Prevalence of dementia disorders in the oldest-old: an autopsy study
- First Online:
- Cite this article as:
- Jellinger, K.A. & Attems, J. Acta Neuropathol (2010) 119: 421. doi:10.1007/s00401-010-0654-5
- 883 Downloads
The prevalence of Alzheimer disease (AD) and vascular dementia (VD) increases with advancing age, but less so after age 90 years. A retrospective hospital-based study of the relative prevalence of different disorders was performed in 1,110 consecutive autopsy cases of demented elderly in Vienna, Austria (66% females, MMSE <20; mean age 83.3 ± 5.4 SD years). It assessed clinical, general autopsy data and neuropathology including immunohistochemistry. Neuropathologic diagnosis followed current consensus criteria. Four age groups (7–10th decade) were evaluated. In the total cohort AD pathology was seen in 82.9% (“pure” AD 42.9%; AD + other pathologies 39.9%), VD in 10.8% (mixed dementia, MIX, i.e. AD + vascular encephalopathy in 5.5%); other disorders in 5.7%, and negative pathology in 0.8%. The relative prevalence of AD increased from age 60 to 89 years and decreased slightly after age 90+, while “pure” VD diagnosed in the presence of vascular encephalopathy of different types with low neuritic AD pathology (Braak stages <3; mean 1.2–1.6) decreased progressively from age 60 to 90+; 85–95% of these patients had histories of diabetes, morphologic signs of hypertension, 65% myocardial infarction/cardiac decompensation, and 75% a history of stroke(s). Morphologic subtypes, subcortical arteriosclerotic (the most frequent), multi-infarct encephalopathy, and strategic infarct dementia showed no age-related differences. The relative prevalence of AD + Lewy pathology remained fairly constant with increasing age. Mixed dementia and AD with minor cerebrovascular lesions increased significantly with age, while other dementias decreased. This retrospective study using strict morphologic criteria confirmed increased prevalence of AD with age, but mild decline at age 90+, and progressive decline of VD, while AD + vascular pathologies including MIX showed considerable age-related increase, confirming that mixed pathologies account for most dementia cases in very old persons. A prospective clinicopathologic study in oldest-old subjects showed a significant increase in both AD and cerebral amyloid angiopathy (CAA), but decrease in VD over age 85, while in a small group of old subjects CAA without considerable AD pathology may be an independent risk factor for cognitive decline.
KeywordsDementing disorders Age-related prevalence Alzheimer disease Vascular dementia Mixed dementia Cerebral amyloid angiopathy Dementia in oldest-olds
Both the prevalence and incidence of dementia exponentially increase with age from 65 to 85 years [1, 35, 75, 76, 80, 81, 84, 101]. However, whether the increase continues after age 90, the fastest growing segment of the population estimated to account for about 12% of the population over age 65 , is unclear . The prevalence of dementia of all causes increases between the seventh and tenth decade from 0.8 to 28.5%, that of Alzheimer disease (AD) (53.7% of the total) from 0.6 to 22.2% and of vascular dementia (VD) (15.8%) from 0.3 to 5.2%  or from 0.2 to 16% over age 80 , while other epidemiologic studies in nonagenarians showed prevalence rates ranging from 48 to 74% [30, 44, 74, 122]. The prevalence of dementias increased from 0.3 to 1% in individuals aged 60–64 years to 42.3–68.3% in those aged 95+, the incidence rates from 0.8 to 4.0/1,000 person/years (p/y) in people aged 60–64 years to 49.8–135.7/1,000 p/y in those elder than 95 years . The incidence rate for dementias of all causes increased from 11.4% per year at age 90–95 to 21.8 per year over age 95 , the prevalence for all dementias over age 85 ranged from 15 to 40% and their incidence between 60 and 100 p/y , being 28% in men and 45% in women, with doubling every 5 years for women, but not men . In the USA, dementia prevalence increased with age from 5% in the eighth decade to 37.4% over age 95 years . The prevalence of VD increased from the seventh to the ninth decade from 0.2 to 16%  or 3.9 to 19.1% , with a double incidence at age 80+ in Afro-Americans as compared to Caucasians . In Canada, the prevalence of dementia ranged from 2.4% for age 65–70 years to 34.5% among those elder than 85 years , and clinically diagnosed AD versus VD increased significantly after age 65 and was higher in the 85+ group than in the younger cohort . In China, the prevalence of all dementias between the seventh and tenth decade increased from 2.5 to 34%. The prevalence for ages 65+ was 4.8% for AD and 1.1% for VD, with age-related increase in AD from 0.5 to 1 to 35%, while the prevalence of VD, after an increase in up to 4% in the ninth decade rapidly decreased after age 90 years [132, 133].
