Positive results in acupuncture research and in AcuiQ’s data

In 1775 the French Académie des Sciences resolved to stop examining proposed solutions to a set of classical problems, and machines claiming perpetual motion, because examining them had become a waste of its officials’ time and energy.

The same test applies to a body of evidence: count the failures it reports. This page applies it to acupuncture research, and then to AcuiQ’s own database.

A machine with no losses

In 1998 Andrew Vickers and colleagues reviewed 252 controlled trials of acupuncture indexed in Medline from 1966 to 1995 and sorted them by country of origin. Every trial from China, Japan, Hong Kong and Taiwan was positive, and 10 of 11 from Russia and the USSR. Their second sample, 405 trials of other interventions, showed the same pattern – 99% positive from China, 97% from Russia, 95% from Taiwan and 89% from Japan, against 75% from England.

Their results include this sentence: no trial published in China or Russia/USSR found a test treatment to be ineffective.

A 2020 systematic review tracked 1,758 acupuncture trials registered with the WHO International Clinical Trial Registry Platform between 1990 and 2018 and searched PubMed for their results on 25 February 2019. It found a full paper for 178 of them, 10%.

What we found in our own database

AcuiQ reads treatment protocols out of the published literature at scale and matches them to the symptoms people search for. In July 2026 we audited every source behind every protocol we held – 3,385 distinct references – and resolved each one to its PubMed record to ask a single question: were the subjects human?

The animal-derived protocols came from studies of rats, mice, rabbits, dogs and cats. One was a systematic review of acupuncture for laminitis in horses. They had been sitting in a database that answers questions from people with symptoms, and they had been sitting there for as long as the database has existed.

A search for bradycardia, a slow heart rate, surfaced it. It returned two protocols, all we had. The first came from a 2013 study of moxibustion at different temperatures on cardiac function in rats. The second came from a 2017 paper on mast cell activation in electroacupuncture against pituitrin-induced bradycardia in rabbits. Both were indexed by the National Library of Medicine with the descriptors that say so.

Where the error entered

Neither of those studies is bad science. Both are careful mechanistic work, faithfully reported, correctly indexed by their publishers, doing exactly what animal research is for. The rabbit paper is a competent investigation of mast cell signalling. It makes no claim about treating a human being and never pretended to.

The failure was ours, and it happened at the moment of aggregation. A point prescription validated in an anaesthetised rabbit carries no dose, no depth, no retention time and no point selection that means anything for a person – and every one of those parameters silently acquired a human interpretation the instant the protocol entered a table next to human trials. Nothing was falsified. A row crossed a boundary without the thing that gave it meaning, and the arithmetic proceeded without complaint.

No stage of our pipeline owned the question of whether the evidence supported the claim, and each stage assumed an earlier one had.

A field that has not agreed where the points are

The field documents its own disagreement about where the points are.

When the World Health Organization set out to standardise acupuncture point locations for the Western Pacific Region, it found international disagreement on the position of roughly a quarter of the 361 classical points. 92 were contested, and 86 of them were standardised over 11 meetings in 3 years. 6 points – LI19, LI20, PC8, PC9, GB30 and GV26 – were published with alternate locations, because the delegations could not agree on where they are. That standard was issued in 2008. Those six points still have two official positions.

Until 10 September 2026 this page said that the 11 meetings were 7 informal consultations and 4 task force sessions. That count comes from Lim’s 2010 note in Evidence-Based Complementary and Alternative Medicine, not from the standard, whose own annexes run to 6 informal consultations, 4 task force meetings and a regional meeting in Tsukuba. We took the secondary count without opening the primary.

Our extraction work meets the downstream version of this daily. Reading 266 papers in a single pass that July, we logged source papers mis-coding their own points against the names they themselves printed:

A paper wrote Hegu as LI14, a different point on the upper arm. These are the field’s most familiar points, mis-numbered in peer-reviewed journals. We catch them only because we resolve every point by its printed name rather than by the code beside it.

Rows that contradict themselves

Some rows contradict themselves: two fields of the same row disagree on the same screen. No coverage metric finds this, because coverage counts what is missing and both fields are present.

XL07 is Lanwei, the Appendix point, an extra point below the knee that exists for one purpose. Its page carried a Cautions block reading avoid in acute appendicitis, set directly above an Indications block reading appendicitis, acute. The auricular Appendix point said the same of itself. So did AT16, the Epilepsy point, cautioning against seizure disorders; EL16, the Groove of Hypotension, cautioning against low blood pressure; HX06, the Anus point, advising against use in haemorrhoids.

