Journal of Neuropathology and Experimental Neurology
Copyright q 2004 by the American Association of Neuropathologists
Vol. 63, No. 10
October, 2004
pp. 1028 1037
Neuropathologic Criteria for Diagnosing Alzheimer Disease in Persons with
Pure Dementia of Alzheimer Type
DANIEL W. MCKEEL, JR., MD, JOSEPH L. PRICE, PHD, J. PHILIP MILLER, ELIZABETH A. GRANT, PHD,
CHENGJIE XIONG, PHD, LEONARD BERG, MD, AND JOHN C. MORRIS, MD
Abstract. Universally accepted neuropathologic criteria for differentiating Alzheimer disease (AD) from healthy brain aging
do not exist. We tested the hypothesis that Bielschowsky silver stained total, cored, and neuritic senile plaques (TSPs, CSPs,
and NSPs, respectively), rather than neurofibrillary tangles (NFTs), best discriminate between the 2 conditions using rigorously
defined nondemented (n 5 7) and AD (n 5 35) subjects with no known co-morbidities. We compared lesions in 3 neocortical
regions, in hippocampal CA1, and in entorhinal cortex in 19 men and 13 women between 74 and 86 years at death. The
Clinical Dementia Rating (CDR) was used to assess degree of cognitive impairment within a year of demise. Neocortical
TSP measures provided the highest correlation with expiration CDR: area under the curve (AUC) 5 0.986 with 97.8%
sensitivity at 90% specificity with an estimated cut-point of 6.0 TSP/ mm2. All SP measures yielded higher estimated AUC
and sensitivity for 90% specificity compared to NFTs. Derived TSP cut-points applied to 149 persons with clinical AD
regardless of their neuropathologic diagnosis yielded a sensitivity of 97% and specificity of 84% for TSPs in the 3 neocortical
areas. Thus cut-points based on both diffuse and neuritic SP in neocortical regions distinguished nondemented and AD subjects
with high sensitivity and specificity.
Key Words: Clinical Dementia Rating; Dementia; Neurofibrillary tangles; Neuropathologic diagnosis; Receiver operating
characteristic; Senile plaques.
INTRODUCTION
Histopathologic examination of the brain establishes
that AD lesions are present in sufficient densities to distinguish AD from age-related neuropathology and allows
detection of other dementing disorders (1). No one set of
histopathologic criteria for AD, however, has been uniformly accepted by neuropathologists (2, 3). Some criteria rely on quantitative or semiquantitative determinations of senile plaque (SP) densities (4, 5), whereas other
criteria emphasize neurofibrillary tangle (NFT) pathology
in addition to plaque burden (6, 7). Controversy also exists about the pathologic significance of SP subtypes.
Whether plaque density measures should include diffuse
SPs or be restricted to neuritic plaques only is unresolved
and this contributes greatly to the substantial variability
among neuropathologists in determining whether AD is
diagnosed (8). There is also the concern that SPs and
NFTs may be only downstream pathologic phenomena
whereas other changes, such as synaptic and neuronal
loss, may be more relevant for the disease process (9).
All currently used sets of neuropathologic diagnostic criteria for AD, however, continue to be based on SP or
NFT densities (or both).
From Departments of Pathology and Immunology (DWM, JCM),
Anatomy and Neurobiology (JLP), the Division of Biostatistics (JPM,
EAG, CX), and the Department of Neurology (LB, JCM), Washington
University School of Medicine, St. Louis, Missouri.
Correspondence to: Daniel W. McKeel, Jr., MD, Associate Professor
of Pathology and Immunology, Dept. of Pathology Box 8118, Washington University School of Medicine, 660 So. Euclid Ave., St. Louis,
MO 63110. E-mail: [email protected]
Supported by Grants 5PO1AG003991 and P050A05681 from the National Institute on Aging, Bethesda, MD.
A major consideration is whether any of these criteria
reliably distinguish AD from the brain changes associated
with nondemented aging. Distinctions between aging and
AD often have been postulated to be quantitative rather
than qualitative. Arriagada et al showed that the lesion
distribution of Alzheimer-type pathologic changes in
nondemented elderly individuals matches that of Alzheimer’s disease and concluded that there is ‘‘a commonality
in the pathologic processes that lead to NFTs and SPs in
both aging and AD.’’ (10). This concept underlies the
age-adjusted threshold of SP and NFT densities adopted
by many sets of neuropathologic diagnostic criteria for
AD, which assume an increase in SP and NFT density
with normal aging. Alternatively, it remains possible that
aging and AD may be qualitatively and quantitatively
different.
Our clinicopathologic studies (11–17) find that the
widespread accumulation of neocortical beta-amyloid
(Ab) in the form of diffuse and neuritic plaques is the
salient neuropathological correlate of the earliest clinically detectable forms of AD (16). Subjects just at the
threshold of dementia already feature large numbers of
neocortical plaques, 85% of which are diffuse (primarily
composed of Ab without abnormal neurites); the remaining 15% are neuritic and possess both Ab and dystrophic
neurites. In the neocortex of nondemented subjects,
plaques are either absent or present as diffuse plaques
only in scattered patches. Thus, both plaque burden (density per millimeter squared) and plaque morphology appear to be useful in distinguishing the earliest detectable
form of AD from nondementing aging. In contrast, the
slow accumulation of NFTs that is restricted to entorhinal
cortex and related medial temporal lobe structures occurs
1028
NEUROPATHOLOGIC CRITERIA FOR DIAGNOSING AD
as a function of age and appears not to reliably distinguish AD from aging (17, 18).