Types of dementia in oldest-olds: clinical cases
Mean age (years)
Percentage of the total cohort
Skoog et al. 
Ebly et al. 
Giannakopoulos et al. 
Rahkonen et al. 
Fitzpatrick et al. 
a: Caucasian, b: Afro-American
Bermejo-Pareja et al. 
Zekry et al. 
Brayne et al. 
Bufill et al. 
Lopez Mongil et al. 
Kobayashi et al.  (all cases)
79.3 ± 6.9
6.3 (FTLD 3.0)
Kobayashi et al.  (age groups)
10.7 (FTLD 6.7)
6.0 (FTLD 1.5)
5.2 (FTLD 3.9)
Types of dementia in oldest-olds: autopsies
Mean age (years)
Percentage of total autopsy cases of each cohort
Mizutani and Shimada 
Seno et al. 
83.6 ± 8.2
Akatsu et al. 
80.7 ± 5.6
White et al. 
Bachetta et al. 
Schneider et al. 
Grinberg et al. 
83.3 ± 5.6
Schneider et al. 
89.8 ± 5.5
79.2 ± 10
Schneider et al. 
89.7 ± 56
Brayne et al. 
Jellinger (present study)
83.3 ± 5.6
The aim of the present hospital-based study was to examine retrospectively the relative prevalence and relevant pathology of dementia of different causes in several old age groups to address some of the unanswered questions about AD and dementia in the oldest-old. The relationship between neuritic AD pathology, volume and distribution of cerebrovascular lesions, and a regression analysis on the effects of age and confounding variables (age, gender, educational level, etc.) were out of the scope of this study. A prospective clinicopathologic study concerned the frequency and severity of cerebral amyloid angiopathy (CAA) in old age and its relationship to AD pathology.
Materials and methods
Classification of cerebrovascular pathology (from )
0: no concomitant cerebrovascular lesions (CVLs)
1+: minimal CVLs
One to two small lacunes, mild-moderate CAA, with/without mild CVLs, mild leukoencephalopathy (LE)
2+: moderate CVLs
>2 lacunes, severe subcortical lacunar state, severe CAA with moderate CVLs, diffuse WMLs
3+: severe CVLs
a Old infarcts, multiple old microinfarcts or hemorrhages, hippocampal sclerosis
b Acute ischemic infarcts, hemorrhages (<1–2 days old)
In 400 demented patients aged 65–101 (mean 84.7 ± 5.8/SD) years (MMSE <20) of the above autopsy series, the occurrence and severity of CAA were examined dividing them into two age groups (60–84 years, n = 185; 85+ years, n = 215).
We also performed a prospective clinicopathologic study of 220 patients in a chronic hospital who had been followed-up to 7 years, with clinical, neuropsychological, and neuroimaging examinations every 6–12 months, the last ones up to 6 months prior to death (mean age at death 88.6 ± 4.5, range 85–103 years, 72% females, 82.8% demented at final examination; mean MMSE 12 ± 7). The patients came from the Prospective Vienna Dementia Study [10, 60] and were not included in the large retrospective autopsy study. In both series, in addition to routine histology/immunohistochemistry and diagnostic evaluation according to current criteria, CAA was examined in several cortical areas (frontal, temporal, parietal, and occipital) using immunohistochemistry for Aβ (clone 4G8), with semiquantitative assessment of its severity according to four degrees .
Statistical methods were the χ2 test for comparing the different types of lesions and grades of CAA in the different age groups.
