Lack of accuracy for the proposed 'Dubois criteria' in Alzheimer's disease: a validation study from the Swedish brain power initiative.

BACKGROUND/AIMS: Our purpose was to investigate whether the new research criteria for Alzheimer's disease proposed in 2007 by Dubois et al. are valid in a naturalistic memory clinic sample.
METHOD: Retrospective diagnostic analyses were carried out to compare the traditional diagnostic criteria for dementia with the new criteria suggested by Dubois et al. No patient had gone through all procedures postulated as additional features in the proposed new Dubois criteria. MATERIAL: Two independent experienced geriatricians re-examined 150 complete patients' records. The study physicians were blinded to any of the results of the core and additional features suggested by Dubois et al. to avoid circular diagnostic bias.
RESULTS: Among our 96 patients with a clinical diagnosis of subjective cognitive impairment and/or mild cognitive impairment, 2 of the patients with subjective cognitive impairment and 5 patients with mild cognitive impairment would classify as pre-dementia Alzheimer's disease according to the Dubois criteria. In our 23 Alzheimer patients diagnosed clinically, only 12 of the cases fulfilled the criteria for Alzheimer's disease suggested by Dubois et al.
INTERPRETATION: The proposed new criteria for Alzheimer's disease are valid in 55% of our patients clinically diagnosed as having full-blown Alzheimer dementia. Additionally, 7.3% 'true' Alzheimer cases will be identified in a group of 96 clinically non-demented patients. Our results show that there is a large heterogeneity in a clinical naturalistic sample of patients with an Alzheimer phenotype.
CONCLUSION: There is a need to further validate the currently existing biomarkers in large unselected samples and avoid the pitfall of workup bias and circular diagnostic processes. Additionally, valid age-specific cut-off values for the diagnostic markers in question have to be defined.