Background. Dementia screening instruments, such as the Cambridge Cognitive Examination (CAMCOG), measure a variety of cognitive functions. However, memory impairment generally is the first sign of Alzheimer's disease (AD). It seems logical, therefore, to use only memory-related items for the early detection of AD. We divided the CAMCOG into a memory section and a non-memory section, and tested the hypothesis that the memory section predicts AD better than the non-memory section. We also provide normative data for both sections.

Methods. Normal subjects (N = 169) and patients with incident AD (i.e. satisfying AD criteria between 1 and 3 years from baseline; N = 25) were participants in the Amsterdam Study of the Elderly (AMSTEL), a population-based longitudinal study on cognitive decline and dementia. Patients with prevalent AD (i.e. satisfying AD criteria at baseline; N = 155) were either recruited in a memory clinic or came from AMSTEL. Normal subjects were cognitively intact at baseline and remained so for at least 3 years. The CAMCOG was administered to all subjects. AD was diagnosed by DSM-III-R criteria.

Results. Logistic regression analysis showed that the memory section was related to prevalent AD, whereas in multivariate analysis the non-memory section was not (after correction for the memory score and demographic characteristics). A similar analysis showed that the memory section predicted incident AD, as did a higher score on the non-memory section. The MMSE did not predict incident AD better than age alone.

Conclusion. For the early detection of AD it is best to use the memory and non-memory sections separately instead of the total CAMCOG score.