Mild cognitive impairment (MCI) is a decline in memory or thinking that tests can measure but that leaves a person able to manage daily life; the amnestic form, in which memory is affected first, is the one most likely to progress, as Deng and Wang note in their 2016 review. Dementia is the stage at which the decline does interfere with daily life. Alzheimer disease is its most common cause, and vascular dementia follows damage to the brain’s blood supply, often after a stroke. Most trials on this page score cognition with the Mini-Mental State Examination (MMSE), a 30-point screening test on which a higher score is better, or with the cognitive subscale of the Alzheimer’s Disease Assessment Scale (ADAS-Cog), which runs from 0 to 70 and on which a lower score is better. Most trials pooled in the reviews below compared acupuncture with a drug or added it to one; few compared it with sham acupuncture, and those few were small.
The Cochrane review
Cochrane has published one review of acupuncture in this area. Acupuncture for vascular dementia, by Weina Peng and colleagues, was published in April 2007; the authors ran a new search in January 2011 and the review has not been updated since. It accepted randomised trials of acupuncture against placebo or no intervention. The authors identified 146 studies and found none that met that test, so the review pooled nothing. They concluded that the effectiveness of acupuncture for vascular dementia was uncertain, and called for double-blind trials with placebo and no-treatment controls. A search of the Cochrane Library in October 2026 found no Cochrane review of acupuncture for Alzheimer disease or for MCI.
Later reviews
The reviews published since then appeared in other journals, and most searched Chinese databases as well as English ones. They differ in what the control group received, and the table keeps those comparisons apart.
| Review | Condition | Searched to | Trials (people) | Comparison | Pooled result | Authors’ rating of the evidence |
|---|---|---|---|---|---|---|
| Lee, Shin and Ernst 2009 | Alzheimer disease | August 2008 | 3 | Drugs | No significant difference in cognition (2 trials, 72 people); daily living favoured drugs | Trials too few; no effect shown |
| Zhou and colleagues 2015 | Alzheimer disease | June 2014 | 10 (585) | Drugs; acupuncture added to donepezil | MMSE 1.05 higher than drugs (6 trials); 2.37 higher added to donepezil (3 trials) | Not graded |
| Wang and colleagues 2020 | Alzheimer disease | January 2019 | 30 (2,045) | Drugs; acupuncture added to drugs | No significant difference against drugs; MMSE 1.94 higher added to drugs, short term | Calls for larger, higher-quality trials |
| Deng and Wang 2016 | Amnestic MCI | July 2015 | 5 (568) | Nimodipine; acupuncture added to nimodipine | MMSE 0.99 higher than nimodipine; 1.09 higher added to it | Generally poor trial quality |
| Kim and colleagues 2019 | MCI | August 2017 | 5 (257) | Electroacupuncture against drugs | MMSE 0.65 and MoCA 0.66 higher | High or unclear risk of bias |
| Yin and colleagues 2022 | MCI | August 2022 | 11 (602) | Drugs (10 trials); sham (1 trial) | MMSE 1.22 higher across both | Low (GRADE) |
| Chen and colleagues 2022 | Vascular dementia | Not stated in the abstract | 34 (2,672) | Not separated in the abstract | MMSE 3.07 higher; no difference in daily living | Low to moderate |
In vascular dementia, the abstract of the review by Chen and colleagues does not say what the control groups received. The authors advised caution in using acupuncture because their analysis rested on low- to moderate-certainty evidence, and asked for large, high-quality trials.
Against drugs
The first review in the table, by Myeong Soo Lee, Byung-Cheul Shin and Edzard Ernst in 2009, searched 17 databases and found three randomised trials of needle acupuncture in Alzheimer disease, two of them against drugs. The authors concluded that the evidence did not show acupuncture to be effective for the condition, while noting how few studies there were. Zhou and colleagues, at Guang’anmen Hospital in Beijing, repeated the search to 2014 with Chinese databases added and reported higher MMSE scores with acupuncture than with drugs. Wang and colleagues, searching to 2019, found no significant difference against drugs such as donepezil and nimodipine, in cognition, daily living or adverse events. In MCI, Deng and Wang and Kim and colleagues both reported higher MMSE scores with acupuncture or electroacupuncture than with drugs, and both warned that the trials behind those figures were of low methodological quality. Kim’s review found that none of its five trials reported adverse events.
