Ginkgo and Memory: 277 Dementia Cases on Ginkgo, 246 on Placebo

Not in the trial designed to answer it. The Ginkgo Evaluation of Memory study followed 3,069 community-dwelling participants aged 72 to 96 for a median of 6.1 years on 120 mg of extract twice daily or identical placebo, and measured cognitive change across five domains. Annual rates of decline did not differ between groups in any of them. For memory the annual z-score decline was 0.043 on ginkgo (95% CI 0.034 to 0.051) and 0.041 on placebo (95% CI 0.032 to 0.050). Attention, visuospatial abilities, language and executive function were the same story. The authors' conclusion is that ginkgo at 120 mg twice daily did not result in less cognitive decline in older adults with normal cognition or with mild cognitive impairment.

No, and the direction of the result is worth knowing. Of 523 participants who developed dementia, 277 were in the ginkgo group and 246 on placebo. The rate was 3.3 per 100 person-years on ginkgo against 2.9 on placebo, giving a hazard ratio of 1.12 with a 95% confidence interval of 0.94 to 1.33 and P=0.21. For Alzheimer's disease the hazard ratio was 1.16 (0.97 to 1.39, P=0.11), and for progression to dementia among those with mild cognitive impairment, 1.13 (0.85 to 1.50, P=0.39). None of these is statistically significant, so this is not evidence that ginkgo causes dementia. What it does mean is that every point estimate landed above 1, which is the opposite of what a preventive would produce.

That was the point of running it. The trial's own framing notes that adequately powered clinical trials testing ginkgo's effect on dementia incidence were lacking, which is why GEM was built: 3,069 participants, five academic medical centres, a median of 6.1 years of follow-up, assessment every six months, and dropout plus loss to follow-up of only 6.3%. This is not a small study that failed to find something. It is a large, long, well-retained study that looked hard and found nothing.

Because the earlier literature was weaker than it looked, and Cochrane says so directly. Its review of ginkgo for cognitive impairment and dementia concluded that many of the early trials used unsatisfactory methods, were small, and publication bias cannot be excluded, and that the evidence ginkgo has predictable and clinically significant benefit for people with dementia or cognitive impairment is inconsistent and unreliable. Small positive studies get published and repeated; large null ones arrive later and travel less far.

GEM tested that explicitly and found no such group. There was no significant effect modification of treatment on the rate of cognitive decline by age, sex, race, education, APOE*E4 allele status or baseline mild cognitive impairment, all with P above 0.05. The APOE*E4 result is the notable one, since that allele is the best-known genetic risk factor for Alzheimer's disease and would be the obvious place to look for a subgroup that benefits. There was not one.

That is not what this evidence says, and the distinction matters. Cochrane's review states that ginkgo biloba appears to be safe in use with no excess side effects compared with placebo, and GEM itself reported that adverse effect profiles were similar in both groups. So this is a page about something that does not work rather than something that harms. The cost of taking it is the money and the false reassurance, not a safety risk. That is a different verdict from supplements where the trial evidence is null and the safety signal is not.

GEM used 120 mg of extract twice daily, which is a standard studied dose rather than a token one, for over six years. This page cannot tell you what a different dose or preparation would do, because that is not what was tested, and it would be dishonest to imply the null result generalises to every formulation ever sold. What can be said is that the largest, longest and best-retained trial of ginkgo for this purpose used a conventional dose and duration, and found nothing on either dementia incidence or cognitive decline.

Because the conclusion does not support one. We place product links on articles where the evidence points at a specific purchase, such as a form and dose a trial actually used to produce a result. This article concludes that a 6.1-year trial in 3,069 people found no effect on dementia and no effect on cognitive decline in any domain. Putting a buy button under that would tell you our buttons mean nothing.