Based on these findings, we hypothesized that plaque
burden, but not tangle burden, distinguishes AD from
nondemented aging. To examine this hypothesis, we
compared these neuropathologic markers of AD in nondemented individuals and clinically defined subjects with
AD (19).
The objectives were to 1) determine the utility of SPs
(cored, neuritic, and total) and NFT densities in distinguishing nondemented aging (CDR 5 0) from AD, uncomplicated by other potentially dementing disorders;
and 2) validate the numeric criteria by deriving cut-points
for SPs and NFTs that best discriminate nondemented
controls and AD cases. The main outcome measures of
the study were the sensitivity and specificity for each
marker, considered separately and when their 3 neocortical scores were averaged, using receiver operating characteristic curves (ROCs) to discriminate between the AD
and non-AD subjects.
MATERIALS AND METHODS
Subject Recruitment and Assessment
Subjects were enrolled in longitudinal research studies at
Washington University’s Alzheimer’s Disease Research Center
(ADRC) in St. Louis, Missouri. Recruitment and assessment
procedures have been described in detail (12); recruitment is
based on media appeals for both healthy and cognitively impaired individuals living in the community. All participants are
assessed annually unless prevented by death, refusal, or relocation far from St. Louis. All procedures are approved by the
Washington University Human Studies Committee.
The inclusionary and exclusionary clinical criteria for AD
used in this study have been validated (10). The gradual onset
and progression of impairment in memory and other cognitive
domains is required for diagnosis, comparable to the ‘‘probable
AD’’ category reported by the National Institute of Neurological and Communicative Disorders and Stroke/Alzheimer Disease and Related Disorders Association (19). Other known neurologic, psychiatric, or medical illnesses that are considered to
have the potential to be the primary cause of dementia are excluded at enrollment. Diagnostic accuracy of our clinical criteria for AD is 93% (12). Nondemented subjects lack known
causes of dementia and have no cognitive or functional impairment.
Determination of nondemented or AD status is based on clinical methods. Experienced neurologists, psychiatrists, geriatricians, and nurse clinicians conduct semi-structured interviews
with the subject and a collateral source who knows the subject
well. A neurologic examination of the subject is also conducted.
The assessment protocol contains items from standard brief
cognitive batteries, including the Mini-Mental State Examination (MMSE) (20) and the Short Blessed test (21). The presence
of dementia and, when present, its severity, is staged with the
Clinical Dementia Rating (CDR) (22). The CDR rates cognitive
function along 5 levels of impairment from none to maximal
(0, 0.5, 1, 2, or 3) in each of 6 domains: memory, orientation,
1029
judgment and problem solving, function in the community,
function at home, and personal care. Only impairment caused
by cognitive dysfunction is rated. The global CDR score is derived from the individual ratings in each domain such that CDR
0 indicates no dementia and CDR 0.5, 1, 2, or 3 indicate very
mild, mild, moderate, or severe dementia. Interrater reliability
for the CDR is established (23, 24). A separate 1.5-hour psychometric battery is administered to all subjects at each assessment as described in detail elsewhere (25).
In all subjects who are autopsied, a validated retrospective
postmortem interview is conducted with the collateral source to
assess the cognitive status of the subject from time of last assessment until death (26). Before the autopsy results are known,
a senior clinician reviews all clinical assessments and the postmortem interview and generates an Expiration Summary. This
Summary yields a final CDR and dementia diagnosis for the
subject.
The sample studied here was comprised of individuals in our
longitudinal cohort who came to autopsy from January 1981
through May 1996 and who were not members of a kindred
with a known causative mutation for AD. This was the same
set of cases reported by Berg et al (12). Additional criteria for
the sample included 1) full morphometric data, 2) last clinical
assessment within 2 years of death, 3) a nondemented or an
AD diagnosis, and 4) no other potentially dementing diagnosis.
One hundred fifty-seven individuals had full morphometric
data. We deliberately undersampled those who were at CDR
level 3 at time of death. Of these, 25 were excluded because
their final clinical assessment was more than 2 years prior to
death. Nine were excluded for nonstandard assessment. Of the
others, 8 had questionable AD diagnoses and 73 had some other
potentially dementing diagnoses. The remaining 42 individuals
served as the sample for this study.
The validity study of the cut-points derived from these 42
individuals was applied to the 149 subjects obtained from the
157 subjects with full morphometric data after excluding 8 patients who were either given non-AD diagnoses at expiration
(clinical diagnoses) or were missing expiration diagnoses. The
pathologic diagnoses assigned to these individuals (85 women,
age range 43.4 to 106.2 years; 64 men, age range 50.5 to 95.7
years) is presented in Table 5
Neuropathology
The methods used for quantifying AD brain lesions have
been described in detail in previous publications (11–13), and
will be described here only briefly.
Lesion counts were made on 5 brain regions defined initially
by CERAD (5), where region 1 5 middle frontal gyrus; region
2 5 superior temporal gyrus; region 3 5 inferior parietal lobule
cortex; region 5A 5 hippocampal subfield CA1; region 5B 5
entorhinal cortex at the level of the lateral geniculate body.