The group of AD cases without further essential pathologies included a number of atypical cases: (1) tangle dominant dementia (TDD) (n = 36 or 3.25% of the total dementia cohort), predominantly occurring in the ninth and tenth decade, associated with neurofibrillary tangles in the allocortical areas (neuritic Braak stages 3–4) with no neuritic plaques, only few diffuse plaques and occasional CAA [55, 57, 63, 68]. (2) The “plaque only” type of AD (n = 15 or 1.4% of the dementia cohort), showing multiple cortical Aβ plaques and neurofibrillary tangles confined to the allocortex or hippocampus (neuritic Braak stages 3–4) without associated Lewy pathology, also predominant in very old age .
Among the group of AD + CVLs, hippocampal sclerosis was seen only at age 80+, accounting for 3.6% of AD cases in this age group. Few cases of hippocampal sclerosis associated with VD and MD were also observed after age 80, representing about 4% of these diagnostic entities (data not shown).
The different subtypes of MIX were AD + MIE (51%), AD + SAE (46%) and AD + SID (3%), the first two combinations most frequently occurring at age 80+. Their frequency ranged from 2.5 to 4% of the total number of demented cases in these age groups, AD + MIE being most frequent at age 90+ (data not shown).
General autopsy findings in VD cases showed signs of hypertension (hypertrophy of the left ventricle, nephrosclerosis, severe generalized and cerebral atherosclerosis, etc.) in 92%, history of diabetes and related morphologic lesions in 61.3%, myocardial infarction in 52%, and cardiac decompensation in 58.4%. The relations of general pathologic findings to the major subtypes of VD showed only mild differences: Both MIE and SAE showed morphologic findings of hypertension (nephrosclerosis, etc.) in 95%, severe coronary atherosclerosis in 93 to 100%, acute or old myocardial infarction in 78% each, severe aortic atherosclerosis in SAE in 100% and in 82% in MIE. The only difference between the two groups was severe atherosclerosis of the basal cerebral arteries which was significantly more frequent in MIE (82%) than in SAE (54%; P < 0.01). Furthermore, in SID there was less severe frequency of myocardial infarction (60%) and of severe atherosclerosis of large cerebral arteries (66%).
General autopsy findings in AD + minor CVLs, mainly those with minor to moderate lacunar state, showed signs of hypertension in 97%, coronary sclerosis in 100%, severe aortic sclerosis in 96%, acute myocardial infarction in 55%, severe cerebral atherosclerosis (LVD) in 60%, and cerebral SVD in 97%. CAA grade 3 to 4+ was present in 40%, CAA 1 to 2+ in 50%, and was absent in 10%. It was not always correlated with AD pathology (neuritic Braak stage). General autopsy findings in AD were not assessed.
Retrospective autopsy data in 400 demented patients (material: 400 demented patients, age 63–101 years)
Age 60–84 (n = 185)
Age 85+ (n = 215)
Prospective clinicopathologic study of subjects aged 85+ [material n = 220 (age 85–103, 72% female), 82.8% demented]
Pathologic diagnoses of non-demented subjects (n = 38)
CAA 67.3% (46% female)
Clinically demented (n = 182)
75% (48% grades 3–4+)
Clinically non-demented (n = 38)
20% (9% grades 3–4+)
Among 148 CAA+
Among 71 CAA−
The incidence of dementia ranges from 0.42 to 2.65% , with doubling every 5 years, particularly in women . In 2005, its prevalence in Europe was about 1.2% of the total population and until 2050 will increase to 3.3% of the total and 10.3% of the population 65 years . The prevalence of dementia increases to 28–56% in centenarians [5, 74], with no further increase after age 95+ [71, 81]. However, the estimates after age 85 are imprecise and inconsistent because of rather small patient numbers in the studies of these age group [20, 54, 71, 96]. Both the prevalence and incidence of VD increase with age, often less severe than AD [30, 114, 124], and may slightly decrease after age 95 . However, these data must be interpreted cautiously because of the lack of definite and uniform clinical and pathologic diagnostic criteria [see 62, 67], the high frequency of mixed pathologies in very old subjects (see Table 2) and occasionally only moderate association between the burden of neuropathology and dementia [105, 112, 124]. Some evidence suggests that this relationship becomes even less clear with increasing age [49, 104].
