The three numbers that explain the acupuncture argument

Two true sentences about acupuncture sound like they describe different treatments.

Here is the first. In a pooled analysis of 39 high-quality trials covering 20,827 patients with chronic pain, acupuncture beat no acupuncture by close to half a standard deviation, and most of that advantage was still there a year after people were randomized. Here is the second. In the same analysis, the advantage over fake acupuncture shrank to about a fifth of a standard deviation, a figure a separate review converted into roughly four millimeters on a 100 millimeter pain scale.

Both sentences are accurate. Neither is spin. The strange part is that the two research camps most often described as opponents agree on that second number almost exactly. What they disagree about is which comparison a patient should care about, and that is a different sort of argument with a different sort of answer.

This article walks through the numbers in the order a reader needs them, then through the parts that get skipped: what the fake needle does, which guidelines name acupuncture and how strongly, what Medicare will pay for and who is allowed to bill it, and what the harm record looks like when you insist on a denominator.

The three response rates that make the facts visible

The central synthesis is an individual patient data meta-analysis by Andrew Vickers and the Acupuncture Trialists’ Collaboration, updated in 2018 in the Journal of Pain. Individual patient data means the researchers obtained the raw numbers for each person in each trial rather than working from published summaries, which lets them reanalyze everything the same way. Trials were only admitted if the method of concealing who got what was unambiguously adequate.

The result: 39 trials, 20,827 patients, four conditions, which were back and neck pain, osteoarthritis, chronic headache and shoulder pain. Among patients whose pain duration was recorded, the median was four years. These were not people with a sore week.

The authors then did something unusually generous. They translated their effect sizes into response rates a non-statistician can hold in mind. Take a typical trial where pain starts at 60 on a 0 to 100 scale. Define a responder as anyone whose pain drops by half or more. On that definition, roughly 30 percent of people respond with no acupuncture, about 42.5 percent respond with sham acupuncture, and about 50 percent respond with real acupuncture.

Read those three numbers slowly, because they contain the whole dispute. The distance from 30 to 50 is what a patient gains by being referred. The distance from 42.5 to 50 is what the needles add on top of the ritual of being needled. The distance from 30 to 42.5 is the ritual itself, and it is the largest single jump of the three.

Durability pointed the same way. Against no-acupuncture controls, the effect eroded by roughly 15 percent at one year, which in most trials fell nine or ten months after treatment ended.

Why both camps quote the same number

In 2009 the BMJ published a systematic review by Matias Vested Madsen, Peter Gotzsche and Asbjorn Hrobjartsson that looked only at trials with all three arms present. Thirteen trials, 3,025 patients, one clear outlier excluded. Their figure for real versus fake was a standardized mean difference of 0.17, which they converted to four millimeters on a 100 millimeter scale, with a range of two to six.

Vickers reported close to 0.2. Madsen reported 0.17. Those are the same finding reached by different methods on overlapping trials. Presenting them as rival estimates invents a conflict that is not there.

What differs is the verdict. The Madsen team noted that a consensus report treats a 10 millimeter reduction on a 100 millimeter scale as a minimal change, so four millimeters sits below the floor. The conclusion Madsen, Gotzsche and Hrobjartsson printed in the BMJ was that the analgesic effect they measured “seems to lack clinical relevance and cannot be clearly distinguished from bias.”

The Vickers group answered the question differently, and said so openly. Their position is that the choice in front of a doctor and a patient is not acupuncture versus a sham procedure nobody sells. It is acupuncture versus not doing it. On that comparison the gap is about half a standard deviation, and they called a referral a reasonable option for someone with chronic pain.

You do not have to pick. You do have to notice which comparison a headline is quoting, because the two produce opposite-sounding stories from identical data.

The fake needle was never blank

The sham device most trials use descends from a 1998 report in The Lancet by Konrad Streitberger and Johannes Kleinhenz. A blunt needle sits in a sheath and telescopes, so the tip presses the skin without piercing it. Sixty volunteers got both the real procedure and the fake one.

The numbers are worth sitting with. Fifty-four of 60 felt a penetration with the real needle. Forty-seven of 60 felt a penetration with the fake one. Thirty-four felt the dull sensation practitioners call deqi with the real needle, and 13 felt it with a needle that never broke the skin. Nobody suspected a thing.

That paper established the device is believable. It did not establish that the device does nothing. Those 13 volunteers are the crack the next twenty years of argument walked through.

Then came the cleanest test anyone has run. In 2009, Daniel Cherkin and colleagues published a four-arm trial of 638 adults with chronic low back pain in Archives of Internal Medicine, funded by the National Institutes of Health. One arm got individualized acupuncture. One got a standardized point prescription. One got a toothpick tapped in a guide tube against the skin, with the twisting and withdrawal mimicked. One got usual care.