31 point rows contradicted themselves this way. A further 147 protocols recommended a point whose own caution named a condition that protocol was treating: LI11 and LV03 in 26 hypertension protocols each, BL15 and CV17 in 45 cardiac protocols, ST41 in 15 protocols for the ankle sprain it warns you off. Every one of those cautions tracks the row’s name rather than its clinical risk, which is the signature of text generated from a label. Before the rewrite described below, 759 of the 773 point rows carried a caution, drawn from 154 distinct strings.

Pregnancy is the largest family and the only one with a resolution already available. LI4 Hegu and SP6 Sanyinjiao appear on forbidden-in-pregnancy lists, and both are indicated, in the same catalogue, for inducing labour. The two fit once one sentence is added: they are forbidden because they are held to move a pregnancy along, so at term the prohibition becomes the indication. The catalogue printed the prohibition and the indication, and not that sentence.

Carr’s 2015 review in Acupuncture in Medicine gathered 15 trials in which 823 women received somewhere between 4,549 and 7,234 treatments at forbidden points, and found rates of preterm birth and stillbirth equivalent to untreated controls. The lists themselves differ from text to text. The lists carry no mechanism and no trial, and our catalogue inherited them beside the indications they contradict.

Two rows were corrupted by our own parser. Source catalogues print indications and contraindications in a single field. Our parser split one on the wrong delimiter, and the auricular Ovary point entered the database as treating a condition named etc.contraindications: pregnancy. The Endocrine point came in treating genitourinary disorders such as prostatitis.contraindications: pregnancy. The contraindication changed sign and was served to readers as something the point was good for; the Endocrine row, having lost the only contraindication it had, read No specific contraindications.

Herb names and species

In 1993 Brussels nephrologists reported a cluster of young women with rapidly progressive kidney fibrosis, all of whom had followed a slimming regimen at the same clinic. The formula was supposed to contain Stephania tetrandra; it contained Aristolochia fangchi, which carries aristolochic acid, and in 2000 the New England Journal of Medicine reported urothelial carcinoma in the same cohort. Belgium recorded more than a hundred cases of the resulting nephropathy.

The substitution happened because both plants are traded as 防己, fang ji. Stephania tetrandra is Han Fang Ji. Aristolochia fangchi is Guang Fang Ji. The Chinese drug name names a role in a formula, which plants from two different families fill, one of them a carcinogen.

Our herb catalogue keys every row on a Latin binomial. Until 9 October 2026 it filed the fang ji entries like this:

Until that date the catalogue had no row for the plant behind the Brussels cluster. Guang fangji was filed under a Stephania and carried a caution saying it probably resembled the other Stephanias, an inference drawn from the binomial. On 9 October we re-anchored the catalogue’s binomials on each entry’s Chinese name, using the taxon Wikidata gives that name. Guang fangji is now filed under Isotrema fangchi, the name GBIF accepts for Aristolochia fangchi.

The re-anchoring withdrew the caution on every entry whose genus changed, which took the aristolochic-acid caution off Guan mutong, now filed under Isotrema manshuriense; on 11 October we restored it and gave Guang fangji one of its own, naming the Brussels cluster. Han fangji moved to Sinomenium acutum, which is not the Stephania tetrandra the Brussels formula was meant to contain, so that row still needs a person to decide what it describes.

The binomial fails in the other direction too. Melia azedarach occupies several rows, the fruit, the bark and the tree, under one shared key. One calls it highly toxic and says to avoid all internal use. Another says use only under professional supervision. Hundreds of herb codes are duplicated like this, and on some the duplicate rows disagree about safety. A lookup keyed on the code keeps whichever row it read last; the conformance register measures how many rows that shadows.

No herb appears in any protocol. The herb pipeline treats a binomial held by more than one row as ambiguous and links it to nothing, and it has not written to the live protocol table. A caution generated from a name describes the name, not the plant, and some herb rows warn against a condition the same row lists as treated, the Appendix-point defect again.

Sources we cannot verify

Of those 3,385 sources, 847 resolve to nothing. Bare DOIs from journals no index carries, free-text citations with no retrievable record, papers that exist only as a line someone typed. They account for 1,034 protocols. We cannot confirm those were human trials. We also cannot confirm they were not.

We kept them, and the decision is a judgement call. They are mostly Chinese-language clinical studies, and deleting a quarter of the sources on suspicion would make the reach of an English-language index the measure of what is true. So they stay, and they are labelled, and this paragraph exists so that nobody has to discover the caveat by reading our source code.