The present large retrospective, hospital-based autopsy study of demented old subjects (90% over age 70) revealed Alzheimer-like pathology in 84%, but “pure” AD without other essential concomitant lesions (CVLs, Lewy pathology, etc.) in only 42.9%. Only about 38% represented classical “plaque and tangle” AD , whereas 4.65% were atypical forms. They included TDD (3.25%), associated with neurofibrillary tangles in the allocortical areas with no neuritic plaques, only few cortical diffuse plaques and occasional CAA. Since both 3R- and 4R-tau are present in the tangles as in the common form of AD, despite differences in ApoE genotyping , TDD is suggested to represent a subtype of AD [57, 63, 68]. Less frequent was the “plaque only” type of AD (1.4% of the total dementia cohort), showing multiple cortical amyloid plaques and neuritic pathology confined to the hippocampus (neuritic Braak stages 3–4, Thal amyloid phase I). Frequently showing cooccurrence of cortical Lewy bodies, it can represent a subgroup of dementia with Lewy bodies (DLB) [85, 120], but in the present cohort only cases without additional Lewy pathology in cortex and amygdala were classified as “plaques only” type . The incidence of both subtypes in various autopsy series ranged from 0 to 10% for TDD [57, 63] and up to 25% in one series for “plaque only” type . As in our study, both subtypes mainly occur in very old people, preferentially in the ninth and tenth decade, whereas they are extremely rare in younger age groups [57, 63, 68, 96]. The reasons for their preferential occurrence in very old subjects is unknown and may not be explained by the frequency of ApoE genotypes (less frequent ApoE ε4 than ApoE ε2) alone, that differ from those in classical AD and are comparable with very old non-demented controls  and AD in nonagenarians .
“Pure” VD (with very low neuritic Braak stages <3, and no other concomitant pathologies) was seen in 10.8% of the present cohort of demented elderly persons. As in other series, its relative prevalence decreased from 15.5% in the seventh decade to 9.0% in the age group 95+, while the prevalence of AD, AD + minor CVLs and MIX significantly increased with age. The general pathologic findings, in particular cardiovascular pathology, in AD with minor CVLs were similar to those in MIX, which showed less severe atherosclerosis of large cerebral vessels than VD, whereas the other cardiovascular lesions were similar.
The contribution of small CVLs with a volume of <10 ml to dementia in the presence of Alzheimer type lesions is under discussion. In elderly patients with subclinical or mild AD and little functional brain reserve (with frequent entorhinal tangles and moderate cortical plaques), either critically located small CVLs, like subcortical lacunes, subcortical and/or cortical microinfarcts due to cerebral hypoperfusion, microangiopathy, or microhemorrhages, may worsen cognitive impairment due to a synergistic interaction of both pathologies [22, 31, 32, 86, 105], while in advanced or full-blown stages of AD concomitant small vascular lesions do not significantly influence the overall state and progress of cognitive decline that is mainly related to the severity and extent of AD pathology overwhelming the effects of cerebrovascular pathology [22, 56, 58, 59, 79]. Recent studies indicate that small subcortical infarcts add to deleterious effects of AD pathology by increasing odds of dementia and lowering memory functions . However, other studies of AD cases with and without CVLs of different extent and location showed no significant differences in MMSE scores between groups and no influence of the extent of vascular pathology on the age of death, severity of AD pathology, or the prevalence of hypertension and myocardial infarcts was observed [26, 62]. In general, patients with CVLs, even comparatively “minor” ones, show significantly lower densities of plaques and tau pathology than “pure” AD cases for every given level of cognitive deficit [12, 129], whereas one study showed significant association between CVLs and neuritic plaques but not tangles . Comparison of 186 AD patients with 13 individuals without cognitive impairment showed that in patients with infarcts, hemorrhages or Parkinson disease tangle and plaque densities were lower despite the presence of dementia . These findings are consistent with those in the NUN study, where patients with autopsy-confirmed AD and CVLs had a higher prevalence of dementia than those without infarcts. The risk of being demented was 20-fold higher in subjects with AD and associated lacunar infarcts, but much lower (sixfold) and statistically non-significant when a large territorial infarct was present . The experience of the CERAD group has corroborated these findings , and in the Optima study, CVLs significantly worsened cognitive impairment in earlier stages of AD . However, others observed no major differences in neurodegenerative lesion load between AD and AD + CVLs, except when these are located in the temporal lobe and hippocampus, suggesting that this location may be important in the pathophysiology of both VD and MIX [27, 103].