At eight weeks, scores on a 0 to 23 disability scale improved by 4.4, 4.5 and 4.4 points in the three stimulation arms, against 2.1 points for usual care. Sixty percent of people in the stimulation arms hit a clinically meaningful improvement, against 39 percent on usual care. Needle placement did not matter. Skin penetration did not matter. The only thing that separated real needling from toothpicks was harm: six adverse experiences in each real arm, and zero in the toothpick arm.

Whether the sham itself is physiologically active remains genuinely unsettled, and three serious analyses disagree. Vickers found a difference between shams that penetrate and shams that do not, at 0.30 standard deviations, which collapsed to 0.07 and lost significance when two outlying trials were removed. Madsen ranked five kinds of placebo by how likely each was to do something physiological and found no relationship at all. A Cochrane review of acupuncture for knee and hip osteoarthritis, led by Eric Manheimer, reported the problem most sharply. The fake procedures that hid the allocation best from participants were also the ones most likely to be doing something physiological in their own right, and limiting the analysis to those shrank the pooled short-term benefit until it was no longer statistically significant.

Where guidelines name acupuncture, and how strongly

Two mainstream guidelines name it. Read both for three things: the indication, the strength of the recommendation, and the evidence quality attached to that specific practice, which is often not the same as the strength.

The American College of Physicians published its low back pain guideline in Annals of Internal Medicine in 2017, built on evidence through April 2015 with updated searches to November 2016. For acute and subacute pain, clinicians and patients are told to select superficial heat on moderate-quality evidence, or massage, acupuncture or spinal manipulation on low-quality evidence. That is a strong recommendation. For chronic low back pain, acupuncture appears in the moderate-quality group alongside exercise, multidisciplinary rehabilitation and mindfulness-based stress reduction, while spinal manipulation sits in the low-quality group in the same sentence. Also a strong recommendation.

A strong recommendation resting on low-quality evidence is not a drafting error. In this grading system it usually reflects a favorable balance of benefit against harm compared with the alternatives, which here were largely drugs. Strong recommendation is not the same thing as strong evidence, and the guideline says which is which.

A joint pain guideline from two oncology bodies, the Society for Integrative Oncology together with the American Society of Clinical Oncology, appeared in the Journal of Clinical Oncology in 2022, drawing on 227 studies published between 1990 and 2021. Its first recommendation names acupuncture for joint pain caused by aromatase inhibitors, a class of breast cancer drugs. A second groups acupuncture with reflexology and acupressure for cancer pain generally. Both sit at moderate strength of recommendation on an intermediate level of evidence, and the panel found the evidence insufficient to recommend anything for children.

Cochrane’s migraine prevention review, updated in 2016 by Klaus Linde and colleagues across 22 trials and 4,985 participants, gives the single best illustration of why the comparison matters. Headache frequency at least halved in 41 percent of acupuncture patients against 17 percent with no acupuncture, which works out to four people treated for one extra benefit. Against sham, the figures were 50 percent and 41 percent, which works out to eleven. Same practice, same review, same outcome measure, and the number needed to treat nearly triples depending on the control arm.

The National Center for Complementary and Integrative Health, whose acupuncture page carries an October 2022 content date, also catalogs the misses. It reports that acupuncture was no more effective than sham for irritable bowel syndrome, that it may have little or no effect on carpal tunnel symptoms against sham, and that it was not shown to beat sham for hot flashes.

What Medicare pays for, and who can bill it

In January 2020 Medicare began covering acupuncture for the first time, and the fine print matters more than the headline.

The national coverage determination took effect on 21 January 2020 and was implemented on 24 June 2020. It covers up to 12 visits in 90 days, with 8 more for patients who are improving, and no more than 20 treatments in a year. Treatment must stop if a patient is not improving or is getting worse.

The covered condition is narrow. Chronic low back pain here means pain of 12 weeks or more with nothing systemic behind it that anyone can identify, so nothing traceable to cancer that has spread, to an inflammatory disease or to an infection, and nothing connected to surgery or pregnancy. Everything else is excluded, and the rule says so flatly. Dry needling is named in that exclusion, for every condition other than chronic low back pain.

Then comes the part almost nobody reports. The rule names who may furnish the treatment. Physicians may. So may physician assistants, nurse practitioners and clinical nurse specialists who meet state requirements and hold a master’s or doctoral degree in acupuncture or Oriental medicine from an accredited school plus a current unrestricted acupuncture license. Auxiliary personnel may, under supervision, meeting the same requirements.

A licensed acupuncturist who is none of those things cannot bill Medicare directly under this rule. If you are on Medicare and you read that acupuncture is covered, the operative question is not whether the condition qualifies. It is who is holding the needles.

The harm record, with denominators attached

Serious harm from acupuncture is rare, and the honest way to show that is to report studies that counted treatments as well as injuries.

