In terminal Alzheimer's disease (AD) the frequency of plaques was found to be reduced in single cases. To test this finding in a larger sample, and in order to determine whether the number of plaques labeled with different markers and the distribution of neurofibrillary tangles are correlated positively to each other and to the degree of dementia, a sample of 134 autopsy brains with and 15 without AD-related pathology has been examined. All of the cases were staged according to Braak and Braak. Both the frequency of plaques immunopositive for beta-amyloid, amyloid precursor protein, and apolipoprotein E and that of microglial cells in the cortex and in the white matter were determined semiquantitatively. The content and distribution of PHF-tau was ascertained by ELISA and immunohistochemistry. Both the clinical dementia rating and the global deterioration scale were used as clinical parameters retrospectively. Correlation coefficients were calculated for all parameters and differences were evaluated statistically. With progressive distribution of neurofibrillary tangles and increasing content of PHF-tau the plaque stages and the degree of cortical microglia reaction increased up to the Braak-stages IV and V, thereafter showing a slightly decreasing tendency in the investigated regions. In end-stage AD resorption of beta-amyloid seems to surpass its deposition. The microglial reaction in the white matter correlated neither with the Braak-stage nor with the accumulation of amyloid. With regard to the degree of dementia, both scales correlated well with the pathological changes. Our data show that neuronal cytoskeletal alterations progressively increase with progressive dementia until the end stage of AD in contrast to the frequencies of plaques and cortical microglial cells, and are therefore preferable for staging purposes.