Total (TSP) and cored (CSP) senile plaques were assessed
quantitatively using a modification of the Hedreen-Bielschowsky silver method (13). Neuritic pathology (i.e. neuritic senile
plaques [NSPs] and extracellular and intraneuronal NFTs) was
also assessed using this modified method. Both intracellular and
extracellular tangles were included in the NFT counts. Total SPs
included all varieties of argyrophilic diffuse and neuritic
J Neuropathol Exp Neurol, Vol 63, October, 2004
1030
McKEEL, JR. ET AL
Fig. 1. Plaque and tangle counting scheme for neocortical regions 1, 2, 3 and entorhinal cortex (region 5B). Lesion density
counts are performed manually in 10 microscopic 1-mm2 fields (5 upper, 5 lower) per region and marker. Each box marker F
represents a single 1-mm2 microscopic field.
plaques. Diffuse plaques are amorphous or finely fibrillar deposits and lack abnormal argyrophilic neurites or central cores.
Neuritic plaques contained abnormal swollen argyrophilic neurites. CSPs were a subset of neuritic plaques and contained
central compact cores. Neuritic plaque densities were determined by means of one silver method on a section on frontal
cortex within 50 mm of the section used for TSP densities with
the other silver method (13).
NSPs were visualized using the modified Bielschowsky silver
method. To be counted as neuritic a plaque had to contain within its boundaries one or more distended, abnormally configured
neuritic process, whether or not a central core was present.
Many NSPs did not have such a central core in a given plane
of section.
NFTs were identified if the classical tangle configuration was
evident. Two types of tangles were included in this lesion category: 1) classical NFT corresponding to argyrophilic, compact,
and fascicular intraneuronal fibrillary structures that were thicker than neurofilaments; or 2) extracellular NFTs with no associated nuclear or cytoplasmic structures that were composed of
fibrillar, loose structures that stained more faintly (tan or light
brown) with silver than did the blacker intraneuronal tangles.
Extracellular NFTs were identified in the neuropil primarily in
regions 5A and 5B, where conversion of intraneuronal to ghost
tangles has been well documented.
SP counts were done using a 310 objective and a 1 3 1 cm2
reticule with 100 subdivisions. Actual field size was calibrated
J Neuropathol Exp Neurol, Vol 63, October, 2004
using a microscope stage micrometer. Ten contiguous nonoverlapping microscopic fields were counted for each marker (Fig.
1). Five fields were from the upper cortex bordering the pial
surface and 5 fields were from the lower cortex bordering the
grey-white matter intersection. SP counts proceeded from the
depth of a gyrus towards the gyral crest. Conversely, NFT density counts proceeded from the gyral crest towards the trough.
The geometric mean of these 10 field densities was computed
to represent the average density. This methodology provides a
measure of average lesion density for ten 1-mm2 microscopic
fields. A total of 10 mm2 thus was assessed for each individual
anatomic region and 30 mm2 was assessed for the neocortical
composite. A combined plaque score derived from averaged
density counts of the 3 neocortical regions was used.
A modified counting paradigm was used to count region 5A
hippocampal CA1 total and cored neuritic plaques and NFTs.
Lesions were counted in ten 1-mm2 microscopic fields proceeding from the medial CA2/CA1 junction of the pyramidal layer
laterally towards the subiculum. An average of the 10 field
counts was then computed as above. This counting method provided coverage of the entire medial to lateral extent of CA1.
Region 5A and 5B counts from the ten 1-mm2 microscopic
fields assessed were combined to form a hippocampal composite score for each type of lesions assessed (TSP, CSP, NFT).
The reliability of identifying TSPs (Hedreen-Bielschowsky
silver method) and NFTs (modified Bielschowsky silver method) was done by blindly recounting (DWM) these markers in
1031
0.8 6 0.5
0.6 6 0.4
0.5 6 0.2
0.8 6 0.4
1.0 6 0.4
10
82.8 6 5.1
4/6
11.2 6 3.5
23.5 6 4.1 (n 5 8)
6
89.6 6 10.9
2/4
14.8 6 4.7
4.5 6 4.2 (n 5 4)
7
80.6 6 7.1
4/3
14.6 6 2.9
0.7 6 1.0 (n 5 6)
1
0.5
Short Blessed test scores range from 0 (no impairment) to 28 (maximal impairment).
Table 1 lists the demographic features of the study
sample. The mean 6 SD age at death of all 35 demented
Subjects
Age
Gender: M/F
Years of education
Short Blessed test score
Interval from last clinical
assessment to death (years)
RESULTS
0
The distribution of plaque and tangle densities in the 10 fields
for each of the 5 regions shows substantial variability and is
skewed to the right. The statistical analysis is done on the transformed density variable for each subject obtained by first taking
the logarithm of each field density (adding 0.05 to each density
to avoid logarithm of 0) and then taking the average of these
logarithms. Values have been transformed back to the original
scale (counts/mm2) for clarity in presentation.
Since the sensitivity and specificity of a diagnostic test with
a continuous scale of values depends on the specific cutoff point
selected, we chose the ROC curve approach for each neuropathologic marker (27–29) in relation with the neuropathologic
diagnosis. The ROC curve is computed by calculating the sensitivity and specificity for each possible cut-point in the distribution. A common statistic to summarize the diagnostic accuracy of a measure is the area under the curve (AUC). An AUC
of 1.0 reflects perfect discrimination, whereas an AUC of 0.5
reflects only a chance association of a measure with the diagnosis. We computed the AUC (and its 95% confidence interval)
for each neuropathologic marker for each of the anatomic locations.
To better compare criteria across markers and anatomic locations, we calculated a smoothed ROC by fitting the binormal
form using the ROCKIT computer program written by Metz
(ROCKIT 0.9B, Beta version, available from Charles E. Metz,
Department of Radiology, University of Chicago). We calculated the estimated sensitivity (and its 95% confidence interval)
for a fixed specificity of 90%. The corresponding cut-point to
achieve such sensitivity and specificity was also calculated
based on the estimated binormal ROC curves and the inverse
of the standard normal distribution function. The complete set
of subjects with the same neuropathologic markers quantified
was then used to assess the degree to which the cut-off scores
performed in a more heterogeneous collection of cases (12).