The largest is a prospective study of 229,230 patients by Claudia Witt and colleagues at Charite University Medical Center in Berlin, published in 2009. Patients averaged 10.2 treatments each, so the study covers roughly 2.3 million needlings. Some 8.6 percent of patients reported at least one adverse effect and 2.2 percent reported one needing treatment. Bleeding or bruising accounted for 58 percent of everything reported. Two patients had a collapsed lung, one needing hospital treatment and one needing observation. The longest-lasting problem was a nerve injury in a lower limb that took 180 days to resolve.

A British prospective survey by Hugh MacPherson and colleagues covered 34,407 treatments and found no serious adverse events, defined as anything requiring hospital admission, prolonging a stay, permanently disabling or fatal. It recorded 43 significant minor events, a rate of 1.3 per 1,000 treatments, including fainting, prolonged symptom flares and prolonged pain or bruising. Three events were avoidable: two patients had needles left in by mistake and one got burns from moxa, the smouldering herb some practitioners use. Only 31 percent of invited practitioners responded and they reported on themselves, so treat it as a floor rather than a count.

A collapsed lung, called pneumothorax, is the signature serious event because it is the one the anatomy predicts. The large series put it near one per million procedures, but those series pool every site including the knee. A case series from a Melbourne emergency department describes three patients who presented between 2014 and 2016 after needling over the shoulder or upper back, with lung collapses of 16 to 26 percent that took 9 to 21 days to resolve. All three were young women with a low body mass index. The authors argue the rate for chest wall needling is likely higher than the pooled figure, and they offer no denominator of their own, so no rate can be calculated from their report.

Two details in that series deserve attention. Two of the three procedures were dry needling, which is often delivered under a different profession’s scope and training, and which Medicare’s rule names among the types of acupuncture it excludes for any condition other than chronic low back pain. Lumping it in overstates acupuncture’s harm record and hides dry needling’s.

Infection is the other documented harm, and its history has a shape. The hepatitis B outbreaks in the published literature predate 2000 and every one involved reused or improperly sterilized needles. Single-use sterile needles are now standard US practice, which is the reason that chapter closed rather than evidence that needling is inherently clean.

The practical signal is narrow and specific. New chest pain or breathlessness in the hours after needling over the upper back, shoulder or chest wall is a reason for urgent medical assessment, not a reason to wait and see. So is a fever or spreading redness at a needle site.

Meridians, points, and what the trials could not locate

Classical acupuncture is organized around meridians, described in the tradition as channels along which vital energy flows, and acupoints positioned along them. Practitioners select points according to that map, and much of a traditional consultation is devoted to deciding which points a particular presentation calls for. Traditional Chinese medicine as a diagnostic system is a larger subject with its own article in this library.

What the trials have not been able to do is find the map in the data. Vickers tested whether outcomes varied with the style of acupuncture, with fixed versus individualized point selection, or with the use of electrical stimulation. None of it made a difference. The only clear modifier of acupuncture’s effect was the number of sessions in trials without a sham arm. Most of the variation between trials came from differences in what control groups received, not from differences in the needling.

The Acupuncture Trialists’ Collaboration, authors of the most favorable major synthesis in this field, wrote in their own opening paragraph that “there is as yet no explanation as to how such effects could persist.” That sentence is in the paper arguing the treatment works. Any article that hands you a confident mechanism is supplying something the field’s own advocates decline to supply.

What to ask before you book

Licensure is real but uneven. A 2018 census by Arthur Fan and colleagues counted 37,886 actively licensed acupuncturists in the United States as of 1 January 2018, a 257 percent increase since 1998, with practice laws in 48 jurisdictions, meaning 47 states plus the District of Columbia. Alabama, Oklahoma and South Dakota had none. California, New York and Florida between them held more than half the national total. That count is now several years old, and the federal complementary health center’s page says only that most states issue licences and that what they require differs from one to the next. The credential inventory across practices like this one is handled elsewhere in this library, and so is the question of what holding a license does not guarantee.

Five questions are worth asking, and the answers tell you more than any brochure:

1. What license do you hold in this state, and what is the number I can verify it with?
2. Will you be using single-use sterile disposable needles, and will I see the package opened?
3. Which points are over my chest wall, upper back or shoulders, and what are you doing to control depth there?
4. How many sessions do you expect before we can tell whether this is working, and what will you consider a failure?
5. Who will you tell about my other treatment, and what do you want to know about my medications and bleeding risk?

The most useful thing to carry away is the habit of asking which comparison a claim rests on, not a verdict. A reported benefit against no treatment answers the question a patient faces. A reported benefit against a sham answers the question about the needles. Any decision about starting, stopping or combining a treatment belongs with the clinician who knows your history, and a new or unusual symptom during a course of treatment belongs with them promptly.

