Multivariate analysis of laminar patterns of neurodegeneration in posterior cingulate cortex in Alzheimer's disease.

Posterior cingulate cortex is the site of earliest reductions in glucose metabolism and qualitatively different laminar patterns of neurodegeneration in Alzheimer's disease (AD). This study used multivariate analyses of area 23 in 72 cases of definite AD to assess relationships between laminar patterns of neurodegeneration, neurofibrillary tangle (NFT) and senile plaque (SP) densities, age of disease onset and duration, and apolipoprotein E (ApoE) genotype. No age-related changes in neurons occurred over four decades in 17 controls and regression analysis of all AD cases showed no relationships between neuron, SP, and tau-immunoreactive NFT densities. Principal components analysis of neurons in layers III-Va and eigenvector projections showed five subgroups. The subgroups were independent because each had a full range of disease durations and qualitatively different laminar patterns in degeneration suggested disease subtypes (ST). Cases with most severe neuron losses (STSevere) had an early onset, most SP, and highest proportion of ApoE epsilon4 homozygotes. Changes in the distribution of NFT were similar over disease course in two subtypes and NFT did not account for most neurodegeneration. In STII-V with moderate neuron loss in most layers, cases with no NFT had a disease duration of 3.5 +/- 0.9 years (mean +/- SEM), those with most in layers IIIc or Va had a duration of 7.3 +/- 1 years, and those with most in layers II-IIIab had a duration of 12.1 +/- 1 years. In STSevere, cases with highest NFT densities in layers II-IIIab also were late stage. Finally, epsilon4 homozygotes were most frequent in STSevere, but four statistical tests showed that this risk is not directly involved in neurodegeneration. In conclusion, multivariate pattern recognition shows that AD is composed of independent neuropathological subtypes and NFT in area 23 do not account for most neuron losses.