suggesting successful blinding. No statistically significant differences were observed between TCA or sham acupuncture, but both groups had significant reductions in J-MAP score compared with the waiting group (−1.1, −1.0 and −0.1, respectively; p<0.001) and similarly in WOMAC pain (−13.7, −14 and −1.7, respectively; p<0.001). Statistically significant differences were observed between the two acupuncture groups in J-MAP pain reduction and satisfaction, favouring the high expectations group (figure 1).

Figure 1

Knee pain scores, by treatment and by expectation.


This is a large, well-conducted trial and deserves close attention. Its message appears to be that TCA was not superior to sham acupuncture, but increased expectation had an effect. Actually, no—there are several reasons to doubt these apparent conclusions. The study might appear to ‘tick all the boxes’ for scientific rigour with randomisation, blinding and careful accounting for all participants—but as we shall see there are several big question marks over its design.

Three acupuncturists delivered high expectation and three low: but their success in influencing the patients' expectation was not measured. So the difference between the groups may not have been due to expectation, but to the practitioner's individual personality, which we already know from Kaptchuk's work can have a much larger effect than any manipulation of expectation. To avoid this, each practitioner's style had to reverse half way through the study, so that acupuncturists giving high expectation were retrained to give low expectation, and vice versa. How successful was this retraining? In other words how much of the style they had got used to persisted in the second half of the study? Clinical trials are a very artificial environment, and it is difficult enough to alter one's behaviour once, let alone twice. These practitioners were selected for their acupuncture not their acting ability. Their success of reversing their style half way through the study was not tested.

Then, the changes in the acupuncture group were rather small compared with other similar studies. There could be several reasons for this: the patients had relatively mild OA, since the inclusion criteria set no minimum duration and rather low pain levels (visual analogue scale (VAS) 3/10); the real acupuncture was given mainly to non-muscular soft tissue; the sham probably had some activity with mild segmental stimulation and the interesting extrasegmental needling to the arm (the effect of arm stimulation in knee pain has rarely if ever been studied); and, because of the design, half the patients receiving real acupuncture would have had low expectations, and we know that expectation facilitates the effects of needling: thus the acupuncture clearly could not realise its full potential.

For all these reasons, it is probably not a surprise that acupuncture did not show up superior to sham and we should not put too much weight on this single result which is, after all, not in agreement with the current Cochrane review's conclusion that acupuncture has a small but significant effect for knee pain compared with sham.

It is tedious, and could be regarded as unscientific, to be always trying to ‘explain away’ negative studies; but it would be equally unscientific to ignore this kind of small but important design feature that is a significant limitation to the interpretation of a study. This is especially important in a trial that might appear to the non-specialist to have been methodologically rigorous: the design flaws are subtle and only understood by specialists.

Chronic shoulder pain

A sham controlled randomised controlled trial (RCT) of acupuncture for chronic shoulder pain (CSP) in 424 outpatients.


Patients with unilateral CSP of between 6 weeks and 2 years, and with an average pain VAS score of at least 50 mm, were recruited in orthopaedic outpatient departments between 1997 and 1999. They were randomly assigned to receive 15 treatments of Chinese acupuncture (verum), sham acupuncture (sham) or conventional, conservative orthopaedic treatment (COT).

The verum treatment consisted of three Ah shi points, and local and distant points according to the channel and site of pain, chosen from the following groups: LU1, LU2 (ventral); LI4, LI11, LI14, LI15 (ventrolateral); TH5, TH13, TH14 (lateral); SI3, SI9 (dorsal). In addition, a distal point chosen from ST36, GB34 and BL58 could be needled while the shoulder was moved. Sham treatment consisted of shallow needling of eight standardised non-points over the tibia, four in each leg. The conventional treatment group received diclofenac daily with 15 sessions of physiotherapy, physical exercise, heat or cold, ultrasound or TENS, but no injections of any kind. The patients were blinded to the type of acupuncture and treated by 31 office-based orthopaedists trained in acupuncture; all received 15 treatments once to three times weekly over 6 weeks.