Sources

1. National Center for Complementary and Integrative Health, National Institutes of Health. “Acupuncture: Effectiveness and Safety (content Last Updated October 2022; sham definition, the no-treatment versus sham comparison, condition-by-condition findings, licensure).” nccih.nih.gov
2. Centers for Medicare & Medicaid Services. “National Coverage Determination: Acupuncture for Chronic Lower Back Pain (cLBP), NCD Manual Section 30.3.3, effective January 21, 2020, implemented June 24, 2020.” cms.gov
3. The Journal of Pain. “Vickers AJ, Vertosick EA, Lewith G, MacPherson H, Foster NE, Sherman KJ, Irnich D, Witt CM, Linde K, for the Acupuncture Trialists' Collaboration. "Acupuncture for Chronic Pain: Update of an Individual Patient Data Meta-Analysis." J Pain, 2018. PubMed. DOI: 10.1016/j.jpain.2017.11.005.” pubmed.ncbi.nlm.nih.gov
4. BMJ. “Madsen MV, Gotzsche PC, Hrobjartsson A. "Acupuncture treatment for pain: systematic review of randomised clinical trials with acupuncture, placebo acupuncture, and no acupuncture groups." BMJ, 2009. PubMed. DOI: 10.1136/bmj.a3115.” pubmed.ncbi.nlm.nih.gov
5. Archives of Internal Medicine. “Cherkin DC, Sherman KJ, Avins AL, et al. "A randomized trial comparing acupuncture, simulated acupuncture, and usual care for chronic low back pain." Arch Intern Med, 2009. PubMed. DOI: 10.1001/archinternmed.2009.65.” pubmed.ncbi.nlm.nih.gov
6. The Lancet. “Streitberger K, Kleinhenz J. "Introducing a placebo needle into acupuncture research." Lancet, 1998. PubMed. DOI: 10.1016/S0140-6736(97)10471-8.” pubmed.ncbi.nlm.nih.gov
7. Cochrane Database of Systematic Reviews. “Linde K, Allais G, Brinkhaus B, Fei Y, Mehring M, Vertosick EA, Vickers A, White AR. "Acupuncture for the prevention of episodic migraine." Cochrane Database Syst Rev, 2016. PubMed. DOI: 10.1002/14651858.CD001218.pub3.” pubmed.ncbi.nlm.nih.gov
8. Cochrane Database of Systematic Reviews. “Manheimer E, Cheng K, Linde K, et al. "Acupuncture for peripheral joint osteoarthritis." Cochrane Database Syst Rev, 2010 (searches through December 2007). PubMed. DOI: 10.1002/14651858.CD001977.pub2.” pubmed.ncbi.nlm.nih.gov
9. Annals of Internal Medicine (American College of Physicians). “Qaseem A, Wilt TJ, McLean RM, Forciea MA. "Noninvasive Treatments for Acute, Subacute, and Chronic Low Back Pain: A Clinical Practice Guideline From the American College of Physicians." Ann Intern Med, 2017. PubMed. DOI: 10.7326/M16-2367.” pubmed.ncbi.nlm.nih.gov
10. Journal of Clinical Oncology (Society for Integrative Oncology and ASCO). “Mao JJ, Ismaila N, Bao T, et al. "Integrative Medicine for Pain Management in Oncology: Society for Integrative Oncology-ASCO Guideline." J Clin Oncol, 2022. PubMed. DOI: 10.1200/JCO.22.01357.” pubmed.ncbi.nlm.nih.gov
11. Forschende Komplementarmedizin. “Witt CM, Pach D, Brinkhaus B, Wruck K, Tag B, Mank S, Willich SN. "Safety of acupuncture: results of a prospective observational study with 229,230 patients and introduction of a medical information and consent form." Forsch Komplementmed, 2009. PubMed. DOI: 10.1159/000209315.” pubmed.ncbi.nlm.nih.gov
12. Acupuncture in Medicine. “MacPherson H, Thomas K, Walters S, Fitter M. "A prospective survey of adverse events and treatment reactions following 34,000 consultations with professional acupuncturists." Acupunct Med, 2001. PubMed. DOI: 10.1136/aim.19.2.93.” pubmed.ncbi.nlm.nih.gov
13. Clinical Practice and Cases in Emergency Medicine (Western Journal of Emergency Medicine). “Grusche F, Egerton-Warburton D. "Traumatic Pneumothorax Following Acupuncture: A Case Series." CPC-EM Vol 1(1) (three cases, Melbourne emergency department, 2014 to 2016).” westjem.com
14. Complementary Therapies in Medicine. “Fan AY, Stumpf SH, Faggert Alemi S, Matecki A. "Distribution of licensed acupuncturists and educational institutions in the United States at the start of 2018." Complement Ther Med, 2018. PubMed. DOI: 10.1016/j.ctim.2018.10.015.” pubmed.ncbi.nlm.nih.gov