A study of almost 600 autopsy cases with dementia revealed significantly lower Braak stages (1–1.8) and higher final MMSE scores (4.6–13) in different types of VD (SAE, MIE, SID) than in “pure” AD (mean Braak stage 5.2; mean final MMSE score 2.1) and in AD with lacunes, small old cortico-subcortical infarcts or hippocampal sclerosis (mean Braak stages 4.6–5.2, mean final MMSE score 5.6–8.8). Mixed dementia showed Braak stages (mean 4.5) and cognitive impairment (final mean MMSE 4.0) similar to AD with small CVLs (mean Braak stage 4.7; final mean MMSE 56.8), but a much more frequent history of stroke [63, 64]. The pathologic substrates of MIX have been reviewed recently [62, 63, 64], and the thresholds for vascular and degenerative lesions for distinguishing “pure” VD from mixed cases have been critically discussed [43, 54].
Among the morphological subtypes of VD, SAE was the most frequent (49%), followed by MIE and SID, and very few cases featured severe CAA without essential neuritic AD lesions (mean Braak stage 1.9), associated with minor CVLs. Therefore, this small group was classified as rare subtype of VD (3.6%). These subgroups of VD showed no essential age-related differences. 92% of confirmed VD cases had hypertension-related cardiovascular pathology, 75% a history of one or several strokes, 52–58% myocardial infarction and/or cardiac decompensation, and 81% of VD patients had a history of previous stroke(s). Almost all of them had cerebral microangiopathy and severe atherosclerosis of the large cerebral vessels. Rockwood et al.  found that 76% of patients with VD had a history of stroke, compared to those with a diagnosis of AD, while among 600 consecutive autopsy cases, we found a history of stroke(s) in 90–98% of cases with “pure” VD compared to 9.8% of those with “pure” AD, while AD cases with subcortical lacunes and those with small old infarcts under 10 ml volume had a history of stroke in 21 and 33%, respectively, whereas cases of MIX, SAE, MID, and SIC had histories of previous stroke(s) in 95 to 100% [63, 64]. Among 175 autopsy cases of VD (mean age 80 ± 8.5 years, 61% averaging 80 years), LVD pathology was seen in 28.6% (pure 4%), SMD in 62.8% (pure 20.5%), hypoxic–ischemic encephalopathy in 31.4% (pure 3.4%), while 72% showed several types of brain pathology, including almost 60% with AD pathology. Myocardial infarction was observed in 45%, coronary and aortic atherosclerosis in 93% .
The present study and other data confirm the existence of “pure” VD in the age group over 70 years, however, with significant age-related decrease in frequency, whereas AD, AD + minor cerebrovascular pathology and MIX showed a significant age-related increase, confirming the presence of mixed cerebral pathologies in the oldest-old [19, 93, 100, 105, 124].
CAA prevalence is consistently higher in demented when compared with non-demented subjects. This supports a significant role for CAA in the pathogenesis of dementia. The prevalence of CAA increases with age and has been reported to be 42.8% in women and 45.8% in men over age 90 in a population-based study . According to population-based studies on average 55–59% of demented patients showed CAA as compared to 28–38% in the non-demented, which was lower than in selected sample studies (demented vs. non-demented 32–100 vs. 0–77% regardless of severity; 0–50% vs. 0–11% for severe CAA) . In autopsy studies of older patients in Japanese geriatric hospitals, its prevalence was 70%  and in a small 95 + cohort CAA was found in 48.6% of the individuals including 8.3% severe forms, most often frontally and associated with small cortical infarcts . Often silent ischemic infarcts are associated with hemorrhagic burden in CAA , whereas spontaneous (non-traumatic) intracranial hemorrhages (excluding microbleeds) have been observed in 2–23% of cases with CAA, but in a large autopsy series were seen in only 5.6% of the total and in 7.2% of definite AD cases . Frequent old microinfarcts in elderly individuals with severe CAA may contribute a vascular component to the cognitive impairment in these patients .