The 50% responder rate for the average pain VAS over the previous 7 days was measured 3 months after the end of treatment (primary end point) and directly after the end of the treatment (secondary end point). Patients without outcome data were regarded as treatment failures in the intention-to-treat analysis. A per protocol analysis was also conducted of patients with follow-up data.


Two thirds of the patients were female, average age was 51 years and the mean pain duration was 11 months. In the intention to treat (n=424) analysis, shown in figure 2, percentages of responders for the primary end point were verum 65% (n=100), sham 24% (n=32) and COT 37% (n=50). For the secondary end point at end of treatment, the responders were verum 68% (n=92), sham 40% (n=53) and COT 28% (n=38). The results are significant for verum over sham and verum over COT (p<0.01) for both the primary and secondary endpoints. The per protocol analysis of the primary (n=308) and secondary endpoints (n=360) yields similar responder results for verum over sham and verum over COT (p<0.01). Descriptive statistics showed greater improvement of shoulder mobility (abduction and arm-above-head test) for the verum group versus the control group immediately after treatment and after 3 months.

Figure 2

Shoulder pain: % responder rates.


This study was not one of the GERAC series from the healthcare insurance industry, but was funded directly by the German Ministry of Education Science and Research. Though generally impeccable, the main weakness of this study was that the blinding was not checked: a large proportion of the control patients may not have been convinced that needling in the leg was a great treatment for shoulder pain.

Interestingly, the results for the primary and secondary outcomes were reversed in the abstract compared with the text and tables—but it makes little difference, both showed a significant effect of acupuncture with a clinically useful response rate, 65%. Also note that this treatment seems widely applicable in shoulder problems, since there was no attempt to make a Western diagnosis of the patients' actual disorder. The trial indicates that Chinese acupuncture is an effective alternative to conventional orthopaedic treatment for CSP.

Endometriosis pain

A crossover study in patients (n=101) with endometriosis.


A total of 101 women aged 20–40 years with endometriosis diagnosed by laparoscopy and with pain score at least five on a VAS scale were randomised into two groups, each receiving two series of 10 acupuncture treatments, twice a week over a period of 5 weeks. Group 1 (n=47) received real acupuncture during the first series, and group 2 (n=54) received non-specific acupuncture. After the first unit of 10 treatments, a washout period of at least two menstruation cycles was set, after which patients received the other intervention.

Real treatment consisted of BL32, ST29, ST36, CV3, SP6 plus up to three extra points according to traditional diagnosis—LR3, LR8, SP9, SP10 and KI10. De qi was elicited, and type of manual stimulation and use of moxa were dictated by each patient's traditional diagnosis. The control group received needling at ‘non-specific’ points: PC9, GB31, LU1, ST8, with no needle stimulation. Pain scores were taken using diaries during a 2-week run-in period and throughout the study period.


In all, 83 of 101 patients finished the study. As shown in figure 3, group 1 showed a significant reduction of pain intensity after the first 10 treatments, but group 2 showed significant pain relief only after the crossover. Changes in SF-36 and pain disability index were similar.

Figure 3

Endometriosis pain visual analogue scale scores.


At first sight the results of this study looks impressive, but we must not abandon rigour! Interestingly, the analysis on this study is not ideal, despite being published in a journal with an impact factor of 1.6. The authors should have tested for the effect of the order of treatment first, and then they should have tested for a difference between the groups. They did neither: they only tested for changes in pain VAS within groups.

In addition, I wonder how credible the control treatment was, particularly to participants in the second series having no needle stimulation, when they had already experienced verum acupuncture. Credibility was not tested: lack of credibility of sham acupuncture in patients who have just had real acupuncture is one of the main reasons why crossover studies of acupuncture are so rare (the other problem is carryover effects leading to differences in baseline measurements).

Postoperative electroacupuncture

RCT (n=30) of daily electroacupuncture for post-sternotomy pain and respiratory function.