CDR at expiration
Statistical Analysis
TABLE 1
Demographic and Clinical Features of the Study Population
The quantitative neuropathologic criteria used to diagnose
AD are derived from National Institute of Aging consensus criteria (4) and thus utilize both neocortical SPs and NFTs. The
average TSP density (number per mm2) of ten 1-mm2 microscopic fields in 6-mm-thick paraffin sections must exceed in at
least 1 standard neocortical region (midfrontal, superior temporal, inferior parietal) the following age-adjusted plaque
scores: less than 50 years, 2 to 5/mm2; 50 to 65 years, 8/mm2
or more; 66 to 74 years, 10/mm2 or more; 75 years or greater,
15/mm2 or more. At least some neocortical NFTs must be present at all ages. Brains from individuals with a CDR of 0.5 at
death or greater with TSP densities that exceed these criteria
are classified as having neuropathologic AD; brains that fail to
meet these criteria are diagnosed as ‘‘no AD.’’
2
Washington University Neuropathologic Criteria for AD
3
83.9 6 6.3
2/1
13.7 6 4.0
16.0 6 2.6 (n 5 3)
3
25 cases. The Pearson correlation coefficients showed reasonable agreement for the independently made sets of counts: r 5
0.95 for TSPs and r 5 0.82 for NFTs.
16
74.8 6 10.7
7/9
13.1 6 3.5
26.0 6 2.3 (n 5 4)
NEUROPATHOLOGIC CRITERIA FOR DIAGNOSING AD
J Neuropathol Exp Neurol, Vol 63, October, 2004
1032
McKEEL, JR. ET AL
subjects (80.4 6 10.5, range 43–105 years) was not different from that of the 7 control subjects (80.9 6 7.1,
range 70–89 years). The CDR 0.5 and 2 groups were
slightly older than the CDR 3 group (p , 0.02). Scores
on the Short Blessed test increased with dementia severity, as expected (too few subjects completed the MMSE,
which was implemented in 1996, to report). There were
no significant differences among the CDR groups for
amount of education or the interval from the last clinical
assessment to death. On average, subjects were last examined clinically within 12 months from the time they
died.
ROC Curve Analysis
Examples of ROC curves are shown in Figure 2. Figure 2A shows data for TSPs in neocortical region 1 and
Figure 2B shows a scatter plot for the same data. Since
1 AD subject had a count of 0, the maximum sensitivity
is 0.97 (34/35). This same sensitivity is maintained for
counts up to 3.5 while the specificity increases to 0.86
(6/7) as 6 of the 7 controls had counts less than 3.5. With
a cut-point of 15.1, all of the controls would be correctly
classified, but the number of AD subjects correctly classified would drop to 0.94 (33/35).
Figure 2C and 2D show similar plots for NFTs in neocortical region 3. Because of the large number of cases
with AD who have counts near 0, using a cut-point of
between 0 and 0.058 results in a sensitivity of 0.71 (24/
34) with a specificity of 0.72 (5/7). Raising the cut-point
to between 0.058 and 0.1 increases the specificity to 1.0,
but drops the sensitivity to 0.65 (22/34). Figure 2E and
2F show the plots for NFTs in hippocampal region 5b
where there is essentially no relationship between the lesion count and the diagnosis.
Table 2 shows the AUC and the corresponding standard error for each of the 3 markers for the anatomical
areas. The AUCs for all types of SP measures are very
high, approaching 1.0. For example, in frontal cortex
TSPs, CSPs, and NSPs fall between the AUC range of
0.950 and 0.971. AUCs for TSPs and CSPs are almost
identical in frontal and temporal cortex, whereas TSPs
have slightly higher AUCs compared to CSPs in the other
Fig. 2. Standard receiver operating characteristic (ROC)
curves are shown in the left panels (A, C, E) and corresponding
lesion density scatter plots are shown in the right panels (B, D,
F). A: ROC for total neocortical plaques in region 1 (midfrontal
cortex). B: Lesion density scatter plot for total neocortical
plaques in region 1 (midfrontal cortex). C: ROC for total neocortical tangles in region 3 (inferior parietal cortex). D: Lesion
density scatter plot for total neocortical tangles in region 3 (inferior parietal cortex). E: ROC for total tangles in region 5B
(entorhinal cortex). F: Lesion density scatter plot for total tangles in region 5B (entorhinal cortex).
TABLE 2
Area under the Curve (AUC) by Anatomic Location for NFTs and SPs
Anatomic location
Tangles
NEOCORTEX
Region 1: dorsolateral middle frontal gyrus (A9/46)
Region 2: superior temporal gyrus (A22)
Region 3: inferior parietal lobule (A39)
Average of Regions 1, 2, 3
0.876
0.925
0.804
0.912
HIPPOCAMPAL
5A: hippocampal CA1 (A28)
5B: entorhinal cortex (A28)
0.836 6 0.102
0.658 6 0.139
J Neuropathol Exp Neurol, Vol 63, October, 2004
6
6
6
6
0.104
0.042
0.073
0.071
Total plaques
0.969
0.970
0.982
0.986
6
6
6
6
0.028
0.026
0.020
0.018
0.972 6 0.024
0.962 6 0.029
Cored plaques
Neuritic plaques
6
6
6
6
0.950 6 0.033
—
—
—
0.971
0.973
0.910
0.970
0.025
0.024
0.060
0.027
0.833 6 0.078
0.934 6 0.067
—
—
1033
—
—
0.6
—
—
0.8
1
55.3
5B: entorhinal cortex (A28)
Estimated sensitivity and 95% confidence interval as calculated from the binormal ROC model.