A retrospective evaluation of 400 autopsies of very old demented subjects of our total group showed significant increase of both definite AD and CAA (general occurrence and severe degrees) between the age groups 60–84 and 85+, with significant decrease in VD and less of other entities after age 85, confirming the results of the above personal and other previous studies.
The suggestion that CAA may be an independent risk factor for cognitive decline in old age [7, 45, 114, 116, 130] was strengthened by a prospective clinicopathologic study of patients over age 85 (82.8% demented) that showed CAA in a total of 67.3%, 42.6% of which severe. Like in the VANTA 85+ study, clinically demented patients showed significantly more frequent and more severe CAA than non-demented ones (75 vs. 20%), and the majority of CAA+ cases had definite AD, although almost in 10% of demented subjects CAA was not associated with essential AD-like lesions. These data suggest that CAA may be an independent risk factor for cognitive decline in old age [7, 45, 114, 116, 130], whereas others did not find significant differences in the prevalence of CAA between demented and non-demented subjects . The former notion is strengthened by the fact that in a small subgroup of demented oldest-olds severe CAA predominantly in the frontal and/or occipital cortex may not be associated with enough severe Alzheimer-like pathology to make the diagnosis of probable or definite AD , while the association of CAA with minor CVLs, in particular cortical microinfarcts/microbleeds and lacunes [21, 117] suggest that it may represent a rare subgroup of VD, which, however, needs further confirmation.
Highly variable combinations of lesion types can lead to indistinguishable clinical conditions, namely in very old subjects. Comorbid pathologies of different etiologies often related to general, in particular cardiovascular changes, frequently converge to produce what appears to be a unitary disease entity, such as AD or VD that often may concur in various forms and intensities. The problem of classification of disorders in very old patients is further complicated by the fact that a high percentage of demented individuals aged 80+ do not meet the pathologic diagnostic criteria of AD, VD, or DLB . This is exemplified by the findings among 180 cases of the Vienna Prospective Dementia Study, in which 24% of the patients had been demented (age at death 85 ± 3.4 years). Autopsy revealed AD in 48%, MD in 19%, VD in 11%, DLB in 9%, whereas 18% did not meet the pathologic criteria of one of these groups (Jellinger, unpublished data). These and other problems represent a big challenge for future research in order to get a better understanding of the etiology and pathogenesis of dementia in the final decades of life. The problem for less association between neuropathology and cognitive state in the oldest-old compared with the young old has been commented recently [49, 54, 108, 124], but the biological background of the increased resistance to AD lesion development in some subjects after age 90 and the impact of microvascular pathology on the development of cognitive impairment in very old people with less severe Alzheimer pathology are still poorly understood . New research suggests the neuropathologic dissociation between those with and without dementia narrows as age increases [19, 91, 92, 104, 110], but these and other problems of the oldest-old need further elucidation by future well designed and controlled, possibly population-based clinicopathologic studies.
The authors thank many colleagues from clinical departments and the Institute of Pathology, Otto Wagner Hospital, Vienna for clinical and autopsy data, and Mr. E. Mitter-Ferstl, PhD for secretarial and computer work. The study was supported by the Society for Support of Research in Experimental Neurology, Vienna, Austria.
Note added in proof
After termination of the present study, it was extended to a total of 1,700 demented elderly persons (671 male, 1,029 female, mean age at death 84.3 ± 6.0 years; autopsies 1981–2008), i.e. adding 590 autopsies, also divided into four age groups. The results were as follows: AD pathology was present in 80%, “pure” AD in 45.7% (36.9% Braak stage 5–6, 8.8% atypical forms), AD and other pathologies in 37.3%, VD in 12.8%, MIX in 5.2%, other degenerative dementias 4.1%, other disorders 1.6%, and negative pathology in 1.2%. The relative prevalence of AD increased from age 60 to 89 years from 38.9 to 48.9% and slightly decreased after age 90 (47.3%); MIX and AD + CVLs increased significantly from 5.2 to 10.6% and from 4.4 to 23.2% respectively (P < 0.001), while VD decreased from 15.0 to 8.7%. AD + LB pathology only slightly increased with age (7.8 to 9.4%), whereas other neurodegenerative disorders and other dementing diseases progressively decreased (21.0 to 0.4% and 5.0 to 0%, respectively). These data correlated well with those of the smaller autopsy series.