Thirty patients, who had undergone a median sternotomy for coronary artery bypass graft, were randomised to either electroacupuncture and pharmacologic analgesia (acupuncture, n=15), or pharmacologic analgesia alone (control, n=15). Electroacupuncture (60 Hz for 20 min) was given daily starting 2 h after extubation. The points used were LI4, LI11, ST36, PC6, LR3. The authors do not state which needles were paired. For analgesia, pethidine hydrochloride and metamizole sodium were given. In each group, daily severity of pain and analgesic intake were recorded and respiratory function assessed at baseline and day 7.


The total postoperative dose of pethidine over 7 days was 87 mg in the acupuncture group and 223 mg in the controls (p=0.002). A similar difference was seen in the metamizole dose (p=0.0001). Despite the lower dose of analgesic drugs, the acupuncture group had significantly lower pain VAS scores on days 3, 5, 6 and 7 (figure 4). There were no differences between the groups for respiratory function.

Figure 4

Postoperative pain visual analogue scale scores.


Sternotomy is common—and painful afterwards. There is increasing evidence that acupuncture can reduce the need for analgesics postoperatively. That may not in itself be cost-effective, but any additional effect in reducing respiratory problems could well be; that was not seen in this study, though there was a trend in favour so a larger study would be justified. In future studies it should be possible to devise a control procedure and achieve some blinding of patients to improve the rigour of the study design.

Congestive cardiac failure

A study (n=17) of acupuncture's effects on exercise tolerance in patients with congestive cardiac failure.


Seventeen stable patients with congestive heart failure (New York Heart Association class II-III), ejection fraction <40%, in sinus rhythm and receiving optimised heart failure medication were randomised into a real acupuncture or a placebo acupuncture group. Outcome was assessed in four ways: standardised 6-min walk, cardiopulmonary function on bicycle ergometer, heart rate variability on 24-h monitoring, and quality of life using SF-36. Acupuncture was given using standardised points, PC4, PC6, HT7, ST36, SP6, KI3 and CV17. In the sham control group, non-penetrating Streitberger needles were used at points 2 cm away from these points to avoid acupressure effects. In both groups, 10 sessions were given over 5 weeks.


There were no significant adverse events including bleeding in these patients—who were taking anticoagulants. There were no changes of the cardiac ejection fraction or peak oxygen uptake, but the mean 6-min walk distance was increased in the real acupuncture group (+32 SD 7 m) but not the placebo acupuncture group (−1 SD 11 m; p<0.01). Changes in individuals are as reproduced in figure 5. Accordingly, post-exercise recovery after maximal exercise and ventilatory efficiency were improved after real but not sham acupuncture. Furthermore, heart rate variability increased after real acupuncture, but decreased after sham acupuncture. The ‘general health’ score and ‘body pain’ score of the quality-of-life questionnaire SF-36 tended to improve after real acupuncture.

Figure 5

Six-minute walking distances (m).


In patients with heart failure, the degree of exercise limitation is not proportional to the severity of impairment of the cardiac function, as one might think. This has led to a hypothesis that raised inflammatory cytokines in the condition cause musculoskeletal fatigue and activate ergoreceptors, increasing the sense of breathlessness. Thus the rationale for this study was acupuncture's potential effects on muscle function and symptoms, not on cardiac function. Even with this small sample, an effect is strongly suggested. As an afterthought, the researchers measured the inflammatory cytokine tumour necrosis factor α retrospectively in those from whom they had blood samples, and detected a greater reduction in patients given real than those given sham acupuncture. This lends some weight to the hypothesis of the possible mechanism and suggests the design of further studies.

Systematic review

Auricular acupuncture for pain relief

A systematic review of 17 RCTs of auricular acupuncture for all kinds of pain.


Five databases were searched up to December 2008. Randomised trials were included if they compared auriculotherapy to sham, placebo or standard care and were published in English. Control groups could be sham acupuncture, standard care or waiting list. Two reviewers independently assessed trial eligibility, and assessed quality according to Agency for Healthcare Research criteria. Standardised mean differences (SMD) were calculated for studies using a pain score or analgesic requirement as a primary outcome.