4.1
70.7
(47.8–87.4)
81.0
(9.7–99.9)
1.4
92.6
(71.0–99.1)
91.7
(74.5–98.3)
22.0
0.7
Averaged over Regions 1, 2, 3
40.6
(1.4–95.7)
3.9
(0.0–78.4)
—
—
0.8
6.0
5.2
8.5
0.8
1.3
Region 3: inferior parietal lobule (A39)
HIPPOCAMPAL
5A: hippocampal CA1 (A28)
—
—
2.0
0.6
—
0.9
92.4
(77.8–98.2)
—
0.9
92.7
(73.5–98.9)
95.2
(82.0–99.2)
73.8
(29.6–96.5)
94.8
(82.3–99.0)
7.9
95.6
(84.1–99.2)
94.2
(80.7–98.9)
97.2
(85.8–99.7)
97.8
(87.0–99.8)
Sensitivity
(%) (range)
Cut-point
No./mm2
Sensitivity
(%) (range)
Cut-point
No./mm2
Total plaques
Sensitivity
(%) (range)
Cut-point
No./mm2
1.1
66.9
(27.6–92.9)
84.8
(61.6–96.0)
63.3
(35.2–85.8)
83.3
(63.1–94.5)
Region 2: superior temporal gyrus (A22)
The data presented strongly suggest that densities of
total neocortical SPs (neuritic plaques and diffuse
plaques) provide very high specificity and estimated sensitivity for the diagnosis of AD. Neither neocortical nor
hippocampal NFTs discriminated nearly as well between
the nondemented and demented AD groups based on the
NEOCORTEX
Region 1: dorsolateral middle frontal gyrus (A9/46)
DISCUSSION
Tangles
A major purpose of this analysis was to establish the
validity of plaque and tangle cutoff values for diagnosing
AD (Table 3). Table 4 shows the observed sensitivity and
specificity when applied to the validity sample of 149
subjects. This heterogeneous group represents the case
mix seen among those seen by our center and who died
with either a clinical diagnosis of AD, or as not demented, regardless of their neuropathologic diagnosis.
Data in Table 5 shows that of the 149 test cases, 11
were not demented and 138 were demented. Definite
(NIA-Reagan high probability) AD was the sole diagnosis in 85 brains while 27 brains had AD pathology
mixed with another secondary dementing disorder. Ten
additional cases had intermediate and low probability
AD. Fourteen cases had either pure or mixed Parkinson
disease (n 5 2), the Lewy variant of AD (n 5 11), or
dementia with Lewy bodies (DLB) (n 5 1) as the sole
dementing diagnosis. Only 1 case had primary vascular
dementia and 1 case had dementia lacking distinctive histopathologic features.
This type of data analysis validates the utility of the
cutoff score and the Washington University ADRC quantitative criteria. The discriminant analysis requirement
that both ‘‘training’’ and ‘‘test’’ data sets are used is also
satisfied.
TABLE 3
Estimated Sensitivity for 90% Specificity
Validity Testing of Cutoff Score Based on Washington
University ADRC Diagnostic Criteria for AD
Cored plaques
The estimated sensitivity along with a 95% confidence
interval for quantitative lesion counts to achieve a specificity of 99.0% is shown in Table 3. Quantitative lesion
density counts to achieve the estimated sensitivity and
90% diagnostic specificity are also given in Table 3.
Using a combined cutoff score of 6.0 TSPs per mm2
of (frontal and parietal) neocortex confers a 97.8% sensitivity for a 90% specificity for establishing a criterionbased quantitative neuropathologic diagnosis of AD.
Quantifying NFTs and using a cut-point of 0.7 in the
same 3 neocortical regions provides only 83.3% sensitivity for the same 90% specificity.
Tables 2 and 3 show that across all 5 regions, TSPs
provide greater estimated AUC and diagnostic sensitivity
at a fixed 90% specificity than do NFTs.
Sensitivity
(%)1 (range)
Diagnostic Sensitivity and Specificity
Anatomic location
Neuritic plaques
3 regions. TSP AUCs are higher compared to AUCs for
NFT across all 5 sampled brain regions.
Cut-point
No./mm2
NEUROPATHOLOGIC CRITERIA FOR DIAGNOSING AD
J Neuropathol Exp Neurol, Vol 63, October, 2004
1034
—
—
—
—
88
78
63
82
81
78
87
93
87
93
HIPPOCAMPAL
5A: hippocampal CA1 (A28)
5B: entorhinal cortex (A28)
81
78
81
—
—
—
94
—
—
—
81
75
94
78
94
97
81
97
88
78
84
84
98
96
97
97
53
68
56
72
NEOCORTEX
Region 1: dorsolateral middle frontal gyrus (A9/46)
Region 2: superior temporal gyrus (A22)
Region 3: inferior parietal lobule (A39)
Averaged over Regions 1, 2, 3
94
88
94
88
Specificity
(%)
Sensitivity
(%)
Specificity
(%)
Sensitivity
(%)
Specificity
(%)
Sensitivity
(%)
Specificity
(%)
Sensitivity
(%)
Anatomic location
Cored plaques
Total plaques
Tangles
TABLE 4
Observed Sensitivity and Specificity in Total Sample using Cut-Points in Table 3
Neuritic plaques
McKEEL, JR. ET AL
J Neuropathol Exp Neurol, Vol 63, October, 2004
estimated AUC and sensitivity with a fixed specificity of
90%. Cutoff scores of 7.9 SPs/mm2 in mid-frontal neocortex (Brodmann area 9/46), 8.5 SPs/mm2 in superior
temporal neocortex (Brodmann area 22), and 5.2 SPs/
mm2 in inferior parietal lobule neocortex (Brodmann area
39/40) will confer 90% specificity and 94% to 98% sensitivity for diagnosing AD. These plaque densities lie
within the range of SP densities diagnostic of AD for
persons 50 to 75 years old at death as recommended in
the original NIA consensus criteria (4). The age-adjusted
neocortical plaque density threshold for persons 75 years
or older was 15/mm2. Average neocortical plaque densities decline modestly rather than increase as a function
of age (12); thus, our cutoff scores should apply to the
diagnosis of AD regardless of age.
