Pathways and Patterns of Cell Loss in Verified Alzheimer’s Disease: A Factor and Cluster Analysis of Clinico-Pathological Subgroups
Pathways and patterns of cell loss in verified A l z h e i m e r ' s disease: a factor and cluster analysis of clinico-pathological subgroups
H. Forstl 0 1
R. Levy 0 1
A. Burns 0 1
P. Luthert 0 1
N. Cairns 0 1
0 Mannheim , Germany
1 1Section of Old Age Psychiatry, 2Department of Neuropathology and 3MRC Alzheimer's Disease Brain Bank, Institute of Psychiatry , London , UK
Thirty-seven patients with neuropathologicaUy verified Alzheimer's disease (AD) have been studied prospectively. A principal components analysis of neuron numbers in cortical and subcortical areas revealed two variables: Variable I with high loadings for the hippocampo-parahippocampo-parietal neuron counts and Variable n with high loadings for coeruleo-frontal cell numbers. Both may reflect functional neuroanatomical connections which may act as pathways of neurodegeneration in AD. A cluster analysis based on these neuron numbers yielded three groups of patients: Cluster A with low hippocampoparahippocampo-parietal cell counts, Cluster B with well-preserved neuron numbers, and Cluster C with low coeruleo-frontal neuron numbers. Differences in clinical features between these patient groups indicated the potential clinical relevance of these clusters.
Alzheimer's disease; neurodegeneration; sub-groups; neuropathology
INTRODUCTION
Alzheimer's disease (AD) manifests a large variety of
clinical symptoms and signs, course characteristics
and neuroradiological or neuropathological changes
(Burns et ai., 1990a)
. Several attempts have been
made to characterize clinical "subtypes" of AD and
to corroborate these distinctions by post-mortem
evidence of different types of underlying pathology
(Berrios, 1985; Jorm, 1985; Bondareff et ai., 1987)
.
Because of the paucity of clinico-pathological studies,
it is still unclear whether the clinical heterogeneity of
AD is related to different patterns of neuro
degeneration which may develop along different func
tional pathways
(Pearson et ai., 1985; Hertz, 1989)
.
Clinical or neuropathological evidence for different
subtypes of AD should be validated by external cri
teria
(Jorm, 1985; Mohr et ai., 1990)
. We have there
fore examined both neuropathological and clinical
variables in patients with verified AD. The following
questions were addressed using principal components
and cluster analysis:
(1) Can the variation of neuron counts in different
brain areas be explained by principal components
related to functional neuroanatomical pathways?
(2) Can different patterns of neuronal loss be
detected which distinguish different clusters of
patients?
(3) Are these potential subtypes of AD associated
with other characteristic clinical features?
METHODS
Demographic characteristics of the patient sample,
details about the prospective clinical examination
and the neuropathological work-up have been pub
lished in previous papers
(Bums et ai., 1990a; Forstl
et al., 1992a)
. The clinical diagnosis of AD according
to NINCDS-ADRDA criteria (McKhan 1984 etal)
was verified neuropathologically in 56 of the first 65
patients from a prospective longitudinal study who
came to post-mortem examination
(Forstl et ai.,
1992a)
. The clinical examination was last adminis
tered within 12 months before death. It included the
Clinical Dementia Rating
(CDR; Berg, 1984)
, the
Cambridge Cognitive Examination
(CAMCOG; Roth
et ai., 1986)
, the Geriatric Mental State Schedule
(GMSS; Copeland et ai., 1976; Gurland et ai., 1976)
and a standardized neurological examination (Bums
et ai., 1990b).
The following tissue blocks were taken from the
brains, fixed in 10% formol saline and embedded
in paraffin wax: frontal lobe (including Area 32),
parietal lobe (Area 7), mediotemporal lobe (para
hippocampal gyrus, hippocampus), mesencephalon
(substantia nigra) and pons
(including the largest
diameter of the locus coeruleus and the dorsal raphe
nucleus; Forst! et al., 1992a)
. Fourteen p.m sections
were stained according to Kluver Barrera and impreg
nated with silver according to Glees and Marsland.
Our earlier analyses had shown significant associ
ations between the clinical features and neuronal
change, therefore we decided to study neurons, and
not plaques and tangles. Large neurons were defined
as Nissl-positive, nucleolated cells with a maximum
diameter of more than 20 p.m in cortex (layer III),
hippocampus (pyramidal cell layer of the CAl field),
substantia nigra and locus coeruleus, and of more
than 25 p.m in the dorsal raphe nucleus. All counts
reported in this paper were carried out visually with
an ocular grid at x 400 magnification. Numbers are
given as counts per mm2 for the cortical and hippo
campal areas and as counts per nucleus per horizontal
section for the brainstem nuclei. The examiner was
blind to the clinical findings. A complete set ofartefact
free slides and stains was available from 37 cases.
Variables accounting for the variance of cell num
bers in different brain areas of the patients with
ver (...truncated)