Seventeen studies with a total of 1007 participants were included (eight perioperative pain, four acute pain and five chronic pain). A range of schedules was used ranging from a single treatment session to indwelling needles for up to 30 days, and techniques varied from needles with manual or electrical stimulation, transcutaneous electrical stimulation, acupressure and laser acupuncture. Six studies were rated as good quality, four fair and seven poor quality.

For pain in general, auriculotherapy was superior to controls (SMD, 1.56 (95% CI 0.85 to 2.26; eight studies); heterogeneity was marked. For perioperative pain, auriculotherapy reduced analgesic use (SMD, 0.54 (95% CI 0.30 to 0.77); five studies) with little heterogeneity.

For acute pain and chronic pain, auriculotherapy also reduced pain intensity compared with controls (SMD for acute pain, 1.35 (95% CI 0.08 to 2.64), two studies; SMD for chronic pain, 1.84 (95% CI 0.60 to 3.07), five studies). Removing the poor quality studies did not alter the conclusions.

Four of the five studies included in the perioperative subgroup used sham acupuncture controls—real points in the concha and sham points in the helix. Meta-analysis for these studies, represented in figure 6, show a significant reduction in analgesic use for the true acupuncture group (SMD 0.63; 95% CI 0.88 to 4.97; four studies).

Figure 6

Auricular acupuncture for postoperative pain.


The most powerful piece of evidence for me is the last paragraph of the results: treatment at appropriate locations is superior to control treatment at inappropriate locations. These four studies were all from the same research unit (Usichenko, Germany) and the evidence would be stronger if different centres had been involved. The real treatment was given to the concha, and the control treatment to the helix. So this result does provide support for a different effect from stimulating different regions of the ear supplied by different nerves—which is not the same as point specificity as such.

The effect sizes shown in this meta-analysis are generally large and clinically useful. Auriculotherapy seems effective for the treatment of a variety of types of pain, especially postoperative pain, so is certainly an area where more research is justified.

Basic research

Functional MRI of laser acupuncture

An investigation into the functional MRI (fMRI) changes seen with laser acupuncture.


Ten healthy subjects were randomly stimulated with a fibreoptic infrared laser on four acupuncture points (LR14, CV14, LR8 and HT7) that this research group has previously used for depression, following the principles of traditional Chinese medicine. Each point was stimulated twice, and one control non-acupoint (a sham point on the abdomen) was stimulated once, in a blocked design (alternating verum laser and placebo laser blocks), while the blood oxygenation level-dependent fMRI response was recorded from the whole brain on a 3T scanner.


Stimulation of most of the points resulted in various patterns of neural activity—though not HT7. Regions with significantly increased activation included the limbic cortex (cingulate gyrus) and the frontal lobe (middle and superior frontal gyrus). Laser acupuncture tended to be associated with ipsilateral brain activation and contralateral deactivation that therefore cannot be simply attributed to somatosensory stimulation.


The authors are to be congratulated on this early study on laser acupuncture. fMRI studies of the blood oxygenation level-dependent type require ‘on/off’ phases of repeated stimulation and rest. This is virtually impossible using a needle because it has to remain in the point during the ‘off’ period—but of course it may still be active. Laser acupuncture has the notable advantage that it gives out no stimulation during the ‘off’ period.

The findings are distinct, but very difficult to interpret in a meaningful way—as so often with fMRI studies of acupuncture. The different points produced neurological responses in different sites, though none affected the prefrontal cortex, which may be one of the most important sites in clinical depression—the condition which these points are supposed to treat. Thinking traditionally, it seems odd that HT7, known as the ‘gate of the mind’ had no effect in this protocol, designed for its psychological effects; whereas the non-point on the abdomen stimulated the parietal region. This supports the lack of point specificity of the Western approach. Of course, on a cautionary note this study was done in healthy volunteers and the changes may turn out to be very different in patients with depression.

View Abstract

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.