We used brain sections from carefully screened, longitudinally assessed research subjects who were clinically
diagnosed as either CDR 0 demented or CDR 0.5 to 3
AD groups. The present results indicate that diffuse
plaques, which have been shown repeatedly to dominate
the neocortical landscape early in the course of very mild
AD (12, 13, 16–18, 30), have great diagnostic import and
should not be ignored.
The identified cutoff values shown in Table 3 were also
consistent with previously formulated Washington University quantitative criteria (11–13) for diagnosis of AD.
These criteria are a modification of the 1985 NIA consensus criteria (4). The main modifications are 1) a modified Bielschowsky silver stain that better visualizes the
whole range of senile plaques, including the diffuse variety (31); and 2) a counting protocol that evaluates the
total number of SPs in 10 contiguous 1 mm2 microscopic
fields in each brain region. This strategy was designed to
assess total average plaque distribution across a 10 square
millimeter expense of cortex, thus precluding a diagnosis
of AD based on only 1 to 3 selected fields. Our counting
protocol differs from the 1985 NIA consensus quantitative method whereby SP density in only a single microscopic field has to meet or exceed criteria, and from the
CERAD semiquantitative counting strategy in which only
NSPs are evaluated and maximal plaque counts in any 3
microscopic fields are averaged and compared to cartoons
that depict mild, moderate, and severe plaque numbers
for 1 field. The recently proposed NIA-Reagan Institute
diagnostic criteria for AD incorporates the CERAD SP
assessment method (6).
We find that neocortical SPs constitute the best marker to differentiate AD, even at the earliest symptomatic
stage (CDR 0.5), from nondemented controls (11–13,
15). Neocortical SPs do correlate moderately (r ; 0.4
to 0.5) with dementia severity as staged by the CDR,
albeit not as robustly as neocortical NFTs. These findings are supported by findings of other investigators who
also employed silver methods to detect SP (32). However, data from many centers, including our own (12),
1035
NEUROPATHOLOGIC CRITERIA FOR DIAGNOSING AD
TABLE 5
Neuropathologic Primary and Secondary Diagnoses in 149 Case Diagnostic ‘‘Test Set’’ for Validity of ROC Cut-Points
Neuropathologic diagnoses (primary and
secondary dementing disorders)
Number of brains
with pathology
Percent of total
brains (n 5 149)
Normal brain, no pathology
No primary, stroke not VaD1 secondary
Pure AD definite (NIA-Reagan high probability)
Mixed AD definite (NIA-R high probability):
● AD definite/high, incidental Lewy pathology
● AD definite/high, hippocampal sclerosis
● AD definite, progressive supranuclear palsy
● AD definite, Other (not otherwise specified)
● AD definite, stroke not VaD1
● AD definite, secondary unknown
Pure Parkinson disease
Mixed Parkinson’s, mixed with stroke not VaD
Pure AD Lewy body variant (AD-LBV)
Mixed AD-LBV with stroke not VaD1
Mixed AD-LBV with Other (not specified)
Dementia with Lewy bodies (DLB), no AD pure
Vascular dementia (VaD) primary
No primary diagnosis, stroke not VaD
Total cases
7
4
85
27
2
3
1
2
18
1
1
1
5
5
1
1
1
1
149
4.7%
2.7%
57.0%
18.1%
(1.3%)2
(2.0%)2
(0.7%)2
(1.3%)2
(12.1%)2
(0.7%)2
0.7%
0.7%
3.3%
3.3%
0.7%
0.7%
0.7%
0.7%
100%
1
2
VaD 5 vascular dementia.
Percentages in parentheses are subcategories under ‘‘Mixed AD definite (NIA-R high probability)’’ main category.
show that neocortical tangle density correlates more robustly than plaque density with measures of dementia
severity (33).
Hippocampal and entorhinal allocortical and transitional cortical SPs also may discriminate CDR 0 subjects
from those with cognitive impairment even at the very
mild Alzheimer-type dementia (DAT) stage, but this effect is less robust than for neocortical CSPs and TSPs
(12).
