We read with interest the article by Deb et al (Reference Deb, Braganza and Norton2000) apparently demonstrating findings contrary to our own (Reference Prasher, Chowdhury and RowePrasher et al, 1997). Overall, we agree with the findings by Deb et al, although clarification on several important points is required.

The principle reason why we did not find a statistically significant association (at the 5% significance level) between apolipo-protein E (ApoE) ϵ 4 and Alzheimer's disease in adults with Down's syndrome was because at that time there was a much smaller sample size of adults with Down's syndrome and dementia available for meta-analysis (102 subjects previously included compared to 158 in Deb et al's report). The three additional reports included in Deb et al's metaanalysis are of significantly la ger samples. However, even with this greater number of subjects available for meta-analysis the power remains at 76%. Given the proportions of ϵ4 in the groups with and without dementia in the Deb et al paper, for a power of 90%, a minimum of 224 adults with Down's syndrome and dementia are required to demonstrate statistical significance at the 5% level. Furthermore, the ϵ4 allele frequency in the different studies varies from 5.9% to 33.4% in subjects with dementia (Reference Deb, Braganza and NortonDeb et al, 2000) and therefore future studies are still required if an association between ApoE ϵ4 genotype and Alzheimer's disease in adults with Down's syndrome is to be established.

Deb et al are incorrect to exclude the study by Wisniewski et al (Reference Wisniewski, Morelli and Wegiel1995) because “ they diagnosed Alzheimer's disease on the basis of neuropathological findings alone”. Wisniewski et al (Reference Wisniewski, Morelli and Wegiel1995) made a diagnosis of dementia (not Alzheimer's disease) by a clinical assessment alone “as judged by the physician following the patient”. However, the inclusion of this study in the present meta-analysis makes little difference to the findings by Deb et al (Reference Deb, Braganza and Norton2000) as only one person with an ϵ4 allele was present.

The increase in risk of developing dementia in adults with Down's syndrome (odds ratio 2.02) appears to be less than that in populations with no learning disability where it can be increased by as much as 30 times for people with two copies of the ϵ4 allele (Reference Swartz, Black and St George-HyslopSwartz et al, 1999). From the allele frequency given by Deb et al (Reference Deb, Braganza and Norton2000) the diagnostic accuracy of ApoE ϵ4 for adults with Down's syndrome and dementia is of some clinical value. The sensitivity is 18% (95% CI 13.5-22%) and specificity 90% (95% CI 88-92%). The absence of an ϵ4 allele strongly suggests the absence of Alzheimer's disease. ApoE genotyping in the Down's syndrome population may possibly be used to screen for dementia.

We conclude, as previously (Reference Prasher, Chowdhury and RowePrasher et al, 1997), that the presence of an ϵ4 allele is neither sufficient nor necessary to cause Alzheimer's disease but ApoE ϵ4 genotype does have a role to play in the presentation of Alzheimer's disease in adults with Down's syndrome. The effect is, however, ‘overwhelmed’ by the excessive amyloidosis due to the triplication of the amyloid precursor gene.