What we changed

The animal-derived protocols in that table are all gone – removed from the database and from the extraction journal behind it. At the time the second deletion was what kept them gone, because the sync replayed the journal over the database. The sync now inserts only journal lines it has not seen before, so a deletion from the database stands on its own, and the journal edit keeps the same papers from being proposed again. Every extraction path now runs a species gate before an article reaches the pipeline, and the rule is stated in the extraction spec itself: human subjects only, and an honest empty result is correct when a paper does not support one.

When this page was first published, a search for bradycardia returned nothing. It now returns one entry, from a 2010 trial in Shenzhen that randomised 60 patients with bradycardia either to a plaster of the herb Wu Zhu Yu (Evodia rutaecarpa) over Neiguan (PC6) and Xinshu (BL15) once a day, or to Xinbao pills. After 4 weeks both groups had a higher average heart rate, with no difference between them. The entry records the two points, the 28 sessions, and that the points were treated with a herbal plaster rather than needled.

Earlier versions of this page listed three entries. The second repeated that prescription with no citation and has been removed. The third came from a 2014 trial of electrical stimulation at 4 points during propofol sedation for bronchoscopy, where bradycardia was a side effect: 6 of 30 patients on propofol alone had it, against none of 30 with stimulation. Nobody in that trial was treated for bradycardia, and the entry is now filed under procedural sedation. The species gate asks whether the subjects were human. It does not ask whether the condition a paper reports is the condition a reader searched for.

24 further protocols turned out to be rows that could never render a recommendation at all – no resolvable points, and in 20 cases a corrupted citation sitting where the point codes belong. They inflated our coverage statistics and offered nothing. They have been removed too. One of them, we noticed on the way out, was a study in arthritic rats that our species audit had missed, because the audit only examined rows that had a reference and that row’s reference field was empty.

The self-contradicting cautions are rewritten – 50 point rows, each correction written by hand. We did not generate them, because a machine can see the contradiction but not which field is wrong. The Appendix point now says that acute appendicitis is a surgical emergency and must never wait on a needle. SP1 says that anticoagulants raise the risk of a haematoma at the needle site, which is a fact about technique, leaving uterine bleeding where it belongs in the indications. The forbidden points carry the prohibition, the reason for it, and Carr’s finding, together, in one sentence. Where a caution was doing two jobs the second one survived: GB21 warns about pregnancy and about pneumothorax, and the tool now refuses to apply any rewrite that would drop an anatomical hazard.

The 147 protocols prescribing a point whose caution named what they treated mostly went with the cautions that produced them.

22 protocols still prescribe a point whose caution names a condition the protocol treats. In 18 of them the caution is pregnancy and the use is during a pregnancy meant to continue. At-term use of the forbidden points counts as their intended use, not as a conflict. These figures are computed as this page is served, with the rules AcuiQ’s database audit runs; the audit itself is an internal script with no page on this site. The count moves up as well as down, because new extractions still arrive carrying the defect, so this line of the ledger stays open.

The two inverted rows are unlinked, the Endocrine point’s real contraindication is restored, and the indication half that came in fused to it – genitourinary disorders, prostatitis – is relinked rather than thrown away with the malformed string. The detector now runs inside the internal database audit, so the next one surfaces on the next pass instead of the next year.

No point row contradicts itself today. The list is drawn from the database as this page is served, with the rules the audit runs.

SP6 is indicated for restless foetus syndrome, a mid-pregnancy use, on a point the same row calls forbidden in pregnancy. The other rows on the list caution against needling over the pregnant abdomen on points indicated for conditions of pregnancy. No text edit resolves these: each needs a clinician deciding which field is wrong, and until one does, it stays on the list above.

What the audit shows

What we can measure is structure. Only 10% of registered acupuncture trials had a full paper in PubMed by 2019. Between 1966 and 1995 no trial published in China or Russia found a test treatment ineffective. A quarter of the classical points were disputed until the 2008 WHO standard, and 6 still carry two official locations. The safety lists forbid treatments the same catalogues prescribe, and herbs are indexed by names that plants from two families answer to.

None of this establishes that acupuncture does not work. That is a different claim, it requires different evidence, and this audit does not touch it. The audit establishes something narrower: the evidence base is not currently in a condition to answer the question either way, and a great deal of what is published as clinical evidence will not survive being asked what species it came from.

The audit continues, and this page is extended as it does. The next passes cover credibility grading, duplicate protocols, and whether trials that registered an outcome published the one they registered. We report what we find in our own data first.

Corrections are welcome and will be credited. Every figure here comes from our own corpus or one of the sources listed below, and if one is wrong we would rather know.