For the most part, the points of disagreement that Drs Blansjaar and van Schaick raise are differences more of emphasis than of substance.

True, the diagnosis of the type of dementia, particularly in late old age, is not always easy. Dementia in later life is probably best described as a syndrome, the emergence of clinical dementia being dependent upon the interplay of two or more pathologies. The ‘Nun Study’ by Snowdon et al (1997) is probably one of the best demonstrations of this. Yet, it is believed that Alzheimer's disease is a major cause of dementia in later life. Even without having seen the patient, one is going to be correct in a diagnosis of Alzheimer's disease, or Alzheimer's disease and cerebrovascular disease, 75% of the time. Rarer diagnoses, such as frontal—temporal dementia and Lewy-body disease, should suggest themselves if they are kept in mind, a careful history taken and the patient followed-up so that departures from the normal symptom progression for Alzheimer's disease are noted. Of course, mistakes in diagnosis will occur, but we think that this will occur insufficiently frequently to compromise the very broad-brush approach to estimating the familial risk of dementia that we have advocated.

As to whether the rate of increase in the incidence and prevalence of dementia begins to slow or goes on increasing exponentially into extreme old age, this is a controversial area, which is, in fact, also highlighted in the two meta-analyses cited by Drs Blansjaar and van Schaick. Jorm & Jolley (1998) suggest that “the incidence rises exponentially up to the age of 90 years”. Gao et al (1998) suggest that “ the acceleration of incidence rates for AD and dementia slows down with the increase in age, although we find no evidence of a rate decline”. Faced with such difficulties of interpretation, we can only commend the clarity of Blansjaar et al's own study (2000), which suggests that the increase in dementia prevalence does not slow down in extreme old age.

We agree that the risk of a first-degree relative of a proband with Alzheimer's disease developing the disorder once they reach the age of 85 may be one in three, if not higher. Perhaps this point could have been made more clearly in our review. The main point we tried to make was that the actual likelihood of surviving to age 85 and developing Alzheimer's disease is lower. We disagree that showing graphs to anxious relatives is “ obfuscating this information”, but we accept that Drs Blansjaar and van Schaick and, indeed, other clinicians may think differently.

In non-Mendelian Alzheimer's disease it is difficult to estimate how much the risk increases as the number of affected first-degree relatives goes up, principally because few studies have addressed this issue. However, the ‘conjugal Alzheimer's disease’ study of Bird et al (Reference Snowdon, Greiner and Mortimer1993), which we cited, and the transmission study of Farrer et al (1990), which we did not cite, indicate that the risk increases substantially. With such pedigrees showing apparently high genetic loading for Alzheimer's disease, we suggested that a psychiatrist seek the advice of a clinical geneticist.

Finally, we agree that it is often reassuring to point out that the course of dementia in late old age is usually more slowly progressive and more benign than dementia occurring in a younger person.