It is now generally accepted that at least a few hippocampal CA1 and entorhinal cortex layer II (ERC-II)
NFTs occur in the brains of nondemented persons, perhaps beginning around age 50. Hippocampal and ERCII NFT densities increase during the course of AD. We
find that CA1 NFTs have a somewhat more discriminant
value between cognitive normality (CDR 0) and AD primarily because ERC-II stellate neurons develop pretangles and tangles earlier than do hippocampal CA1 pyramidal neurons (7). Significant entorhinal layer II neuron
loss also occurs very early in CDR 0.5 AD (18). Perhaps
because of their early accumulation, ERC-II tangles correlate only weakly with CDR although they build up progressively in that location. One difficulty in assessing the
natural history of ERC-II and CA1 NFTs in AD is their
propensity to truncation by proteolytic digestion accompanied by loss of argyrophilia and PHF tau labeling of
antibodies directed at the protease-susceptible N- and Ctermini (34). These extracellular ‘‘ghost’’ tangles may be
more difficult to count than intraneuronal NFTs, which
label robustly with many silver methods and PHF tau
directed antibodies. Berg et al have shown that neocortical NFTs occur at very low densities in neocortex of
CDR 0.5 AD subjects to the extent that such cases form
a significant proportion of our ‘‘plaque predominant’’ or
‘‘plaque only’’ AD (12).
Our results underscore the clearcut histopathologic differences in plaque and tangle markers between normal
aging, as represented by the marker profile of the nondemented (CDR 0) control group, and the AD-equivalent
marker profile of the very mild AD (CDR 0.5) group.
The marker densities for the very mild DAT group confirm that individuals at the CDR 0.5 stage in this sample
have AD. The present data reinforce and strengthen the
conclusions reached by Morris et al in 1996 (16) and by
Price and Morris in 1999 (18) who proposed that a subset
of nondemented persons whose brains harbored numerous and widespread neocortical plaques (superimposed
on aging-related hippocampal tangles) had ‘‘preclinical’’
AD. The preclinical case histopathologic marker profile
was almost identical to that of the CDR 0.5 very mild
AD group.
The present findings must also be viewed in light of
new and emerging diagnostic criteria for diagnosing AD
histopathologically such as the 1997 NIA-Reagan Institute has proposed (6, 35, 36). This important development has emphasized a probabilistic approach (low, intermediate, and high likelihood) to AD diagnosis that
emphasizes the importance of SPs and NFTs being pathologic lesions. The new diagnostic criteria also emphasize using neurofibrillary pathology according to the hierarchical model described by Braak and Braak in 1991
J Neuropathol Exp Neurol, Vol 63, October, 2004
1036
McKEEL, JR. ET AL
(7). Use of these criteria, however, would result in many
of CDR 0.5 and some CDR 1 cases being classified as
‘‘low probability’’ AD despite convincing clinical and
psychometric data that indicate that they have recognizable progressive dementia, distinct from cognitive changes associated with nondemented aging (37). This suggests
that NFT-based neuropathologic criteria will be weighted
toward diagnosing AD in its moderate and severe stages,
because neocortical neurofibrillary pathology—although
present—is relatively modest in early-stage AD (i.e. very
mild and mild DAT).
ACKNOWLEDGMENTS
We acknowledge the expert histopathologic assistance of Deborah
Carter and the contribution of Linda Baldwin who typed the manuscript.
We also thank the Clinical, Biostatistics and Neuropathology Cores of
the Washington University Alzheimer’s Disease Research Center
(ADRC) for assessing subjects, providing statistical expertise, and providing brain specimens for analysis.
REFERENCES
1. Duyckaerts C, Dickson DW. Neuropathology of Alzheimer’s disease. In: Dickson DW, ed. Neurodegeneration: The molecular pathology of dementia and movement disorders. Basel: ISN Neuropath Press, 2003: 47–65
2. Bancher C, Paulus W, Paukner K, Jellinger K. Neuropathologic
diagnosis of Alzheimer disease-Consensus between practicing neuropathologists. Alzheimer Dis Assoc Disord 1997;11:207–19
3. Wisniewski HM, Rabe A, Zigman W, Silverman W. Neuropathological diagnosis of Alzheimer’s disease. J Neuropathol Exp Neurol
1989;48:606–9
4. Khachaturian ZS. Diagnosis of Alzheimer’s disease. Arch Neurol
1985;42:1097–105
5. Mirra SS, Heyman A, McKeel DW, et al. The Consortium to Establish a Registry for Alzheimer’s Disease (CERAD). Part II. Standardization of the neuropathologic assessment of Alzheimer’s disease. Neurology 1991;41:479–86
6. The National Institute on Aging, Reagan Institute and Working
Group on Diagnostic Criteria for the Neuropathological Assessment
of Alzheimer Disease. Consensus recommendations for the postmortem diagnosis of Alzheimer’s disease. Neurobiol Aging 1997;
18:S1–2
7. Braak H, Braak E. Neuropathological stageing of Alzheimer-related
changes. Acta Neuropathol 1991;82:239–59
8. Mirra SS, Gearing M, McKeel DW, Jr., et al. Interlaboratory comparison of neuropathology assessments in Alzheimer’s disease: A
study of the Consortium to Establish a Registry for Alzheimer’s
Disease (CERAD). J Neuropathol Exp Neurol 1994;53:303–15
9. Terry RD, Masliah E, Salmon DP, et al. Physical basis of cognitive