Added to drugs
Zhou, Wang and Deng each pooled trials that gave acupuncture alongside a drug and compared that with the drug alone, and each found higher cognitive scores in the combined group. In these trials the control group had no needling and no sham, so the comparison cannot separate the needles from the extra attention and contact that come with them. Sham acupuncture describes how trials try to make that separation.
Against sham
Yin and colleagues included sham acupuncture as a control and graded their evidence with GRADE. Their search to August 2022 found one sham-controlled trial among eleven, with 32 participants, and their subgroup result against sham rests on that trial alone. All eleven trials were run in China, and none reported follow-up after treatment ended. The authors rated the cognition outcomes low and the memory and daily-living outcomes low to critically low.
An overview of 35 reviews by He and colleagues, searched to October 2020, assessed their methods with AMSTAR 2, a checklist for systematic reviews, and rated most of them critically low. Of the 73 outcomes the overview graded, 50 were low or very low in certainty. The authors suggested acupuncture as an option where drugs are contraindicated and called for better evidence.
Trials
A 2017 trial at the First Teaching Hospital of Tianjin University of Traditional Chinese Medicine randomised 87 people with mild to moderate Alzheimer disease to acupuncture three times a week for 12 weeks or to donepezil, 5 mg a day rising to 10 mg. The basic prescription was CV17 (Danzhong), CV12 (Zhongwan), CV06 (Qihai), ST36 (Zusanli), TB05 (Waiguan) and SP10 (Xuehai), with further points at the acupuncturist’s discretion. ADAS-Cog scores were lower in the acupuncture group at the end of 12 weeks of follow-up, and global ratings favoured it; daily living and behavioural scores did not differ. Four people stopped donepezil because of adverse events and none stopped acupuncture. The assessors were blinded and the acupuncturists were not, the authors chose not to use a sham, and the trial was registered after it had run.
A 2015 pragmatic trial in Tianjin, led by Guang-Xia Shi and Cun-Zhi Liu, compared up to 21 sessions over 6 weeks plus routine care with routine care alone in vascular dementia. It randomised 48 people; 20 more who declined randomisation were given acupuncture and followed as a third group. The prescription centred on GV20 (Baihui), Sishencong (XH01A to XH01D), GV24 (Shenting) and PC06 (Neiguan), with points on the trunk and legs. MMSE scores did not differ significantly between the three groups; only when the two acupuncture groups were pooled did their score exceed the control group’s. Daily-living scores improved more than the control group’s in the non-randomised group and in the pooled groups, and quality of life did not differ.
A 2025 trial by Yin and colleagues randomised 72 people with amnestic MCI to acupuncture or a waiting list and reported larger ADAS-Cog changes with acupuncture at weeks 12 and 24, alongside brain-imaging findings. A waiting list is a no-treatment control, so it does not account for expectation.
A 2025 pilot from the Korea Institute of Oriental Medicine compared acupuncture with sham. It randomised 30 people with MCI to 24 sessions over 12 weeks at GV20, Sishencong, CV12 and, on both sides, ST36, HT07 (Shenmen), KD03 (Taixi) and SP06 (Sanyinjiao), or to a non-penetrating sham needle at 14 sites away from recognised points. ADAS-Cog scores fell in both groups and the difference between them was not significant. A memory subscore leaned towards acupuncture without reaching significance. Blinding held, and no adverse events were linked to treatment. The authors called the study a feasibility test and asked for a larger trial.
What guidelines say
NICE’s dementia guideline, NG97, published on 20 June 2018, states at recommendation 1.4.5: “Do not offer acupuncture to treat dementia.” NICE last reviewed the guideline on 24 October 2025 and decided not to update its recommendations, so the advice is current. Clinical guidelines compares other bodies’ positions on acupuncture.
What the reviews could not settle
The one Cochrane review found no trial to include, and its search is from 2011. The later reviews pool trials against drugs and trials that add needling to drugs, and neither design separates a specific effect of needling from the attention that comes with it. The sham-controlled evidence is the single trial in Yin’s 2022 review and the Korean pilot, and the trials on this page followed people for weeks rather than years; Where the trials come from discusses what a trial’s country tells a reader. None of the MCI reviews on this page pooled progression from MCI to dementia. Cognitive impairment after stroke has its own literature, covered in Stroke recovery.
AcuiQ’s cognitive impairment, mild cognitive impairment, dementia, Alzheimer’s disease and vascular dementia pages list the protocols individual studies prescribed, each with its citation.