Skip navigation

Dementia - diagnosis and treatment

In Bandolier 40 we examined a trial of a new dementia drug, donepezil, and criticised it because of the unhelpful and preliminary way the information had been presented. So this month it is a pleasure to report a really well designed study of Ginkgo biloba extract which is much easier to interpret. But first our Finnish correspondent brought to Bandolier's attention an important investigation of how different ways of diagnosing dementia produce wildly differing results.


A detailed study of systems of diagnosing dementia has shown a 10-fold variation in prevalence in 1879 subjects, from 3.1% to 29% [1].

A Canadian study of health and ageing surveyed 10,263 people aged 16 years or more. Of these 1879 had a full clinical examination, including a neuropsychological examination. Records were then examined independently by a neurologist, a neuropsychologist and a nurse to agree whether each patient met criteria included in each of a number of different diagnostic systems:

  • Diagnostic and Statistical Manual of Mental Disorder (DSM)-III
  • DSM-IV
  • International Classification of Disease (ICD)-9
  • ICD-10
  • Cambridge Examination for Mental Disorders of the Elderly (CAMDEX)


The mean age of the 1879 included in the study was 80 years, and two-thirds lived in the community. The results for the different schemes in this population are shown in the Figure.

The lowest prevalence was 3.1% with ICD-10, and the highest was DSM-III with 29.1%. Clinical consensus in the Canadian study was 20.9%. Only 20 people were given a diagnosis of dementia by all six systems.


Bandolier 18 reported on a systematic review pointing out the efficacy of Ginkgo biloba in cerebral insufficiency. Now a trial of an extract of Ginkgo biloba has shown efficacy in patients with dementia [2]. It is a cracking study, beautifully reported, and has exactly the sort of information Bandolier found lacking earlier in a trial of donepezil ( Bandolier 40 ).


This was a randomised, double-blind study in which patients with dementia were given extract of Ginkgo (called EGb, now licensed in Germany) as a 40 mg tablet before each main meal, or matched placebo. The planned duration was 52 weeks with assessments at baseline, and 2, 12, 26 and 52 weeks.


Patients were 45 years or older with a diagnosis of uncomplicated dementia according to the DSM-III-R and ICD-10 criteria, with either Alzheimer's disease or multi-infarct dementia.


There were three main outcomes. The first was ADAS-Cog, an Alzheimer Disease Assessment scale. The total score can range from 0 to 70, and the higher the score, the poorer the performance (we got that wrong in Bandolier 40 ). A daily living and social behaviour score (GERRI) was completed by family members. The third was a Clinical Global Impression of Change (CGIC) score, completed by the physician.


A total of 549 patients were screened, 327 were randomised, and data from 309 patients was used for analysis. There was a high rate of withdrawal after baseline, and 50% of patients on Ginkgo completed 52 weeks, compared with 38% with placebo.

Results of Ginkgo extract in dementia at one year
Improved on Gingko Improved on placebo Relative benefit (95% CI) NNT (95% CI)
All dementia patients
ADAS-Cog 2 points better 48/96 30/104 1.7 (1.2 to 2.5) 4.7 (2.9 to 13)
ADAS-Cog 4 points better 26/96 15/104 1.9 (1.1 to 3.3) 7.9 (4.2 to 67)
GERRI improved 33/89 20/88 1.6 (1.02 to 2.6) 7.0 (3.3 to 97)
Alzheimer patients
ADAS-Cog 2 points better (lower) 40/75 21/75 1.9 (1.3 to 2.9) 4.0 (2.5 to 9.9)
ADAS-Cog 4 points better (lower) 22/75 10/75 2.2 (1.1 to 4.3) 6.3 (3.5 to 32)
GERRI improved 26/75 13/65 1.9 (1.1 to 3.4) 5.3 (2.9 to 28)
In the Ginkgo group the ADAS-Cog score was unchanged at 52 weeks compared with a mean increase (worsening) of 1.5 points for placebo patients. The GERRI scores worsened on average with placebo, but showed a continuing small improvement with Ginkgo. These differences were highly significant between groups. There was no significant difference for the physician score.


Categorical results were also available. So the number of patients with 2 and 4 point ADAS-Cog improvement, and improved and worsened GERRI scores were provided. This allowed calculation of NNTs, which are shown in the Table for all patients and for just those with Alzheimer's disease. For ADAS-Cog scores showing an improvement of 4 points or more, the NNT was 7.9 (4.2 to 67) for all patients and 6.3 (3.5 to 32) for those with Alzheimer's disease.

This means that about seven patients have to be treated with 120 mg of Ginkgo extract daily for one year for one of them to have an improved ADAS-Cog score of four points which they would not have had with placebo. For a two point improvement, about four patients have to be treated for one year. For a patient's family member to notice an improvement in their daily living and social behaviour about 7 patients have to be treated for one year.


High-quality clinical research in dementia will never be easy, but the authors of this study have done a brilliant job in making the study and its results transparent and usable. They comment on the clinical value of changes in the scores, particularly the ADAS-Cog score. They say that an improvement of four points may be equivalent to a six month delay in progression of the disease. But the trial does not permit conclusions about sustained effects, even though at one year it was longer than many studies in dementia. And because of the way in which they have made conservative decisions about the way data were treated, Ginkgo extract may be undervalued by their treatment.

But there is another point. The difference in ADAS-Cog score between treatment and placebo of 1.5 points after one year should be compared with 3.2 after treatment with donepezil for 14 weeks (though12 weeks seems to be the time of peak improvement). There is a need for someone to define what is a useful outcome, and how long it must be sustained. As more treatments for dementia and Alzheimer's disease appear, the need for some common sense to allow the right drugs to be used for the right patients at the right time increases.

What constitutes dementia to begin with? As the authors of the study on diagnosis comment, a diagnosis of dementia might mean not being able to drive, or make a will, or manage one's own affairs. Prevalence estimates differing by a factor of 10 have huge implications for individuals and for health care planners. A paper published a few years ago [3] showed a 15-fold variability in Alzheimer prevalence. The authors concluded that methodological issues were to blame, but without being able to identify the sort of diagnostic criteria that were important. Bandolier thought that maybe the fact that the largest difference was between southern California and China might have given them a clue.

The ability to make an accurate diagnosis underscores the estimation of efficacy of any treatment. The implications are enormous.


  1. T Erikinjuntti, T Østbye, R Steenhuis, V Hachinski. The effect of different diagnostic criteria on the prevalence of dementia. New England Journal of Medicine 1997 337: 1667-74.
  2. PL Le Bars, MM Katz, N Berman et al. A placebo-controlled, double-blind, randomized trial of an extract of Ginkgo biloba for dementia. JAMA 1997 278: 1327-32.
  3. M Corrada, R Brookmeyer, C Kawas. Sources of variability in prevalence rates of Alzheimer's disease. International Journal of Epidemiology 1995 25: 1000-5.

previous or next story in this issue