Skip navigation
Link to Back issues listing | Back Issue Listing with content Index | Subject Index

Alzheimer's disease treatments


A major problem with studies of Alzheimer's disease has been the diagnosis of dementia, and the choice of measurements of dementia progression. These often lack any immediate meaning, and lend themselves to dismissal as surrogate measures. Reporting these measures as means or with mean changes also tends to demean the results. A short but thoughtful analysis on a number needed to treat basis makes the case for Alzheimer treatments much stronger [1].


The review sought any randomised, double-blind placebo comparison with more than 10 patients and of duration longer than one day in patients exclusively with Alzheimer's disease. Outcomes where the percentage or proportion of patients who responded to treatment at some defined level were used to calculate NNTs. Crossover studies and those with enriched enrolment were excluded.


There were five reports that met the inclusion criteria, two on rivastigmine, and one each on donepezil, tacrine and huperzine-A. The results based on ADAS-Cog movements are shown in the Table.

Table: Selected NNTs for Alzheimer's Disease treatments

Outcome Drug Dose (mg) Weeks Active (%) Placebo (%) Relative Benefit (95% CI) NNT (95% CI)
Improved 4 or more on ADAS-Cog Rivastigmine 1-4 26 36/242 (15) 38/239 (16) 0.9 (0.6 to 1.4) N/A
Improved 4 or more on ADAS-Cog Rivastigmine 6-12 26 58/242 (24) 38/239 (16) 1.5 (1.1 to 2.2) 12 (6.6 to 103)
Prevent decline of at least 7 on ADAS-Cog Rivastigmine 6-12 26 139/149 (93) 138/197 (70) 1.3 (1.2 to 1.5) 4.3 (3.3 to 6.4)
Improved 7 or more on ADAS-Cog Donepezil 5 24 23/152 (15) 12/153 (8) 1.9 (1.0 to 3.7) 14 (7 to 572)
Improved 7 or more on ADAS-Cog Donepezil 10 24 38/150 (25) 12/153 (8) 3.2 (1.8 to 5.9) 5.7 (3.9 to 11)
Improved 4 or more on ADAS-Cog Tacrine 160 30 26/64 (40) 29/116 (25) 1.6 (1.1 to 2.5) 6.4 (3.3 to 81)

With higher doses the numbers needed to treat are about 5 over periods of six months or more. An improvement in the ADAS-Cog score of seven points is equivalent to reversing a typical year's cognitive deterioration an Alzheimer's disease.


This short little paper makes you think. The questions about NNTs are about what the comparison is, the duration, the dose, and most importantly how meaningful is the outcome. These authors challenge us about the meaningfulness of the outcomes. They also show that the studies have similar NNTs for other outcomes. And we should remember that ginkgo biloba produces similar NNTs for similar outcomes ( Bandolier 48). The problem most people have, though, is whether the outcomes are meaningful. What is needed here is a new beginning.

First we need an independent definition about what change on what score is meaningful. That definition should include views of healthcare professionals and carers. Defining clinically meaningful harm should be part of this process.

Second we need the individual patient data from the clinical trials assessed against the new definition. That would mean that manufacturers would have to make the data available, because it is unlikely that the information is available from published sources.

Third we need a sensible assessment of costs and benefits. Not a theoretical treatise based on guesswork, but a real world assessment based on evidence.

Now that would be a first.


  1. G Livingstone, C Katona. How useful are cholinesterase inhibitors in the treatment of Alzheimer's disease? A number needed to treat analysis. International Journal of Geriatric Psychiatry 2000 15: 203-207.
previous or next story in this issue