alterations in Alzheimer’s disease: Synapse loss in the major correlate of cognitive impairment. Ann Neurol 1991;30:572–80
10. Arriagada PV, Marzloff K, Hyman BT. Distribution of Alzheimertype pathologic changes in nondemented elderly individuals matches the pattern in Alzheimer’s disease. Neurology 1992;42:1681–88
11. Berg L, McKeel DW, Jr.,Miller JP, Baty J, Morris JC. Neuropathological indices of Alzheimer’s disease in demented and nondemented persons aged 80 years and older. Arch Neurol 1993;50:349–58
12. Berg L, McKeel DW, Jr.,Miller JP, et al. Clinicopathologic studies
in cognitively healthy aging and Alzheimer disease: Relation of
histologic markers to dementia severity, age, sex, and apolipoprotein E genotype. Arch Neurol 1998;55:326–35
J Neuropathol Exp Neurol, Vol 63, October, 2004
13. McKeel DW, Jr.,Ball MJ, Price JL, et al. Interlaboratory histopathologic assessment of Alzheimer neuropathology: Different methodologies yield comparable diagnostic results. Alzheimer Dis Assoc
Disord 1993;7:136–51
14. Morris JC, McKeel DW, Fulling K, Torack RM, Berg L. Validation
of clinical diagnostic criteria for Alzheimer’s disease. Ann Neurol
1988;24:17–22
15. Morris JC, McKeel DW, Jr., et al. Very mild Alzheimer’s disease:
Informant-based clinical, psychometric, and pathological distinction
from normal aging. Neurology 1991;41:469–78
16. Morris JC, Storandt M, McKeel DW, Jr., et al. Cerebral amyloid
deposition and diffuse plaques in ‘‘normal’’ aging: Evidence for
presymptomatic and very mild Alzheimer’s disease. Neurology
1996;46:707–19
17. Price JL, Davis PB, Morris JC, White DL. The distribution of tangles, plaques and related immunohistochemical markers in healthy
aging and Alzheimer’s disease. Neurobiol Aging 1991;12:295–312
18. Price JL, Morris JC. Tangles and plaques in nondemented aging and
‘‘preclinical’’ Alzheimer’s disease. Ann Neurol 1999;45:358–68
19. McKhann G, Drachman D, Folstein M, Katzman R, Price D, Stadlan EM. Clinical diagnosis of Alzheimer’s disease: Report of the
NINCDS-ADRDA Work Group under the auspices of Department
of Health and Human Services Task Force on Alzheimer’s disease.
Neurology 1984;34:939–44
20. Folstein MF, Folstein SE, McHugh PR. Mini-mental State: A practical method for grading the cognitive state of patients for the clinicians. J Psychiatr Res 1975;12:189–98
21. Katzman R, Brown T, Fuld P, Peck A, Schechter R, Schimmel H.
Validation of a short orientation-memory-concentration test of cognitive impairment. Am J Psychiatry 1983;140:734–39
22. Morris JC. Clinical Dementia Rating. Current version and scoring
rules. Neurology 1993;43:2412–14
23. Burke WJ, Miller JP, Rubin EH, et al. Reliability of the Washington University Clinical Dementia Rating. Arch Neurol 1988;45:
31–32
24. Morris JC, Ernesto C, Schafer K, et al. Clinical dementia rating
training and reliability in multicenter studies: The Alzheimer’s
Disease Cooperative study experience. Neurology 1997;48:1508–
10
25. Hibbard LS, McKeel DW, Jr. Multiscale detection and analysis of
the senile plaques of Alzheimer’s disease. IEEE Transactions on
Biomedical Engineering 1995;42:1218–25
26. Davis PB, White H, Price JL, McKeel DW, Robins LN. Retrospective postmortem dementia assessment: Validation of a new clinical
interview to assist neuropathologic study. Arch Neurol 1991;48:
613–17
27. Hellmich M, Abrams KR, Sutton AJ. Bayesian approaches to metaanalysis of ROC curves. Medical Decision Making 1999;19:252–64
28. Ikeda M, Ishigaki T, Yamauchi K. Review of the new methodology
of analyzing receiver operating characteristic curves. Medinfo
1998;9:837–40
29. McNeil BJ, Keeler E, Adelstein SJ. Primer on certain elements of
medical decision making. N Engl J Med 1975;293:211–15
30. Ogomori K, Kitamoto T, Tateishi J, Sato Y, Suetsugu M, Abe M.
ß-Protein amyloid is widely distributed in the central nervous system of patients with Alzheimer’s disease. Am J Pathol 1989;134:
48–54
31. Hibbard LS, McKeel DW, Jr.,Arnicar-Sulze TL, Burrell LD. Automated detection and morphological analysis of senile plaques in
SDAT. J Neurosci Methods 1994;52:175–89
32. Bierer LM, Hof PR, Purohit DP, et al. Neocortical neurofibrillary
tangles correlate with dementia severity in Alzheimer’s disease.
Arch Neurol 1995;52:81–88
33. Arriagada PV, Growdon JH, Hedley-Whyte ET, Hyman BT. Neurofibrillary tangles but not senile plaques parallel duration and severity of Alzheimer’s disease. Neurology 1992;42:631–39
NEUROPATHOLOGIC CRITERIA FOR DIAGNOSING AD
34. Novak M. Truncated tau protein as a new marker for Alzheimer’s
disease. Acta Virol 1994;38:173–89
35. Ball M, Braak H, Coleman P, et al. Consensus recommendations
for the postmortem diagnosis of Alzheimers disease. Neurobiol Aging 1997;18:S1–S2
36. Hyman BT. Editorial on consensus recommendations for the postmortem diagnosis of Alzheimer disease from the National Institute
on Aging and the Reagan Institute Working Group on diagnostic
1037
criteria for the neuropathological assessment of Alzheimer disease.
J Neuropathol Exp Neurol 1997;56:1095–97
37. Goldman WP, Storandt M, Miller JP, et al. Mild cognitive impairment represents early-stage Alzheimer’s disease. Arch Neurol 2001;
58:397–405
Received December 30, 2003
Revision received May 7, 2004
Accepted June 3, 2004
J Neuropathol Exp Neurol, Vol 63, October, 2004